Large-scale simulations or scientific experiments produce petabytes of data per run. This poses massive challenges for I/O and storage when scientific analysis workflows are run manually offline. Unsupervised deep learning-based techniques to extract patterns and non-linear relations from these large amounts of data provide a way to build scientific understanding from raw data, reducing the need for manual pre-selection of analysis steps, but require exascale compute and memory to process the full dataset available. In this paper, we demonstrate a heterogeneous streaming workflow in which plasma simulation data is streamed directly to a Machine Learning (ML) application training a model on the simulation data in-transit, completely circumventing the capacity-constrained filesystem bottleneck. This workflow employs openPMD to provide a high level interface to describe scientific data and also uses ADIOS2, to transfer volumes of data that exceed the capabilities of the filesystem. We employ experience replay to avoid catastrophic forgetting in learning from this non-steady state process in a continual manner and adapt it to improve model convergence while learning in-transit. As a proof-of-concept, we approach the ill-posed inverse problem of predicting particle dynamics from radiation in a particle-in-cell (PIConGPU) simulation of the Kelvin-Helmholtz instability (KHI). We detail hardware-software co-design challenges as we scale PIConGPU to full Frontier, the Top-1 system as of June 2024 Top500 list.

Plasma wakefield accelerators use tabletop equipment to produce relativistic femtosecond electron bunches. Optical and X-ray diagnostics have established that their charge concentrates within a micrometre-sized volume, but its sub-micrometre internal distribution, which critically influences gain in free-electron lasers or particle yield in colliders, has proven elusive to characterize. Here, by simultaneously imaging different wavelengths of coherent optical transition radiation that a laser-wakefield-accelerated electron bunch generates when exiting a metal foil, we reveal the structure of the coherently radiating component of bunch charge. The key features of the images are shown to uniquely correlate with how plasma electrons injected into the wake: by a plasma-density discontinuity, by ionizing high-Z gas-target dopants or by uncontrolled laser-plasma dynamics. With additional input from the electron spectra, spatially averaged coherent optical transition radiation spectra and particle-in-cell simulations, we reconstruct coherent three-dimensional charge structures. The results demonstrate an essential metrology for next-generation compact X-ray free-electron lasers driven by plasma-based accelerators. Imaging the visible light emitted from accelerated electron bunches reveals important information about the three-dimensional charge structure of the bunches, which strongly influences the performance of free-electron lasers.

Abstract Plasma wakefield accelerators use tabletop equipment to produce relativistic femtosecond electron bunches. Optical and x-ray diagnostics have established that their charge concentrates within a micrometer-sized volume, but its sub-micrometer internal distribution, which critically influences gain in free-electron lasers or particle yield in colliders, has proven elusive to characterize. Here, by simultaneously imaging different wavelengths of coherent optical transition radiation (COTR) that a laser-wakefield-accelerated e-bunch generated when exiting a metal foil, we elucidated the structure of the coherently-radiating component of bunch charge. Key features of the images correlated uniquely with how plasma electrons injected into the wake: by a plasma-density discontinuity, by ionising high-Z gas-target dopants, or by uncontrolled laser-plasma dynamics. With additional input from electron spectra, spatially-averaged COTR spectra, and particle-in-cell simulations, we reconstructed coherent 3D charge structures. The results demonstrate essential metrology for next-generation X-ray free-electron lasers.

We formulate a data independent latent space regularisation constraint for general unsupervised autoencoders. The regularisation rests on sampling the autoencoder Jacobian in Legendre nodes, being the centre of the Gauss-Legendre quadrature. Revisiting this classic enables to prove that regularised autoencoders ensure a one-to-one re-embedding of the initial data manifold to its latent representation. Demonstrations show that prior proposed regularisation strategies, such as contractive autoencoding, cause topological defects already for simple examples, and so do convolutional based (variational) autoencoders. In contrast, topological preservation is ensured already by standard multilayer perceptron neural networks when being regularised due to our contribution. This observation extends through the classic FashionMNIST dataset up to real world encoding problems for MRI brain scans, suggesting that, across disciplines, reliable low dimensional representations of complex high-dimensional datasets can be delivered due to this regularisation technique.

Pollution with chloroethenes threaten groundwater resources worldwide. Cis-Dichloroethene (cDCE) and Trichloroethene (TCE) are widespread pollutants that often occur together at contaminated sites, either as primary discharges or as degradation products of anaerobic dechlorination. In this study, comprehensive microcosm experiments were conducted with groundwater samples of seven sites contaminated with chloroethenes. In total, twelve wells with different pollutant concentrations and chloroethene compositions were sampled, and aerobic microcosms including sterile controls were set up. The results revealed interactions as well as interferences between cDCE and TCE. First, co-metabolic cDCE degradation with TCE as growth substrate was detected for the first time in this work. Transformation yields Ty' (molar ratio of co-substrate degraded to primary substrate degraded) of the degradation process were determined and showed a linear relationship with the cDCE/TCE concentration ratio. At low cDCE/TCE ratio, aerobic metabolic TCE degradation can result in complete cDCE removal due to co-metabolic degradation. Secondly, interfering effects were detected at notable cDCE levels resulting in deceleration of TCE degradation and residual concentrations which were also correlating linearly with the cDCE/TCE concentration ratio. These findings are significant for investigating chloroethene contaminated sites and planning remediation strategies. In particular, the efficiency biological remediation methods in the presence of both pollutants can be evaluated more precisely through the knowledge of interactions and interferences. Our study emphasizes that co-contaminants and possible effects of contaminant mixtures on the degradation rates of individual substances should be considered in general.

Trichloroethene (TCE) is one of the most prevalent contaminants in groundwater pollution worldwide. Aerobic-metabolic degradation of TCE has only recently been discovered at one field site. It has significant advantages over aerobic co-metabolism because no auxiliary substrates are required, and the oxygen demand is considerably lower. This study investigated the intrinsic degradation potential as well as the stimulation potential by bioaugmentation in microcosm experiments with groundwater from seven different sites contaminated with chloroethenes. An enrichment culture metabolizing TCE aerobically served as inoculum. The groundwater samples were inoculated with liquid culture in mineral salts medium as well as with immobilized culture on silica sand. Additionally, some samples were inoculated with groundwater from the site where the enrichment culture originated. The microcosms without inoculum proved the occurrence of aerobic TCE-metabolizing bacteria stimulated by the supply of oxygen in 54% of the groundwater samples. TCE degradation started in most cases after adaptation times of up to 92 d. The doubling time of 24 d indicated comparatively slow growth of the aerobic TCE degrading microorganisms. Bioaugmentation triggered or accelerated TCE-degradation in all microcosms with chlorothene concentrations below 100 mg L-1. All inoculation strategies (liquid and immobilized enrichment culture or addition of groundwater from the active field site) were successful. Our study demonstrates that aerobic-metabolic TCE degradation can occur and be stimulated across a broad hydrogeologic spectrum and should be considered as a viable option for groundwater remediation at TCE-contaminated sites.

Understanding and control of Laser-driven Free Electron Lasers remain to be difficult problems that require highly intensive experimental and theoretical research. The gap between simulated and experimentally collected data might complicate studies and interpretation of obtained results. In this work we developed a deep learning based surrogate that could help to fill in this gap. We introduce a surrogate model based on normalising flows for conditional phase-space representation of electron clouds in a FEL beamline. Achieved results let us discuss further benefits and limitations in exploitability of the models to gain deeper understanding of fundamental processes within a beamline.

In this work we propose a deep neural network based surrogate model for a plasma shadowgraph - a technique for visualization of perturbations in a transparent medium. We are substituting the numerical code by a computationally cheaper projection based surrogate model that is able to approximate the electric fields at a given time without computing all preceding electric fields as required by numerical methods. This means that the projection based surrogate model allows to recover the solution of the governing 3D partial differential equation, 3D wave equation, at any point of a given compute domain and configuration without the need to run a full simulation. This model has shown a good quality of reconstruction in a problem of interpolation of data within a narrow range of simulation parameters and can be used for input data of large size.

The phase‐field crystal (PFC) model allows for the resolution of atomic‐scale structures on diffusive time scales. It is based on an approximation of the two‐particle direct correlation function in the free energy, which provides the symmetry of the lattice structure. Various approaches have been proposed to model common lattices. We here only focus on two‐dimensional crystals, review the different approaches and demonstrate that one, which is based on affine linear deformations, violates rotational invariance. The consequences of this physical inconsistency are shown numerically using a pseudospectral algorithm for the higher order equations.