Excellence Cluster Universe

Application of deep learning has proved useful in many scientific domains and has also gained increased interest as a tool for weather and climate modeling in recent years. Deep Learning weather models have already demonstrated competitive prediction performance to state-of-the-art methods while hybrid models and emulators have shown some promise for climate simulation. However, the realism of simulated climate variability, and climate modes of pure deep learning models trained only on observational or reanalysis data, has not received as much attention.As one example of these models, we investigate DLESyM, an autoregressive deep learning model based on the U-Net architecture and originally trained on ERA5 reanalysis data from 1981 to 2017 (REF1). Unlike many weather-generating deep learning models, DLESyM does not draw on sea-surface temperatures as boundary conditions, but learns to generate ocean surface patterns. Its applications could, therefore, extend to free-running simulations. The original authors showed its ability to generate stable climate simulations for time-spans up to three millenia, with the absence of spurious drifts and unphysical smoothing in the annual cycle. Here we test how realistic the simulated climate variability of DLESyM is, focusing on interannual to centennial spatio-temporal modes of internal climate variability. We seek to identify whether it is able to generalize to the underlying physical processes of the climate system, or if it is only capable of reproducing spatio-temporal statistical patterns of its training data. We compare the unforced variability of the deep learning model to that in equilibrium simulations out of General Circulation Models out of the Coupled Model Intercomparison Project phase 6 (CMIP6 GCMs), and palaeoclimate reconstructions (REF2). We focus on regional and global power spectra of surface temperatures, and gradients between land and ocean, tropics and extratropics, as well as the high latitudes. To assess the model’s ability to generalize outside the distribution of the training data we perform simulations from varying initial conditions, and comparing them with the output of CMIP6 GCMs. Based on this we discuss potentials and limitations of such a purely data-driven model for climate simulations and future climate risk assessment, where characteristics beyond mean state and slow changes become relevant. REF1 Cresswell-Clay, N., Liu, B., Durran, D. R., Liu, Z., Espinosa, Z. I., Moreno, R. A., & Karlbauer, M. (2025). A deep learning Earth system model for efficient simulation of the observed climate. AGU Advances, 6, e2025AV001706. https://doi.org/10.1029/2025AV001706REF2 Laepple, T., Ziegler, E., Weitzel, N. et al. (2023) Regional but not global temperature variability underestimated by climate models at supradecadal timescales. Nat. Geosci., 16, 958–966. https://doi.org/10.1038/s41561-023-01299-9

Human vision is characterized by frequent eye movements and constant shifts in visual input, yet our perception of the world remains remarkably stable. Here, we directly demonstrate image-specific foveal feedback to primary visual cortex in the context of saccadic eye movements. To this end, we used a gaze-contingent fMRI paradigm, in which peripheral saccade targets disappeared before they could be fixated. Despite no direct foveal stimulation, we were able to decode peripheral saccade targets from foveal retinotopic areas, demonstrating that image-specific feedback during saccade preparation may underlie this effect. Decoding was sensitive to shape but not semantic category of natural images, indicating feedback of only low-to-mid-level information. Cross-decoding to a control condition with foveal stimulus presentation indicates a shared representational format between foveal feedback and direct stimulation. Moreover, eccentricity-dependent analyses showed a U-shaped decoding curve, confirming that these results are not explained by spillover of peripheral activity or large receptive fields. Finally, fluctuations in foveal decodability covaried with activity in the intraparietal sulcus, thus providing a candidate region for driving foveal feedback. These findings suggest that foveal cortex predicts the features of incoming stimuli through feedback from higher cortical areas, which offers a candidate mechanism underlying stable perception.

User-generated interactive mechanisms ("gimmicks") in metaverse platforms offer unique experiences, but their complexity hinders player exploration. Existing navigation agents, reliant on static metadata, cannot comprehend these dynamic gimmicks. We propose the "Conversational Knowledge Acquisition Loop," a method for agents to dynamically learn gimmicks through player conversation. A preliminary study showed that our agent successfully acquired knowledge via conversation and applied it to solve analogous gimmicks.

Abstract There is a growing interest in lithium niobate (LN) as a material for integrated optics and metaphotonics. However, the limited strength of the electro‐optic effect remains a challenge for the realization of optical tunability in LN‐based nanophotonic systems. In this paper, a novel approach is introduced to enhance optical tuning. The proposed mechanism exploits the coupling between the photovoltaic effect induced in ferroelectric meta‐atoms under light illumination and a surrounding liquid crystal (LC) medium. Specifically, the LCs can be locally reoriented by the fringe photovoltaic fields generated by the meta‐atoms, whose spatially‐dependent dielectric anisotropy is modelled with a tensor approach. The LC re‐alignment in the vicinity of the meta‐atoms ultimately results in the self‐tuning of the metasurface optical response. The optical tuning of fundamental electric and magnetic metasurface resonances is numerically demonstrated via multiphysics simulations with and without the presence of an external voltage. This can result in high quality factor optical resonances, and their spectral shifts beyond what is achievable with the sole electro‐optic effect. For the first time to the authors' knowledge, the photovoltaic effect is introduced as a tuning strategy in dielectric metasurfaces, opening new opportunities for reconfigurable nanophotonic devices, such as light modulators, optical limiters, and optical sensors.

Implementing an autonomous agent on a social VR platform where many users share space requires diverse information. In particular, it is required to recognize the distance from other users, their orientation toward each other, the avatar's pose, and text and voice messages, and to behave accordingly. This paper proposes an API to obtain the above information on “Cluster,” a multi-device social VR platform in operation, and an agent that uses the API. We have implemented this API using a network proxy. The agent using this API can connect to ChatGPT [7] and have a conversation in real time. We measured the latency required for the conversation and confirmed that the response time was about 1 second.

In our Galaxy, star formation occurs in a variety of environments, with a large fraction of stars formed in clusters hosting massive stars. OB stars have an important feedback on the evolution of protoplanetary disks orbiting around nearby young stars and likely on the process of planet formation occurring in them. The nearby massive association Cyg OB2 is an outstanding laboratory to study this feedback. It is the closest massive association to our Sun and hosts hundreds of massive stars and thousands of low-mass members, both with and without disks. In this paper, we analyze the spatial variation of the disk fraction (i.e., the fraction of cluster members bearing a disk) in Cyg OB2 and study its correlation with the local values of far-ultraviolet (FUV) and extreme-ultraviolet (EUV) radiation fields and the local stellar surface density. We present definitive evidence that disks are more rapidly dissipated in the regions of the association characterized by intense local UV fields and large stellar density. In particular, the FUV radiation dominates disk dissipation timescales in the proximity (i.e., within 0.5 pc) of the O stars. In the rest of the association, EUV photons potentially induce a significant mass loss from the irradiated disks across the entire association, but the efficiency of this process is reduced at increasing distances from the massive stars owing to absorption by the intervening intracluster material. We find that disk dissipation due to close stellar encounters is negligible in Cyg OB2 and likely to have affected 1% or fewer of the stellar population. Disk dissipation is instead dominated by photoevaporation. We also compare our results to what has been found in other young clusters with different massive populations, concluding that massive associations like Cyg OB2 are potentially hostile to protoplanetary disks but that the environments where disks can safely evolve in planetary systems are likely quite common in our Galaxy.

Abstract We introduce a surface finite element method for the numerical solution of Navier–Stokes equations on evolving surfaces with a prescribed deformation of the surface in the normal direction. The method is based on approaches for the full surface Navier–Stokes equations in the context of fluid‐deformable surfaces and adds a penalization of the normal component of the velocity. Numerical results demonstrate the same optimal order of convergence as proposed for surface (Navier–)Stokes equations on stationary surfaces. The approach is applied to high‐resolution three‐dimensional scans of clothed bodies in motion to provide interactive virtual fluid‐like clothing.

The AMADEUS Collaboration conducts research aimed to experimentally investigate the low-energy K− hadronic interactions with light nuclei like hydrogen, helium, and carbon, in order to provide new constraints to the antikaon-nucleon strong interaction studies in the non-perturbative quantum chromodynamics regime. K− nuclear absorption, both at-rest and in-flight, are explored using the unique low-momentum and monochromatic kaon beam from the DAΦNE collider interacting with the KLOE detector components, a detector characterized by high acceptance and excellent position and momentum resolutions. This paper presents an overview of the AMADEUS results.

Mapping the eco-evolutionary factors shaping the development of animals’ behavioural phenotypes remains a great challenge. Recent advances in ‘big behavioural data’ research—the high-resolution tracking of individuals and the harnessing of that data with powerful analytical tools—have vastly improved our ability to measure and model developing behavioural phenotypes. Applied to the study of behavioural ontogeny, the unfolding of whole behavioural repertoires can be mapped in unprecedented detail with relative ease. This overcomes long-standing experimental bottlenecks and heralds a surge of studies that more finely define and explore behavioural–experiential trajectories across development. In this review, we first provide a brief guide to state-of-the-art approaches that allow the collection and analysis of high-resolution behavioural data across development. We then outline how such approaches can be used to address key issues regarding the ecological and evolutionary factors shaping behavioural development: developmental feedbacks between behaviour and underlying states, early life effects and behavioural transitions, and information integration across development.

Heterostructured Au/CuS nanocrystals (NCs) exhibit localized surface plasmon resonance (LSPR) centered at two different wavelengths (551 and 1051nm) with a slight broadening compared to respective homostructured Au and CuS NC spectra. By applying ultrafast transient absorption spectroscopyweshow that a resonant excitation at the respective LSPR maxima of the heterostructured Au/CuS NCs leads to the characteristic hot charge carrier relaxation associated with both LSPRs in both cases. A comparison of the dual plasmonic heterostructure with a colloidal mixture of homostructured Au and CuS NCs shows that the coupled dual plasmonic interaction is only active in the heterostructured Au/CuS NCs. By investigating the charge carrier dynamics of the process,wefind that the observed interaction is faster than phononic or thermal processes (< 100fs). The relaxation of the generated hot charge carriers is faster for heterostructured nanocrystals and indicates that the interaction occurs as an energy transfer (we propose Landau damping or interaction via LSPR beat oscillations as possible mechanisms) or charge carrier transfer between both materials.Ourresults strengthen the understanding of multiplasmonic interactions in heterostructured Au/CuS NCs and will significantly advance applications where these interactions are essential, such as catalytic reactions.