A quantitative study of space charge solute segregation at grain boundaries in TiO, is conducted, using a new STEM method for the measurement of aliovalent solute accumulation. It is shown that the electrostatic potential at grain boundaries can be varied in sign and magnitude with doping, oxygen pressure, and temperature, and that the isoelectric point lies in slightly donor-doped compositions for samples annealed in air. The experimental results closely fit the space charge model in Part I. Space charge solute segregation is found even in defect regimes of high electron concentration. Approximately one in ten grain boundaries are "special" in exhibiting no detectable segregation; in one such instance a twin boundary is identified. Among boundaries with significant amounts of segregation, clear differences in potential also exist. From the potential determined in acceptor-and donor-doped compositions, the Frenkel energy (assumed to be lower than the Schottky energy in TiO,) can be separated into its individual terms. An average value for the titanium vacancy formation energy of gVn = 2.4 eV and an upper limit to the titanium interstitial formation energy of g , = 2.6 eV are obtained.while the solubility of Nb,O, is reported to be 0.7-3 m~l%'-~ in the temperature range used in this study.Another major distinction between this and previous studies is the use of a highly quantitative method for measuring aliovalent solute segregation. We have developed a scanning transmission electron microscope (STEM) method for accurately measuring the total accumulation of solute at an interface. From the net amount of acceptor or donor segregation as a function of composition and temperature, the relationship between the electrostatic potential and the lattice defect structure has been systematically studied. These results are reported here, and are quantitatively compared with the pxedictions of the model in Part I. Experimental Procedure ( I ) Sample PreparationPowders were coprecipitated from aqueous solutions of TiCI, (Johnson Matthey 99.999%) to which NbCl, (Puratronic 99.999%), AIC13.6H,0 (Puratronic 99.9995%), or Ga(NO,), (Puratronic 99.999%) was added in the desired concentrations. Polyethylene or Teflon labware and 18-MCl deionized water was used throughout powder processing to minimize impurities. A number of additional precautions, discussed in greater detail in Ref. 6, were necessary because of the high volatility of TiCI,, its highly exothermic reaction with water, and the hygroscopic and/or reactive nature of the dopant salts. The basic process' consists of first preparing an aqueous solution of TiCI, by mixing the chloride with water and ice. The dopant salts were dissolved into this solution and homogenized by stirring. To this colorless, precipitate-free solution, a 1 : 1 solution of NH,OH:H,O was slowly added while stimng to precipitate the metal hydroxides, reaching a solution pH of about 10. The precipitate was allowed to digest while stirring for at least 1 h, and then thoroughly washed by repeated dilution...

The electrical potential difference between the interface and the bulk in TiO, is obtained as a function of temperature, oxygen pressure, and acceptor or donor doping from a space charge model that explicitly includes the high-temperature lattice defect chemistry. Using defect equilibrium constants for TiO, from previous literature studies, it is shown that for a space charge determined by ionic defect equilibration with the interface, the potential is negative in undoped and acceptor-doped TiO, and positive at high donor concentrations. The isoelectric point lies in the donor-doped regime at high temperatures due to the contribution of defects from reduction, even for fairly oxidizing ambients.

During the oxidation of redu ed single crystals of Al-doped rutile, unusual anisotropies in color-boundary migration have been observed that are opposite to those predicted from published diffusion data. Analysis of the redox kinetics and of surface segregation (using ESCA and AES) shows that rapid transport of minority Al interstitials in the c-axis direction occurs under an oxidation potential, resulting in a surface segregation layer inhibiting further reoxidation. This surface segregation is nonequilibrium in nature, is driven by oxidation, and bears similarities to the phenomena of kinetic demixing in ionic systems. The results show that minority defects can play critical roles in demixing at the local scale; in their absence this system would not be expected to demix. This thus appears to be an additional mechanism for nonequilibrium interfacial segregation in ionic systems. [