Unraveling the geometry of three-dimensional devices with Raman spectroscopy

In the recent years, novel three-dimensional device architectures have emerged to offer a more efficient control of the channel with the gate. However, the quest for better electrostatic control has led to a tremendous increase in processing complexity requiring more metrology steps with tighter specs. A critical metrology step is the measurement of the device geometry, which is commonly carried out with scatterometry, also called optical critical-dimension (OCD) measurements. In this technique, the reflectivity of an array of devices is measured over a broad range of incident wavelengths and the geometry is reconstructed by fitting the measured reflectivity spectra using an ad-hoc model. However, albeit highly sensitive to most geometrical parameters of the device, OCD lacks material-specific information and solely relies on the change in reflected intensity (i.e. 1 parameter) as the geometry changes. Measuring a complex three-dimensional object with many different materials may therefore lead to correlations between the different geometrical parameters such that reconstruction becomes ambiguous and non-unique.

This project aims at measuring the geometry of complex three-dimensional objects using Raman spectroscopy, a technique which offers material-specific information while also leveraging the unique sensitivity of the optical interactions to any variations in device geometry. Raman spectroscopy is indeed based on the nonlinear interaction of incident light with the crystal lattice of the materials under study. The wavelength and intensity of the emitted Raman-shifted light (i.e. 2 parameters) are independent probes of respectively the nature and volume of the material where the interaction has taken place, which lifts the ambiguity of the standard OCD reconstruction. In other words, measuring the different Raman-scattered light components over a broad spectrum of incident wavelengths allows the unique reconstruction of the complex object geometry with a very high precision.

In order to build the understanding required to achieve this challenging goal, this project proposes to use a combined experimental and theoretical study on structures with an increasingly complex geometry. Starting with the direct problem, the candidate will look at and model the Raman response of arrays of one-dimensional (1D) structures such as arrays of Si fins and metal lines with different dimensions (pitch, width, height,...), before transitioning to more complex 2D structures such as patterned Si/SiGe superlattices. Finally, full three-dimensional devices will be considered. The experiments will be complemented with COMSOL and QuantumATK simulations. The candidate will also look at different options to solve the inverse problem, including fitting algorithms and more elaborate machine-learning solutions.