Probing spin polarization : point contacts and tunnel junctions

The topic of this thesis is the unbalance of the electron spin in currents, a parameter known as spin polarization. Spin polarization is the most important parameter within the field currently known as spintronics, and is responsible for the high magnetic field-sensitivity of modern magnetoresistive devices. In this thesis two different methods are used to measure and study spin polarization. These two methods are point contact Andreev reflection spectroscopy and spin-polarized tunneling. In both methods a superconductor serves as a detector for the spin polarization. When a bias voltage forces electrons to flow from a normal metal (N) into a superconductor (S), the magnitude of the current is dependent on the spin polarization. This offers the opportunity to deduce the degree of spin polarization directly from the measurement of the current, or, more precisely, from the measurement of the conductance. N and S can be in metallic contact, as is the case in point contact Andreev reflection spectroscopy, or they can be separated by an ultra-thin insulator that serves as a tunnel barrier, as is the case in spin-polarized tunneling. Since the electron transport in N/S point contacts and N/barrier/S tunnel junctions have a different nature, different physical mechanisms are responsible for the spin-dependence of the conductance. In point contacts, the spin-dependence originates from the Andreev reflection process. This is a conduction process in which two electrons with opposite spin are paired into a Cooper pair. In tunnel junctions, the spin-dependence is due to the Zeeman splitting of the spin-up and spin-down state energy in the superconductor induced by an externally applied magnetic field. Point contacts Point contact Andreev reflection spectroscopy is the topic of part I of this thesis. The point contacts are obtained at low temperature by pressing a superconducting tip onto a sample of the normal metal. The conductance-voltage relation measured in case of a highly transparent N/S interface and a high polarization of the current, looks very similar to the one measured when the interface has a relatively low transparency. In chapter 3 it is pointed out that this similarity can potentially lead to a misinterpretation of the experimental data. Specifically, the suppression of the zero-bias conductance due to a low interface transparency may be incorrectly interpreted as proof for a highly spin-polarized current. Anomalous conductance minima at finite bias voltage observed for some contacts can be explained by a phase transition of the contact from the N/S state to the N/N state induced by the current-generated field and heating. For a series of contacts it is observed that the bias voltage at which the phase transition occurs, is systematically dependent on the contact resistance. In chapter 5 it is shown that a simple analysis of such a series can give insight in the microscopic contact geometry. It is shown in chapter 6 that the spin polarization measured with point contact Andreev reflection spectroscopy, decays exponentially with the amount of scattering at the N/S interface. This decay is fully explained by a simple model involving spin-flip scattering in an extended interface region. Tunnel junctions The topic of part II of this thesis is the spin polarization of the tunnel current in S/barrier/N tunnel junctions. The polarization is measured by the well-established spinpolarized tunneling technique, also known as superconducting tunneling spectroscopy. The junctions are prepared by a combination of room-temperature sputter-deposition and plasma oxidation. From the results presented in chapter 10, it is apparent that it is relatively difficult to prepare high-quality junctions with aluminum or vanadium superconducting electrodes deposited on top of the barrier. On the other hand, by depositing a magnesium seed prior to the deposition of aluminium, high-quality junctions with a superconducting aluminum bottom electrode are prepared reproducibly with satisfactory large transition temperature and critical field. The topic of chapter 11 is a recently recognized aspect of spin-polarized tunneling transport, namely the dependence of the spin polarization on the insulator used for the tunnel barrier. To study this dependence, Al/barrier/Co(Fe) junctions are prepared with various barrier materials. A significant difference between the tunneling spin polarization in junctions with AlOx and MgO barriers is observed. With cobalt and iron electrodes, a tunneling spin polarization of roughly +40 % is found with an AlOx barrier and one of roughly +30 % with a MgO barrier. Observations support that this difference is intrinsically determined by the barrier material. With HfOx and TaOx barriers, the Zeeman splitting in the conductance-voltage characteristic is quenched. This is caused by the large spin-orbit scattering rate induced by the heavy hafnium and tantalum atoms that are probably located at the Al/barrier interface. Chapter 12 describes a unique investigation of the role of manganese diffusion in the thermal stability of the tunneling spin polarization. Thermal stability is crucial for successful incorporation of tunnel junction devices into existing semiconductor technology. The diffusion of manganese is widely believed to be responsible for the collapse of the magnetoresistance effect of magnetic tunnel junctions observed above 250 ±C. Tunneling spin polarization is the parameter responsible for the magnetoresistance effect. The thermal stability of this parameter is directly investigated using the spin-polarized tunneling technique. To this end Al/AlOx/Co and Al/AlOx/Co90Fe10 junctions are used both with and without an additional FeMn layer deposited on top of the ferromagnetic electrode. The diffusion of Mn atoms from the FeMn layer to the AlOx barrier interface is confirmed independently with X-ray photoelectron spectroscopy. All the junctions, including their tunneling spin polarization, are found to be intrinsically stable up to 500 ±C, showing that the diffusion of manganese has no observable influence. The tunneling spin polarization, however, is preserved only when annealing is performed under sufficiently clean conditions. Under less clean conditions, it degrades severely above 250 ±C, similar as observed for the tunnel magnetoresistance effect. The degradation of the polarization is attributable to the diffusion of impurities from the environment into the junctions