Challenges for semiconductor spintronics.pdf

nature physics | VOL 3 | MARCH 2007 | /naturephysics 153 REVIEW ARTICLE DAVID D. AWSCHALOM1 AND MICHAEL E. FLATTé2 1Center for Spintronics and Quantum Computation and Department of Physics, University of California, Santa Barbara, California 93106, USA 2Optical Science and Technology Center, Department of Physics and Astronomy, and Department of Electrical and Computer Engineering, University of Iowa, Iowa City, Iowa 52242, USA e-mail: awsch@; michael_fl atte@ Semiconductor information-processing devices are among the most sophisticated, complex high-performance structures. As the construction cost of a new fabrication line approaches $3.5 billion, and 25% of its tools are obsolete and need replacement every three years, it is reasonable to question the appeal of alternate methods of information processing. Is it even realistic to imagine that an entirely distinct method of performing logical operations might be competitive? Metallic spintronic devices, such as hard disk read heads and magnetic random access memory (MRAM) are one of the most successful technologies of the past decade, with scaling trends outdoing even CMOS — can semiconductor analogues provide enough new functionality to warrant interest? Electro-optic modulators are also a successful room-temperature technology. What is the advantage of replacing these devices with magnetic analogues? Quantum computing seems to be progressing rapidly with atoms and superconductors — why use spins in semiconductors? Metallic spintronic devices1, originating from the discovery of giant magnetoresistance (GMR)2,3 in 1988 and the subsequent development of the spin valve4, can be understood by assuming that any current of spins is carried by two ‘types’ of carriers, spin-up and spin-down. Th e two-channel picture of spin transport proposed by Mott5 explains the behaviour of magnetoresistive devices6–10, including GMR and tunnelling magnetoresistance11 (TMR), as well as spin injection into metals