Aberration-Corrected STEMs Open New Vistas
Alexander E. Braun, Senior Editor -- Semiconductor International, 9/1/2006
Since the electron microscope's inception in the 1930s, resolution limits imposed by electron lens aberrations have complicated its evolution. Compared with their visible-light counterparts, electron lenses are some 50× inferior. So while the ordinary microscope attained its resolution limit — close to the wavelength of visible light — about a century ago, this never quite happened for electron microscopy.
Unsurprisingly, aberration correction became electron microscopy's Holy Grail because, although the wavelength used is several orders of magnitude smaller than that of light, it had not been possible to approach the wavelength of the electrons used (~0.03 Å), which is considerably smaller than the average atom's diameter of 0.1-0.5 nm. Because e-beam resolution is limited to about 50 wavelengths, true atomic resolution was only achieved in the past couple of decades.
Today, complex sets of electron optics correct for dominant aberrations, and limitations have been pushed to the next higher order of aberrations. Over this short time, resolution has been doubled, making the atomic structure of materials easier to see, and providing more contrast than before so that finer details are more visible.
The Electron Microscopy Group of the Materials Science and Technology Division at Oak Ridge National Laboratories (ORNL, Oak Ridge, Tenn.), under the direction of Stephen Pennycook, has been investigating an unanticipated result from aberration correction, analogous to what happens with a common camera: When the aperture is increased, the depth of focus (DoF) is reduced. In a STEM's case, because it can now focus on one specific depth within the sample, it becomes possible to step through the sample's entire thickness merely by changing focus. The effort was aimed at doubling lateral resolution, and since depth resolution improves as the square of the aperture size, a depth resolution 4× better resulted.
 |
| A 3-D representation of the HfO2/SiO2/Si interface structure. It was created by setting intensity thresholds in the 3-D data set. The silicon substrate was color-coded in gold, and the HfO2 film marked yellow. Single Hf atoms in the interface layer were coded separately in green, black, red and blue. Some surface roughness of the HfO2 film was observable at the interface with the SiO2 layer. (Source: ORNL) |
These atoms can be located with an X-Y precision of better than 1 Å, and a depth (Z) precision of better than 5 Å or 0.5 nm. This makes it possible to scan along a long segment, count stray hafnium atoms, and map their positions. Interestingly, no hafnium atoms were present right at the Si/SiO2 interface. After considerable theoretical modeling, the ORNL group concluded that the reason was the strain in the SiO2 when it attaches to the silicon lattice. The oxide structure is compressed, and all the “cages” (the silicon-oxygen/silicon-oxygen rings) in the oxide's structure become smaller and, because hafnium is larger, it is energetically forced to remain at least 2.5 Å away from the silicon. Determining hafnium atom density allows data correlation with bulk transport properties, such as leakage and mobility.
Although aberration-corrected microscopes of this type are available, the ORNL group's prototypes still hold two world records: the direct imaging of a crystal lattice at sub-angstrom resolution, and the spectroscopic identification of a single atom within a bulk material.
The group is investigating the electronic structure around isolated atoms. Besides obtaining an image, it is possible to do spectroscopy. If a particular atom is focused on in 3-D and the transmitted electrons are put through a spectrometer, it is possible to identify the atom from its characteristic absorption edge and extract the surrounding band structure. This enables the probing of electronic structures in 3-D around these features to measure their effect on the surroundings; for example, determining whether there are certain states that might short circuit a device. Currently, however, this requires a large dose of electrons, which might damage the sample by breaking its bonds.
Get more information on inspection, measurement and test.
