Annika Bonhoff

This work investigates extreme ultraviolet (EUV)-induced photoemission in materials relevant for EUV lithography, combining synchrotron-based X-ray photoelectron spectroscopy (XPS and ARPES) measurements with Monte Carlo photoemission modeling. Gold (Au) is used as a reference material to benchmark the model, while ruthenium (Ru) thin films represent a technologically important, but less well-characterized capping material for EUV mirrors and masks. For Au, a comparison between experiment and simulation indicates a good understanding of the underlying physics: core-level peak positions and the shape of the secondary electron region are well reproduced. Remaining discrepancies in the intensity ratio between core levels and secondary electron background are resolved by explicitly including valence-band emission. For Ru, XPS and ARPES demonstrate that in-situ argon (Ar) sputtering effectively removes carbon contaminations and reveals Ru-specific spectral features. While the simulation reproduces Ru core-level energies, it fails to capture it’s complex secondary electron shape, indicating that the underlying non-material-specific inelastic scattering data are inaccurate for Ru. These results establish Au as a benchmark for the model and identify material specific inelastic scattering data as the key limitation for the Ru modelling. A dedicated EUV laboratory system with integrated ARPES is planned to measure such material specific loss functions under realistic EUV hydrogen plasma conditions and to extend these studies to EUV optics, multilayers, and mask materials.

The authors present their recent work regarding a lab-scale EUV reflectometry and scatterometry setup for metrology on thin-films, nanoscale gratings and latent images in a positive tone, chemically amplified EUV photoresist. Optical metrology methods like scatterometry are applied for quality control of nanoscale structures e.g., test structures in the photolithographic process. Applying EUV radiation for scatterometry allows to push resolution limits to the sub-nanometer scale, while providing short measurement times and reliable results. Furthermore, EUV scatterometry proofs capable of characterizing latent images in photoresists that are created during exposure which is a crucial tool in the metrology on photoresists materials and processes. In the experimental study, samples with a line grating as well as a latent image of a line grating in a chemically amplified resist are measured by EUV scatterometry. Furthermore, unstructured samples are prepared: one unexposed, the other exposed over the full surface at nominal dose-to-clear of the resist. These two samples are measured by EUV reflectometry to characterize the changes of optical constants of the resist due to exposure. The study demonstrates the capabilities of EUV scatterometry to characterize nanogratings and latent images of nanograting patterns in resists with sensitivity to the chemical contrast of exposed and unexposed resist areas as well as surface structures. Accuracies on geometrical features go as low as the single nanometer and sub-nanometer range.

Diffractive neural networks (DNNs) as physical representations of artificial neural networks have demonstrated remarkably flexible potential as optical systems for laser beam shaping but dynamic reconfiguration typically relies on active spatial light modulators (SLMs). We present a method to train DNNs so that rotating one or more passive phase masks switches the device among multiple, completely distinct beam-shaping functionalities, enabling dynamic beam control without SLMs. The method can also be realized experimentally with reflective DOEs produced with a unique direct laser writing approach.

Sensor positions have a significant impact on the performance of optical sensing systems but often suffer from constraints due to partially blocked targets. In the scope of the EU and BMBF cofunded project “Vision Optics with Integrated VCSELs and Autofocal Lenses” (VIVA), VCSELs with integrated photodiodes (ViP) are placed in the frame of glasses to track the gaze direction of the wearer. Here, we present an optimization tool for positioning these sensors in the frame of the glasses considering constraints due to eyelashes and eyelids. To accurately map the solution space of possible sensor positions, all geometric constraints are simplified for numerically efficient raytracing. The sensor placement area is then discretized. For each position, rays are sent in the direction of the target in a grid and the quality of the position is evaluated by the number of rays hitting the target under a desired angle. The evaluation results are used to select an optimized combination of sensor positions. This allows the sensing system to evaluate the pose of the eye for a variety of gaze directions and glasses positions.

We present a novel approach to designing continuous diffractive optical elements (DOEs) for laser beam shaping using diffractive neural networks (DNNs) with trainable polynomial phase masks. This method enables the creation of single phase masks and systems of cascaded phase masks that achieve consistently high beam shaping accuracy, regardless of the initial guess. Additionally, we demonstrate how the approach can be specifically adapted to incorporate the manufacturing conditions of a new type of continuous reflective DOEs, which we verify experimentally in a setup incorporating two DOEs.

The iterative Fourier transform algorithm (IFTA) is the most widely used algorithm for the generation of phase masks for laser beam shaping in the field of laser material processing. But its simplicity and efficiency also come with heavy limitations. We here present an overview of our research into application adapted laser beam shaping beyond the capabilities of traditional phase retrieval algorithms for beam shaping. The presented algorithms enable, among others, the explicit optimization of phase masks for the generation of target light volumes and for tailored temperature distributions within a work piece. Furthermore, we demonstrate the consideration and pre-compensation of non-linear effects in an optical amplifier for laser beam shaping with subsequent amplification.

Traditional approaches to optics design often involve lengthy iterative processes, highlighting an area in which modern machine learning approaches can provide substantial assistance. The potential of reinforcement learning, an agent-based machine learning method, is being explored for automated optics design. Without relying on example systems for training data, an agent is trained to analyze the current state of an optical system and optimize it through its actions to meet desired specifications. The effectiveness of the agent’s decision-making depends on its ability to observe the current state of the optical system. Apart from lens properties, various other information about the system and rays traversing the system may be beneficial for the agent to determine optimal actions. Thus, the impact of including third- and fifth-order Seidel aberrations as well as ray incidence heights and angles in the observation space is examined. This adaptation consistently results in higher rewards, with third-order aberrations making the most significant contribution.

In Laser Powder Bed Fusion (LPBF), process productivity, stability and quality depend largely on the intensity distribution used. While current state of the art manufacturing machines use Gaussian or ring-shaped intensity profiles, the latest developments in beam shaping optical components allow for spatial and temporal intensity distributions with a much larger number of degrees of freedom. Nevertheless, the large parameter space associated with this makes it difficult to determine advantageous intensity distributions through experimental investigations or classic process simulations. In this work, a reduced process model is presented that enables the determination of such intensity distributions by solving the inverse heat conduction problem for LPBF based on the conjugate gradient method with an adjoint problem. For this purpose, an FEM-based heat conduction simulation is used that models the temperature distribution of an LPBF process by taking temperature-dependent material parameters into account. In particular, the powder bed is modeled by applying effective thermal properties and an adapted volumetric heat source that models the scattering within the powder bed. This allows for investigating the influence of process-sided asymmetries in the material properties as they occur for example during multi-track experiments. It is shown that these asymmetries have an influence on the optimized intensity distribution and should be considered for the integration of optical systems when using LPBF machines with the newest beam shaping technologies.

Spatial light modulators (SLMs) based on liquid crystal on silicon (LCoS) are powerful tools for laser beam shaping as they can be used to dynamically create almost arbitrary intensity distributions. However, laser beam shaping with LCoS-SLMs often suffers from beam shaping artifacts in part caused by unconsidered properties of the LCoS devices: astigmatism that stems from the non-normal incidence of the laser beam on the SLM and the effect commonly referred to as the ’0-th diffraction order’ that is caused by both the crosstalk between neighboring pixels and the direct reflection at the cover glass of the SLM. We here present a method to consider and compensate for these inherent properties of LCoS devices by treating the SLM as a diffractive neural network.