Repository for Publications and Research Data
News from the ETH Library
Recently Added
Direct Polymer-on-Polymer Grafting of Polyolefins under Visible Light
Item type: Journal Article
Kim, Hongsik; Wang, Hyun Suk; Yun, Namkyu; et al. (2026)
Polyolefins dominate global plastic production but resist chemical transformation, leading to persistent waste accumulation. Developing new strategies that can repurpose these waste materials into higher-value products is therefore essential. Although conventional polar small-molecule grafting improves functionality, the resulting densely substituted polyolefins often become softer materials due to lower crystallinity and strength. Here, we report a visible-light-driven radical method that directly grafts diverse vinyl polymers onto polyolefins without the need for catalysts or initiators. The approach is broadly applicable to both pristine and postconsumer polyolefins on a multigram scale. Despite substantial functionalization that markedly increases polarity, the grafted polyolefins retain crystallinity, thermal stability, and mechanical robustness owing to their sparse yet extended polar polymer side chains. As a proof of concept, we demonstrate exceptional adhesion performance, with shear strengths approaching an order of magnitude higher than those of commercial hot-melt adhesives. This work establishes a general principle for polymer-on-polymer grafting of commodity plastics, expanding the conceptual space of polyolefin modification.
Deep interpretable ensembles
Item type: Journal Article
Kook, Lucas; Götschi, Andrea; Baumann, Philipp F. M.; et al. (2026)
Ensembles improve prediction performance and uncertainty quantification by aggregating predictions from mul tiple models. In deep ensembling, these individual models can be black box neural networks or more interpretable models, such as neural additive models. However, interpretability of the ensemble members is generally lost when computing ensemble predictions. This is a crucial drawback of deep ensembles in high-stake decision fields, which interpretable models are desired. We propose a novel transformation ensemble which aggregates prob abilistic predictions with the guarantee to preserve interpretability and yield uniformly better predictions than the ensemble members on average. Transformation ensembles are tailored towards a class of normalizing flows, called deep transformation models, but are applicable to a wider range of probabilistic neural networks. In ex periments on several publicly available data sets, we demonstrate that transformation ensembles perform on par with classical deep ensembles in terms of prediction performance, discrimination, and calibration. In addition, we demonstrate how transformation ensembles capture both aleatoric and algorithmic uncertainty, and produce minimax optimal predictions under certain conditions.
Ligand-Engineered Methylammonium Lead Bromide Nanoplatelets: Single-Photon Emission and Strong Light-Matter Coupling
Item type: Journal Article
Sekh, Taras V.; Kim, Taehee; Sabisch, Sebastian; et al. (2026)
Lead halide perovskite nanoplatelets (LHP NPLs) are of immense interest in the materials science and optoelectronics communities owing to their strong quantum confinement leading to narrow emission peaks, thickness-dependent tunable photoluminescence, and large exciton binding energies. Thus far, their further development and photophysical investigations at ensemble and single-particle levels have been impeded by suboptimal ligand passivation, inferior environmental and colloidal stability compared to their 3D nanocrystal counterparts, and limited compositional diversity. Here, we report highly monodisperse methylammonium lead bromide (MAPbBr3) NPLs (11.3 ± 2.3 × 1.7 ± 0.5 nm) with tunable ligand chemistry and enhanced photoluminescence quantum yields of up to 80%. NPLs capped with zwitterionic ligands exhibit improved stability upon air exposure and sequential purification. Nuclear magnetic resonance spectroscopy on ligand-exchanged NPLs, capped with either phosphocholine- or phosphoethanolamine-type ligands, confirms the partial replacement of the pristine ligands. Owing to the size and shape uniformity of synthesized MAPbBr₃ NPLs, they readily form assemblies of stacked face-to-face NPLs, as observed in both colloidal dispersions and films, giving rise to concentration-dependent multicolor emission. The temperature dependence of the NPL emission exhibits a nonmonotonous trend, attributed to the highly anisotropic confinement and the consequent exciton–phonon coupling. We also observed photoluminescence from single MAPbBr₃ NPLs at both room and cryogenic temperatures, revealing highly polarized fine-structure emission lines with single-photon purity exceeding 80%. The high uniformity and optical transparency of MAPbBr₃ NPL films enabled their integration into optical cavities, where they exhibited strong light-matter coupling with a substantial Rabi splitting of 200 meV.
HereSay: Gamifying Urban Crowdsourcing Through Affordance-Based Data Collection
Item type: Conference Paper
Baer, Manuel F.; Fonseca Alves, Luma; Laato, Samuli (2026)
We present HereSay, a gamified mobile crowdsourcing application that transforms urban observation into engaging gameplay whilst collecting valuable data for participatory planning. The app addresses the challenge of sustaining citizen engagement in urban planning through five integrated gamification mechanics: progressive unlocking, daily streaks, experience points, levels and leaderboards. Unlike traditional planning tools that solely rely on functional categorisation, HereSay adds affordance-based prompts to capture how citizens actually perceive and use urban spaces, with particular emphasis on urban landscape perception. Built using Flutter and Firebase for cross-platform accessibility, the application requires only standard smartphone capabilities. Initial evaluation with seven users demonstrates strong engagement with gamification features and successful motivation to explore urban spaces. The application showcases a scalable solution for inclusive citizen participation that bridges the gap between informal experiential knowledge and structured planning data.
Landslide susceptibility modeling by interpretable neural network
Item type: Journal Article
Youssef, Khaled; Shao, Kevin; Moon, Seulgi; et al. (2023)
Landslides are notoriously difficult to predict because numerous spatially and temporally varying factors contribute to slope stability. Artificial neural networks (ANN) have been shown to improve prediction accuracy but are largely uninterpretable. Here we introduce an additive ANN optimization framework to assess landslide susceptibility, as well as dataset division and outcome interpretation techniques. We refer to our approach, which features full interpretability, high accuracy, high generalizability and low model complexity, as superposable neural network (SNN) optimization. We validate our approach by training models on landslide inventories from three different easternmost Himalaya regions. Our SNN outperformed physically-based and statistical models and achieved similar performance to state-of-the-art deep neural networks. The SNN models found the product of slope and precipitation and hillslope aspect to be important primary contributors to high landslide susceptibility, which highlights the importance of strong slope-climate couplings, along with microclimates, on landslide occurrences.