Conference Agenda

The constantly shrinking transistor dimensions in microchips demand exceptional imaging performance, particularly in 3D, where achieving high resolution is especially challenging. In this talk, I will present novel 3D X-ray imaging methods that have enabled us to reach a world-record X-ray imaging resolution of 4 nanometers. I will focus on recent advances in synchrotron-based X-ray microscopy, specifically ptychographic X-ray computed tomography. Our latest work addresses nanometer-scale experimental instabilities using a method called burst ptychography, in which multiple short-exposure frames are acquired at each scan position. Additionally, imaging at such resolution typically requires very small sample sizes due to depth-of-field limitations, which we overcame using backpropagation tomography. These innovations enabled 3D imaging of a microchip sample ten times larger than the depth of field, at a resolution of 4 nanometers. I will discuss the broader implications of these advancements for nanoscale metrology and highlight the growing need for high-throughput imaging. In this context, I will also present our progress on data-driven acquisition strategies for high-speed, dose-efficient data collection.

For the purpose of Automatic Differentiation 3D imaging, a method which is simultaneously reliable and computationally performant is highly desired.

Most methods to solve Maxwell’s equations are either poorly performant, or numerically unstable. One method which has broken this dichotomy is the modified Born series (MBS) by Vellekoop et al.. This method formulates a simple iteration which stably converges without sacrificing per-iteration performance. However, in the case of highly-scattering materials, the iteration must be altered to ensure convergence, generally degrading the convergence rate.

After studying the modified Born series, we derived a new variant iteration, making use of a Cayley transform to guarantee stability regardless of scattering strength. Beside this improved stability being desireable in and of itself, this property can be leveraged to potentially outperform the modified Born series, in cases of scattering from strongly-scattering media.

Extreme ultraviolet (EUV) or soft X-ray scatterometry is a powerful technique enabling contactless non-destructive structural characterization of patterned nano-stacks with down to sub-nanometer precision1. The technique requires coherent EUV / soft X-ray radiation and is thus typically limited to a synchrotron or free-electron laser facilities. With the continual improvement of table-top high-harmonic-generation (HHG) sources over the last few decades, this approach has become feasible in a lab-scale environment. Here we present EUV scatteromertric results obtained using the coherent 92 eV HHG source in imec’s AttoLab on grating-type samples highly relevant for semiconductor industry. Roughness characterization was performed via direct detection of scattered light utilizing a synthetic aperture and high-dynamic-range detection coupled with noise suppression techniques. Structural parameters (CD, pitch, height, angles, layer and interlayer diffusion thicknesses) were obtained by fitting the sample model to match simulated diffraction patterns to the experimental ones. The latter were acquired through a wide angular scanning (ranging 0 to 65° from grazing). The simulated patterns were calculated by propagating light (using JCMsuite software) through a model of the structured sample. The results demonstrate sub-nanometer precision and are in good agreement with AFM and TEM measurements.

Accurate surface measurements are crucial for industrial applications. Multiwavelength digital holography (MDH) is a cutting-edge technology that extends the range of holographic measurement to the mesoscopic scale while maintaining sub-micron axial accuracy. MDH achieves this by digitally "beating" precisely measured phasers at two wavelengths to create phase maps at longer wavelength scales. However, laser mode instability and wavelength uncertainty can cause errors in the scaling factor. This leads to misestimations and unwanted phase jumps in the heightmap. Therefore, MDH typically requires highly stable lasers.

Our algorithm addresses this problem by detecting wavelength shifts during post-processing. It uses the residuals in the spatial frequency of the phasers that arises due to the misestimation of wavelength to compensate for the scaling factors and synthetic phase maps. This ensures an accurate, phase-jump-free heightmap. We find this technique essential for making MDH viable with environmental wavelength shifts and more affordable laser sources while maintaining accuracy and precision.

Microscopy with table-top high-harmonic generation (HHG) sources enable high-resolution imaging with excellent material contrast, due to the short wavelength and numerous element-specific absorption edges available in this spectral range. However, accurate characterization of dispersive samples in terms of composition and thickness remains challenging due to the limitations of lens-based optics in this spectral range. Here, we performed spectrally resolved lensless imaging using multiple high harmonics. The diffractive shearing interferometry reconstruction serves as a foundational step for element-sensitive metrology, while ptychographic reconstruction enabled the retrieval of high-precision spectral imaging and quantitative thickness mapping. Our non-destructive method offers a powerful tool to extract both the material composition and layer thicknesses of complex nanostructured samples.