Despite the continued demand for Flash RAM memory for portable data storage and other applications, intensive research into the alternative magnetic RAM continues. MRAM offers the potential for faster write times than Flash memory and greater permanence.

Magnetic tunnel junctions form the elementary elements of MRAM with typical junctions comprising two ferromagnetic electrodes separated by an insulating layer acting as the tunnelling barrier. Magnesium oxide, MgO junctions promise better performance than those using alumina insulation, yet experimentally this has not been realised; partially due to a mismatch between the models and the reality of laying down monolayers of material, but it is also due to poor understanding of the boundary regions between the electrode and insulating layer.

Experiments on a series of MgO junctions with polarised neutron reflectivity [Laloe et al., Appl. Phys. Lett. (2008) 93 012505] has shown that magnetically dead layers form at the interfaces and that the quality of the layers is critical to the efficient operation of these devices.

A series of samples of Fe/MgO and Co/MgO layers was made with molecular beam epitaxy. Polarised neutron reflectivity was used to measure the magnetization, magnetic moment and atomic density for each layer.

Magnetically dead layers between 1 and 3 monolayers formed at each interface arising from diffusion and the formation of compound layers such as FeO. Evaporation from a MgO source crystal was found to yield a tunnel layer with better stoichiometry and epitaxy than that obtained by evaporating magnesium in an oxygen atmosphere.

To achieve spin-coherent tunnelling of electrons through the oxide barrier clearly requires excellent interfaces and minimal intermixing between the layers of the tunnel junctions. This is certainly a challenge for large volume MRAM production.