Spinning with silicon - Silicon spintronics

Materials World magazine
1 Jan 2008

Electrical engineering is built on the understanding that electrons have two fundamental properties, mass and electric charge. However, physicists have known for over 80 years that electrons also have an intrinsic magnetic moment, which, for historical reasons, is called ‘spin’. The use of electron spin to control charge transport is known as spintronics. This field has made a major impact on daily life by driving increases in hard drive information storage density over the last decade (resulting in the 2007 Nobel Physics prize going to Albert Fert and Peter Grünberg for the discovery of giant magnetoresistance). However, the discipline has not yet influenced traditional information processing, where circuits of silicon transistors dominate.

Time for change

Transistor scaling laws are approaching their physical limits, meaning a new paradigm must take over. Spin can encode information in its ‘up’ or ‘down’ orientation, just like the presence or absence of charge encodes information in today’s integrated circuits of transistors. But, unlike charge, spin is not conserved over time, so if all the electron spins in an ensemble are oriented (polarised) in the ‘up’ direction, the spins will become random and information will be lost over the spin’s lifetime. Therefore, it is important to build an information-processing paradigm based on a long spin lifetime, so information can propagate from the input of a large circuit to the output, over great distances.

Fortunately, silicon, the material backbone of the semiconductor microelectronics industry, also has the longest spin lifetime of any bulk semiconductor. This potentially makes transition from electronics to spintronics more feasible for chipmakers, silicon poses problems for spintronic devices. In the last 10 years, significant progress has been made with spin injection, transport, manipulation, and detection – the bare minimum required for spintronic logic devices – in other semiconductors. However, until the team at the University of Delaware, USA, demonstrated it, achieving the same results in silicon had been a long-standing challenge.

Spin solution

Delaware researchers used the difference in scattering rates for spins oriented either parallel to the magnetisation of a thin-film ferromagnetic metal layer, or anti-parallel (in the same direction, or opposite to it). The electrons that which do not scatter (called ballistic electrons) in the ferromagnetic thin film, are therefore spin-polarised, with more of one spin direction than the other. The ballistic electrons are collected by the conduction band of the silicon transport layer in the devices. The ferromagnetic thin film acts as the ballistic electron analogue of an optical polariser, scattering or absorbing, electron spins of one orientation over the other.

Just as two absorbing optical polarisers in a series can modulate the transmission of light, a second ferromagnetic thin film can analyse spin polarisation after movement through the silicon. This is acheived by measuring the number of ballistic electrons being passed through when the magnetisations of injector and detector are parallel (spin ‘up’ is both injected and detected) or anti-parallel (spin ‘up’ is injected, but ‘down’ is detected – like crossed polarisers in the optical experiment), using ferromagnetic films with different magnetic switching fields. This results in the so-called ‘spin-valve’ measurement with a 350-micron-thick silicon transport layer.

This technique is performed when a magnetic field is applied in the same plane as ferromagnetic magnetisation (and therefore spin orientation). A more important demonstration of spin transport through a semiconductor is spin precession, where the spin orientation rotates around an applied perpendicular magnetic field, similar to the way the axis of rotation of a gyroscope revolves around the direction of gravity when it is not aligned vertically.

Because the measured ballistic current is dependent on the projection of the final spin angle on the fixed magnetisation measurement axis, changing the perpendicular magnetic field magnitude (and the precession frequency) will result in oscillations that are increasingly damped out at high fields. This is due to a fixed uncertainty in transit time caused by random thermal diffusion (spin dephasing). The oscillation spacing indicates the transit time, so by using this with spin polarisation data, the amount of spin that decays due to non-conservation over time can be determined. The spin lifetimes can be measured at a range of temperatures and are more than 0.5 microseconds at 60K.

Although this may seem short, with a present-day microelectronic processor clock speed approaching 10GHz, it is sufficient for several thousand logic operations. With microelectronic device features approaching 10nm, spin transport over 350 microns (35,000 times longer) indicates that large silicon spintronic circuits are possible. Of course, these measurements were taken at relatively low temperatures, and a lot of engineering is required to bring them to ambient conditions.

Delayed development

One significant achievement of this work is demonstration of a spin field effect transistor. Applying an external voltage between two electrical terminals changes the output current flowing through two other terminals. This uses spin precession and transit time control by changing the accelerating electric field. The system was first suggested (in a different form) 17 years ago in a seminal paper that ignited the semiconductor spintronics field, and had not yet been realised in a bulk semiconductor device, let alone in silicon.

Because the influence of silicon microelectronic devices is partly derived from their ability to be laterally patterned and linked into integrated circuits, lateral spin transport devices are being researched, and will be combined to create spin interactions constituting spin logic gates. Many researchers are pursuing nanotechnology by reducing the length-scales of the device technologies. However, scaling up to demonstrate even longer transport lengths (possibly in the centimetre range), to enable complex silicon-based spintronic circuits, will become the basis for a new age of information processing devices.


Further information:

Dr Ian Appelbaum, Assistant Professor, Electrical and Computer Engineering, University of Delaware, Newark, DE 19716, USA. Tel: +1 302 831 3295, email: appelbaum@ee.udel.edu.