187 High density magnetic recording, magnetic random access memories, displacement and current detection, contactless switching and electronic compass applications all require magnetic field sensors with unprecedented sensitivity. The spin-valve effect (giant magnetoresistance) found in 1988 in magnetic metallic multilayers provided a new form of magnetoresistance with promising possibilities. The name indicates its physical origin: the magnetic layers form magnetic field controllable valves for electrons with parallel spins, leading to low resistance under application of a magnetic field. The resistance is usually measured using a four point technique, with the current in plane. Now, ten years later, the first hard disk incorporating this effect has reached the marketplace. Its read head shows a spin-valve effect of about 5% allowing for a capacity of 16.8 Gbyte. The use of the much larger perpendicular spin-valve effect was limited because of the ultra low perpendicular resistance of the atomically thin multilayers. Subject of this thesis is the spin valve transistor, which we introduced 1995. It is able to circumvent the low resistance dilemma by its direct mean free path dependence in the output. In the spin valve transistor a spin-valve multilayer serves as a base region of an n-silicon metal base transistor structure: via an emitter barrier hot electrons are injected into the spin-valve, a collector barrier accepts only non-scattered electrons by means of its angle and energy selectivity. Since the number of collected electrons depends exponentially on the spin dependent mean free path in the spin-valve base, and since the mean free path changes with field due to the perpendicular spin-valve effect, the collector current change with magnetic field is much larger than the resistance change in current-in-plane measurements and can be effectively used for magnetic field detection. Since the electron energy of the injected electrons can be varied either using different Schottky barriers or bias variation using a tunnel emitter, the spin-valve transistor allows electron spectroscopy to be performed in spin-valves. Because the electron energy range is of the same order as the spin-split 3d bands of the ferromagnetic metals Co, Fe and Ni, a direct investigation of the spin-dependent band structure which causes the spin-valve effect is possible. An extension of the two channel model has been formulated, showing that the collector current change can be extremely large, for example, implementation of a 5% spin-valve (measured in plane) may lead to practically useful output current variations of more than 500%. Moreover, thickness variation e.g. using wedges, allows determination of spin dependent mean free paths. In this work effects up to 400% have been measured at 77K and 15% at room temperature. Values larger than 100% at room temperature are expected with proper spin-valves: it was a major problem to grow a good antiferromagnetically coupled Co/Cu spin-valve sandwich on top of oxide free silicon. The room temperature hot electron spin valve effect was observed in a Si-Pt-Co/Cu/Co-Si spin valve transistor. The magnetic field and temperature dependence are found to be in accordance with expectations. It is suggested to try band engineered inverse spin valves to observe clear band structure related effects in the collector current. This can be done using a Schottky emitter or using a tunneling emitter. The first shows a narrow electron injection spread (kT), whereas the latter the energy can be changed by varying the applied bias, at the cost of increased energy spread (of the order of 0.2 eV for a 2eV barrier). For applications, the absolute collector current change 'I can be optimized by manipulating the thickness and improving the structural quality of the spin-valve layers. An uncoupled spin-valve sandwich such as Co-Cu-NiFe is more appropriate for this purpose than an antiferromagnetically coupled systems. The signal to noise ratio of the spin valve transistor is surprisingly large in spite of the relatively small collector current. This is both related to the absence of thermal noise in the collector current and the small input noise currents of (transimpedance) amplifiers. For low field sensing advantages of the spin valve transistor are the applicability of very thin magnetic layers increasing magnetic efficiency and layer decoupling possibility by thick spacer layers. Disadvantage is that growth of the spin-valve has to take place on oxide free semiconductors and that power dissipation is larger than in current-in-plane measurements. For sensor designs, self heating is the limiting factor of the signal/noise ratio and must be calculated accordingly. To verify the noise model put forward in this thesis, noise measurements of the collector current may be performed. Preparation of the spin valve transistor has been performed by using a new technique: direct bonding of substrates during deposition of a metal. A bond interface is absent due to recrystallisation of the metal film, even at room temperature. The bonding technique may be used for many vertical transport device ideas such as magnetic tunnel junctions with crystalline (semiconductor) barriers or proximity effect experiments, but also forms a permanent link for demanding adhesion applications, in which an ultrathin, permanent, room temperature, chemically resistant, thermally and electrically conducting and, using magnetic materials, magnetic flux conserving link is required.
- SMI-MAT: MATERIALS
- SMI-TST: From 2006 in EWI-TST