Search

Efficient physics simulations are essential for numerous applications, ranging from realistic cloth animations or smoke effects in video games, to analyzing pollutant dispersion in environmental sciences, to calculating vehicle drag coefficients in engineering applications. Unfortunately, analytical solutions to the underlying physical equations are rarely available, and numerical solutions require high computational resources. Latest developments in the field of physics-based Deep Learning have led to promising efficiency improvements but still suffer from limited generalization capabilities and low accuracy compared to numerical solvers. In this work, we introduce Metamizer, a novel neural optimizer that iteratively solves a wide range of physical systems with high accuracy by minimizing a physics-based loss function. To this end, our approach leverages a scale-invariant architecture that enhances gradient descent updates to accelerate convergence. Since the neural network itself acts as an optimizer, training this neural optimizer falls into the category of meta-optimization approaches. We demonstrate that Metamizer achieves unprecedented accuracy for deep learning based approaches - sometimes approaching machine precision - across multiple PDEs after training on the Laplace, advection-diffusion and incompressible Navier-Stokes equation as well as on cloth simulations. Remarkably, the model also generalizes to PDEs that were not covered during training such as the Poisson, wave and Burgers equation. Our results suggest that Metamizer could have a profound impact on future numerical solvers, paving the way for fast and accurate neural physics simulations without the need for retraining.

3D reconstruction of dynamic scenes is a long-standing problem in computer graphics and increasingly difficult the less information is available. Shape-from-Template (SfT) methods aim to reconstruct a template-based geometry from RGB images or video sequences, often leveraging just a single monocular camera without depth information, such as regular smartphone recordings. Unfortunately, existing reconstruction methods are either unphysical and noisy or slow in optimization. To solve this problem, we propose a novel SfT reconstruction algorithm for cloth using a pre-trained neural surrogate model that is fast to evaluate, stable, and produces smooth reconstructions due to a regularizing physics simulation. Differentiable rendering of the simulated mesh enables pixel-wise comparisons between the reconstruction and a target video sequence that can be used for a gradient-based optimization procedure to extract not only shape information but also physical parameters such as stretching, shearing, or bending stiffness of the cloth. This allows to retain a precise, stable, and smooth reconstructed geometry while reducing the runtime by a factor of 400-500 compared to $\phi$-SfT, a state-of-the-art physics-based SfT approach.

Considering subglacial liquid water, a significant extension of the classical Habitable Zone is obtained. Elaborating on the model of Wandel (2023) it is shown how an atmosphere and liquid water could survive on tidally locked planets closely orbiting an M-dwarf host, extending the Habitable Zone boundary inwards. In addition, subglacial liquid water could extend the Habitable Zone beyond the outer boundary of the conservative Habitable Zone as well. These two results enhance the circumstellar region with a potential for liquid water well beyond the conservative boundaries of the classical Habitable Zone. It is argued that the probable recent JWST detection of atmospheric water vapor on the rocky Earth-sized exoplanet GJ 486 b, along with earlier detections of water on other planets orbiting M-dwarf stars gives an empirical answer to the much-argued question, of whether such planets can support liquid water, organic chemistry and eventually life. It is shown how water on terrestrial planets closely orbiting M-dwarf stars may sustain in a subglacial melting layer. Finally, the model is applied to a few exoplanets demonstrating how water detection may constrain their atmospheric properties., Comment: 9 pages, 2 figures, accepted for publication in the Astronomical Journal

There is growing interest in Bayesian clinical trial designs with informative prior distributions, e.g. for extrapolation of adult data to pediatrics, or use of external controls. While the classical type I error is commonly used to evaluate such designs, it cannot be strictly controlled and it is acknowledged that other metrics may be more appropriate. We focus on two common situations - borrowing control data or information on the treatment contrast - and discuss several fully probabilistic metrics to evaluate the risk of false positive conclusions. Each metric requires specification of a design prior, which can differ from the analysis prior and permits understanding of the behaviour of a Bayesian design under scenarios where the analysis prior differs from the true data generation process. The metrics include the average type I error and the pre-posterior probability of a false positive result. We show that, when borrowing control data, the average type I error is asymptotically (in certain cases strictly) controlled when the analysis and design prior coincide. We illustrate use of these Bayesian metrics with real applications, and discuss how they could facilitate discussions between sponsors, regulators and other stakeholders about the appropriateness of Bayesian borrowing designs for pivotal studies.

This paper focuses on the specification of the weights for the components of mixture priors.

A long-standing issue in astrobiology is whether planets orbiting the most abundant type of stars, M-dwarfs, can support liquid water and eventually life. A new study shows that subglacial melting may provide an answer, significantly extending the habitability region, in particular around M-dwarf stars, which are also the most promising for biosignature detection with the present and near-future technology., Comment: 6 pages, 2 figures, Nature Communications

A new solution to the Fermi Paradox is presented: probes or visits from putative alien civilizations have a very low probability until a civilization reaches a certain age (called the Contact Era) after the onset of radio communications. If biotic planets are common, putative advanced civilizations may preferentially send probes to planets with technosignatures, such as radio broadcastings. The contact probability is defined as the chance to find a nearby civilization located close enough so that it could have detected the earliest radio emissions (the radiosphere) and sent a probe that would reach the Solar System at present. It is found that the current contact probability for Earth is very low unless civilizations are extremely abundant. Since the radiosphere expands with time, so does the contact probability. The Contact Era is defined as the time (since the onset of radio transmissions) at which the contact probability becomes of order unity. At that time alien probes (or messages) become more likely. Unless civilizations are highly abundant, the Contact Era is shown to be of the order of a few hundred to a few thousand years and may be applied not only to physical probes but also to transmissions (i.e. SETI). Consequently, it is shown that civilizations are unlikely to be able to inter-communicate unless their communicative lifetime is at least a few thousand years., Comment: 12 pages, 5 figures, accepted for publication in the Astrophysical Journal

The development and approval of new treatments generates large volumes of results, such as summaries of efficacy and safety. However, it is commonly overlooked that analyzing clinical study data also produces data in the form of results. For example, descriptive statistics and model predictions are data. Although integrating and putting findings into context is a cornerstone of scientific work, analysis results are often neglected as a data source. Results end up stored as "data products" such as PDF documents that are not machine readable or amenable to future analysis. We propose a solution to "calculate once, use many times" by combining analysis results standards with a common data model. This analysis results data model re-frames the target of analyses from static representations of the results (e.g., tables and figures) to a data model with applications in various contexts, including knowledge discovery. Further, we provide a working proof of concept detailing how to approach analyses standardization and construct a schema to store and query analysis results., Comment: 11 pages, 4 figures, 1 table, submitted to the journal Scientific Data

The call for patient-focused drug development is loud and clear, as expressed in the 21st Century Cures Act and in recent guidelines and initiatives of regulatory agencies. Among the factors contributing to modernized drug development and improved health-care activities are easily interpretable measures of clinical benefit. In addition, special care is needed for cancer trials with time-to-event endpoints if the treatment effect is not constant over time. We propose the predictive individual effect which is a patient-centric and tangible measure of clinical benefit under a wide variety of scenarios. It can be obtained by standard predictive calculations under a rank preservation assumption that has been used previously in trials with treatment switching. We discuss four recent Oncology trials that cover situations with proportional as well as non-proportional hazards (delayed treatment effect or crossing of survival curves). It is shown that the predictive individual effect offers valuable insights beyond p-values, estimates of hazard ratios or differences in median survival. Compared to standard statistical measures, the predictive individual effect is a direct, easily interpretable measure of clinical benefit. It facilitates communication among clinicians, patients, and other parties and should therefore be considered in addition to standard statistical results.

Partial Differential Equations (PDEs) are notoriously difficult to solve. In general, closed-form solutions are not available and numerical approximation schemes are computationally expensive. In this paper, we propose to approach the solution of PDEs based on a novel technique that combines the advantages of two recently emerging machine learning based approaches. First, physics-informed neural networks (PINNs) learn continuous solutions of PDEs and can be trained with little to no ground truth data. However, PINNs do not generalize well to unseen domains. Second, convolutional neural networks provide fast inference and generalize but either require large amounts of training data or a physics-constrained loss based on finite differences that can lead to inaccuracies and discretization artifacts. We leverage the advantages of both of these approaches by using Hermite spline kernels in order to continuously interpolate a grid-based state representation that can be handled by a CNN. This allows for training without any precomputed training data using a physics-informed loss function only and provides fast, continuous solutions that generalize to unseen domains. We demonstrate the potential of our method at the examples of the incompressible Navier-Stokes equation and the damped wave equation. Our models are able to learn several intriguing phenomena such as Karman vortex streets, the Magnus effect, Doppler effect, interference patterns and wave reflections. Our quantitative assessment and an interactive real-time demo show that we are narrowing the gap in accuracy of unsupervised ML based methods to industrial CFD solvers while being orders of magnitude faster., Comment: AAAI 2022 (Main Track)

Physically plausible fluid simulations play an important role in modern computer graphics and engineering. However, in order to achieve real-time performance, computational speed needs to be traded-off with physical accuracy. Surrogate fluid models based on neural networks have the potential to achieve both, fast fluid simulations and high physical accuracy. However, these approaches rely on massive amounts of training data, require complex pipelines for training and inference or do not generalize to new fluid domains. In this work, we present significant extensions to a recently proposed deep learning framework, which addresses the aforementioned challenges in 2D. We go from 2D to 3D and propose an efficient architecture to cope with the high demands of 3D grids in terms of memory and computational complexity. Furthermore, we condition the neural fluid model on additional information about the fluid's viscosity and density which allows simulating laminar as well as turbulent flows based on the same surrogate model. Our method allows to train fluid models without requiring fluid simulation data beforehand. Inference is fast and simple, as the fluid model directly maps a fluid state and boundary conditions at a moment t to a subsequent fluid state at t+dt. We obtain real-time fluid simulations on a 128x64x64 grid that include various fluid phenomena such as the Magnus effect or Karman vortex streets and generalize to domain geometries not considered during training. Our method indicates strong improvements in terms of accuracy, speed and generalization capabilities over current 3D NN-based fluid models.

Structural reconstruction of plant roots from MRI is challenging, because of low resolution and low signal-to-noise ratio of the 3D measurements which may lead to disconnectivities and wrongly connected roots. We propose a two-stage approach for this task. The first stage is based on semantic root vs. soil segmentation and finds lowest-cost paths from any root voxel to the shoot. The second stage takes the largest fully connected component generated in the first stage and uses 3D skeletonization to extract a graph structure. We evaluate our method on 22 MRI scans and compare to human expert reconstructions., Comment: Accepted final version. In 25th International Conference on Pattern Recognition (ICPR2020)

We explore the ultrafast generation of spin currents in magnetic multilayer samples by applying fs laser pulses to one layer and measuring the magnetic response in the other layer by element-resolved x-ray spectroscopy. In Ni(5~nm)/Ru(2~nm)/Fe(4~nm), the Ni and Fe magnetization directions couple antiferromagnetically due to the RKKY interaction, but may be oriented parallel through an applied magnetic field. After exciting the top Ni layer with a fs laser pulse, we find that also the Fe layer underneath demagnetizes, with a $4.1 \pm 1.9$\% amplitude difference between parallel and antiparallel orientation of the Ni and Fe magnetizations. We attribute this difference to the influence of a spin current generated by the fs laser pulse that transfers angular momentum from the Ni into the Fe layer. Our results confirm that superdiffusive spin transport plays a role in determining the sub-ps demagnetization dynamics of synthetic antiferromagnetic layers, but also evidence large depolarization effects due to hot electron dynamics, which are independent of the relative alignment of the magnetization in Ni and Fe., Comment: This article may be downloaded for personal use only. Any other use requires prior permission of the author and AIP Publishing

Fast and stable fluid simulations are an essential prerequisite for applications ranging from computer-generated imagery to computer-aided design in research and development. However, solving the partial differential equations of incompressible fluids is a challenging task and traditional numerical approximation schemes come at high computational costs. Recent deep learning based approaches promise vast speed-ups but do not generalize to new fluid domains, require fluid simulation data for training, or rely on complex pipelines that outsource major parts of the fluid simulation to traditional methods. In this work, we propose a novel physics-constrained training approach that generalizes to new fluid domains, requires no fluid simulation data, and allows convolutional neural networks to map a fluid state from time-point t to a subsequent state at time t + dt in a single forward pass. This simplifies the pipeline to train and evaluate neural fluid models. After training, the framework yields models that are capable of fast fluid simulations and can handle various fluid phenomena including the Magnus effect and Karman vortex streets. We present an interactive real-time demo to show the speed and generalization capabilities of our trained models. Moreover, the trained neural networks are efficient differentiable fluid solvers as they offer a differentiable update step to advance the fluid simulation in time. We exploit this fact in a proof-of-concept optimal control experiment. Our models significantly outperform a recent differentiable fluid solver in terms of computational speed and accuracy., Comment: to be published at International Conference on Learning Representations (ICLR) 2021

Superconductivity and charge density waves (CDW) are competitive, yet coexisting orders in cuprate superconductors. To understand their microscopic interdependence, a probe capable of discerning their interaction on its natural length and time scales is necessary. We use ultrafast resonant soft x-ray scattering to track the transient evolution of CDW correlations in YBa$_{2}$Cu$_{3}$O$_{6+x}$ following the quench of superconductivity by an infrared laser pulse. We observe a non-thermal response of the CDW order characterized by a near doubling of the correlation length within $\approx$ 1 picosecond of the superconducting quench. Our results are consistent with a model in which the interaction between superconductivity and CDW manifests inhomogeneously through disruption of spatial coherence, with superconductivity playing the dominant role in stabilizing CDW topological defects, such as discommensurations., Comment: Accepted version. 34 pages, 11 figures, Main text and Supplementary Materials

Following its flyby and first imaging the Pluto-Charon binary, the New Horizons spacecraft visited the Kuiper-Belt-Object (KBO) (486958) 2014 MU69 (Arrokoth). Imaging showed MU69 to be a contact-binary, made of two individual lobes connected by a narrow neck, rotating at low spin period (15.92 h), and having high obliquity (~98 deg), similar to other KBO contact-binaries inferred through photometric observations. The origin of such peculiar configurations is puzzling, and all scenarios suggested for the origins of contact-binaries fail to reproduce such properties and their likely high frequency. Here we show that semi-secular perturbations operating only on ultra-wide (~0.1-0.4 Hill-radius) KBO-binaries can robustly lead to gentle, slow-speed binary mergers at arbitrarily high obliquities, but low rotational velocities, that can reproduce MU69's (and similar oblique contact binaries) characteristics. Using N-body simulations, we find that ~15% of all ultra-wide binaries with cosine-uniform inclination distribution are likely to merge through this process. Moreover, we find that such mergers are sufficiently gentle as to only slightly deform the KBO shape, and can produce the measured rotation speed of MU69. The semi-secular contact-binary formation channel not only explains the observed properties of MU69, but could also apply for other Kuiper/asteroid belt binaries, and for Solar/extra-solar moon systems., Comment: Submitted on 27 June 2019

Magnetic resonance imaging (MRI) enables plant scientists to non-invasively study root system development and root-soil interaction. Challenging recording conditions, such as low resolution and a high level of noise hamper the performance of traditional root extraction algorithms, though. We propose to increase signal-to-noise ratio and resolution by segmenting the scanned volumes into root and soil in super-resolution using a 3D U-Net. Tests on real data show that the trained network is capable to detect most roots successfully and even finds roots that were missed by human annotators. Our experiments show that the segmentation performance can be further improved with modifications of the loss function., Comment: 6 pages, 5 figures, in the 28th European Symposium on Artificial Neural Networks

Ultrafast optical pumping of spatially nonuniform magnetic textures is known to induce far-from-equilibrium spin transport effects. Here, we use ultrafast x-ray diffraction with unprecedented dynamic range to study the laser-induced dynamics of labyrinth domain networks in ferromagnetic CoFe/Ni multilayers. We detected azimuthally isotropic, odd order, magnetic diffraction rings up to 5th order. The amplitudes of all three diffraction rings quench to different degrees within 1.6 ps. In addition, all three of the detected diffraction rings both broaden by 15% and radially contract by 6% during the quench process. We are able to rigorously quantify a 31% ultrafast broadening of the domain walls via Fourier analysis of the order-dependent quenching of the three detected diffraction rings. The broadening of the diffraction rings is interpreted as a reduction in the domain coherence length, but the shift in the ring radius, while unambiguous in its occurrence, remains unexplained. In particular, we demonstrate that a radial shift explained by domain wall broadening can be ruled out. With the unprecedented dynamic range of our data, our results provide convincing evidence that labyrinth domain structures are spatially perturbed at ultrafast speeds under far-from-equilibrium conditions, albeit the mechanism inducing the perturbations remains yet to be clarified.

We use femtosecond resonant soft x-ray scattering to measure the ultrafast optical melting of charge-order correlations in La$_{1.875}$Ba$_{0.125}$CuO$_4$. By analyzing both the energy-resolved and energy-integrated order parameter dynamics, we find evidence of a short-lived nonequilibrium state, whose features are compatible with a sliding charge density wave coherently set in motion by the pump. This transient state exhibits shifts in both the quasielastic line energy and its wave vector, as expected from a classical Doppler effect. The wave vector change is indeed found to directly follow the pump propagation direction. These results demonstrate the existence of sliding charge order behavior in an unconventional charge density wave system and underscore the power of ultrafast optical excitation as a tool to coherently manipulate electronic condensates., Comment: 8 pages, 9 figures, 39 references

The steady advances in computer performance and in programming raise the concern that the ability of computers would overtake that of the human brain, an occurrence termed "the Singularity". While comparing the size of the human brain and the advance in computer capacity, the Singularity has been estimated to occur within a few decades although the capacity of conventional computers may reach its limits in the near future. However, in the last few years, there have been rapid advances in Artificial Intelligence (AI). There are already programs that carry out pattern recognition and self-learning which, at least in limited fields such as chess and other games, are superior to the best human players. Furthermore, the quantum computing revolution, which is expected to vastly increase computer capacities, is already on our doorstep. It now seems inevitable that the Singularity will arrive within the foreseeable future. Biological life, on Earth and on extraterrestrial planets and their satellites, may continue as before, but humanity could be 'replaced' by computers. Older and more advanced intelligent life forms, possibly evolved elsewhere in the universe, may have passed their Singularity a long time ago. Post Singularity life would probably be based not on biochemical reactions but on electronics. Their communication may use effects such as quantum entanglement and be undetectable to us. This may explain the Fermi paradox or at least the "Big Silence" problem in SETI., Comment: 9 pages, 2 figures, accepted by International Journal of Astrobiology (in press)

Phase transitions driven by ultrashort laser pulses have attracted interest both for understanding the fundamental physics of phase transitions and for potential new data storage or device applications. In many cases these transitions involve transient states that are different from those seen in equilibrium. To understand the microscopic properties of these states, it is useful to develop elementally selective probing techniques that operate in the time domain. Here we show fs-time-resolved measurements of V Ledge Resonant Inelastic X-Ray Scattering (RIXS) from the insulating phase of the Mott- Hubbard material V2O3 after ultrafast laser excitation. The probed orbital excitations within the d-shell of the V ion show a sub-ps time response, which evolve at later times to a state that appears electronically indistinguishable from the high-temperature metallic state. Our results demonstrate the potential for RIXS spectroscopy to study the ultrafast orbital dynamics in strongly correlated materials., Comment: 12 pages, 4 figures

The Kepler data show that habitable small planets orbiting Red Dwarf stars (RDs) are abundant, and hence might be promising targets to look at for biomarkers and life. Planets orbiting within the Habitable Zone of RDs are close enough to be tidally locked. Some recent works have cast doubt on the ability of planets orbiting RDs to support life. In contrast, it is shown that temperatures suitable for liquid water and even for organic molecules may exist on tidally locked planets of RDs for a wide range of atmospheres. We chart the surface temperature distribution as a function of the irradiation, greenhouse factor and heat circulation. The habitability boundaries and their dependence on the atmospheric properties are derived. Extending our previous analyses of tidally locked planets, we find that tidally locked as well as synchronous (not completely locked) planets of RDs and K-type stars may support life, for a wider range of orbital distance and atmospheric conditions than previously thought. In particular, it is argued that life clement environments may be possible on tidally locked and synchronously orbiting planets of RDs and K-type stars, with conditions supporting Oxygenic Photosynthesis, which on Earth was a key to Complex life. Different climate projections and the biological significance of tidal locking on putative complex life are reviewed. We show that when the effect of continuous radiation is taken into account, the Photo-synthetically Active Radiation (PAR) available on tidally locked planets, even of RDs, could produce a high Potential Plant Productivity, in analogy to mid-summer growth at high latitudes on Earth. Awaiting the findings of TESS and JWST, we discuss the implications of the above arguments to the detection of biomarkers such as liquid water and oxygen, as well as to the abundance of biotic planets and life., Comment: 23 pages, 5 figures, accepted for publication in the International Journal of Astrobiology. arXiv admin note: text overlap with arXiv:1802.00141

We study the habitability of the two 1-2 Earth-mass planets, recently detected by the CARMENES collaboration, around the ultra-cool nearby M dwarf Teegarden's Star. With orbital periods of 4.9 and 11.4 days, both planets are likely to be within the Habitable Zone and tidally locked. They are among the most Earth-like exoplanets yet discovered. Applying an analytical habitability model we find that surface liquid water could be present on both planets for a wide range of atmospheric properties, which makes them attractive targets for bio-signature searches. The prospects of the planets retaining such an atmosphere over their history are discussed., Comment: 7 pages, 2 figures, published in ApJL