Portland, Ore. A team of research scientists has used an improved electron microscope to confirm a long-held theory concerning the structural nature of doped atomic-scale surfaces. The discovery also promises to give material designers the capability of predicting the composition of materials without having actually to fabricate samples.
The scientists from Oak Ridge National Laboratory (ORNL), Pixon LLC and the Japan Society for the Promotion of Science recently produced images of an atom at resolutions as fine as 0.7 angstrom, a new world record. In doing so, they found out why trace amounts of dopants have such drastic effects on a material's properties.
"It's been one of the world's long-standing unsolved mysteries, how the grains of ceramics form," said ORNL Fellow Stephen Pennycook about the surface of his silicon-nitride test material. "A tiny bit of dopant has a huge effect on a material's properties, but we did not know why."
Silicon nitride forms perfect crystalline surfaces when fabricated in its pure form, but when doped with the slightest amount of a heavy element such as lanthanum its properties change drastically. What atomic-scale architectural features cause such drastic changes? According to Pennycook, his atomic-scale images reveal that "whiskers" are the answer.
"Nobody knew just where the atoms sat," said Pennycook. "There have been theories, of course, but seeing is believing."
Theories predicted that, like catalysts, dopants grew perpendicular to the surface like whiskers, thereby coating the surface with atoms that provide structural strength that is disproportionate to their trace amounts. In semiconductors, such dopants have similar drastic effects on carrier mobility.
"We have confirmed the theory of how grains form our images really show where the atoms are, right on the surface where we predicted, barely an atomic spacing away," said Pennycook. "The ability to see the atoms reveals the structure of a material."
Next, Pennycook's research group plans to tackle semiconductors to reveal what atomic structural changes affect carrier mobility. In particular, the group plans to carefully characterize the layer-by-layer deposition of gate oxides, in an attempt to reveal the atomic structures that make advanced gate dielectrics have such high capacitance (so-called "high-k" dielectrics).
Pennycook, who works for the Condensed Matter Sciences group at ORNL (Oak Ridge, Tenn.), collaborated on the work with Gayle Painter and Paul Becher of ORNL's Metals and Ceramics Division. Visiting researcher Naoya Shibata, as well as researcher William Shelton of ORNL's Computer Sciences and Mathematics Division and engineer Tim Gosnell of Pixon LLC, also took part in the research.
The key to ORNL's world record lies in its unique 300-kilovolt Z-contrast scanning-transmission electron microscope (Stem). A recent retrofit by Pixon improved its resolution down to 0.7 angstrom, or seven-hundredths of a nanometer.
"When we first got the scanning-transmission electron microscope," it had a resolution of "1.3 angstroms," said Pennycook.
At that resolution, atoms (which measure about 2 angstroms) were blurry images. With the doubling of the microscrope's resolution, ORNL was able to clearly image individual atoms.
"With the help of a U.S. company, which has built a corrector that compensates for the intrinsic aberrations in an electron lens, we now have a resolution of around 0.7 angstrom a factor of two better which means our images are four times sharper, meaning we can now pick out individual atoms," said Pennycook.
Resolution: 0.7 angstrom
Shibata, a Fellow of the Japan Society for the Promotion of Science, actually imaged the atoms with the aberration-correction technology provided by Pixon (Setauket, N.Y.).
In action, the Stem works by hosing a material with a beam of electrons. The beam's diameter defines the resolution of the images in this case 0.7 angstrom. Atoms scatter the beams, and a detector picks up how much the beam is scattered. Scattering is more pronounced and thus easier to see for elements with large atomic numbers (heavy elements).
Luckily, dopants typically have large atomic numbers (compared with the relatively light elements like nitrogen and silicon).
"We don't just image heavy elements; we have imaged light elements, too. For instance, we have seen individual nitrogen atoms on the silicon-nitride surface, but we can see heavy elements more easily, and there are quite a few applications where the heavy atoms do the work and the light atoms are just the carriers," said Pennycook. "We are now looking at the atomic layers of high-dielectric oxides, to see how structure affects carrier mobility."
For the future, Pennycook's group will be aiming at creating three-dimensional images. As the resolution of their Stem was increased, its depth of field accordingly decreased. But by performing multiple scans, Pennycook hopes to program a computer to combine the separate scans into a 3-D image.
"We will do multiple scans at different focuses and have the computer compose the 3-D images of semiconductor gate oxides," said Pennycook. "Then we can actually see where the atoms are and how they are lining up."
Pennycook also predicts that as theory is bolstered by the direct evidence of atomic-scale images, the designer of the material will gain more skill in picking the right composition without having to resort to endless trial and error.
"With new confidence in theory, we hope to model materials on a computer screen and predict their properties without having to actually fabricate and characterize a large number of samples, which is very expensive and time-consuming," Pennycook said.
Just the right dopants
In the past, researchers have had to try out many different combinations of ingredients until a good material was discovered, but designers had no confidence that their material was the best they could make, much less optimal. Pennycook's images, however, could pin down the dopants that make for an optimal material.
For instance, theorists predicted that lanthanum was a better dopant than lutetium, because the former was predicted to hug the surface, resulting in long, thin grains, whereas lutetium was predicted to produce short, fat grains. Now the theory has been confirmed by observation.
ORNL is a part of the U.S. Department of Energy. The DOE's Office of Basic Energy Sciences supported the ORNL work with Materials Sciences and Engineering, another DOE division. UT-Battelle LLC manages Oak Ridge National Laboratory for the DOE.