Innovation in ion beam analysis for nanoelectronic materials

Nanoelectronics relies more and more on novel materials and architectures for technology advancements in terms of power, cost and area. The device functionalities depend on the composition and the amount of impurities of the device materials, whereby controlling the concentrations with high accuracy is essential to tailor the device performances. Besides, it is crucial to characterize devices with relevant dimensions and substrates as it is demonstrated that compositions may be size dependent, and atom areal densities may be substrate dependent. Ion beam analysis, in particular  elastic recoil detection (ERD) and Rutherford backscattering spectrometry (RBS), is recognized to offer compositional quantification as a function of depth in the target. However, in ion beam analysis, mass and depth resolutions are limited by the detector performance. Lateral resolution is hampered by the broad primary beam dimensions, whereby conventional RBS analyses are on blanket layers. Sensitivity to atom areal densities is limited by the counting statistics and background noise.

The objective of this thesis work is to extend ion beam analysis towards next generation nanoelectronics devices through four major advancements, namely high mass resolution, sensitivity, lateral and depth resolution.

The mass resolution in elastic recoil detection is limited by the detector resolution performance, whereby neighboring elements in the periodic table have overlapping recoil distributions. This hampers the quantification of the atom areal densities for recoils with small mass difference. A mass discrimination procedure is developed which deconvolves the overlapping recoil signals. We show that 1 amu mass resolution in ERD can be attained with the mass discrimination algorithm.

In Rutherford backscattering spectrometry, the sensitivity to low amounts of materials is enhanced by reducing the background due to pile-up effects, hence by increasing the signal-to-noise ratio. Pile-up background strongly depends on the count rate, whereby improving this background requires low count rates. The count rate is reduced either through the segmentation of the detector active area and the data acquisition by multiple devices or through the dispersive power of a magnetic spectrometer used to deviate the substrate signal outside of the detector area, thus suppressing the dominant backscattering yield in a conventional RBS spectrum. With such advancements, it is feasible to probe the defectivity at the early stages of area selective atomic layer depositions on plasma treated substrates.

Furthermore, RBS is extended towards the analysis of confined nanostructures with lateral dimensions of as low as 16 nm. The broad beam is utilized to probe simultaneously a multitude of periodically repeated nanostructures embedded in a foreign matrix, whereas the mass difference between the elements in the fins and in the matrix is used to isolate the information from the nanostructure. The extracted compositions average over the ensemble of probed devices, thereby providing a statistically relevant analysis.

Finally, the RBS depth resolution is conventionally limited at 10-15 nm by the detector energy resolution. In a magnetic spectrometer, ions are spatially dispersed as a function of their magnetic rigidities, thus energies, whereby high energy resolution is enabled by a position sensitive detector which records the ion position on the focal plane with high spatial resolution.

The superior energy resolution allows a depth resolution of 2.7 nm of cobalt. When the detector resolution is improved, other sources of energy broadening must be considered, namely the primary beam broadening, the geometrical broadening and the energy spread induced by sample modifications. These contributions are discussed and improvements are proposed to minimize the energy broadenings towards high energy resolution.