ETH Zurich

Dynamic sparsity is intrinsic to biological computing and is key to its extreme power efficiency. Edge computing systems can improve their energy efficiency and reduce response latency by exploiting this neuromorphic principle. The neuromorphic approach for the extraction of acoustic features replaces conventional ADC and DSP with biological cochlea-inspired filters and event generators implemented in mixed-signal circuits. The resulting sparse feature events drive inference in dynamic-sparsity-aware neural network accelerators to reduce computational load and memory access. The demonstration of edge keyword spotting shows the dynamic savings in power. Exploiting dynamic sparsity at all levels will be the next step toward the design of intelligent devices for the edge.

Over the past decade, lead halide perovskites (LHPs) have become a vibrant thrust in the field of direct conversion X‐ray and gamma‐ray radiation detectors, offering promising cost‐effective and robust alternatives to traditional semiconductors. This review article chronicles the significant strides made since the inception of this field, emphasizing the material, structural, and functional advancements. It begins with an overview of the fundamental properties of perovskites that render them suitable for high‐energy radiation detection, such as their high atomic number, prominent charge carriers’ mobility and lifetime, and high resistivity. The review highlights key developments in material synthesis and processing techniques that have enhanced these detectors' stability, efficiency, and scalability. Furthermore, the review discusses the evolution of device architectures from single‐channel photodiodes to complex multi‐pixel arrays for imaging applications. The conclusion is focused on the remaining challenges that hamper the immediate progression of LHP radiation detectors to higher technology levels. This review is intended as a resource for academic researchers and industry stakeholders, summarizing the first decade of LHP detectors and forecasting the trajectory of this promising field, while remembering that forecasting the future trajectory, though challenging, is guided by current technological trends.

Recent advances in research have made global sensitivity analysis of very large and highly linear life cycle assessment systems feasible. In this paper, we build on these developments to include sensitivity analysis of correlated parameters and nonlinear models. We augment numerical uncertainty propagation with Monte Carlo simulations (i) to include propagation of uncertainty from uncertain variables in parameterized inventory datasets; (ii) to account for correlations between process inputs and outputs and in particular incorporate the carbon balance of combustion activities; (iii) to employ published time‐series data instead of static values for electricity generation market mixes in Europe; (iv) to ensure that inputs which are supposed to reach a fixed total (e.g., the percentage contributions of power sources to an electricity mix) actually do so consistently by using the Dirichlet distribution. We then iterate on existing global sensitivity analysis protocols for high‐dimensional systems to improve their computational performance. To correctly calculate sensitivity rankings for correlated inputs, we use SHapley Additive exPlanations as feature importance metrics with gradient boosted trees. Our results for a case study of climate change impacts of an average Swiss household confirm that neglecting correlations limits the validity of uncertainty and sensitivity analysis. Our methodology and correlated sampling modules are given as open source code.

The pharmaceutical industry is increasingly shifting to decentralized clinical trials (DCTs) conducted at the patient's home, sometimes including trial material home delivery. The traditional clinical trial (CT) is conducted at the investigational site. Research suggests that centralized and decentralized trials have a large carbon footprint, with DCTs potentially providing patient‐centric solutions. However, leaders must determine how to integrate environmental, economic, and social sustainability pillars into their portfolios and subsequent downstream trial‐level decisions. An online survey was designed and deployed via Eidgenössische Technische Hochschule's (ETH, Swiss Federal Institute of Technology) SurveySelect software to capture perceptions of priorities and tradeoffs when deciding between a DCT and a traditional CT for each pillar. The survey closed on 31st January 2023. A total of 447 participants responded. The findings revealed that the overall cohort prioritized greenhouse gas emissions (22.4%) for environmental impact, trial probability of success (15%) for economic considerations, and patient convenience (23.3%) for social criteria. Overall, the DCT setting was perceived as more sustainable in all pillars. Participants reported tradeoffs centered on patient engagement and bringing new medicines to the market. The results from this survey provide initial insights into international multistakeholder perceptions of the priorities and tradeoffs when choosing between a traditional CT and DCT. The synthesized perceptions inform three key recommendations: the need (1) for simulation studies to guide holistic decision‐making across all pillars as empirical data accumulates, (2) to protect the environment, and (3) to protect the supply chain. As empirical data accumulates, these recommendations provide directionality for further research.

Transition metal dichalcogenides (TMDs) are layered two‐dimensional semiconductors explored for optoelectronic applications, ranging from light‐emitting diodes to single‐photon emitters. Such devices require monolayer TMDs, typically obtained through mechanical exfoliation followed by manual identification with a brightfield optical microscope. While this procedure provides high‐quality crystals, the identification is time‐intensive, low‐throughput, and prone to human error, which significantly limits TMD research. Here, we report a simple and fully automated approach for high‐throughput identification of TMD monolayers using photoluminescence microscopy. We demonstrate the identification of up to three monolayers per second, offering a 10000× efficiency increase compared to conventional search and verification is demonstrated. This ability allows us to measure geometric and photoluminescence‐intensity features of more than 2400 monolayers and bilayers of WSe2, MoSe2, and MoS2. Due to these large numbers, we can study and quantify material properties previously inaccessible. For example, we show that the mean photoluminescence intensity from a monolayer correlates with its size due to reduced emission from its edges. Further, we observe large variations in brightness (up to 10×) from WSe2 monolayers of different batches produced by the same supplier. Therefore, our approach not only increases fabrication efficiency but also enhances sample quality for optoelectronic devices of atomically thin semiconductors.

Python has become the de facto language for scientific computing. Programming in Python is highly productive, mainly due to its rich science-oriented software ecosystem built around the NumPy module. As a result, the demand for Python support in High-Performance Computing (HPC) has skyrocketed. However, the Python language itself does not necessarily offer high performance. This work presents a workflow that retains Python’s high productivity while achieving portable performance across different architectures. The workflow’s key features are HPC-oriented language extensions and a set of automatic optimizations powered by a data-centric intermediate representation. We show performance results and scaling across CPU, GPU, FPGA, and the Piz Daint supercomputer (up to 23,328 cores), with 2.47x and 3.75x speedups over previous-best solutions, first-ever Xilinx and Intel FPGA results of annotated Python, and up to 93.16% scaling efficiency on 512 nodes. Our benchmarks were reproduced in the Student Cluster Competition (SCC) during the Supercomputing Conference (SC) 2022. We present and discuss the student teams’ results.

Photonic emulators have enabled the study of many solid-state and quantum optics phenomena, such as Anderson localization, topological insulators and non-Hermitian dynamics. Current photonic emulators are generally limited to bosonic behaviour with local interactions, but the use of synthetic dimensions offers a pathway to overcome this constraint. Here we investigate the flow of liquid light in modulated fast-gain ring lasers, and we establish a platform for emulating quench dynamics within a synthetic photonic lattice with equal densities across the reciprocal space. We apply an artificial electric field to the lattice and introduce a slow timescale to the flow, given by Bloch oscillations. Despite the dispersion and dissipation in our system, which desynchronize the Wannier–Stark ladder states, we were able to directly observe coherent oscillations facilitated by the fast gain. Additionally, we quenched a steady state of a coupled system onto an uncoupled one, which revealed coherent interactions between the decaying modes. These coherent dynamics resulted from the liquid state of light, which rapidly suppressed fluctuations at the shortest timescale of the system. This platform enriches our understanding of collective dynamics in the non-perturbative regime and improves our ability to control and generate coherent, multi-frequency sources.

The Jinshan Gorge in the middle Yellow River, northern China, currently connecting the upstream Hetao Graben to the downstream Weihe Graben, is hypothesized to have recently developed, which implies changing incision rates along the gorge. However, different integration processes have been proposed, such as paleolake regression in the Weihe Graben or integration with a paleolake in the Hetao Graben. These different mechanisms imply different timings, from the Late Miocene to the Late Pleistocene. In this study, we model variations in channel profiles and incision in response to different integration processes using the stream power law. We show that variations due to paleolake regression in the Weihe Graben or integration with paleolake in the Hetao Graben are inconsistent with the incision history preserved by river terraces and reconstructed by inverse analysis of tributary profiles. Instead, a recent increase in the slip rate of the Weihe Graben's boundary faults can explain the main features. Inverse analysis of channel profiles along the Jinshan Gorge suggests that relative uplift rates were a steady ∼0.04 mm/yr before the Early Pliocene, and increased to ∼0.16 mm/yr at present, especially since the Middle‐Late Pleistocene. Our analysis is supported by further data from paleoseismic, drilling and paleostress studies.

Aminopolycarboxylate chelates are emerging as a promising class of electrolyte materials for aqueous redox flow batteries, offering tunable redox potentials, solubility, and pH stability through careful selection of ligands and transition metal ions. Despite their potential, the impact of molecular structure modifications on the electronic and electrochemical properties of these chelates remains underexplored. Here, we examine how introducing a hydroxyl group, often employed for its solubilizing properties, to the backbone of CrPDTA, a reference chelate material, significantly changes the thermodynamics and kinetics of the chelate's redox process. We correlate changes in molecular and electronic structures to different electrochemical responses resulting from the hydroxyl addition and show that the introduction of this functional group leads to a distortion in the octahedral coordination of chromium. Furthermore, increased anisotropic spin density and nonintegral oxidation state changes in the Cr metal center result in a larger barrier for electron transfer in CrPDTA‐OH. It is demonstrated that preserving a hexacoordinate chelate structure across a broad pH range is crucial for efficient flow battery application and it is emphasized that ligand modifications must avoid distorting the octahedral coordination of the transition metal.

Myeloid malignancies exhibit considerable heterogeneity with overlapping clinical and genetic features among subtypes. We present a data-driven approach that integrates mutational features and clinical covariates at diagnosis within networks of their probabilistic relationships, enabling the discovery of patient subgroups. A key strength is its ability to include presumed causal directions in the edges linking clinical and mutational features, and account for them aptly in the clustering. In a cohort of 1323 patients, we identify subgroups that outperform established risk classifications in prognostic accuracy. Our approach generalises well to unseen cohorts with classification based on our subgroups similarly offering advantages in predicting prognosis. Our findings suggest that mutational patterns are often shared across myeloid malignancies, with distinct subtypes potentially representing evolutionary stages en route to leukemia. With pancancer TCGA data, we observe that our modelling framework extends naturally to other cancer types while still offering improvements in subgroup discovery.

Copper (Cu) is a marine micronutrient whose distribution and budget remain incompletely understood. Here, we present a section of dissolved Cu isotope compositions (δ⁶⁵Cu) across the North Atlantic (GEOVIDE cruise, GEOTRACES GA01). High δ⁶⁵Cu are observed in surface waters and co‐vary with carbon uptake rates, indicating light Cu removal by biological activity or complexation of heavy Cu by organic ligands. Beneath the surface, low δ⁶⁵Cu may be partially caused by remineralization. Below 1,500 m, an increase in δ⁶⁵Cu points to removal by particulate scavenging. At greater depths, reversible scavenging, driven by high vertical particulate exports, could explain the increase in Cu concentrations between the surface and deep ocean, mostly in the eastern part of the transect. Investigation of external sources and sinks reveals that anthropogenic aerosols and benthic processes locally supply isotopically light Cu to the ocean, whilst hydrothermal activity above the Reykjanes ridge does not seem to represent a significant source. A striking feature is the low δ⁶⁵Cu observed between 300 and 1,500 m from the Iberian margin to the Icelandic basin, which coincides with elevated non‐conservative dissolved neodymium fractions (Ndxs). This comparison suggests that margin inputs are a source of light Cu to the ocean, and that this Cu can be transported over long distances. The Iberian margin is a hotspot of internal tides and their energy triggers sediment resuspension, leading to particle dissolution and Cu release. These results suggest that continental margins contribute significantly to the missing source of light Cu in the ocean.

Drought events are becoming more severe and recurrent over Europe. Changes in temperature and rain patterns can affect soil nutrient mobility and availability, modulating the biomass and activity of soil microbial communities. Here, we investigated the effects of drought on extracellular polymeric substances (EPS) and microbial biomass carbon (MBC) and nitrogen (MBN) in differently managed cropping systems. An on-field drought simulation experiment using rain-out shelters was conducted as part of a long-term field experiment cultivated with winter wheat, comparing cropping systems with contrasting fertilization strategies and crop protection measures: A biodynamic system and a mixed conventional system with no pesticide application, and a purely minerally fertilized conventional system, with conventional pesticide use. The implemented drought lasted for three months, starting at plant tillering stage and ending at ripening stage. No watering was performed on the drought treatment during that period. Soils were sampled at stem elongation, flowering, and ripening. EPS-carbohydrates and EPS-proteins significantly increased by approximately 20% due to induced drought but remained roughly constant from stem elongation to ripening under drought. Mean EPS-carbohydrates to EPS-proteins ratio was 1.9. MBC and MBN remained largely unaffected by drought. The ratio of both EPS fractions to microbial biomass was lowest in the biodynamic system and highest in the minerally fertilized conventional system, indicating that rhizodeposits and mucilage were predominantly diverted into microbial biomass, rather than into microbial EPS. This might be an important reason for the higher soil fertility of the biodynamic system.

The idea that human needs should be secured for all people is largely uncontroversial, and recent research demonstrates that decent living standards could be secured for all, globally, with far lower energy and resource use than today. However, how the energy requirements of decent living vary across populations is poorly understood – particularly in high-income countries—and important questions regarding inequality remain unexplored. Here we show how, with a fairer distribution of energy, Switzerland could dramatically reduce energy consumption while securing wellbeing for all. We advance previous work on energy and wellbeing by decomposing an established net-zero scenario into the energy required to support human needs, and that related to affluence or excess. We estimate decent living energy in 2050 at 19.5 gigajoules per capita (18–26 gigajoules in varying subnational contexts), making it only ~13% of Switzerland’s 2019 energy footprint, and ~23% of that projected in the net-zero scenario. This highlights the theoretical potential for affluent countries to move towards a more just, egalitarian global distribution of energy and resource consumption, while securing wellbeing for their own citizens.

Recently there has been much effort in developing a quantum generalisation of reference frame transformations. Despite important progress, a complete understanding of their principles is still lacking. Here we derive quantum reference frame transformations for a broad range of symmetry groups from first principles, using only standard quantum theory. Our framework, naturally based on incoherent rather than coherent group averaging, yields reversible transformations that only depend on the reference frames and system of interest. We find more general transformations than those studied so far, which are valid only in a restricted subspace. Our framework contains additional degrees of freedom in the form of an “extra particle”, which carries information about the quantum features of reference frame states. We study the centrally extended Galilei group specifically, highlighting key differences from previous proposals.

Within-species diversity of microorganisms in food systems significantly shapes community function. While next-generation sequencing (NGS) methods have advanced our understanding of microbiomes at the community level, it is essential to recognize the importance of within-species variation for understanding and predicting the functional activities of these communities. This review highlights the substantial variation observed among microbial species in food systems and its implications for their functionality. We discuss a selection of key species in fermented foods and food systems, highlighting examples of strain-level variation and its influence on quality and safety. We present a comprehensive roadmap of methodologies aimed at uncovering this often overlooked underlying diversity. Technologies like long-read marker-gene or shotgun metagenome sequencing offer enhanced resolution of microbial communities and insights into the functional potential of individual strains and should be integrated with techniques such as metabolomics, metatranscriptomics, and metaproteomics to link strain-level microbial community structure to functional activities. Furthermore, the interactions between viruses and microbes that contribute to strain diversity and community stability are also critical to consider. This article highlights existing research and emphasizes the importance of incorporating within-species diversity in microbial community studies to harness their full potential, advance fundamental science, and foster innovation.

Recent advances in next-generation sequencing have opened up new possibilities for applying the human microbiome in various fields, including forensics. Researchers have capitalized on the site-specific microbial communities found in different parts of the body to identify body fluids from biological evidence. Despite promising results, microbiome-based methods have not been integrated into forensic practice due to the lack of standardized protocols and systematic testing of methods on forensically relevant samples. Our study addresses critical decisions in establishing these protocols, focusing on bioinformatics choices and the use of machine learning to present microbiome results in case reports for forensically relevant and challenging samples. In our study, we propose using operational taxonomic units (OTUs) for read data processing and generating heterogeneous training data sets for training a random forest classifier. We incorporated six forensically relevant classes: saliva, semen, skin from hand, penile skin, urine, and vaginal/menstrual fluid, and our classifier achieved a high weighted average F1 score of 0.89. Systematic testing on mock forensic samples, including mixed-source samples and underwear, revealed reliable detection of at least one component of the mixture and the identification of vaginal fluid from underwear substrates. Additionally, when investigating the sexually shared microbiome (sexome) of heterosexual couples, our classifier could potentially infer the nature of sexual activity. We therefore highlight the value of the sexome for assessing the nature of sexual activities in forensic investigations while delineating areas that warrant further research. IMPORTANCE Microbiome-based analyses combined with machine learning offer potential avenues for use in forensic science and other applied fields, yet standardized protocols remain lacking. Moreover, machine learning classifiers have shown promise for predicting body sites in forensics, but they have not been systematically evaluated on complex mixed-source samples. Our study addresses key decisions for establishing standardized protocols and, to our knowledge, is the first to report classification results from uncontrolled mixed-source samples, including sexome (sexually shared microbiome) samples. In our study, we explore both the strengths and limitations of classifying the mixed-source samples while also providing options for tackling the limitations.

Establishing the sustainability of chemical recycling of plastics demands prioritizing realistic feedstocks while advancing catalyst design informed by sustainability-driven frameworks, as envisioned in the ‘Plastic-to-X’ concept. The main outcome of this work is bridging experimental and system-level approaches to explore the interplay of catalyst composition, structure, and hydrogenolysis performance using high-density polyethylene ( M w = 100, 200 kDa) in virgin and in plastic caps form. By incorporating nickel as a modifier to the ruthenium active phase (100% gas yield), we developed titania-supported Ru-Ni alloy nanoparticles (ca. 5 nm) producing 25% of liquid (C 6 –C 45 ) products. Structure sensitivity is elucidated through experiments and simulations, disclosing that energetically favorable backbone scission is preceded by dehydrogenation and hydrogenation cycles over defective alloy sites. Integrating these findings with life cycle and technoeconomic analyses, we predict the potential for profitable processing of plastic caps using the optimal catalyst (2.5 wt% Ru and 5 wt% Ni) with significantly reduced CO 2 emissions even when using green hydrogen. Furthermore, and within the ‘Plastic-to-X’ framework, we identify a minimal average chain length threshold of C 11 for product distributions as a key metric to reconcile environmental and economic objectives.