Spin electronics is a new research area, which not only uses the electron’s charge but also its spin. By using the electron spin dependent properties of magnetic materials one can make devices with a new functionality. This has lead to magnetoresistive devices that can change their resistance by 10 to 50% in small magnetic fields, such as giant magnetoresistance (GMR) devices and the magnetic tunnel junction (MTJ). This large resistance change can be used in applications such as read heads or serve as memory elements in a magnetic random access memory (MRAM). This thesis describes two devices: the spin-valve transistor (SVT) and the magnetic tunnel transistor (MTT). The SVT has an unique property, namely its huge relative collector current change of more than 300% in small magnetic fields at room temperature. This unique property by itself is not enough to warrant the applicability of the SVT. The other properties that are important for the applicability of the SVT are described in this thesis. An alternative to the SVT, the MTT, will also be discussed. The SVT is a hybrid device that generally has an n-Si/ Pt / Ni82Fe18/ Au/ Co/ Au/ n-Si structure. The Pt / Ni82Fe18/ Au/ Co/ Au multi layer is the base and the two semiconductors on each side are the emitter and the collector respectively. The SVT is used in the common base configuration, where the emitter barrier (Si/Pt) is forward biased and the collector diode is zero or reversed biased. A flow of electrons from the silicon over the Schottky barrier into the metal base starts when the emitter is forward biased. These electrons have an excess energy compared to the Fermi level of the base and move in the direction of the collector. The electrons that are scattered in the base will lose their energy or momentum and make up the base current. Only those electrons that reach the collector with the right momentum and a high enough energy can enter the collector. The collector current is thus extremely sensitive to the scattering conditions in the base. The scattering conditions in the Ni82Fe18 and the Co layer are different for the spin-up and spin-down electrons. This makes the total scattering dependent on the relative magnetizations of the Ni82Fe18 and the Co layer. The collector current is largest when the magnetizations are parallel (I P C ) and smallest when the magnetizations are anti-parallel aligned (I AP C ). The relative change in collector current is called the magnetocurrent (MC = (I P C - I AP C )/I AP C ). This PhD research started with the development of a reliable process for fabricating spin-valve transistors. The introduction of this process together with the introduction of an ultra-high vacuum metal-evaporation system and the right choice of materials resulted in the SVT’s that exhibit an MC of more than 300% at room temperature. This thesis starts with a study on the size dependence of the magnetic and electrical properties of the SVT. We extended the previously mentioned process by using silicon on insulator (SOI) wafers, a combination of dry and wet etching techniques and SU8 (a negative tone photoresist) as an insulator layer. We were successful in producing SVTs with lateral dimensions that ranged from 300µm by 300µm to 10µm by 10µm. As expected, we saw no influence of the dimensions on the Schottky barrier height. Moreover the reverse current scaled down linearly with area. Both observations show that we have high-quality Schottky diodes. The key property of the SVT, its MC, showed no size dependence and remained constant around 240% for all dimensions. The transfer ratio is the ratio between the applied emitter current and the measured collector current. This ratio showed a slight decrease for transistors with dimensions below 25µm by 25µm. This is attributed to a deterioration of the emitter efficiency. The maximum possible emitter current decreases with transistor dimensions. The limiting factor is the maximum possible current density in the spin-valve base, which is 1.7 × 10 7 A/cm 2 . This value agrees with electromigration failure of spin valves. We have shown that it is possible to scale the lateral dimensions of the SVT down to 10µm by 10 µm. In my view further scaling down is limited to the physical height needed for the emitter, which includes the depletion width for the Schottky barrier and the doping profile needed for the Ohmic contact. To characterize the noise sources of the SVT we studied the frequency spectrum of three types of transistors that differed only in the type of metal base. The measurement showed that the frequency spectrum of the transistor with only non-magnetic layers in the base was completely dominated by shot noise in the frequency range of the measurement (10 Hz to 100kHz). The inclusion of one or more magnetic layers lowered the collector current and thus the level of the shot noise. It did not however change the nature of the noise or add noise (of magnetic origin) to the collector current. The collector current spectral density (SI) changes linearly with IC in a quasi-static magnetic field as expected for shot noise. We have however not observed 1/f noise in our measurements, not even at the switching fields of the spin valve. With this knowledge we can calculate the signal to noise ratio (SNR) of the SVT. The SNR increases with increasing MC and also with the absolute value collector current. From the basic relation IC = α IE we see that we can increase the collector current by either increasing the emitter current (IE) or the transfer ratio (alpha). We saw before that the emitter current has an upper limit imposed by device breakdown, therefore the way to enlarge IC is to improve the transfer ratio. We started to improve α by enlarging the energy difference between the emitter and collector barrier. The transfer ratio increased with increasing energy difference due to the larger number of states available at the collector semiconductor when electrons arrive with a higher energy. The transfer ratio also improves when materials with longer attenuation lengths are used in the base, i.e. Au instead of Pt. The influence of the SVT’s structural quality on the transfer ratio is demonstrated by the optimum in collector current versus Pt layer thickness. Furthermore, by varying the thickness of the NiFe layer we were able to prove that there is a maximum in the absolute current change for a certain thickness, due to the trade-off between transfer ratio and MC. The same study yielded a value for the attenuation of an interface, which is a factor of 0.55. The influence of crystal orientation on the transfer ratio was found to be negligible. Temperature effects on the transfer ratio are weak and are due to the spatial distribution of Schottky barrier heights and thermal spin wave scattering. Summarizing, we improved the transfer ratio by a factor of 118 from a Si/Pt/NiFe/Au/Co/Pt/Si SVT compared with a Si/Au/NiFe/Au/Co/Cu/Si transistor, while the MC remained constant above 200% and showed only small and non-systematic changes. The latter implies that the collection of both the spin-up and spin-down electrons can be improved, resulting in an increase in collector current without affecting the MC. The best results so far for SVTs are with a Si/ Au (20Å)/ Ni82Fe18 (30Å)/ Au (70Å)/ Co (30Å)/ Au (40Å)/ Si SVT, it has a transfer ratio of 1.2 × 10 4 and an MC of 230%. Further improvement of the transfer ratio might result from better control over the quality of the complete SVT structure. Another option is to use a tunnel barrier on the emitter side. This not only allows one to further enlarge the energy difference between the injected electrons and the collector Schottky barrier, but also opens up the possibility to remove layers from the base if a ferromagnetic emitter electrode is used, as in an MTT. Magnetic tunnel transistors have been successfully realized with the use of in situ shadow mask technology. Already we achieved a transfer ratio equal to that of SVTs, while the MCof the MTT is above 100%. The MTT has a Si/ Co/ Al2O3/ CoFe/ IrMn/ Ta structure. We have shown that the MTT can be used to determine a lower limit for the tunnel spin polarization of a ferromagnet/insulator interface. With a MTT this lower limit can be determined in a large temperature and tunnel-barrier bias range. The transfer ratio measured versus tunnel-barrier bias continues to increase, due to the larger number of available states at the collector at higher energies. More research is needed to explain the tunnelbarrier bias dependence of the MC. We expect that MTTs can be improved by using evaporation techniques rather than sputter techniques. Furthermore the quality of the collector diode can be improved with a corresponding increase in transfer ratio by choosing the right materials. The comparison of the SVT and MTT with tunnel junctions in terms of signal, noise, scalability, frequency response, robustness and of course the ability to study the properties of spin-polarized hot-electrons in magnetic materials justifies the further research of SVTs and MTTs. Last modified: May 16, 2002 by Hans.
- SMI-TST: From 2006 in EWI-TST
- SMI-MAT: MATERIALS