Light/Matter Interaction to Improve Semiconductor Interfaces
Alexander E. Braun, Senior Editor -- Semiconductor International, 5/8/2008 7:44:00 AM
Research carried out at North Carolina State University (Raleigh) on the interaction of light with matter promises a better understanding of interfaces between different materials, which may lead to the development of new semiconductor technologies and far more efficient, less power-hungry devices.
The work, led by Distinguished University Professor David Aspnes, focuses on second-harmonic generation (SHG), or how light wavelengths are shortened by their interaction with materials. The researchers, including post-doctoral Research Associate Eric Adles, have developed a superior understanding of non-linear optics. This leads to better interpretations of non-linear optical data, and could enable adoption of non-linear optical techniques in metrology. For the semiconductor sector, this could enable better selection and processing of materials that bond to silicon, such as high-k dielectrics.
“We’ve taken a bond model that goes back quite a way in the past and put all the pieces together, including some new ones such as diffraction,” Aspnes explained. “The result is a much more solid basis for interpreting non-linear optical data at the atomic scale. Bond models go back nearly a century, and earlier versions have already proven useful.” The research leader refers to work done some years before on silicon-SiO2 interfaces. At the time, the data could not be understood at the atomic level, so they were presented as 14 tensor coefficients and 11 Fourier coefficients. Using an earlier version of the model, the researchers described everything in terms of three parameters. As Adles put it, “When you reduce 14 to three, you’re making progress.”
Stimulus for the present investigation came from work by Professor Michael Downer’s group at the University of Texas at Austin. Downer’s group examined silicon nanospheres in glass and reported considerable enhancement of SHG signals. Adles noted that, “Any time this happens in non-linear optics, there is reason for excitement because these signals tend to be very weak.”
Among the questions being pondered by the North Carolina State group are why do the enhancements occur, and what can be learned from them? “We started by investigating second-harmonic generation in glass, which is the host material, and exploring several aspects that had not been adequately treated before,” Aspnes said. “For instance, we identify three basic mechanisms that contribute to the second-harmonic signal and obtain analytic expressions for them. Analytic expressions provide considerably more insight than numerical analysis done on a computer.”
The results demonstrate that the treatment of non-linear optics is conceptually simpler than that of linear optics — something not widely realized or appreciated until now. According to Adles, the reason follows from the fundamental four-step process of optics, which was first developed for linear optics in 1912 and 1915, and led to the Ewald-Oseen extinction theorem. “First is to evaluate the electric field at a particular bond site that results from the incoming beam, the laser, for instance. This field exerts a force on the bond charge, and step two is to calculate the acceleration resulting from this force. Step three is to evaluate the radiation from the accelerated charge, and four is to put the radiation from all charges together with diffraction theory. What then results is a description of what is observed, a reflected beam for linear optics and harmonic signals for non-linear optics.”
 |
| A 3-D perspective view of the outermost layer of a silicon crystal of <111> orientation, with oxygen atoms attached to the vertical silicon bonds. The black circles are silicon, the green oxygen. Second-harmonic generation (SHG) is indicated for one bond by a red arrow representing the incident radiation and a blue arrow representing the SHG signal originating from the bond. (Source: North Carolina State University) |
SHG is ideal for the study of silicon dielectric interfaces because strong signals require an ordered array of asymmetric bonds. This means that the contributing bonds must have dissimilar atoms at each end and appear in a regular array. Thus, SHG signals from bulk silicon and disordered dielectric overlayers are both extremely weak, leaving the dominant signal from the interface.
The approach of the North Carolina State group is general for non-linear optical processes. However, odd harmonics, such as the third, are not particularly useful for interface analysis because the above symmetry constraints do not apply and strong signals arise from the bulk. Fourth-harmonic signals are even weaker than second-harmonic signals and, in addition, require special equipment to deal with wavelengths shorter than 200 nm. The researchers believe that a more promising approach is to investigate Raman scattering, which is another diagnostic tool that has not found use in metrology.
