Structural Effects of Lanthanide Dopants on Alumina

Lanthanide (Ln3+) doping in alumina has shown great promise for stabilizing and promoting desirable phase formation to achieve optimized physical and chemical properties. However, doping alumina with Ln elements is generally accompanied by formation of new phases (i.e. LnAlO3, Ln2O3), and therefore inclusion of Ln-doping mechanisms for phase stabilization of the alumina lattice is indispensable. In this study, Ln-doping (400 ppm) of the alumina lattice crucially delays the onset of phase transformation and enables phase population control, which is achieved without the formation of new phases. The delay in phase transition (θ → α), and alteration of powder morphology, particle dimensions, and composition ratios between α- and θ-alumina phases are studied using a combination of solid state nuclear magnetic resonance, electron microscopy, digital scanning calorimetry, and high resolution X-ray diffraction with refinement fitting. Loading alumina with a sparse concentration of Ln-dopants suggests that the dopants reside in the vacant octahedral locations within the alumina lattice, where complete conversion into the thermodynamically stable α-domain is shown in dysprosium (Dy)- and lutetium (Lu)-doped alumina. This study opens up the potential to control the structure and phase composition of Ln-doped alumina for emerging applications.


Results and Discussion
In this study, the effects of sparse Ln 3+ dopants [Lanthanum (La), Praseodymium (Pr), Dysprosium (Dy), Neodymium (Nd), Erbium (Er), Thulium (Tm), Lutetium (Lu), Gadolinium (Gd) and Ytterbium (Yb)] on the phase transition and structural geometry of alumina are reported. The concentrations of Ln-dopants were kept at a constant level of Ln 0.002 Al 1.998 O 3 (e.g. the weight concentration of Er-dopant was ~0.03271 wt% (or 400 ppm) in the alumina lattice) to avoid any potential formation of LnAlO 3 and/or Ln 2 O 3 phases. Furthermore, in light of the great importance of Ln-doped alumina and the challenges associated with the analysis of substituent species in crystalline Al 2 O 3 materials, five possible locations were proposed for the Ln-dopants, and experimental efforts were made to find the most chemically reasonable location. A suite of state-of-the-art structural probes was utilized to test the hypothesis that the local structure and positioning of Ln-ions could largely influence the structural transition and phase composition of doped alumina due to their interactions with the Al 2 O 3 host lattice. 1D magic angle spinning (MAS) and 2D 27 Al multiple quantum magic angle spinning (MQ-MAS) solid-state nuclear magnetic resonance (ss-NMR) spectroscopy techniques, together with high resolution X-ray diffraction (Synchrotron 11-BM at the Advanced Photon Source of Argonne National Laboratory), were employed to examine the Ln-dopant effect on Al 2 O 3 phase composition. The structural characteristics and phase evolution of Ln-doped Al 2 O 3 were further confirmed through scanning electron microscopy (SEM) and high temperature differential scanning calorimetry (HT-DSC). A wide range of different spectroscopic techniques was employed to enable a detailed understanding of the local structure of Ln 3+ ions in Al 2 O 3 and provide critical insight into the rational design of alumina materials.
All of the Ln-doped alumina samples referenced in this study were prepared by mixing an acidic solution of aluminum nitrates, magnesium nitrates, and lanthanide nitrates in stoichiometric amounts (as shown in Fig. 1a, targeting a composition of Ln 0.002 Al 1.998 O 3 with 250 ppm of Mg 2+ ) with a basic solution. The addition of Mg 2+ to the composition was believed to have an influence over structural distortions that could assist in dissolving Ln 3+ into the Al 3+ octahedral sites. There is significant literature detailing the role of MgO additives for enhancing the sintering behavior of alumina, suggesting that the MgO acts as a grain growth inhibitor, leading to a more uniform grain structure [27][28][29][30] . However, these studies examined significantly larger quantities of MgO than are reported here. After calcination of the resulting powder, Ln-doped alumina was obtained and further confirmed using a suite of advanced spectroscopy techniques. In addition, the location of the Ln dopants in alumina has been a topic of debate in the past and is still ongoing 21,[27][28][29][30] . Five possible locations were identified for the Ln-dopants to reside (Fig. 1b): (i) intercalating in the alumina lattice by substituting some aluminum cations, (ii) creating LnAlO 3 or Ln 2 O 3 phases which are entirely separate from the alumina phase, (iii) forming a shell of LnAlO 3 or Ln 2 O 3 around the alumina core particles, (iv) residing in the vacant octahedral or tetrahedral sites, or (v) residing on the grain boundaries of the alumina lattice 21,31,32 .
Substitution of larger atomic radius La atoms into the Al sites of the alumina lattice would have potentially resulted in a larger bond distance of La-O (2.3-2.8 Å) as compared to that of Al-O (1.8 Å). However, such distortions were not reflected in the spectroscopy studies (shown in a later section), and it seemed unlikely that the La atoms were substituted for the Al atoms in the alumina lattice, as suggested in scenario (i). Furthermore, scenarios (ii) and (iii) were ruled out due to the amount of dopant used in this study, as ~0.03271% (by weight) of the total molecular weight of Ln 0.002 Al 1.998 O 3 was not enough for the formation of separate phases or layers of LnAlO 3 /Ln 2 O 3 . This was also confirmed through spectroscopy studies in this report 8 . In addition, assuming a homogeneous lattice with sparse Ln cations well dispersed in the matrix, the distortion that occurred from one cation would not interfere with other dopants. Since a short and fast calcination schedule was used to process the material, there was not enough time for the cations to migrate towards one another and form a secondary phase.
To understand the location of the Ln-dopants and structural changes developed in Al 2 O 3 , Er-doped alumina was selected as the first prototype doping example. The Al 2 O 3 , Er-doped Al 2 O 3 , pure α -Al 2 O 3 , and pure θ -Al 2 O 3 samples were characterized using 1D 27 Al Magic Angle Spinning (MAS) ss-NMR with spin rate of 10 kHz (Fig. 2a). To isolate pure phases of α -Al 2 O 3 and θ -Al 2 O 3 , the calcination temperature was adjusted such that only α -and θ -Al 2 O 3 were detected by X-ray diffraction. The 27 Al chemical shift depended strongly on the number of coordinating oxygens 33 . The α -Al 2 O 3 was represented by hexagonal packing of oxygen anions. The distribution of Al cations for this phase was limited to octahedral sites, and was represented by a single peak at ~13.9 ppm in the MAS spectrum. The θ -Al 2 O 3 phase, on the other hand, was represented by monoclinic arrangements of the oxygen and distribution of Al cations in tetrahedral and octahedral sites. This arrangement led to two distinct inhomogeneously broadened peaks at ~59.3 and ~8.8 ppm for tetrahedral and octahedral sites, respectively. Additionally, there was a downfield shoulder at ~2.7 ppm on the peak at ~8.8 ppm for θ -Al 2 O 3 . The 1D 27 Al MAS ss-NMR spectrum of Al 2 O 3 represented a weighted summation of the spectra of pure α -and θ -phase Al 2 O 3 . Upon doping Al 2 O 3 with 400 ppm Er 3+ , the MAS spectrum changed relative to undoped Al 2 O 3 . Most notably, there was significant attenuation, which reduced the intensity of the peak at ~13.9 ppm.
Multi-dimensional ss-NMR was used to further explore these results. Since 27 Al is a quadrupolar nuclei (S = 5/2), the signals in 1D 27 Al ss-NMR spectra were broadened by quadrupolar coupling, as well as other anisotropic spin interactions, including chemical shift anisotropy and dipolar coupling 34,35 . The MQ-MAS experiment 36 provided a tool to discern these interactions. This multi-dimensional experiment correlated the spectrum without second order quadrupolar anisotropic effects in the indirect dimension (y-axis, F1) while the spectrum was broadened by these anisotropic effects in the direct dimension (x-axis, F2), making the MQ-MAS experiment useful for resolving overlapping peaks. The 2D 27 Al MQ-MAS spectra were collected on undoped and Er 3+ -doped Al 2 O 3 ( Fig. 2b and c). There were three clear peaks in the MQ-MAS spectrum of Al 2 O 3 at ~59.3, 13.9 and 8.8 ppm, respectively, in the F2 dimension. The latter peak included a downfield shoulder at ~2.7 ppm. The intensity of the peak at ~13.9 ppm in F2 in the MQ-MAS spectrum was reduced significantly upon doping with Er 3+ , which was in agreement with the results of the 1D MAS spectra. Taken together, these results suggested that, given the ~0.03271% weight percentage of dopant, changes in the NMR spectra were not attributed to Er 3+ induced changes to anisotropic spin interactions, but an alteration in alumina phase composition caused by the doping procedure (Fig. 1a).
To reveal the Er-dopant effect on the phase composition and structure of alumina, HT-DSC was utilized to investigate the phase transition of alumina before and after Er-doping. Despite limited solubility, Ln-dopants effectively altered the phase transition temperature of alumina, as more thermal energy was necessary to convert to the α -Al 2 O 3 phase, the most thermodynamically stable phase, after Er-doping 37,38 . The phase transition took place at a temperature (T c ) of ~1280 °C for the pristine Al 2 O 3 (Fig. 3a) and ~1325 °C for the Er-doped Al 2 O 3 , suggesting that doping alumina with Er increased the transition temperature significantly. SEM images of Er-doped Al 2 O 3 (Fig. 3b) calcined at 1300 °C, which was lower than the alpha alumina transition temperature for Er-doped Al 2 O 3 , showed a mixture of particle morphologies. Theta alumina has been known to have a smaller particle size than alpha alumina, as the grains will not grow until the final alpha alumina phase has been developed. It has been theorized that the larger grains represent alpha alumina and the smaller grains represent theta alumina. These observations further confirmed the NMR spectra of Er-doped Al 2 O 3 discussed above. Loading of large Er 3+ ions into alumina forced the lattice to expand, distorting the neighboring tetrahedral and octahedral locations. Consequently, the distorted lattice conversion into octahedral sites and rearrangement into stable structures for the phase transformation from θ -Al 2 O 3 to α -Al 2 O 3 required a higher thermal energy, leading to a phase transformation temperature shift.
In order to confirm the hypothesis that Ln-dopants altered the phase transition temperature in alumina, and to distinguish the doping mechanisms between scenarios (iv) and  (Fig. 4). The 2D MQ-MAS spectrum of Pr-doped Al 2 O 3 (Fig. 4) included three clear transitions at ~59.3, ~13.9 and ~8.8 ppm, respectively. The 2D MQ-MAS of La-and Yb-doped Al 2 O 3 also showed three transitions, albeit with weaker downfield peaks. The 2D MQ-MAS results for Tm-, Nd-, Lu-, Dy-, Gd-and Yb-doped Al 2 O 3 also exhibited two upfield peaks at various intensities ( Fig. 4c-h). Additionally, the 2D MQ-MAS further highlighted the presence of θ -Al 2 O 3 in various concentrations as a second minor component for Yb, Gd, La, Nd and Tm-doped Al 2 O 3 . The presence of the dopants in the lattice generated stress in the geometry of arranged alumina, leading to changes in the system configurations. It should be noted that Pr-doped alumina showed a significant increase in θ -domains (~59.3 ppm, Fig. 4b), whereas the Dy and Lu dopants showed no indication of θ -domains ( Fig. 4c and i). Since the ratio of Al to Ln was constant, it was assumed that an increase in one type of domain was followed by the decrease of another type of domain by the same amount. Whether or not the dopant favored formation of the θ -domain or α -domain at a given temperature was dictated by the size of the dopant and the size of the site it was being doped into.
The final confirmation of the presence of α -and θ -phases in various ratios for each Ln-dopant was determined using data obtained from HR-XRD. This quantitatively reinforced conclusions drawn based on the ss-NMR results for the Ln-doped alumina. HR-XRD spectra of each Ln-dopant indicated notable differences (Fig. 5a) in α -and θ -peaks, with Dy-and Lu-doped alumina exhibiting only α -phase, and Pr-doped alumina revealing the θ -phase in extremely high quantities. It should be noted that all of these doped Al 2 O 3 samples were prepared at the same calcination temperature of ~1300 °C for 30 minutes. As observed in the case of Er-doped alumina, the transition temperature of α -Al 2 O 3 increased to ~1325 °C. Therefore, the θ -Al 2 O 3 domains in the sample prepared at ~1300 °C were expected to be present, since the material was calcined at a temperature lower than the transition temperature for alpha alumina. These spectra indicated the presence of uneven phase percentages in each Ln-doped Al 2 O 3 sample. The graphical representations of phase populations for each dopant were generated using Rietveld refinement data from the HR-XRD (Fig. 5b), which were analyzed using X-ray refinement software, GSAS-II 39 . They provided a quantitative summary of the phases in each sample doped with Ln-elements. With the exception of Dy-and Lu-doped alumina, the θ -Al 2 O 3 phase was present in all of the other samples. This suggested that the Dy dopant decreased the α -θ phase transition temperature when compared to the pure alumina phase. The alpha phase transformation temperatures of Dy-and Lu-doped alumina were expected to be significantly lower than ~1300 °C, since a complete transformation of α -phase was observed. Based on this hypothesis, the presence of θ -domains in all other samples clearly suggested that the dopants increased the transition temperatures above ~1300 °C. This temperature requirement was much higher in the case of Pr-doped alumina, as a significantly high population of θ -Al 2 O 3 domains were observed after calcining at ~1300 °C. The Ln ionic size effect appeared to play a role in the transition temperature shift. There was a notable variation in the phase population of θ -and α -domains (Fig. 5c). It was possible that elements from the second half of the Lanthanide series (Gd, Dy, Tm, Lu, Yb), which were smaller in size due to Lanthanide contraction, could have led to a lower percentage of θ -phase when compared to elements from the first half of the series (La, Pr). The smaller dopant sizes could have resulted in less distortion of the neighboring lattice, leading to a lower energy requirement for atomic rearrangement. Furthermore, the Ln-dopant effects were purely intrinsic, and did not lead to formation of any undesirable LnAlO 3 or Ln 2 O 3 phases, which was confirmed by both ss-NMR and HR-XRD spectra. The refinement summary from HR-XRD also provided the geometry and dimensions of the domains, indicating that the ionic size effect of Ln-dopants (Fig. 5c, and Table S1 of Supporting Information) was responsible for delaying the phase transformation of θ -domains into α -domains. The theta alumina particle morphology was anisotropic (ellipsoid-like) whereas the alpha alumina particles were isotropic (cube or sphere-like). The SEM images of two representative Ln-doped alumina samples (Dy and Tm doped alumina, Fig. 5d) were consistent with the structural conclusions from the refinement summary, verifying that contrasting morphologies of the samples were clearly shown. The full SEM images for all Ln-doped alumina samples are shown in the Supporting Information. The Pr-doped alumina sample revealed spherical particles within the elliptical-rich matrix due to significantly large populations of theta alumina, while the Gd-doped alumina sample showed elliptical particles in the spherical-rich matrix, and the Dy-and Lu-doped alumina samples exhibited only spherical particles due to their pure α -phases. Similarly, the other samples were analyzed using SEM, providing rough estimates of their phases and phase populations, which aligned well with the high resolution X-ray diffraction analysis. These dopants changed the basic structure of the host lattice in such a way that the transition temperatures of the polymorphs required more energy to form. With all doped materials being processed under the same furnace conditions (1300 °C for 30 minutes), the materials developed different theta and alpha phase ratios. Since each phase of alumina exhibited a different geometry as dictated by its structure, and the presence of a certain percentage of those phases in the sample resulted in a unique geometry (as shown in the SEM images), it further suggested that not all particles formed alpha alumina during the calcination cycle (i.e. grains were of the theta polymorph). Dopants residing at the grain boundaries were unlikely to generate enough lattice distortions to demonstrate such a notable differences in the phase transformation temperatures. This eliminated scenario (v), leaving scenario (iv) as the only possible location for the Ln-dopants, indicating that they resided in the one-third vacant octahedral sites since the small dimensions of the vacant tetrahedral sites could not accommodate the large-sized Ln 3+ dopants.

Conclusions
In summary, Ln doping (400 ppm) in alumina had a significant influence over structural and phase evolution. Ln dopants served as structural promoters to increase the phase transformation temperature (θ → α ) by a notable magnitude, which delayed the onset of alumina lattice phase transformation. Ln doping in alumina enabled control of the phase population, with no additional unwanted phases (such as LnAlO 3 or Ln 2 O 3 ) observed. Additionally, it was determined that the Ln dopants resided in the vacant octahedral locations within the alumina