December 17, 2007
By: by Alexander E. Braun, Senior Editor, Semiconductor International
"Engineering the strain across a silicon transistor channel 35 nm across is no mean feat," stated Robert Geer, associate professor of nanoscience at the University at Albany's College of Nanoscale Science and Engineering (Albany, N.Y.). While companies like IBM (Yorktown Heights, N.Y.), AMD (Sunnyvale, Calif.) and Intel (Santa, Clara, Calif.) do this routinely, yet to be solved is how to quickly and reliably measure strain in a nanoscale transistor. A way of doing this is by cutting a piece of it and using a TEM to measure the distance between individual silicon atoms. While this approach works well, it is time and labor intensive, as well as destructive. "An easier way would be to shine a light on it and look for a change in the reflected light's wavelength," Geer said, adding that some of the reflected photons surrender or receive part of their energy through their interaction with vibrating silicon atoms. "These are Raman-shifted photons, and their wavelength changes are directly related to the silicon crystal's strain level. This makes measuring silicon device strain as easy as shining light on it," he said.
The problem is that light can only be focused to a spot with a diameter on the order of its wavelength, and a 35 nm channel in a transistor is smaller than the wavelength of light used to measure the reflected photons' Raman shift, making it impossible to receive sufficient resolution. To get around this, Geer is working with tip-enhanced Raman scattering (TERS) techniques. "Apertureless near-field microscopy can reduce the spot size on a surface from hundreds to tens of nanometers," he said.
The concept is straightforward. The illumination of a nanoscale metal particle by the correct wavelength light creates collective surface change oscillations - surface plasmons. The electrical field associated with these oscillates at the incident light's frequency, but decays exponentially as it moves away from the particle surface, creating a nanoscale optical spot. By mounting the particle on the tip of a scanning probe microscope and illuminating it, the resulting nano-optical spot can be used to collect optical spectra from a nanoscale structure.
The Albany group has been successful in using TERS to gain significant spatial resolution improvements over standard optical techniques. "We're working with a tip-enhanced Raman where we use these illuminated tips like little local nanoflashlights to produce more scattering from just the area right around the tip," Geer said. "Measurements of relaxation in patterned silicon test structures show that the method works well. When we measure structure relaxation in these structures, which are relatively large compared to a transistor, we get exactly what one would expect from finite element modeling."
The challenge is that resolution must be pushed down to that of a transistor feature. "For OEMs like Applied Materials and, of course, companies like IBM, Intel and AMD, that's key - tuning the process to measure how uniform your stress engineering is from device to device." Using carbon nanotube (CNT) probes, Geer has obtained resolutions between 50 and 100 nm. "Going below that is a challenge because we're investigating which materials to make the CNT probe tips out of. We've used micro- and nano-machined materials such as silver and gold. This has worked well, but you still need a high degree of control over the tip's exact structure."
Over the next six months, the Albany Nanotech group will look at systems to do these measurements in a multi-wavelength mode; that is, a perfected tip will supply local information on the strained silicon's Raman spectra. "We're looking for good resolution when scanning over the transistor, but also want a 3-D picture and, for that, we must look deeper into the silicon," Geer said. Presently, options being investigated are ultraviolet (UV) wavelengths for the surface and visible wavelengths, whether blue, green, red or near-infrared (IR) to look deeper into the silicon. "A comparison would be going from the single-wavelength ellipsometry used years ago to the kind of spectral version that is now standard for thin-film analysis," he said.
Successfully applying nanotech to produce sufficiently small probes is crucial to the results. "Without a reliable and controllable probe, this will remain a lab instrument and not a tool that can be used on a day-to-day basis," Geer said. "An area we're investigating uses a carbon nanotube metal composite to produce a well-defined probe in nanoscale dimensions - 25, 50 nm across - to give us the desired resolution." The capability to exercise a very high level of control over how these probes are produced is key to good results. A nanowire or some type of self-assembling nanowire will be used for the probes. "Fortunately, much of the groundwork has already been done, and nanowires have already been attached to probes," Geer said. "We're looking for a very special one with a certain metallic component in it, which offers the size control that is obtainable with self-assembled materials, such as CNTs."
This workable CNT probe is beginning evaluation, and Geer is optimistic that it will accomplish what he and his colleagues expect in the way of a measurement breakthrough.