A new spin on compound semiconductor devices

Moving electrons with magnetism

An electron can be viewed as a spinning sphere of charge. For those of us who can dredge up those old memories from our undergraduate electromagnetics courses, a spinning charge generates a magnetic field. In the case of the electron, this magnetic field is referred to as its spin, with the direction of the magnetic field pointing up or down depending on the direction of the electron s rotation. Of course an electron cannot be treated as a classical sphere, but it is a quantum mechanical object in which the spin (or magnetic strength) is quantized to a value of one half its angular momentum - a fixed quantity. But still, a magnetic field can be used to manipulate the electron.

The key to controlling an electron s motion is by the use of external magnetic fields that can align the direction of the electron s spin. While this behavior has yet to be exploited in commercial electronic devices, it is the key element in state of the art read heads within magnetic hard-disk drives, through the use of the giant magnetoresistance (GMR) effect. When a layered magnetic thin-film structure consisting of alternating ferromagnetic and nonmagnetic layers is exposed to an external magnetic field, the spin directions of the electrons in the ferromagnetic layers align. This minimizes the spin-dependent scattering of carriers moving through the material and thereby reduces the thin film s resistance. When the spins in the material are anti-aligned, scattering is maximized and resistance is increased. These changes in resistance can now be obtained with very small changes in magnetic fields, hence the name GMR. The very small magnetic domains in hard disks can therefore be read, since these small domains induce large resistance changes in a GMR hard-disk drive head.

GMR-based devices

To extend the exploitation of GMR to electronic devices, an obvious approach would be to introduce the ferromagnetic layers between two semiconductor layers, in an analogous manner to a metal-base transistor, to create what is called a spin-valve transistor (Kim et al.).

Researchers at the University of Twente in the Netherlands have recently reported such a device in which the spin-valve transistor is formed by using an n-type Si emitter and collector, and a ferromagnetic/nonmagnetic-layered base consisting of a Pt/Ni0.8Fe0.2/ Au/Co/Au stack (Monsma et al.). Using a Si/Pt Schottky diode at the emitter, operating under forward bias, ballistic electrons are injected into the emitter at an energy of 0.9 eV. When these high-energy electrons enter the base, they experience scattering dependent on the alignment of the electron s spin within the base material. Therefore, just as in GMR-based disk drives, transport through the base s metallic stack can be manipulated by an applied magnetic field that realigns the electron s spin. By using modest external magnetic fields ranging up to field strengths of 60 Oe, it was observed that the relative collector currents in these spin-valve transistors could be increased by 240%. While such a device as currently configured has few practical applications, it clearly demonstrates the manipulation of transport behavior through an electronic device as a result of external magnetic fields and the subsequent control of electron spin.

An obvious question when considering such a device is in the fabrication sequence used to place the metallic layer between the emitter and collector. In an approach similar to that of the wafer bonding techniques used to insert an SiO2 layer beneath a thin Si layer and a Si substrate, the metallic spin-valve base layer is deposited on a n-type Si substrate. A second Si wafer is bonded to the metallic layer, and then etched back to form the emitter layer. Unfortunately, such bonding techniques double substrate costs, increase processing complexity, and decrease overall device yields.

Compound semiconductor spintronics

Despite the fact that these first spin-valve transistors were fabricated in the Si-material system, there was nothing inherently unique to Si that made these devices work. In fact, it was the use of Si that dictated the necessity of implementing the wafer bonding approach to fabricate the metallic base, due to the lack of suitable ferromagnetic materials that could be epitaxially grown on Si. The ability to epitaxially grow a spin-valve transistor structure, or other electronic devices that exploit the behavior of an electron s spin rather than its charge (collectively referred to as spintronic devices), would simplify processing and open up a potentially wider range of device types and performance capabilities.

While many epitaxial metals and semiconductors can be grown on compound semiconductor substrates, only those that exhibit ferromagnetic behavior at room temperature would be practical for spintronic applications. Fortunately, investigators are beginning to identify such materials, including Fe and Fe-Co alloys on GaAs, and zincblende CrSb/GaAs/CrSb layers, along with (Ga,Mn)As and (In,Mn)As alloys. In addition, it has been shown that in some Mn-doped III-V materials the normally paramagnetic Mn ions can exhibit ferromagnetic behavior in the presence of light, which offers another element of device control beyond that of simply using magnetic fields (Wolf et al.).

The extensive epitaxial capabilities developed for the manufacture of compound semiconductor devices now make these devices ideally suited for the study and realization of many spintronic devices. The first spin-valve transistor may have been fabricated in silicon, but the overall subject of spintronics is definitely an area of research, and hopefully eventual commercial products, that could be dominated by compound semiconductors.

Further reading Kim et al. 2002 Trans. Elec. Dev. 49(5) 847.
Monsma et al. 1988 Science 281 407.
Wolf et al. 2001 Science 294 1488.