PROJECT LEADER
Caroline Morgan
Wayne State University

PROJECT PARTICIPANTS
Panagiotis Papoulias
Wayne State University

Joseph Schick
Villanova University

Peter Kratzer, Matthias Scheffler
Fritz-Haber-Institut der Max-Planck-Gesellschaft (Germany)

Joseph Landman
SGI and Wayne State University

Petra Specht, Eicke Weber
UC Berkeley

REFERENCES

Schick, J.T., C.G. Morgan, and P. Papoulias. 2000. First-Principles Study of a Nearest Neighbor Arsenic Interstitial Pair in GaAs. Phys. Rev. B (submitted).

Kratzer, P., C.G. Morgan, and M. Scheffler. 1999. Model for Nucleation in GaAs Homoepitaxy derived from First Principles. Phys. Rev. B 59:15246.

Morgan, C.G., P. Kratzer, and M. Scheffler. 1999. Arsenic Dimer Dynamics during MBE Growth: Theoretical Evidence for a Novel Chemisorption State of As₂ Molecules on GaAs Surfaces. Phys. Rev. Lett. 82:4886.

Kratzer, P., C.G. Morgan, and M. Scheffler. 1998. Density-Functional Theory Studies on Microscopic Processes of GaAs Growth. Progress in Surface Science 59:135.

Movement on Semiconductor Surfaces

wide variety of electronic devices are based on multilayered structures composed of semiconductors, whichare used in the manufacture of electronic devices including diodes, transistors, integrated circuits, and--perhaps some day soon--nanodevices. One such semiconductor is gallium arsenide. Layers of gallium arsenide are commonly produced by epitaxial growth--the growth on a substrate of a crystalline substance that mimics the orientation of the substrate. Although static structures of the gallium-arsenide surface have been extensively studied, the kinetic processes involved in growth at this surface have been largely ignored until now. Caroline Morgan of Wayne State University, along with an international group of researchers, are shedding light on atomic-scale dynamics at the gallium-arsenide surface.

Figure 1. Trench Dimer Reconstruction

The trench dimer reconstruction of a GaAs surface--arsenic atoms (grey), gallium atoms (white)--with an incoming arsenic ad-dimer at the preferred "physisorbed" distance above the surface. It keeps the bond length of the isolated molecule (2.1 Ã…), rather than expanding to the bond length of the surface dimers (2.5 Ã…), and there is very little distortion of the GaAs surface below. GaAs growth by molecular beam epitaxy, under moderately arsenic-rich conditions, produces this reconstruction, which grows by filling the trenches and growing the new mountains of the next layer up.

Understanding the fundamental kinetic processes of epitaxial growth is an important step toward optimizing the characteristics of these semiconductor layers by altering the growth conditions. Morgan has developed a model for how the nucleus of a new atomic layer is formed, or nucleation, in gallium-arsenide homoepitaxial growth. The group has also studied the interaction and clustering of adsorbed gallium atoms and the adsorption of arsenic molecules onto clusters of adsorbed gallium atoms. They found that the stable nuclei consist of bound pairs of gallium atoms, which can originate either from dimerization or from an indirect interaction mediated through the substrate reconstruction.

"We found arsenic adsorption to be strongly exothermic on sites with a square array of four gallium dangling bonds," Morgan said. "Putting together all of our results for the possible initial steps of growth, we concluded that growth of the surface is most likely to begin with a partial filling of the trenches of the surface reconstruction, followed by the nucleation of a new atomic layer, which occurs preferentially in regions where the trenches are already partially filled."

ARSENIC BINDING

Morgan's latest discoveries examine where the arsenic molecule binds on the gallium-arsenide surface. If the surface is bare, and no gallium atoms come down, then the arsenic goes on top of the arsenic atoms at the surface. Arsenic prefers the hilltops, not the trenches (Figure 1). If the temperature is low enough, then the arsenic molecule is stable enough to remain on the mountain, or hilltop. "We now understand what happens at low temperature," Morgan said. "Incoming arsenic molecules can bind to the arsenic atoms in the top mountain layer during growth--and if growth occurs too rapidly, some extra arsenic atoms get buried and remain in the crystal."

Morgan also wanted to know how the trench dimer structure grows: how atoms come down on the surface, move around, and finally become incorporated into the crystal. If gallium comes down on the surface, where does it like to go? In trenches, as her research group discovered, it forms stable dimers, and tends to insert itself into the bonds between neighboring arsenic atoms.

"The gallium atom can move up and down the trench," Morgan said. "But it's quite stable after it goes over an energy barrier, breaks a bond between two neighboring arsenic atoms, and inserts itself into that bond. Two gallium atoms in the same trench can find each other eventually and create a dimer. This is a very stable structure. Once enough gallium atoms cluster together to fill the gallium layer in the bottom of the trench, if other gallium atoms fall on the surface nearby, they must remain on the mountain. Incoming arsenic binds to gallium atoms in the trenches and gallium atoms on the mountains more strongly than to other arsenic atoms."

Furthermore, gallium arsenide doesn't make a completely flat surface during growth. When the gallium layer in the bottom of the trench is completely filled and the arsenic layer at the top of the trench is partially filled, new gallium atoms collect on the mountaintops. Then new arsenic molecules come down on the gallium atoms on the mountaintops to start a new mountain layer. Though the trenches are still not completely filled, new mountains will begin to form. This explains why the new mountains form on top of the old mountains, and not on top of the old trenches, as has been observed experimentally (Figure 2).

Over the next year, Morgan's group also plans to extend its work on the identification of arsenic-containing defects in low-temperature-grown, very arsenic-rich gallium arsenide with the collaborating experimental groups of Petra Specht and Eicke Weber at UC Berkeley, and Martina Luysberg at the Forschungszentrum Juelich (in Germany). For this collaboration, Morgan's group has been investigating how the microscopic structure of subsurface defects such as arsenic interstitials and interstitial-containing complexes are modified by the presence of the surface. They are generating simulated scanning tunneling microscopy (STM) images of items such as important defects in gallium arsenide grown by molecular beam epitaxy (MBE) at low substrate temperatures, for comparison with experimental STM images obtained by these groups.

MOLECULAR SLABS

Morgan used a "slab geometry" to model the ad-dimer/substrate systems, with slabs containing seven or eight atomic layers, and an additional bottom layer of pseudohydrogen atoms. When an arsenic molecule comes down on the surface, it is generally not in exactly the right position and orientation to bind immediately to the dangling bonds of the gallium-arsenide surface. Instead, the group discovered that the molecule initially comes into a region above the surface corresponding to an intermediate plateau in the total energy (Figure 1).

In this region, the ad-dimer behaves as if it is physisorbed: it keeps the bond length of the isolated molecule (2.1 Ã…), rather than expanding to the bond length of the surface dimers (2.5 Ã…), and there is very little distortion of the gallium-arsenide surface below. As long as the physisorbed ad-dimer is located over the mountains of the surface, it floats about 2.5 Ã… above the surface. However, its vertical position over the trench is not so well defined, since it can float into the trench without encountering an energy barrier.

If no gallium adatoms are present on the surface, the group found that the strongest binding of the arsenic ad-dimer occurs when it comes down on top of the mountain, breaking the bonds of two adjacent arsenic surface dimers and inserting itself into these bonds.

COMPUTING POWER

Morgan's group used the Car-Parinello-type molecular dynamics code developed by Matthias Scheffler's group at the Fritz-Haber-Institut der Max-Planck-Gesellschaft, which is based on density-functional theory, and is considerably more accurate than less computationally demanding methods. The code takes into account the effect of the electrons in the atomic core on the outer electrons, which participate in the bonding, by means of pseudopotentials, which can be generated and tested using additional software written by Martin Fuchs at the Fritz-Haber-Institut. Morgan used a parallelized and extremely fast version of this molecular dynamics code, which runs on the Cray T3E at SDSC.

Typical runs require about 64 processors, with the data sets saved in 1 GB files, though some of the larger runs required 128 processors. Such larger runs, corresponding to larger energy cutoffs or more atoms, for example, are necessary to improve the accuracy of results for the configurations determined in initial studies to be the most important. In all, the molecular dynamics calculations on the Cray T3E required Fortran90 and the SHMEM libraries, parallel processing (with jobs running optimally on 128 processors for larger jobs, 64 processors for small and medium-sized jobs), 4 GB memory, and 4 GB disk space per active (or queued) run, and a large-capacity, rapid-access archival storage system to store files in between runs.

Other supercomputer support for this project includes a Cray T3E at the DoD NAVO Major Shared Resource Center, a Cray T3E at the Fritz-Haber-Institut in Berlin, the Max Planck Supercomputing Center at Garching, as well as several smaller machines at Wayne State University.

"I have confidence that we understand what's going on here," Morgan said. "We're reproducing what is experimentally accessible and exactly what is going on microscopically." --AV *