Since the discovery of the spin Hall magnetoresistance (SMR) effect in 2013, there has been much discussion about the relative roles of the ferromagnetic insulator (FMI) and non-magnetic metal (NM) in its underlying mechanism. In most cases Y3Fe5O12 (YIG) is the FMI used in such SMR experiments. YIG has also been a bit of a hot topic material in spintronics for other reasons too recently, being used in spin caloritronics experiments (spin Seebeck effect and spin Peltier effect) and for spin pumping. This week’s Pick of the Week, “Modulation of spin currents with a ferromagnetic insulator” by Villamor et al. in PRB suggests another use of YIG, this time as a magnetic gate in a spin-valve device, using some of the principles applicable to SMR.
The most recent Pick of the Week article looked at the spin transistor, discussing some of the difficulties in realising such a device. For a Datta Das style device, spin-orbit coupling (SOC) plays both a necessary and destructive role; control of the device requires manipulation of spin current with SOC from an electrical gate; however SOC reduces the lifetime of spins, allowing them to lose their information before they reach the detector. The use of ballistic spins injected from quantum point contacts is one way to circumvent to SOC problems in a spin transistor. Another approach, as shown in this week’s pick by Villamor and colleagues, is to find another method of manipulating the spin current that does not require SOC. This therefore allows the spins to propagate along a channel made from a low SOC material, and hence have longer spin diffusion lengths.
Magnetic control of spin
The idea of a magnetic gate has previously been used in an electrical transistor type device to modulate electrical current. Here, Villamor utilises the magnetic gate modulate spin current. Both the electrical and spin current magnetic gate devices share the same principle; that the current in a channel which is electrically disconnected from the magnetic gate varies with the orientation of the gate material’s magnetization. The device for Villamor’s experiment is simple. A NM Cu channel is deposited on-top of a YIG substrate. Ferromagnetic (FM) Co electrodes of different coercively to one another are then deposited towards either end of the Cu channel. The Cu channel and Co contacts form a lateral spin valve (LSV). By sourcing current from one of the Co electrodes to the nearest end of the Cu channel (i.e. in a direction away from the other electrode), a pure spin current diffuses towards the other electrode, and can be detected as a non-local voltage (NLV) between that electrode and the other end of the Cu channel. By switching the relative magnetic orientations of the Co electrodes, the NLV of the LSV changes, a result that has previously been documented. The uniqueness of Villamor’s device is that the NLV can be additionally tuned by YIG underlayer, regardless of the Co electrode orientations.
YIG is an especially useful material for the purpose of Villamor’s experiment for two reasons. Firstly, it is electrically insulating, and so does not shunt any of the current, which could cause an anisotropic magneoresistance effect. Secondly, YIG is a soft magnet, and its coercivity is less than Co. Therefore, by applying a suitable external magnetic field to the device, the magnetization of YIG can be controlled without affecting the magnetizations of the Co electrodes. Rotating a 250 Oe field in the device plane is shown to cause the NLV to oscillate with a cos(2θ) symmetry (or sin(2θ) depending on the Co electrode orientations), where θ is the angle between the external field (YIG magnetization) and Co electrode magnetization. Reference experiments prove that the modulation of the spin current is attributable to the magnetization orientation of the gate, and hence magnetic gating of the spin current is achieved.
Simpler than SMR
So, what is the cause of the effect? It is in-fact very similar to SMR. In SMR, spin-current (generated from the spin Hall effect in a large SOC NM) can be absorbed by an adjacent FMI layer via a spin transfer torque mechanism that depends on the relative orientations of the spins forming the spin current and the magnetization of the FMI. When these vectors are parallel, none of the spin current is absorbed by the FMI, and this leads to a reflection of the spin current back into the NM, and subsequently a change in the charge current due to the inverse spin Hall effect (ISHE). As SOC is weak in Cu and the detection method is a NLV measurement, ISHE from reflected spin current is not an important factor for Villamor’s device. Only the first stage of the SMR mechanism occurs (amount of spin current absorbed into YIG depends on YIG’s magnetization), so it is in many ways a simpler concept than SMR. The electrical NLV is a measure of spin current, and so its variation with YIG magnetization reflects the extent to which the YIG is reducing the spin current in the Cu channel.
As well as the demonstration of magnetic gating of spin current, the paper highlights that the device can also be used (in conjunction with numerical analysis) to obtain a value of the spin-mixing conductance (SMC) at a NM/FMI interface. This is unique as SMC values had only previously been obtained in experiment involving the ISHE, i.e. a large SOC NM material is needed. Villamor records a largest magnetic spin gating effect of his device of 8.3%, which is significantly lower than the signals of last week’s spin transistor; however this value can be improved with a longer NM channel, thinner NM layer, and interfaces with higher spin-mixing conductances.
Stick with Datta Das
I chose this article for this week’s Pick of the Week as it is a nice complement to last week’s article, demonstrating an alternative mechanism to tune spin current. The technological advances that could be made with a spin transistor are enormous, and so it is important to consider different ways in which spins can be controlled. However, the mechanism presented here by Villamor, interesting as it is, would (I expect) be not be as useful for real world application when compared with last week’s. Firstly, the size of the high/low states is lower. Secondly, the materials involved are not compatible with today’s semiconductor based electronics. Thirdly, I would assume that NM substrates would be considerably more expensive than standard semiconductor substrates. Finally, I guess that the energy cost associated with rotating the YIG magnetization (which would probably be needed to be done with a field rather than some type of current induced torque) would be larger than applying a voltage to an electric gate. Of course, I am no expert when making these claims, and I hope that work can be done to make magnetic gating of spin current a more attractive prospect for future technologies.
Estitxu Villamor, Miren Isasa, Saul V ¨ elez, Amilcar Bedoya-Pinto, Paolo Vavassori, Luis E. Hueso (2015). Modulation of pure spin currents with a ferromagnetic insulator Phys Rev B, 91