James N. Hilfiker, Nina Hong, Rafał Korlacki, Jeffrey S. Hale, Joel Mohrmann, Jeremy Van Derslice, J.A. Woollam Company, Lincoln, NE
For decades, spectroscopic ellipsometry (SE) has been a prominent technique for non-destructive and accurate measurement of thin film thicknesses and optical constants. To understand recent advances, we first need to consider the conventional SE measurement technology. Most SE tools probe the sample of interest using light with wavelengths from the ultraviolet (UV) to the near-infrared (NIR) or even the mid-infrared (IR). SE data are often collected in seconds. The data analysis has become routine for the thickness and refractive index of single-layer coatings, and with some effort, can be extended to more complex structures. Now, let’s examine several new advances in instrumentation and software that are enhancing SE capabilities.
While SE performed at UV to NIR wavelengths is fast, SE measurements in the mid-IR can take hours to achieve an adequate signal-to-noise ratio. Recently, quantum cascade lasers (QCL), with many orders of magnitude more brightness than standard blackbody radiation sources, have been integrated into IR-SE. The extra light allows much faster measurements, which is particularly useful for dynamic data collection, anisotropic characterization requiring multiple sample orientations, and uniformity mapping.
Standard SE measurements describe the transformation of polarized light by the sample. A more complete description of light is considered by the Stokes-Mueller formalism. With this capability, both the cross-polarization and depolarization of light can be quantified. One such example is LiNbO3, used in photonic applications, where the uniaxially anisotropic refractive index can be determined via sample-rotated Mueller matrix (MM) SE measurements. MM-SE is also used for chiral materials and non-symmetric crystal materials. MM-SE measurements have even found applications in many semiconductor processing steps, where critical dimensions (CD) can be determined via the specular scatter measured by MM-SE for 3D memory and logic device structures.
Finally, we will examine how machine learning is transforming the approach to SE data analysis. Will computers put us out of a job? Only time will tell, but maybe we had better start saving for an early retirement.