Research Area Summary

We investigate the physics of clean and adsorbed surfaces of semiconductors and metals. Reduced dimensionality plays an important role at surfaces, profoundly influencing electronic and magnetic properties. We also study the interfaces between materials, which are at the heart of technologically important phenomena such as grain-boundary formation, band-offset engineering, and spin injection.


Chains of Gold on Silicon

Figure 1. Proposed "double honeycomb chain" structure of Si(111)5x2-Au. Large circles are Au, small circles are Si. The elementary 5x2 unit cell is outlined. Each unit cell contains two honeycomb chains (HC) based on the outlined hexagons, one of alternating Au and Si atoms, the other of all Si. Three additional Si adatoms, with 5x4 periodicity, are also shown.
Figure 1. Proposed "double honeycomb chain" structure of Si(111)5x2-Au. Large circles are Au, small circles are Si. The elementary 5x2 unit cell is outlined. Each unit cell contains two honeycomb chains (HC) based on the outlined hexagons, one of alternating Au and Si atoms, the other of all Si. Three additional Si adatoms, with 5x4 periodicity, are also shown.

When electrons are confined to two spatial dimensions, a variety of striking physical phenomena can result. Examples, once exotic but now familiar, include the Quantum Hall Effect in a 2-dimensional electron gas (2DEG) and high-temperature superconductivity in layered cuprates. Besides their widespread technological applications, both 2DEGs and high-critical-temperature materials now constitute important testing grounds for theories of electrons in reduced dimensionality.

In one spatial dimension electrons have been predicted to exhibit even more spectacular behavior. For example, the concept of single-particle excitations must be replaced by one of collective excitations. For 1-dimensional (1D) metals even the concept of individual quasiparticles becomes untenable, and gives way to separate spin and charge excitations ("spinons" and "holons").

At least, that's the theory. In practice, 1D metals have been difficult to realize and study experimentally. One general difficulty is that a 1D metal with a half-filled band is unstable due to a Peierls distortion, which renders it insulating. Carbon nanotubes are an important exception, but they must be probed individually because their chirality cannot yet be controlled.

Metallic chains adsorbed on rigid substrates offer a solution to both problems: (1) If the substrate is sufficiently rigid, the energy penalty for distortion may preempt the Peierls instability; (2) If the chains all self-assemble identically, then collective probes such as photoemission become feasible. Since the mid-1970s, it has been known that gold forms nearly perfect chain-like structures on several different faces of silicon, including Si(111) and its nearby tilted ("vicinal") orientations. Early LEED experiments on Au/Si(111) revealed a substantial rearrangement of the substrate, but the large unit cell made it impossible to determine a structural model. Starting in the early 1990s, many research groups have used a variety of surface probes---STM, x-ray diffraction, reflectance spectroscopy, ARPES, inverse photoemission, and core-level spectroscopy---to provide numerous constraints on the structure of Si(111)-Au, but they have not revealed the structure itself. In 1999 high-resolution photoemission results were reported in the journal Nature on a closely related system, Au on vicinal Si(111). The data were argued to be the long-sought "smoking gun" evidence for the spin-charge separation predicted earlier on theoretical grounds. Without a model for the physical structure of Si(111)-Au, however, any detailed comparison between theory and experiment remained moot.

In this work we proposed a complete structural model for Si(111)-Au which, if correct, will open the door to a more fundamental physical understanding of this physically realized 1D metal. The model successfully explains all the main experimental features observed with STM and photoemission. These include the nature and origin of bright protrusions seen in STM; the dispersion of the strong surface band seen in ARPES; and the change in dimensionality of that band from its top to its bottom. The model derives its stability from an unusual "self-doping" mechanism which drives the formation of a period-quadrupling superlattice of Si adatoms; this adatom superlattice is consistent with experimental observation.

The model has wider implications as well. Its structure is modular, consisting of interlocking building blocks that can be easily rearranged on other crystal faces. In this way, we have recently proposed new structural models to explain the observed stability of Au chains on Si(335), Si(553), Si(557), and Si(775). Some of these 1D systems are known to be metallic, while others are insulating. Since their discovery, little progress has been made explaining even such basic differences. Now, with promising structural models finally in hand, new insights into the fundamental questions of how electrons really behave in one dimension are within grasp.
Principal Investigator: Steve Erwin


Growth of Iron on Gallium Arsenide

Figure 1. Potential energy surface (calculated within the local-density approximation) describing diffusion of atomic Fe on GaAs(001).
Figure 1. Potential energy surface (calculated within the local-density approximation) describing diffusion of atomic Fe on GaAs(001).

The integration of magnetism into microelectronics is widely expected to occur over the next decade or two. An important component of this effort is the successful growth of high-quality epitaxial ferromagnetic films on semiconductor substrates. The most promising candidate system is iron on gallium (or indium) arsenide, because of the excellent lattice match of the two bulk materials. Over the past decade, the magnetic properties of the grown Fe films have greatly improved, but problems remain. For example, Fe is usually deposited on GaAs(001) at ~200 degrees C to promote layer-by-layer growth, but this temperature also promotes the diffusion of As into the growing Fe film. The As impurities form antiferromagnetic Fe-As complexes and thus quench the magnetism near the interface, degrading device performance. Two strategies have been developed to address this: using surfactants (such as sulfur) to suppress As diffusion, and using the Ga-terminated surface of GaAs(001) instead of the more common As-terminated surface. An additional complication of growth at relatively high temperatures---to date unexplored---is the possible role of entropy in stabilizing adsorbate and interface morphologies having low-energy vibrational modes. At present, there is little definitive information concerning the optimal combination of substrate termination, growth temperature, or the use of surfactants.

We use density-functional theory to describe the initial stages of Fe film growth on GaAs(001), focusing on the interplay between chemistry and magnetism at the interface. Four features appear to be generic: (1) At submonolayer coverages, a strong chemical interaction between Fe and substrate atoms leads to substitutional adsorption and intermixing. (2) For films of several monolayers and more, atomically abrupt interfaces are energetically favored. (3) For Fe films over a range of thicknesses, both Ga- and As-adlayers dramatically reduce the formation energies of the films, suggesting a surfactant-like action. (4) During the first few monolayers of growth, Ga or As atoms are likely to be liberated from the interface and diffuse to the Fe film surface. Magnetism plays an important auxiliary role for these processes, even in the dilute limit of atomic adsorption. Most of the films exhibit ferromagnetic order even at half-monolayer coverage, while certain adlayer-capped films show a slight preference for antiferromagnetic order.
Principal Investigator: Steve Erwin