CrO 2 has the highest measured spin transport polarization (P), yet produces low values of magnetoresistance (MR) in spin valves. This work tests the limits of CrO2 surface stability as a culprit of spin loss in magnetoresistive devices. Epitaxial CrO2 films are grown on TiO2 single-crystal substrates from Cr8O21 precursor in flowing oxygen and transferred directly to UHV without exposure to air. The surface structure is observed by low-energy electron diffraction as a function of temperature. The main result is that it is possible under specific conditions to stabilize the CrO2 (001) and (110) surfaces for growth of multilayer devices. Under UHV conditions CrO2 (110) is stable to ∼99 °C while (001) is stable to ∼150 °C with evidence of faceting. In some cases well ordered CrO2 surface can be recovered by annealing in oxygen.

Presently, the best epitaxial thin films of CrO2 are made by chemical-vapor deposition (CVD) in a two-zone furnace with oxygen flow from a CrO3 precursor. The growth mode has previously been described as CrO3 vaporizing in the first zone, and thermally decomposing at higher temperature in the second zone onto a substrate. In the more recent works, the focus has been on the properties of the obtained layers rather than on deposition mechanisms. In the present experimental work, we attack the epitaxial growth of CrO2 by two completely different methods, namely, molecular-beam epitaxy (MBE) and CVD. We focus on the CVD process itself, and show the importance of an intermediate compound, Cr8O21, for the growth of CrO2 films. We show that it is not necessary to start the CVD from CrO3; instead, one can prepare Cr8O21 ex situ, and use it directly for the growth of high-quality CrO2 epitaxial layers, avoiding any contamination caused by the decomposition of CrO3 to Cr8O21. We discuss in parallel our failed attempts to deposit CrO2 from either CrO3 or Cr and oxygen plasma by MBE and our experiments with the CVD process, and conclude that CrO3 does not decompose directly to CrO2 and oxygen, as was expected. We propose a hypothesis that the role of Cr8O21 in the CVD process is to exude unstable molecules of CrO4, and that the reaction on the substrate is the decomposition CrO4→CrO2+O2.

Neutron diffraction studies of Fe3O4/NiO superlattices reveal that the field dependence of domains in the antiferromagnetic NiO is correlated with the presence or absence of exchange biasing. Measurements of the full width at half maximum of the (111) NiO reflection show that after cooling in zero field, the antiferromagnetic domain size both parallel and perpendicular to the growth axis is sensitive to the strength of the applied field. In contrast, after cooling from room temperature in a field of 5 T, the domain size is generally smaller than the zero-field-cooled value and does not vary with field. These data suggest that exchange biasing originates from domain walls frozen into the antiferromagnet upon field cooling.

In order to understand the interplay between exchange coupling and magnetic structure, we have examined the magnetic ordering of a series of epitaxial Fe3O4/NiO superlattices using polarized neutron diffraction techniques. As expected, the net ferrimagnetic moment of the Fe3O4 layers aligns parallel to an applied magnetic field. The antiferromagnetic NiO spins order into alternating antiparallel 〈111〉 planes as in bulk, but the direction of the spins in the planes are determined by field preparation. The NiO moments tend to align perpendicular to the field. In addition, the relative population of the NiO domains varies as the field is raised. The changes in the antiferromagnetic spin order relative to bulk seem to result from magnetic coupling with the Fe3O4 moments.