Technical Field The present invention relates to apparatus and methods for obtaining electrical energy from thermoelectric effects, and to methods of manufacturing apparatus for this purposes.

Background

The UK has committed to meet an 80% reduction in greenhouse emissions by 2050. It is recognised that this will likely stem from a diverse portfolio of energy sources (renewable and existing) and development of technologies for energy storage, conversion and usage. As the majority of the UK's total energy consumption can be attributed to heating (48%) and transport (38%) these are clearly significant targets for change. As such, technology designed to make use of the vast supply of currently untapped waste heat will have a significant impact. For example, in the UK, 18% of the total energy consumption in 2011 was used for industrial processes, of which it is likely that 20-50% was lost as waste heat. More strikingly however, 27% of total UK energy consumption in 2011 could be attributed to road transport where an average of 40-60% is typically expelled as waste heat. These examples represent a viable target for improvement and a potential application for energy harvesting technologies.

Thermoelectric generators, based on the Seebeck effect, are typically too inefficient to be of appreciable use for energy harvesting. Investigation of such devices has been limited because of the interdependence of the thermal and electric conductivity of such devices places a fundamental limit on their efficiency . The present disclosure takes a new approach to related technologies which address similar problems. This approach differs in that it is based on quantum phenomena, such as electron spin, spin currents, and spin injection. Accordingly, to put these ideas in context, this background discussion includes:

• Classical ideas of charge current, Ampere's circuital law, and the Lorentz force.

• Spin, and spin current, and the idea of spin-orbit coupling .

· The classical Hall effect, the Hall angle and Hall resistivity .

• The so-called anomalous Hall effect, skew scattering, the Spin Hall effect, and the concept of spin Hall angle. Charge Current: Conventional electrical current is a net flow of charge, which may be referred to as charge current. This may be brought about by the flow of negative charge carriers (such as electrons) , or by a flow of positive charge carriers (such as holes) . Charge currents give rise to magnetic fields - an idea that is succinctly expressed in Ampere's circuital law.

Spin and Spin Current: Electrons have an intrinsic form of angular momentum known as spin. The spin of any given electron is denoted by a spin quantum number - which indicates whether that electron is in a "spin-up" or "spin-down" state. The spin may also be associated with a so-called spin magnetic moment, and an electron in a "spin-down" state may be thought of as having an opposite spin magnetic moment from an electron in a "spin-up" state. Conceptually, it can be appreciated that just as charge carriers can give rise to charge current, it can also give rise to spin current - a net flow of spin.

Spin-Orbit Coupling: Where charge currents exist there may be a degree of interaction between the magnetic field associated with the charge current and the spin magnetic moment of the charge carriers which give rise to that current. Perhaps the most widely known manifestation of this effect is spin-orbit coupling. The Hall Effect, Hall Angle & Hall Resistivity: Other effects arise due to the interaction of charge currents and magnetic fields. One example is the Hall Effect. The Hall effect is the production of a voltage difference across a current carrying electrical conductor in the presence of a magnetic field that is perpendicular to the current. This so called "Hall voltage" arises across, e.g. perpendicular to, the direction of the electric current carried by the conductor.

In other words, orthogonal electric and magnetic fields generate, in a conductor, a secondary electric field that is normal to both those primary fields. In a classical particle model of charge transport it can be said that, as a result of the Lorentz force deflection, charges build up on the edges of the conductor until the field due to the non-uniform charge distribution exerts a force equal to the deflecting Lorentz force of the magnetic field.

By way of example, if an electric field is applied to a conductor in a positive x-direction and a magnetic field is applied in the positive z-direction, electrons that would (in the absence of applied magnetic field) travel in the negative x-direction are Lorentz force deflected in a negative y-direction. This causes an accumulation of charges on the negative-y and positive-y edges of the conductor. This charge build up in turn gives rise to a secondary electric field, the Hall Field, in the negative-y direction. As a result, the equipotentials on the conductor are not perpendicular to the current flow (as would ordinarily be the case in the absence of an applied magnetic field) . Instead, the equipotentials are rotated by an angle Θ _Η which may be referred to as the "Hall angle" .

Because the current and equipotentials are not perpendicular to each other, it can be said that in the presence of a steady magnetic field the resistivity has both a longitudinal component and a transverse component. The longitudinal resistivity, p _XX is the ratio of applied primary electric field to the current density in the direction of that primary electric field. The transverse resistivity (also known as the "Hall resistivity") is the ratio of the secondary electric field (the "Hall field") to the current density in the direction of the primary electric field. It can be seen therefore that the Hall resistivity, p _xy , and the Hall angle are inherently related parameters. Both relate the secondary Hall field to the primary longitudinal field (but of course the resistivity takes account also of the longitudinal current density in the conductor) . Both parameters are dependent upon on the applied perpendicular magnetic field H _z .

Anomalous Hall Effect - Spin Hall Effect: The Hall angle or Hall resistivity depends upon the applied perpendicular magnetic field. This dependence is different in ferromagnetic conductors as compared to non-magnetic conductors. When observed in ferromagnetic materials, an additional contribution to the Hall effect can be detected. This additional contribution is related to the magnetization of the material and is known as the "anomalous Hall effect". In non-magnetic conductors, p _xy increases linearly with Hz. This is as one would expect from the Lorentz force. In ferromagnetic conductors however p _xy initially increases steeply in weak H _z , but saturates at a large value that is nearly H _z - independent .

This contribution to the measured Hall voltage is proportional to magnetization, but cannot be explained as being due to the magnetic field produced by magnetization. This anomalous contribution has been termed the spin Hall effect. The spin Hall Effect consists in spin accumulation at the lateral boundaries of a current-carrying conductor, the directions of the spins being opposite at the opposing boundaries. For a cylindrical wire the spins wind around the surface. The boundary spin polarization is proportional to the current and changes sign when the direction of the current is reversed. In the normal Hall effect charges of opposite signs accumulate at the sample boundaries due to the action of the Lorentz force in magnetic field. The spin Hall Effect is somewhat similar. However, there are significant differences. In particular, no magnetic field is needed for spin accumulation.

A complete explanation of the spin Hall Effect would go far beyond the present discussion and in the interests of brevity has been omitted here. What is relevant however is that three distinct contributions to the spin Hall Effect have been identified - these contributions are: (i) intrinsic (ii) skew scattering, and (iii) side-jump scattering. The second of these, spin skew scattering, is thought to be associated with spin-orbit coupling. This spin-orbit coupling gives rise to a spin dependence of scattering during charge transport in a magnetised conductor and, in some circumstances, is thought to be a dominant contribution to the spin Hall Effect. Under other circumstances side-jump scattering and intrinsic effects may be significant. Without wishing to be bound by any particular theory, we note that all three of these contributions may be associated with spin-orbit coupling

In skew scattering the final momentum direction after a scattering interaction is different for spin-up charge carriers as compared to spin-down charge carriers. In other words, spin- orbit interactions give rise to asymmetric scattering of spin polarised electrons. If a flow of polarised electrons hits a target, it will deviate in direction depending upon the spin polarisation of that flow. If however the flow of electrons is not polarised, this same spin asymmetry will result in a separation of spin-up and spin-down electrons. Spin up electrons will mostly go to the right, and spin down electrons to the left. Accordingly, a spin current will appear perpendicular to the direction of the incoming charge current. This is what is observed in the so called spin Hall effect. It can be thought of as a charge current induced accumulation of spin inhomogeneity perpendicular to the direction of the charge current.

Conversely, an inhomogeneous spin density results in an anomalous Hall voltage. Another way to express this converse idea might be that a spin current can give rise to a charge current, perpendicular to the direction of that spin current. This is the so called inverse Spin Hall effect. The present disclosure aims to improve energy efficiency and reduce energy wastage.

Summary of Invention

Aspects and embodiments of the present invention aim to address the above described technical problems (and other related problems) and are set out in the appended claims.

Brief Description of Drawings

Some embodiments will now be described, by way of example only, with reference to the accompanying drawings in which:

Figure 1 is a schematic diagram of an apparatus for power harvesting;

Figure 2 is a schematic diagram of an example of an apparatus such as that shown in Figure 1; Figure 3 is a schematic diagram of an apparatus for use m a testing method for determining the spin Hall angle of a material; and

Figure 4 provides examples of x-ray crystallography data obtained from measurements of a magnetic layer suitable for use in the apparatus of Figure 1 and Figure 2.

In the drawings like reference numerals are used to indicate like elements .

Specific Description

Embodiments of the disclosure enable a spin current to be injected from a magnetic material under the influence of a temperature gradient into an attached non-magnetic material. The non-magnetic material is arranged to convert that spin current into a charge current to enable electrical power to be harvested from the temperature gradient.

Figure 1 includes a schematic elevation, Fig 1. (a) , and an end view (Fig 1. (b) ) of an apparatus 1 for power harvesting.

The apparatus 1 comprises a magnetic layer 7 which carries a spin conversion structure 9. In the illustration in Figure 1, the magnetic layer 7 is coupled to a substrate 15, such as a piece of glass or a semiconductor device. This apparatus 1 is configured to convert a temperature gradient across the magnetic layer 7 into a charge current through the spin conversion structure. To achieve this, spin polarised current is provided from the magnetic layer 7 into the spin conversion structure 9 in response to the temperature gradient.

Without wishing to be bound by any particular theory, it is believed that the spin Seebeck effect in the magnetic layer gives rise to this spin polarised current in the direction of the temperature gradient (e.g. towards the spin conversion structure) . The spin polarised current is then injected into the spin conversion structure 9. In the spin conversion structure, the spin Hall angle of the material of the spin conversion structure causes this spin current to give rise to a spatially inhomogeneous spin distribution. As a result of the inverse spin Hall effect, this inhomogeneous spin distribution causes an anomalous Hall field, which in turn drives charge current in the spin conversion structure.

It can be seen in Figure 1 that the spin conversion structure 9 is carried on the magnetic layer 7. The spin conversion structure 9 comprises at least two electrically conductive legs, 11, 13 coupled together in series. The electrically conductive legs 11, 13 comprise a material such as an alloy. This may comprise a host metal, such as copper, doped with an impurity or dopant. The first of these two legs 11 is doped with a first dopant, and the other leg 13 is doped with a second dopant (different from the first dopant) . The first dopant is selected for the host metal so that the alloy has a negative spin Hall angle. In contrast, the second dopant is selected for the host metal so that the alloy has a positive spin Hall angle.

Only two legs 11, 13 are illustrated in Figure 1, however it will be appreciated in the context of the present disclosure that many more legs may be used. In general, the spin conversion structure usually comprises doped legs 11, 13 arranged in series so that adjacent legs are doped differently. The doping is chosen so that each leg 11 has a spin Hall angle of opposite sign to its neighbour 13.

The host metal used in the spin conversion structure 9 may be copper, the first dopant may be bismuth, and the second dopant may be iridium. For example, the negative spin hall angle legs may comprise copper doped with bismuth at a level of approximately 1%. Other host metals and other dopants may be used. To determine whether a host material and dopant combination is suitable for use in apparatus of the present disclosure a method, such as that explained with reference to Figure 3 may be used. Other methods may also be used.

The spin Hall angle, Θ _ΞΗ , may be defined based on the ratio of the transverse spin conductivity to the longitudinal charge conductivity, thus: a x.y

0 _SH <x at

Where a _xx is the longitudinal charge conductivity, a _xy is the transverse spin conductivity (also known as spin Hall conductivity) . In other words, the spin Hall angle is proportional to the ratio of the transverse spin conductivity to the longitudinal charge conductivity. Therefore a negative spin Hall angle may arise when the transverse spin conductivity is of opposite sign to the longitudinal charge conductivity, for example when the transverse spin conductivity is negative. For example, the transverse spin conductivity of the legs doped with the first dopant may be negative, and the transverse spin conductivity of the legs doped with the second dopant may be positive. The transverse spin conductivity is the ratio of the current density in the direction of the primary electric field to the secondary electric field due to spin inhomogeneity (the "anomalous Hall field") . Methods of measuring the spin Hall angle are also described herein. The magnetic layer 7 may be ferrimagnetic or ferromagnetic and may comprise an iron containing material. This material may be polycrystalline . The magnetic layer 7 may also act as a thermal insulator. In some embodiments the thermally insulating nature of the magnetic layer may improve the efficiency with which the device can convert a heat flow to a charge current. If this layer 7 comprises iron, it is typically present in the form of Iron- II- III Oxide (Fe ₃ 0 ₄ ) . This magnetic layer may have a coercive field of at least 50 Oersted (about 400Am ^_1 ) . The magnetic layer is magnetised to give rise to a field that is aligned with the plane of the layer, that is to say the net magnetisation through (e.g. normal to) the layer is less than the magnetisation along (e.g. parallel to) the layer 7. The magnetic layer 7 may also be an electrical insulator. In some embodiments, voltage generation is proportional to the heat flow through the structure.

Although not illustrated in Figure 1, in some embodiments a capping layer may be provided on the magnetic layer 7 between the magnetic layer 7 and the spin conversion structure 9. Where one is present, this capping layer comprises a conductive material, for example a metal such as gold. The thickness of this layer may be less than the spin diffusion length in the material (e.g less than 30 nm for the example of gold) .

Spin diffusion length may be measured by weak localisation measurements, conduction electron spin resonance, Andreev reflection/point contact and spin absorption measurements. At room temperature: spin diffusion length in gold is approximately 30 nm, wheras in platinum it is approximately 2nm. The actual value will depend upon level of impurities (dominant at low temperature) and upon temperature (phonons) .

The capping layer may provide a diffusion barrier between the magnetic layer and the spin conversion structure. The capping layer may comprise Gold and/or Platinum (Pt) , or perhaps Niobium (Nb) . Such materials have spin diffusion lengths of the order of 2nm, and 30nm, respectively. The values relate to room temperature - spin diffusion length decreases with increasing temperature . After testing we have found that a gold capping layer of about 5 nm thickness is very effective. Embodiments may comprise capping layers of about 2 nm to 5 nm. This has been found sufficient to inhibit interdiffusion of oxygen into the spin conversion structure. The capping layer is generally thick enough to inhibit interdiffusion of oxygen between the magnetic layer and the spin conversion structure. For example, the thickness of the capping layer may be greater than the diffusion length of oxygen atoms in the electrically conductive material of the capping layer. In general, the capping layer is at least 1 nm thick, for example at least 2 nm thick. It may also be less than 10 nm thick, for example less than 5 nm thick.

The function of the magnetic layer 7 is to generate spin polarised current in response to a temperature gradient across the magnetic layer. That is to say, in response to a temperature a gradient through (e.g. normal to) the major surface of the layer .

The function of the capping layer, if one is included, is to channel spin polarised charge current from the magnetic layer 7 into the spin conversion structure. The capping layer may also inhibit interdiffusion of oxygen from the magnetic layer into the spin conversion structure. The function of the negative spin Hall angle legs 11 of the spin conversion structure 9 is to provide an anomalous Hall voltage (spin Hall voltage) in a first direction in response to the spin current from the magnetic layer. The function of the positive spin Hall angle legs 13 of the spin conversion structure is to provide an anomalous Hall voltage (spin Hall voltage) in a second direction, different from the first direction, in response to the spin current from the magnetic layer 7. The legs 11, 13 may be antiparallel (e.g. negative spin Hall angle legs aligned in the opposite direction to the positive spin Hall angle legs) , or they may be arranged in a zig-zag or wave pattern. Any such configuration will suffice, provided that it gives rise to a series of reversals along a direction aligned with the plane of the magnetic layer and perpendicular to the direction of the net magnetisation of the magnetic layer. The legs are connected together, head to tail, to provide this alternating series. As a result the spin Hall voltages of the legs combine additively together so that the spin conversion structure as a whole can generate a useful voltage across the series of legs.

In operation, a temperature gradient is provided through the magnetic layer 7. This may arise from flow of waste heat generated by an electrical or electronic component such as a power transistor or microprocessor. Such a heat source may provide (or be thermally coupled to) the substrate 15. In the magnetic layer 7, the temperature gradient generates a spin polarised current that flows in the direction of (for example parallel to the direction of) the temperature gradient. Without wishing to be bound by any particular theory, it is believed that this effect is associated with the spin Seebeck effect. This spin polarised current is transferred from the magnetic layer to the legs 11, 13 of the spin conversion structure. The first legs 11 (negative spin Hall angle) of the spin conversion structure generate an inverse spin Hall voltage in a first direction, while the second legs 13 (positive spin Hall angle) generate an inverse spin Hall voltage in a second direction, opposite to the first direction. As noted above, the first legs 11 and second legs 13 are arranged so that the opposing voltages in alternate legs combine additively (opposite voltages applied in opposing geometric directions add constructively) . In other words, the path traced by the spin conversion structure undergoes a series of reversals in direction. In combination with the reversals in spin Hall voltage in adjacent legs 11, 13, these geometric reversals mean that the spin Hall voltages of the legs of the spin conversion structure combine additively. As a result the total voltage across a series of such legs comprises the sum of the (magnitudes of the) spin Hall voltages on each of the individual legs 11, 13 of the structure 9. This can enable appreciable electrical power to be drawn from the spin conversion structure 9 in the form of a charge current.

It can be seen therefore that apparatus of the present disclosure, such as that explained above with reference to Figure 1 may be fixed to any heat transferring substrate (such as a heat source or heat sink) to provide a temperature gradient across the magnetic layer. Embodiments of such devices have been investigated. Such investigation suggests that a 10 Kelvin temperature difference through a device having 10x10 cm area may provide a Voltage of 0.13 mV and a power of 0.06nW. The same 10 Kelvin temperature difference with a 100 leg (10x10 cm) spin conversion structure (thermopile) may provide a voltage of 13 mV and a power of 8.4nW. It will be appreciated in the context of the present disclosure that these numerical examples are presented purely by way of example, and different device configurations may give rise to differing results.

Figure 2 shows another apparatus 100 according to the present disclosure. As in Figure 1, the apparatus illustrated in Figure 2 comprises a magnetic layer 7, and a spin conversion structure 9 disposed on the magnetic layer 7. In addition the apparatus 100 illustrated in Figure 2 comprises a capping layer 17 interposed between the magnetic layer 7 and the spin conversion structure 9. The apparatus 100 shown in Figure 2 also comprises a first power coupling 200, and a second power coupling 220. These are illustrated as being carried by the magnetic layer, but they may also be carried by the spin conversion structure. These power couplings 20, 22 may comprise the same material as the spin conversion structure 9, or they may be provided by any other conductor .

The spin conversion structure 9 comprises a plurality of legs 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22 joined together by a plurality of conductive bridges 24, 26, 28, 30, 32, 34. These trace a tortuous path across the capping layer 17 to provide reversals in direction between consecutive legs. One end of the spin conversion structure 9 is connected to the first power coupling 200 and the other end of the spin conversion structure 9 is connected to the second power coupling 220.

The spin conversion structure 9 of the apparatus shown in Figure 2 comprises a plurality of doped legs 2-22. As noted above, these doped legs comprise a host material, such as a metal, doped with a material which changes the spin Hall angle of that host material. Some of the doped legs are doped so that the spin hall angle is positive (referred to herein as "positive doped legs"), and others of the legs are doped so that the spin Hall angle is negative (referred to herein as "negative doped legs") . In Figure 2, the negative doped legs and the positive doped legs are arranged in parallel on the capping layer. The positive and negative doped legs are joined together by bridges of host material. The bridges are arranged so that the conductive path provided by the spin conversion structure across the capping layer comprises a series of reversals along a direction aligned with the plane of the magnetic layer and perpendicular to the direction of the net magnetisation of the magnetic layer. The bridges connect the legs together, head to tail, to provide this alternating series.

In more detail, a first negative doped leg 2 is connected to the first power coupling 200. The first negative doped leg 2 extends in the y direction across the surface of the capping layer 17. A bridge 24 of host material extends transverse to the first negative doped leg 2 (in the x direction) to connect the other end of the first negative doped leg 2 to a first positive doped leg 4. The first positive doped leg 4 extends from this bridge 24 of host material back in the negative y direction, parallel to the first negative doped leg 2. The other end of this positive doped leg 4 is joined by another bridge 36 of host material to a second negative doped leg 6. As with the first negative doped leg 2, this second negative doped leg 6 extends in the y direction until it meets another bridge of host material 26 which connects its other end to a second positive doped leg 8. As shown in Figure 2, this pattern of a negative doped leg, a transverse bridge, and a positive doped leg opposition to the negative doped leg, and so on can be repeated to fill the area available on the capping layer 17. As shown in Figure 2 this repeating pattern extends the spin conversion structure 9 in the x direction across the capping layer 17. The last bridge in the sequence 28, positioned at the far right hand edge (maximum x) of this pattern, is connected to a (slightly longer than) double length positive doped leg 12. The other end of this double length positive doped leg 12 is connected to a bridge of host material 30 which extends in a negative x-direction. This bridge is then connected to a negative doped leg 14, a further (negative x direction) bridge 42, a positive doped leg 16, a bridge 32, and so on. Eventually, by making a series of reversals, this enables the spin conversion structure 9 to return across the capping layer to the second power coupling 220. It can be seen in Figure 2 that in this embodiment, in addition to having a series of reversals, the spin conversion structure 9 is arranged as a loop between the two power couplings 200, 220. For example, the loop may not be fully closed, and the power couplings 200, 220, may be disposed at the opening to that (nearly) closed loop. This is one way in which the power couplings can be arranged to reduce the temperature difference between the power couplings 200, 220 thereby avoiding additional Seebeck contributions to the voltage that would detract from the efficiency of the spin conversion structure. It will be appreciated in the context of the present disclosure that, depending on the polarity of the voltage (+/-), these additional Seebeck contributions to the voltage may not detract from the efficiency of the spin conversion structure.

An apparatus 100 such as that illustrated in Figure 2 may be manufactured by depositing the magnetic layer on to a substrate, for example using thin film deposition techniques such as magnetron sputtering, chemical vapour deposition (CVD) , physical vapour deposition (PVD) . One example of PVD is pulsed laser deposition (PLD) . The magnetic layer 7 can then be covered with a capping layer. The magnetic layer may be magnetised either during or after deposition, for example by the use of an electromagnet. It will be appreciated in the context of the present disclosure that, the coercive field of the magnetic layer may enable this magnetisation to be maintained. It will also be appreciated that the magnitude of volume magnetisation in plane is not necessarily directly related to the magnitude of spin current generated by the layer. The magnetic layer may be magnetised in plane that is to say the component of the magnetisation in the plane of the layer may be greater than the component of its magnetisation normal to the plane.

The capping layer 17 may also be deposited by one of the above mentioned deposition techniques onto the magnetic layer. The spin conversion structure 9 is then laid down onto the capping layer 17. This may be done by physical vapour deposition, or by any other deposition technique. For example, one option for manufacturing the spin conversion structure is to use physical vapour deposition. In this option, interchangeable masks can be applied to the structure to allow co-deposition of the main alloy (e.g. copper) . The provision of dopant (e.g. Bi, Ta, Ir) could then be controlled independently so that the selected dopant could be co-deposited into the relevant legs of the spin conversion structure. This may enable fabrication to be scaled-up to enable mass production of power harvesting devices of the disclosure . The apparatus 100 may be deposited directly onto a substrate or directly onto a heat source. The substrate onto which the magnetic layer is deposited may comprise a part of another larger apparatus, such as an electronic or electrical device. Examples of substrates onto which the magnetic layer can be deposited include glass, fused silica, silicon, SrTi0 ₃ .

Such substrates may then be mechanically secured to another heat transferring surface - for example by being clamped or adhered to it, for example using a glue. Where the substrate is to be attached to a heat source a thermally stable glue may be used, or it may be clamped with thermal grease to aid thermal contact between heat source and device, or the apparatus may be deposited directly onto the heat source. In some embodiments, the magnetic layer and spin conversion structure (with interposed capping layer if included) may be deposited directly onto the heat transferring substrate. Investigation of such structures has provided embodiments which, for a 10 Kelvin temperature difference across 10x10 cm area provide Voltage V=570 mV; power P =1.2 mW. For 10 Kelvin temperature difference across a 100 leg (10x10 cm area) spin conversion structure (thermopile) Voltage V=57 V; power P=1.56 W. In an embodiment, for a 1 Kelvin temperature difference across a 10x10 cm area, Voltage V=57 mV; power P=12 . For 1 Kelvin temperature difference across a 100 leg (10x10 cm area) spin conversion structure (thermopile), V=5.7 V; P=2.7 mW. The temperature differences mentioned above may occur predominantly (or, perhaps entirely) across the magnetic layer . In addition to a bilayer structure (e.g. spin conversion structure and magnetic layer with or without an interposed capping layer) some embodiments may be arranged in a 'sandwich' structure in which the spin conversion structure is sandwiched between magnetic layers. It will thus be appreciated that a plurality of such bilayer structures (e.g. each comprising a spin conversion structure and magnetic layer with or without an interposed capping layer) , may be stacked up on top of one another to provide a plurality of such layers. Accordingly, a single heat current can be passed through the stack (through the vertically stacked multiple layers) and the electrical energy generated by the spin conversion structure in each device can then be combined (for example by connecting them in series or in parallel) . In an embodiment, the legs of the spin conversion structure may comprise a pure alloy into which a less than 5% substitution (doping) leads to a change in the spin Hall angle. Without wishing to be bound by any particular theory, this change in spin Hall angle may arise due to the skew scattering mechanism. As noted above, in skew scattering the spin-orbit interaction gives rise to spin dependant scattering with differing momentum directions for spin up and spin down electrons. An example of such a spin conversion structure comprises pure copper doped with between 0.5% and 1.5% bismuth. The spin Hall angle of such copper doped with bismuth may be large and negative, for example about - 24%. This has been tested and we have observed a negative spin hall angle (e.g. reversal of voltage polarity compared to Fe304:Pt devices) . Our experiments indicate that the capping layer may be useful in this context as this effect in copper doped with bismuth seems to be sensitive to presence of oxygen, which may migrate from the magnetic layer where a capping layer is not present. Bismuth and iridium have been mentioned as examples of dopants where the host material comprises copper. It will be appreciated however that these are merely examples.

In addition to or as an alternative to iridium other dopants which can be used to produce a positive spin Hall angle in copper include platinum. It will be appreciated in the context of the present disclosure however that these dopants are merely examples, and that any dopant which gives rise to skew scattering to provide a positive contribution to the spin Hall angle of the host material may be used to provide the positive doped legs. Indeed, any dopant which provides an extrinsic contribution to the spin Hall angle, such as by skew scattering may be used. Examples of such dopants may give rise to side-jump effects as mentioned above.

In addition to or as an alternative to Bismuth other dopants can be used to produce a negative spin Hall angle. Any dopant which gives rise to skew scattering to provide a negative contribution to the spin Hall angle of the host material may be used to provide the negative doped legs. Such dopants can be identified by routine testing using a method such as that described with reference to Figure 3.

Although the use of dopants may enable a wide range of different host materials to be used, it will be appreciated in the context of the present disclosure that if different materials are used for the positive and negative legs of the spin conversion structure then no dopant is necessary. In particular, the spin conversion structure may comprise legs of material having positive spin hall angle, and legs of material having negative spin hall angle arranged as illustrated in Figure 1 or Figure 2. A combination of doped and undoped material may be used for the legs - e.g. the positive legs may be provided by a material which inherently (e.g. intrinsically) provides a positive spin Hall angle (such as Pt or Pd) , and the negative legs may be provided by doping of a host material. Likewise, in some embodiments the negative legs may be provided by a material which inherently (e.g. intrinsically) provides a negative spin Hall angle (such as W) , and the positive legs may be provided by doping of a host material .

Different spin orbit scattering may be observed depending on the host material, the dopant, and the dopant density. Testing methods which can be used to identify suitable materials for the positive and negative legs of the spin conversion structure are described below. Examples under testing and evaluation include:

Host material - Gold; Dopant - N - it is believed that other light elements may also work well. Host material - Copper; Dopant Pt, Bi, Tait is believed that, where copper is used as the host material effective dopants are likely to be heavier elements than the dopants used for Gold. One test that can be used to determine the suitability of a material (whether doped or not) to provide the positive or negative legs of the spin conversion structure is described here with reference to Figure 3. This test measures the spin Hall angle of the material. Other options for measurement include ferromagnetic resonance techniques in which a non-magnetic conductor (the test material) is placed in contact with a ferromagnet with precessing magnetisation. Measurements of the test material can be used to determine the spin Hall angle of the material - however in these testing methods some knowledge of other quantities (such as spin diffusion length) may be required. As another alternative, the sign of the spin Hall angle can be measured using a spin Seebeck device. For example, such a spin Seebeck device may comprise a magnetic layer of the same material and the same thickness as to be used in the apparatus 100. This layer may also be applied to the same substrate as to be used in the apparatus 100.

A single leg of the material to be tested can be applied to the surface of the magnetic layer (e.g. by deposition as described above) . A temperature gradient can then be applied through the magnetic layer. The voltage generated in this sample of material in response to this temperature gradient can then be measured. The sign of this voltage in the direction perpendicular to the magnetisation of the magnetic layer (e.g. is it right or left handed with respect to the magnetisation direction) can be used to determine whether the material is suitable for use as a positive leg, or a negative leg. This same test can be used to select the thickness of the legs of the spin conversion structure.

Figure 3 illustrates an apparatus which can be used in an alternative measurement method for this purpose. This is a so- called spin absorption method of measurement.

The apparatus in Figure 3 comprises an electrically conductive, non-magnetic, bar 400, which may comprise copper. This is arranged at right angles to two ferromagnetic bars 402, 404 magnetized along their length. These magnetic bars 402, 404 may comprise a material such as permalloy.

The magnetic bars 402, 404 are laid across the conductive bar 400 in electrical contact with it, and spaced apart along its length.

Also laid across the conductive bar, between the two magnetic bars is a bar of the test material 406. This too is at right angles to the conductive bar 400 and in electrical contact with it .

A charge current source 408 is conductively coupled to a contact at one end of a first one 402 of the two magnetic bars, and another at an end of the conductive bar 400 adjacent to that magnetic bar. The current source 408 can therefore provide a charge current through the magnetic bar 402 and through the conductive bar 400 between these two contacts.

A voltmeter 410 is coupled to the ends of the bar of test material 406 for measuring the voltage across the bar 406 (e.g. between its two ends transverse to the direction of the conductive bar) .

To perform the test, a spin current is generated in the conductive bar by operating the current source to pass charge current through the first magnetic bar strip and into the nonmagnetic conductive bar. This produces spin accumulation, which produces spin accumulation, which results in a pure spin current in the non-magnetic conductive bar. Most of this spin current may be absorbed into the bar of test material. Accordingly, spin accumulation in the test material gives rise to a voltage across the bar of test material (due to the inverse spin Hall effect) . This voltage can then be measured and used to determine the spin Hall angle To correct for systematic error, the voltage and current leads can be swapped to determine spin Hall effects, and the second Py strip is monitored to determine the level of absorption (of the spin current) into the test material. Although this method may be used to determine the actual spin hall angle of the material, the sign of the spin hall angle can be determined using a spin Seebeck device as explained above.

As noted above, the magnetic layer may comprise a thermal insulator. For example, the thermal conductivity (at room temperature) of the magnetic layer may be less than 20 Wm-lK-1, for example less than 15 Wm-lK-1, for example less than 10 Wm-IK- 1, for example less than 5 Wm-lK-1 for example more than 2 Wm- lK-1, for example more than 3 Wm-lK-1, for example between 3 Wm- lK-1 and 6 Wm-lK-1. The magnetic layer may exhibit a coercive field of at least 50 Oe, for example at least 100 Oe, for example at least 200 Oe . Examples of materials suitable for use in the magnetic layer include YIG and Fe ₃ 0 ₄ . The coercive field should be large enough that the film will not demagnetise from stray field or thermal gradients. The magnetic layer may comprise YIG and Fe304, Bi -doped YIG, (La,Ca)Mn0 ₃ , (La,Sr)Mn0 ₃ , GaMnAs , LaY ₂ Fe ₅ 0 ₁₂ . It is likely that the magnetic layer will include both Iron and Oxygen, but this is not essential. The Oe (oersted) is 1000/4n {-19.5774715 ) amperes per meter, in terms of SI units

As described with reference to Figure 1, the present disclosure may provide an apparatus comprising a magnetic layer and a spin conversion structure. The magnetic layer may comprise epitaxially grown Fe ₃ 0 ₄ . Such layers may be about 50nm thick, and they may be grown on substrates such as SrTi0 ₃ . In some embodiments the magnetic layer may be deposited using physical vapour deposition techniques such as Pulsed Laser Deposition (PLD) . Such substrates may be crystalline or amorphous, for example they may comprise glass . The layers may be prepared in vacuum (e.g. with a base pressure of 5xl0 ^"9 mbar) by pulsed laser deposition. The laser used for such PLD may comprise a frequency doubled Nd:YAG laser (Quanta Ray GCR-5) with a wavelength of 532 nm and 10 Hz repetition rate.

Where the magnetic layer is deposited onto a substrate the substrate may be heated during deposition, for example using a SiC wander track heater.

The films were deposited onto 10 x 10 mm and 22 x 22 mm amorphous glass slides, which were outgassed at 400 °C prior to the growth of the Fe ₃ 0 ₄ layer at the same substrate temperature. The samples were then left to cool in vacuum until they reached room temperature, at which point the Pt layer was deposited. The FeO _x and Pt layers were deposited from stoichiometric Fe ₂ 0 ₃ (Pi-Kern purity 99.9%) and elemental Pt (Testbourne purity 99.99%) targets. The target to substrate distance was 110 mm. The Fe ₃ 0 ₄ layer thickness may be approximately 80 ± 10 nm. The legs of the spin conversion structure may have a thickness of between 2 nm and 16 nm. In some examples, the magnetic layer may have a thickness of the order of (e.g. selected based on, e.g. equal to) the spin diffusion length. It will be appreciated in the context of the present disclosure that in the magnetic layer the magnon free path length may be used instead of the spin diffusion length. The spin diffusion length may be approximately 2-5 nm for Pt, approximately 30 nm for Bi -doped Cu, and hundreds of nm for pure Cu . In preparation for manufacture of the devices of the present disclosure, magnetic layers comprising Fe ₃ 0 ₄ were deposited. Some parameters of these films are illustrated in Figure 4.

Figure 4 shows the results of X-ray diffraction measurements of 80 nm thick Fe ₃ 0 ₄ on 0.3 mm thick glass slides. Figure 4. a shows x-ray diffraction peaks associated with <111>, <311>, <222>, <333>, <622> and <444> crystal planes. Figure 4. (b) shows a close-up of the 311 and 222 peaks seen in Figure 4. (a) . Figure 4. (c) shows a plot of the resistivity of the magnetic layer as a function of temperature, and Figure 4. (d) shows X-ray reflectivity data of a magnetic layer of thickness 79 nm.

X-ray diffraction (XRD) measurements of these layers indicates formation of <111> or <311> texture that is highly sensitive to the direction of the plume during PLD. It was also observed that, due to the instability of the Fe304 phase, some ocFe is present. In addition, resistivity measurements confirmed the presence of the Verwey transition at 117 K, which is a good indicator of film quality. An example of the X-ray Reflectivity (XRR) fits used to obtain film thickness is given in Fig. 4.1. (d) . These layers were deposited on 0.3mm thick glass slides in thicknesses ranging from 60 nm to 90 nm, and inverse Spin Hall effect was observed in conductive layers carried on these magnetic layers. We have also prepared and characterised apparatus which comprises an 80 nm thick Fe ₃ 0 ₄ magnetic layer with a 5 nm thick platinum spin conversion structure. In another example we prepared and characterised a structure having the following layers Fe3O4(80 nm) :Au(5 nm) :CuBi (30 nm) : Au(5nm) - where the numbers in brackets indicate the thicknesses of the respective layers. The first of these two structures exhibits positive spin Hall angle in the platinum layer. The second indicates negative spin Hall angle in the copper doped with bismuth layer. Accordingly, in an embodiment there is provided an apparatus in which the positive legs comprise platinum, and the negative legs comprise copper doped with bismuth.

The spin conversion structures 9 described herein may comprise copper alloy films which may be deposited using physical vapor deposition (PVD) such as sputtering, chemical vapor deposition (CVD) , or electro-chemical deposition (ECD) techniques, or a combination of such techniques. Heat- labile soft materials such as plastics may be used as substrates in some embodiments. In these embodiments a spray-coating technique called "ferrite plating" may also be used to fabricate a magnetic layer in the form of a ferrimagnetic ferrite thin film. Conventional ferrite- film preparation techniques, including sputtering, liquid phase epitaxy, and pulsed laser deposition, may require high temperature processes (ranging from 400° C to 800°C) for crystallizing. Accordingly, a different approach may be used for the formation of ferrite films on heat-labile soft materials, such as plastics. Such a ferrite plating method is based on chemical reaction processes, and thus does not need any high temperature processes, enabling the coating of ferrite films on a variety of substrates including flexible films, which may comprise plastics and/or glass.

The apparatus described herein may be manufactured by such conventional techniques, and may be sold as an integrated part of a larger product, such as an electronic device, a piece of network apparatus such as a bridge switch or router, a mobile telephone, a processor chip, a refrigerator or any other or electrical device in which significant temperature gradients may exist and comprising substrates to which the magnetic layer may be secured (either directly or indirectly) .

Some methods of manufacture have been described, but other methods may also be used. For example, the apparatus may be manufactured by way of additive manufacture, also known as ^X 3D printing' whereby a three-dimensional machine readable model of the apparatus is supplied, in machine readable form, to a ^X 3D printer' adapted to manufacture the apparatus. This may be by additive means such as extrusion deposition, Electron Beam Freeform Fabrication (EBF) , granular materials binding, lamination, photopolymerization, or stereolithography or a combination thereof. The machine readable model comprises a spatial map of the object to be printed, typically in the form of a Cartesian coordinate system defining the object's surfaces. This spatial map may comprise a computer file which may be provided in any one of a number of file conventions. One example of a file convention is a STL (STereoLithography) file which may be in the form of ASCII (American Standard Code for Information Interchange) or binary and specifies areas by way of triangulated surfaces with defined normals and vertices. An alternative file format is AMF (Additive Manufacturing File) which provides the facility to specify the material and texture of each surface as well as allowing for curved triangulated surfaces. The mapping of the apparatus may then be converted into instructions to be executed by 3D printer according to the printing method being used. This may comprise splitting the model into slices (for example, each slice corresponding to an x-y plane, with successive layers building the z dimension) and encoding each slice into a series of instructions. The instructions sent to the 3D printer may comprise Numerical Control (NC) or Computer NC (CNC) instructions, preferably in the form of G-code (also called RS-274), which comprises a series of instructions regarding how the 3D printer should act. The instructions vary depending on the type of 3D printer being used, but in the example of a moving printhead the instructions include: how the printhead should move, when / where to deposit material, the type of material to be deposited, and the flow rate of the deposited material .

The apparatus as described herein may be embodied in one such machine readable model, for example a machine readable map or instructions, for example to enable a physical representation of said apparatus to be produced by 3D printing. This may be in the form of a software code mapping of the apparatus and/or instructions to be supplied to a 3D printer (for example numerical code) . The above embodiments are to be understood as illustrative examples. Further embodiments are envisaged. It is to be understood that any feature described in relation to any one embodiment may be used alone, or in combination with other features described, and may also be used in combination with one or more features of any other of the embodiments, or any combination of any other of the embodiments. Furthermore, equivalents and modifications not described above may also be employed without departing from the scope of the invention, which is defined in the accompanying claims.