Research Area Summary
The Center studies a broad range of magnetic materials, from hard magnetic materials to dilute magnetic semiconductors. Some of these materials form the basis of current magnetoelectronic technologies, while others -- both soft and hard magnets -- are being studied for future applications. In addition, this research focuses on semiconductors that are rendered magnetic by either intrinsic or extrinsic effects. These materials, which offer the advantages of semiconductors combined with the non-volatile properties of magnetic materials, are the materials foundation for future "spintronics" technologies.
Magnetism in Metals
We are interested in a broad class of physical problem related to magnetism in metals. This includes such diverse issues as microscopic understanding of magnetic anisotropy, nuclear magnetic resonance in metals, and the theory of spectroscopic probes of spin polarization. In the last few years the main topics of our research in this area were itinerant ferromagnetism near a quantum critical point and frustrated low-dimensional magnetism. Strongly correlated magnets, especially oxides, are also a subject of intensive investigation. These subjects are intimately related to spin-fluctuation properties and, as such, to unconventional superconductivity.
Principal Investigators: Igor Mazin, Michelle Johannes
In the late 1980s and early 1990s, researchers in Japan and the United States showed that semiconductors could be made magnetic by doping them with manganese. An early finding was that II-VI semiconductors (such as ZnSe) were antiferromagnetic when doped with Mn, while III-V semiconductors (such as GaAs) were ferromagnetic. This suggests a competition between two interactions: a weak antiferromagnetic interaction due to superexchange, and a ferromagnetic interaction due to an indirect exchange mechanism mediated by free carriers. When a divalent dopant such as Mn substitutes for a Group III atom such as Ga, a hole is introduced; this is the reason GaAs becomes ferromagnetic while ZnSe does not.
In 2001, Jonker and coworkers at NRL developed the first ferromagnetic semiconductor based on an elemental host, Mn-doped germanium. To help understand the origins of ferromagnetism in this material, we used density-functional theory to compute the effective coupling strength between Mn spins as a function of their separation. Interestingly, this coupling is antiferromagnetic for nearest-neighbor Mn atoms, but ferromagnetic for all separations greater than this. To calculate the Curie temperature (as a function of Mn concentration), we then used a simple percolation theory based on our computed coupling strengths. The predicted Curie temperature increases approximately linearly with Mn concentration, in agreement with experiment, but the absolute temperatures are too large by roughly a factor 4-5.
The reason for this overestimate is two-fold. First, the local-density approximation to density-functional theory generally overestimates spin coupling strengths by similar factors. Second, the actual materials (both Mn-doped Ge and Mn-doped GaAs) are found to be strongly compensated; that is, their measured hole concentrations are much lower than expected based on the measured Mn concentrations. Recently we proposed that the source of this compensation is Mn in interstitial sites (rather than substitutional sites). We have shown that even though the substitutional site is strongly preferred in the bulk crystal, during the growth process Mn atoms can follow a very low-energy pathway starting from the gas phase and come to rest at a subsurface interstitial site (see Fig. 1). In the bulk crystal environment interstitial Mn is an electron donor, with each interstitial Mn compensating two substitutional Mn.
We are also conducting theoretical research on other types of magnetic semiconductor systems. Our current projects include: a computational search for promising new ferromagnetic semiconductors in the chalcopyrite family; a first-principles investigation of the structure of Fe/GaAs interfaces; a study of Co-doped TiO2, a relatively new dilute magnetic semiconductor with Curie temperatures reported to be 700 K or higher; studies of magnetic impurities in diamond; and a theoretical assessment of spin injection from the magnetic semiconductor CdCr2Se4 into semiconductors such as Si and GaAs.
Principal Investigator: Steven Erwin
Intrinsic Magnetism in Silicon
The integration of single-spin magnetoelectronics into standard silicon technology may soon be possible, if experiments confirm a new theoretical prediction by physicists in the Center for Computational Materials Science and the University of Wisconsin-Madison. We predict that a family of well-known silicon surfaces, stabilized by small amounts of gold atoms, can be intrinsically magnetic despite having no magnetic elements. None of these surfaces has yet been investigated experimentally for magnetism, but the new predictions are already supported indirectly by existing data.
The idea of creating magnetism in a nonmagnetic material by manipulating its structure has long intrigued scientists. The hope of realizing this idea in silicon has been widely discussed for decades, but so far none of these speculations has held up under scrutiny.
Silicon provides a unique entry point for combining magnetoelectronics based on single spins with standard electronics technology. If a single-spin device can be built on a silicon wafer, input and output electronics can be directly integrated with the magnetic part of the device. This has been an obstacle for current spintronics approaches. For example, spin injection from a metal into silicon is very inefficient unless the metal/semiconductor interface is carefully optimized.
Our predictions have the advantage that nature itself guides, by a self-assembly process, the formation of long chains of polarized electron spins with atomically precise structural order. This integration of structural and magnetic order is crucial for future technologies based on single spins at the atomic level. The magnetic silicon surfaces, one of which is illustrated in Fig. 1, naturally form steps which are stabilized by chains of gold atoms (yellow). According to the our calculations, some of the silicon atoms at the step edges have unpaired electrons that are fully spin polarized and probably magnetically ordered at sufficiently low temperatures.
The atom chains on the Si(553)-Au surface were discovered in the group of Franz Himpsel at the University of Wisconsin-Madison. Several other groups worldwide have been investigating such "one-dimensional" silicon surfaces in recent years.
This theoretical work suggests several experiments, such as spin-polarized scanning tunneling microscopy (STM), to test their predictions directly. But there is already indirect experimental evidence to support the possibility of magnetism at silicon surfaces. Two research groups, at Yonsei University in Korea and at Oak Ridge National Laboratory in the US, have found that Si(553)-Au develops periodic "ripples" with two different periodicities at low temperatures. One ripple occurs along the silicon step edges with three times the normal periodicity, and the other along the gold chains with two times the normal periodicity. The theoretically predicted STM image, shown in Fig. 2, reproduces this pattern perfectly. Moreover, this pattern only emerges when magnetism is allowed in the calculation. When magnetism is "turned off" in the theory, the ripples completely vanish. Thus the observation of threefold and twofold ripples offers indirect - if preliminary - confirmation of magnetism.
Linear chains of spin-polarized atoms provide atomically perfect templates for the ultimate memory and logic, in which a single spin represents a bit. One potential application is a "spin shift register" recently proposed theoretically by Gerald Mahan, a theoretical physicist at Pennsylvania State University. Another application is the storage of information in single magnetic atoms. Our work also predicts that the magnitude, and even the sign, of the spin coupling can be changed by doping electrons or holes into surface states. The closely related Si(111)-Au surface can be electron-doped by adsorbates (for example, silicon adatoms) on the surface. By varying this adsorbate population one can perform band-structure engineering with extraordinary precision. The possibility of tuning surface magnetism on Si(553)-Au and its relatives using surface chemistry suggests a fascinating new research direction.