azing at orderly rows of green PacMan-like images on a computer screen, Steve Pennycook can "see" individual columns of silicon atoms and the holes between them. In other materials, he can see defects one millionth the thickness of a human hair.
Pennycook, a scientist in ORNL's Solid State Division, does not have X-ray vision. Instead he has "electron vision"--he uses a powerful electron microscope designed to apply his "Z-contrast imaging" technique to give the sharpest direct images yet of atoms in a solid.
"We now have the ability to use the electron microscope to simply photograph the arrangements of atoms inside materials," says Pennycook. "Our system is more economical than conventional microscopes and provides images that are much easier to interpret."
A 300-kilovolt scanning transmission electron microscope (STEM) built by VG Microscopes of England has been installed at ORNL's Solid State Division. It is being prepared for operation later this year. Earlier tests of it in England indicated that it can see, or resolve, features as small as 1.3 angstroms. It will complement ORNL's 100-kilovolt STEM, which can image atoms 2.2 angstroms apart. A unit of measure, an angstrom is one ten-billionth of a meter and about one millionth of the diameter of a human hair.
Once in operation, the ORNL microscope should have higher resolution than the 1-million-volt Atomic Resolution Microscope at DOE's Lawrence Berkeley Laboratory, a conventional electron microscope that can resolve individual atoms that are only 1.6 angstroms apart. A 1.3 million volt microscope being built by a Japanese company has achieved 1-angstrom resolution, "but the images produced by these conventional instruments generally cannot be directly interpreted by the eye," Pennycook says.
"The key advantage of the Z-contrast images is that we can interpret them by eye," he adds. "In other words, unexpected structures are immediately apparent."
Furthermore, using special computer programs developed for radio astronomy, Pennycook believes the new ORNL microscope "can break the 1-angstrom barrier" and achieve a resolution on the order of 0.8 angstrom.
Using the 300-kilovolt microscope, Pennycook says, ORNL researchers and collaborators should be able to see copper atoms in high-temperature superconducting material made of oxides of yttrium, barium, and copper. With the 100-kilovolt microscope, scientists can see only the wider-spaced columns of yttrium and barium atoms, which have higher atomic numbers (heavier nuclei based on numbers of protons).
"We should also be able to better observe interfaces--the ways that different crystalline grains of materials come together at grain boundaries," Pennycook says. "Examples are metals, ceramics, semiconductors, and superconductors. This information on grain boundaries is important for many applications of high-tech materials. Examples are increasing electron speed in semiconductors and compact electronic devices and improving the ability of high-temperature superconductors to carry electrical current."
The ORNL method for achieving ultrahigh resolution costs about $2 million, roughly 10% of the cost of building and operating a conventional microscope having equivalent resolution. The conventional approach requires a large building and a team of operators to run and maintain the microscope. The ORNL approach requires only one operator and a normal-size room.
Pennycook says that the interpretation of atomic images is improved by his use of a ring-shaped detector, which picks up electrons scattered through large angles (like foul balls in a baseball game) rather than those moving through the central hole (like balls hit "up the middle" in rapid succession).
"The number of electrons scattered by the sample is directly related to the composition," Pennycook says. "Because atoms of higher atomic number, or Z, scatter more electrons, they produce a brighter image. We call this technique
Z-contrast imaging because, although atoms are always white in the image, the heavier an atom is, the whiter it appears."
The 300-kilovolt microscope has higher resolution than the 100-kilovolt one because its higher accelerating voltage produces electrons of a shorter wavelength. The smaller the wavelength and the more tightly the electron beam can be focused, the sharper the image. Pennycook plans to extend the range of materials he images to include more complex semiconducting and superconducting materials and metals and ceramics.--Carolyn Krause
Chemical Analysis on Atomic Level Achieved
Unknown elements and the ways they link up to each other in materials can now be identified at the atomic level using a microscope technique developed at ORNL. This approach, described in the November 9, 1993, issue of Nature, could lead to more effective high-temperature super- conducting materials for energy-saving applications.
Atomic resolution chemical analysis, a major long-term goal of analytical electron microscopy, has recently been achieved in ORNL's Solid State Division by Steve Pennycook and his colleague Nigel Browning. Using Pennycook's successful Z-contrast technique for imaging the atomic-scale structure of materials in a direct manner, scientists now use a scanning transmission electron microscope to analyze the chemical composition of a material at the atomic level. In this way, they can identify atoms of unknown elements and determine the details of their chemical bonding.
The Z-contrast image itself has chemical sensitivity in that the higher the atomic number, the brighter the image of that atomic column. However, the image cannot by itself reveal the identity of unknown chemical species.
"The Z-contrast technique's great advantage is that it provides a unique image of the atomic structure that can be interpreted directly," Pennycook says. "Even with a completely unexpected structure, the location of the heavier atomic columns can be seen directly."
Because the Z-contrast image uses only those electrons scattered through large angles, the rest of the electrons may be analyzed by an electron spectrometer. It measures the energy the electrons give up to the specimen they pass through.
This technique, called electron energy loss spectroscopy (EELS), can identify unknown chemical species from the fingerprint of energy loss. In addition, the fine structure on the loss peaks in the electron energy spectrum can be used to determine the number and directions of chemical bonds between neighboring atoms.
One of the challenges of microscopy is to determine the oxygen content in grain boundaries--areas where different crystalline grains of a solid come together--of high-temperature superconducting material made of oxides of yttrium, barium, and copper (YBa2Cu3O7).
According to Pennycook, "If there is enough oxygen in the grain boundaries, the material may act as a superconductor, but if there is not enough, it is an insulator." Chemically speaking, the material is superconducting if its formula is YBa2Cu3O7, but it is an insulator if its formula is YBa2Cu3O6. This subtle difference in oxygen content can be determined by the EELS combined with Z-contrast imaging.
ORNL studies have so far revealed the first clear link between structure and oxygen content. Pennycook and Browning have observed one particular type of grain boundary that has no oxygen deficiency.
"This is a particularly exciting observation," he says, "because of the importance of correlating local structure with local superconducting properties. It will greatly assist our scientific understanding of the effect of defects on superconducting properties and may lead to significant technological advances and applications.
For his research Pennycook received the 1992 Materials Research Society Medal and several Department of Energy materials science awards. --Carolyn Krause
First codes run on 512 nodes
Challenge is a key word at ORNL's Center for Computational Sciences. For a year its computer experts struggled to get ORNL's new Intel Paragon XP/S 35 running effectively. ORNL's earlier Paragon machine--the XP/S 5, which has 66 processors, or nodes--had been working soon after the Intel Corporation shipped it to ORNL for testing. But the newer machine was beset by problems largely because of the complications in extending to 512 nodes.
The challenge was recently met by running two large ORNL-developed codes on 512 nodes. Both codes were developed under the Materials Grand Challenge Project of the Partnership in Computational Science (PICS) consortium. The Materials PICS includes ORNL, Brookhaven National Laboratory, and Ames Laboratory.
Grand challenges are huge problems requiring numerous calculations to progress toward a solution. Parallel computers like the Intel Paragon machines have numerous nodes that make calculations simultaneously. They are called supercomputers because if, each node is assigned a small part of a huge problem, the supercomputer can provide solutions in a few hours, whereas previously weeks or even months were required.
The first code to be run on the XP/S 35 was developed by Bill Shelton of the Engineering Physics and Mathematics Division and Malcolm Stock and Yang Wang, both of the Metals and Ceramics Division. This code calculates the physical properties of disordered materials, such as metallic alloys as well as intermetallic and ceramic compounds, on the basis of a fundamental view of the electron "glue" that holds together the atomic nuclei in matter. The code is useful both for understanding physical properties of disordered materials such as electrical resistivity and for predicting the temperatures, pressures, and other conditions under which elements will mix as well as the types of ordered and disordered compounds these elements will form. Such information is critical to the design of new alloy systems and development of new theories.
Another code developed at ORNL under the Materials Grand Challenge Project calculates densities of electrons and surface energies to identify the preferred binding sites for a single germanium atom on a reconstructed silicon surface. The code was developed by Victor Milman, now of the Solid State Division (SSD), and a team of researchers at Cambridge University in England after electron microscope images made by Dave Jesson of SSD suggested that a germanium atom could exchange positions with a silicon atom on a surface. The images were obtained using Steve Pennycook's Z-contrast technique described in the previous article. A colorful visualization of these calculations appears on the front cover of this issue of this issue of the Review.
"Because experimental information on germanium diffusion on the silicon surface is indirect," says Milman,"theoretical modeling is the only way to understand this interaction between germanium and silicon atoms. This large first-principles calculation is possible only due to the availability of the massively parallel computers."
Such calculations combined with information from Pennycook's Z-contrast electron microscope images (see previous highlights) may lead to an understanding at the atomic level of how to grow light-conducting silicon-germanium films. Such "optically active" silicon-germanium films would revolutionize the manufacture of advanced semiconductor devices that incorporate light-wave guides. These devices will increase the speed of information flow because light signals are faster than electrical currents. Wave guides for compact disk players and other uses are currently made of gallium arsenide. Because gallium arsenide is more expensive than silicon and involves the use of the toxic metal arsenic, silicon-germanium films could better meet the need of the computer, communications, and entertainment industries for faster and cheaper semiconductor devices.
These materials challenges will more likely be met if ORNL succeeds at the challenges of operating a new class of parallel machines with an increasing number of nodes. ORNL in 1994 will be testing an XP/S 75 (1024 nodes) and an XP/S 150 (2048 nodes). Won't it be grand to run ORNL codes on more and more nodes?--Carolyn Krause
From Waste to Warheads: