Spin-Torque and Spin Polarisation in Epitaxial Magnetic Silicides

Iron and silicon are two of the most abundant elements in the earth's crust. Nevertheless, the simplest chemical compound of these two elements, iron monosilicide (FeSi), possesses bizarre electronic and magnetic properties that have confounded researchers for decades. At low temperatures it is a non-magnetic semiconductor with a narrow gap. On warming, most materials become harder to magnetise: FeSi becomes easier, and it transforms into a heavy electron metal. Although known experimentally for over four decades, the proper theoretical description of this is still not settled.

When cobalt is substituted for iron things get even more interesting. Theory predicts (and indirect experiments on bulk crystals seem to confirm) that each Co atom contributes one current carrying electron, and also one electron spin's worth of magnetism, suggesting a perfectly polarised magnetic semiconductor - and what is more, one based on Si. Indeed, we have recently been able to prepare thin films of this material on commercial silicon wafer that appear to be epilayers: single crystals where every atom is in register with the lattice defined by the substrate. Spin polarisation is the key figure of merit for all spintronic materials, with all spintronic effects growing as the polarisation increases. Having a high polarisation material that is silicon-based is therefore a very tantalising prospect. In the first part of our project we will confirm the nature of our thin films and their structural, magnetic, and electronic properties. We will also investigate a simpler and quicker way of forming films known as sputtering. We will then go on to make the first direct measurements of the spin polarisation of this remarkable material, and moreover, do so in the technologically vital thin film form on Si wafer.

The magnetism is truly remarkable in another way, however. The crystal structure of this material is very unusual in that it lacks mirror symmetry, and so an obscure effect that is suppressed in almost every other magnetic material comes into play: the so-called Dzyaloshinskii-Moriya interaction. Instead of the usual uniform state in a ferromagnet, this term causes the spins to spiral around each other in a helix. This can be brought to a uniform saturated state in a large enough magnetic field, but on the way another largely forgotten piece of theoretical physics comes into play. There is an intermediate state formed from a lattice of magnetic vortices called skyrmions, a topological structure first invented to describe fields of pi-mesons in the 1960s. Last year it was shown (using bulk crystals of a related compound, manganese monosilicide) that because of this special topology, these swirling magnetic structures can be set into motion by a current flowing through the crystal at a current density around one million times smaller than that needed to move a vortex in a conventional magnetic material. We shall seek these magnetic skyrmion objects in our silicide wafer samples and measure the current density needed to move them.

Unfortunately, this material is only magnetic at temperatures a few tens of degrees above absolute zero, and all magnetic properties are lost well before room temperature is reached. Nevertheless, replacing silicon with its neighbour in the periodic table, germanium, can also transform iron silicide into a helimagnetic metal, with complete replacement preserving this structure up to a temperature a few degrees above zero Celsius. We shall complete our project by doping this material with cobalt and see if the critical temperature can be pushed above room temperature to technologically useful values.

Spintronics is the use of the electron's magnetic property-called its spin-to store and process information in the same way that charge is used in conventional microelectronics. The topic hit the headlines in 2007 when the Nobel Prize was awarded to Fert and Grünberg for their discovery of giant magnetoresistance. This effect allows the construction of exquisitely sensitive magnetic field sensors, and its impact was felt in the data storage industry where such sensors are used as the playback heads in ultrahigh density hard disks. This underpins portable storage in MP3 players such as Apple's iPod, as well as providing the huge amounts of extremely cheap storage needed to provide social media, such as Facebook, Twitter and, Youtube, free to users.

Even so, society's need for digital data seems insatiable, and modern data centres are major consumers of electrical power. Hard disks are used whenever large volumes of data need to be stored because their cost per bit is far lower than any other technology. Their major disadvantage, however, is that they rely on moving parts, which attracts large penalties in terms of both power consumption and reliability when compared to solid-state alternatives, especially at the data densities projected in the future. There are a number of proposals for 'storage class' magnetic memories that could form a solid state replacement for a hard disk at equivalent cost, which will repeat the success story of spintronics in the data storage industry. Many of these work on the principle of moving magnetic domain walls around using spin-polarised current pulses, for instance the widely-publicised 'racetrack memory' concept from IBM. One of the main barriers to realising this technology is that very high current densities are required to move the domain walls in conventional magnetic materials. Such domain walls can also be used to represent digital data in innovative magnetic logic schemes.

The recent discovery of skyrmion motion in magnetic silicides at current densities one million times smaller than in conventional materials is therefore an extremely significant one. The Joule heating by such a current is smaller by a twelve orders of magnitude, so that operations, such as moving bits around in a racetrack, that previously required one joule of energy will now require only one picojoule. These effects have yet to be measured in any material that is amenable to the usual planar processing techniques of the microelectronic industry, and demonstrating such low current densities is one of the main ways in which we envisage this project having impact. This is all the more important since we have already demonstrated the ability to prepare epilayers of these materials on commercial silicon wafer, the predominant semiconductor material.

Nevertheless, we recognise that the magnetism in these silicide materials is a low temperature effect, restricted to tens of Kelvins. There are reports in the literature of materials where these skyrmion structures have been observed at up to 260 K where silicon is replaced by its near neighbour germanium, and element also widely used in conventional CMOS electronics. We will therefore pursue closely related compounds to see if operation at even higher temperatures is possible, with room temperature operation as our ultimate goal. Even if this does not prove to be possible, these materials are excellent model systems for studying spin-torque and related effects in materials with strong Dzyaloshinskii-Moriya interactions and novel spin textures and topologies that can be used to guide future materials discovery efforts of other magnetic compounds with non-centrosymmetric lattices that show room temperature magnetism, and also their technological exploitation. The robust topology of skyrmion structures makes them ideal for the representation of digital data in solid state memories or in more advanced architectures where data storage and processing are combined.