The rapidly growing field of spintronics has recently attracted much attention. Spintronics is electronics in which the spin degree of freedom has been added to conventional chargebased electronic devices. A magnetic tunnel junction (MTJ) is an example of a spintronic device. MTJs consist of two ferromagnetic layers separated by a thin insulating barrier. The tunnel current that flows through the barrier depends on the relative alignment of the magnetization in the ferromagnetic layers. As a consequence of this dependence, the MTJ exhibits two different resistance values that distinguish a logical ‘0’ and a logical ‘1’, corresponding to anti-parallel and parallel magnetization. Due to these two distinct states, the MTJs can be used as magnetic memory elements, and serve as bits for information storage in magnetic random access memories (MRAMs). MRAMs can lead to instant-on computers and longer battery lifetimes for mobile devices, which gives MRAMs the potential to replace the current RAM technologies. However, for the current RAM technologies to be replaced by MRAMs, the dimensions of the MTJs have to decrease to sub - 100 nm in order to achieve a high enough areal density and to match the semiconductor technology. Therefore, the research in this Thesis aims at fabricating sub - 100 nm MTJs and investigating the influence of the reduced dimensions on the modification of magnetic and electronic properties. The MTJ is incorporated in an engineered multilayer stack to promote stability and reproducibility of the magnetic and electric response of the MTJ. For the structuring of these layers, we have used top-down nanofabrication techniques to produce the sub - 100 nm MTJs. For MTJs with a surface area of less than 0.01 µm2, the Al2O3 barrier has an approximate thickness of a nanometer to ensure an appropriate tunnel current. Therefore, we have concentrated on the plasm oxidation of sub -nm thin Al layers to produce Al2O3 barriers. We have shown that over-oxidation of sub -nm thin Al2O3 barriers of MTJs can be observed in real-time using in situ differential ellipsometry measurements. The change in ellipsometry signal of Al layers grown on CoFe films, is proportional to the amount of oxidized metallic material. As a result, the derivative of this signal is a direct measure of the oxidation rate. Further analysis of this oxidation rate allowed us to determine the onset of the CoFe oxidation. We have found the onset to be proportional to the deposited Al layer thickness. The amount of CoO determined from in situ X-ray Photoelectron Spectroscopy data on identical samples is found to be proportional to that obtained from ellipsometry. In short, this means that the point in time on which over-oxidation starts can be precisely determined. This is a critical necessity in producing exact, well-functioning MTJs. We have used electron beam (EB) lithography to pattern the sub - 100 nm features. With EB lithography features of only a few nanometer in size can be defined, hence EB lithography enables the exploration of the fundamental boundaries of magnetic and electric scaling properties. For the structuring of large-area samples with ultra-dense arrays of sub - 100 nm MTJs the throughput is limited because of the sequential writing process. However, with the developed special high-speed EB writing strategy we could pattern a sample area of 16 mm2 with ultra-dense arrays of sub - 100 nm elliptical features in 10 minutes. The strategy is employed to define the pattern of hard Ta masks with sub - 100 nm features. The Ta masks are etched at -50 ¿C in a SF6/O2 plasma to an etch depth that can be controlled with nanometer precision. Ar+ ion beam milling is used to transfer the pattern and to produce dense arrays of sub - µm MTJs. Insight in the magnetic switching behavior of nanoscale MTJs as a function of the size, shape and thickness is vital for MRAM application. Especially, the collective properties of high areal density arrays of MTJs are of interest, because magnetostatic coupling mechanisms between elements can be a limiting factor for applications. In order to understand the effects of geometry and coupling mechanisms, the switching of 5 nm thick polycrystalline nanoscale Co dots is examined using SQUID measurements and the switching of sub - µm MTJs is studied with MOKE measurements. An array consists of approximately 108 elements with a width ranging from 50 to 300 nm, and a length to width aspect ratios of 1.5 to 2.5, arranged on a rectangular lattice. The measured switching fields of the Co dots and MTJs were low compared to predictions using the Stoner-Wohlfarth model. The deviations of the Stoner-Wohlfarth behavior could be explained in term of interdot coupling and edge roughness. Comparison with the outcome of OOMMF simulation of the switching of a single dot revealed that the interdot coupling has a major influence on the magnetic switching behavior of arrays of nanoscale magnetic elements. This implies that for the feasibility of ultra-high areal density arrays of nanoscale MTJs for information storage new strategies are needed. For example, a more complex toggle MTJ multilayer stacks can be used for the nanostructuring of sub - µm MTJs. Faceting of the etch mask due to physical sputtering of the mask material is a problem during deep etching of ultra-high bit density arrays of sub - µm MTJs for MRAMs. Besides that, chlorinated etch residues can reduce the magnetization of patterning magnetic materials substantially, and therefore constitutes a considerable concern. To get more insight into the magnetization losses, CoFeB dots were etched in a high ion density Cl2- based plasma with a width ranging from 0.3 to 6.4 µm. The magnetic properties of the CoFeB dots were measured by SQUID magnetometry. The sub-µm CoFeB dots showed significant magnetization reductions, despite H2O rinsing. Scanning electron microscopy (SEM) studies revealed that etching in a Cl2-based plasma caused faceting of the masks, leading to sloped sidewalls. SEM pictures were used to determine the geometric volume which was compared to the effective magnetic volume resulting from the magnetometry measurements. The SEM data are in good agreement with the magnetometry data, and a chloride penetration depth of only a few nanometer could be derived, indicating that the postetch rinsing is sufficient to prevent considerable corrosion of the CoFeB dots. This means that the chlorinated etch residues could be removed from the samples, without severely effecting the magnetic properties. The I-V characteristics of nanometer thin AlOx barriers are measured by applying a voltage over a 1×1 µm2 square MTJ pillar located at the cross-point of the bottom and top electrode. The I-V response of the MTJs showed in principle three different characteristics, that is an ohmic like response, a response resembling breakdown, and a tunneling response. Approximately 30 % of the measured 32 MTJs showed a tunneling response and had a resistance-area product between approximately 20 and 50 k µm2. Furthermore, the conductance showed roughly a parabolic behavior implying an asymmetrical barrier. As a consequence, we used the Brinkman formula to fit the experimental data. The fits yielded average barrier heights between 1.2 and 2.3 eV, and an asymmetry parameter ranging from 0.3 to 0.8 eV, which are close to reported observations in literature. Resistance measurements yielded no significant magnetoresistance for the 1 µm2 MTJs. Probably, as a consequence of interface roughnesses, a strong N´eel coupling originates, through which an independent switching of the two magnetic layers of the MTJ is hindered. As an alternative electrical characterization technique, the conductive atomic force microscopy (c-AFM) technique can used to measure the local electrical transport properties of nanoscale MTJs. We have explored this c-AFM technique and performed I-V measurements by applying a constant bias voltage to the bottom electrode and measuring the tunnel current through the barrier. However, due to resist remains no significant bias voltage dependency was observed.
|Qualification||Doctor of Philosophy|
|Award date||26 Feb 2008|
|Place of Publication||Eindhoven|
|Publication status||Published - 2008|