Theme 2: Spin accumulation and spin pumping

An electric current consists of a flow of electrons in a certain direction. Usually the spin of these electrons is randomly oriented. But under certain circumstances it is possible to obtain a flow of electrons with their spin oriented in a single direction. This current is said to be spin polarised (see Fig. 1). The generation and control of such spin polarised currents are central to the phenomena that we wish to investigate. If two spin polarised currents with opposite spin polarisation cross an interface in opposite directions the charge current cancels out (since the same number of electrons move from right to left as from left to right), but there is a net spin flow from one side of the interface to the other. We aim to understand how efficiently spins may be transferred across different interfaces and how far they may propagate through different materials. There is still no simple means by which to measure the spin-polarisation of a current, and spin accumulation has never been directly observed.

Figure 1. Electrical charge current without (left) and with (right) spin polarisation

a) Measurement of local spin accumulation by XMCD during DC spin injection

A DC current will be passed through a Co/Cu bilayer nanopillar (Fig. 2). The current will be spin polarised in the Co layer and we want to measure the spin diffusion length in the Cu layer using XMCD. This can be done using two different methods. One method would be to measure the integrated spin polarisation in the copper layer, using samples with different Cu thickness, or with a thin polarisation destroying layer within the Cu layer placed at different depths. In this method the XMCD will be tuned to detect the polarisation of the Cu. Another method is to use a thin doping layer of Mn which will be polarised by the spin polarised current. Then the XMCD signal of the Mn layer will be measured using the chemical sensitivity of the XMCD. By placing the Mn layer at different depths in the thick Cu layer the degree of spin polarisation in the Cu can be determined. To improve the signal to noise ratio the sample has to consist of an array of nanopillars to use the full diameter of the x-ray probe beam (some 100 µm).

Figure 2. Sketch of the Co/Cu nanopillar. The degree of spin polarisation is indicated to the right.

b) Detection of precessional spin pumping with magneto-optical and XMCD probes 

A precessing magnetisation will pump spins into the neighbouring nonmagnetic layers and thus increase the apparent damping in the magnetic layer. This mechanism can be used to excite a non-equilibrium spin population near the interface in the nonmagnetic layer which will lead to a spin current in the nonmagnetic layer. Using this method spin polarised currents can be achieved by exciting the magnetisation in the magnetic layer at its resonance frequency. This rf pumping of the magnetisation has already been demonstrated at the University of Exeter. To excite the magnetisation in the GHz regime the magnetic layer has to be included in a high-frequency waveguide structure that will be connected to a microwave signal generator. The advantage of this method is that one can excite spin polarised currents in large area samples without patterning the sample into an array of nanopillars, as needed in part a). 

c) Theory 

The calculation of the accumulation with and without a current (London) is manageable and there are interesting predictions such as the spin density should be independent of the spacer thickness in the regime where the thickness is less than the spin diffusion length. The London group has developed a self-consistent theory of transport of spin and charge in an arbitrary magnetic multilayer. The theory allows us to calculate in a unified way, spin currents and local spin densities at zero bias, corresponding to equilibrium exchange coupling, as well as transport spin and particle currents, and transport spin densities in any atomic plane of the structure under applied bias. This can all be done for a fully realistic ab-initio band structure of the multilayer, and codes for calculating spin currents in Co/Cu multilayers have already been successfully implemented. We propose to calculate the spin current due to electrons injected from a transition metal ferromagnet to a nonmagnet and determine the spatial profile of the accumulated spin density at the interface between a ferromagnetic and a paramagnetic metal due to a dc current or due to precession in the ferromagnetic layer. We shall also calculate the transport spin accumulated in a nonmagnetic spacer layer sandwiched between two ferromagnetic layers under external bias. We predict that the accumulated spin in the spacer layer should be nonzero over a distance of the order of the spin diffusion length. To test this theoretically we shall include, using perturbation theory, spin-orbit coupling in the spacer which controls the spin diffusion length in this model. We shall develop in parallel a coherent potential approximation theory of this effect in collaboration with our colleagues at Nagoya University, Japan.