Reflectometry is an incredibly powerful tool, allowing accurate metrology of nanometer-scale structures. However, in some cases the reflectometry signal has a weak sensitivity to parameters of interest. To overcome this challenge additional, independent measurement channels are collected using multiple polarizations and measuring angles.
Dark-field detection is a novel approach for collecting spectral information. This approach is especially useful for measurement of samples that are embedded inside a uniform bulk, such as Through-Silicon-Via (TSV).
In order to isolate and highlight the signal related to reflection from TSV side walls, the dark-field method blocks all light, which is specularly reflected from the wafer top surface. Only light that has entered the Via is collected for analysis. This method provides high-sensitivity TSV profile measurements including side walls and bottom characteristics.
The dark-field scheme highlights specific features of the measured structure, and is completely insensitive to others. It allows isolation geometrical parameters of interest, avoiding the complications arising from spectral correlations with uninteresting parameters.
By removing signal contributions, which are not beneficial to the analysis (i.e. not sensitive to the parameters of interest), the noise related to this signal is also removed. While signal to noise ratio (SNR) is not generally improved in the dark-field scheme, the ratio between the beneficial part of the signal and the noise is greatly improved.