Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “parameter space”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 1,729 records · Page 96

Dark Energy Survey Year 3 results: Cosmology with moments of weak lensing mass maps

We present a cosmological analysis using the second and third moments of the weak lensing mass (convergence) maps from the first three years of data (Y3) data of the Dark Energy Survey. The survey spans an effective area of 4139 square degrees and uses the images of over 100 million galaxies to reconstruct the convergence field. The second moment of the convergence as a function of smoothing scale contains information similar to standard shear 2-point statistics. The third moment, or the skewness, contains additional non-Gaussian information. The data is analyzed in the context of the Λ CDM model, varying five cosmological parameters and 19 nuisance parameters modeling astrophysical and measurement systematics. Our modeling of the observables is completely analytical, and has been tested with simulations in our previous methodology study. We obtain a 1.7% measurement of the amplitude of fluctuations parameter S 8 ≡ σ 8 ( Ω m / 0.3 ) 0.5 = 0.784 ± 0.013 . The measurements are shown to be internally consistent across redshift bins, angular scales, and between second and third moments. In particular, the measured third moment is consistent with the expectation of gravitational clustering under the Λ CDM model. The addition of the third moment improves the constraints on S 8 and Ω m by ∼ 15 % and ∼ 25 % compared to an analysis that only uses second moments. We compare our results with Planck constraints from the cosmic microwave background, finding a 2.2 – 2.8 σ tension in the full parameter space, depending on the combination of moments considered. The third moment, independently, is in 2.8 σ tension with Planck, and thus provides a cross-check on the analyses of 2-point correlations.

79 ASTRONOMY AND ASTROPHYSICS↗

SU-F-T-153: Experimental Validation and Calculation Benchmark for a Commercial Monte Carlo Pencil Beam Scanning Proton Therapy Treatment Planning System in Water

Purpose: Eclipse proton Monte Carlo AcurosPT 13.7 was commissioned and experimentally validated for an IBA dedicated PBS nozzle in water. Topas 1.3 was used to isolate the cause of differences in output and penumbra between simulation and experiment. Methods: The spot profiles were measured in air at five locations using Lynx. PTW-34070 Bragg peak chamber (Freiburg, Germany) was used to collect the relative integral Bragg peak for 15 proton energies from 100 MeV to 225 MeV. The phase space parameters (σx, σθ, ρxθ) number of protons per MU, energy spread and calculated mean energy provided by AcurosPT were identically implemented into Topas. The absolute dose, profiles and field size factors measured using ionization chamber arrays were compared with both AcurosPT and Topas. Results: The beam spot size, σx, and the angular spread, σθ, in air were both energy-dependent: in particular, the spot size in air at isocentre ranged from 2.8 to 5.3 mm, and the angular spread ranged from 2.7 mrad to 6 mrad. The number of protons per MU increased from ∼9E7 at 100 MeV to ∼1.5E8 at 225 MeV. Both AcurosPT and TOPAS agree with experiment within 2 mm penumbra difference or 3% dose difference for scenarios including central axis depth dose and profiles at two depths in multi-spot square fields, from 40 to 200 mm, for all the investigated single-energy and multi-energy beams, indicating clinically acceptable source model and radiation transport algorithm in water. Conclusion: By comparing measured data and TOPAS simulation using the same source model, the AcurosPT 13.7 was validated in water within 2 mm penumbra difference or 3% dose difference. Benchmarks versus an independent Monte Carlo code are recommended to study the agreement in output, filed size factors and penumbra differences. This project is partially supported by the Varian grant under the master agreement between University of Pennsylvania and Varian.

60 APPLIED LIFE SCIENCES↗

A Constraint on Primordial B-modes from the First Flight of the Spider Balloon-borne Telescope

We present the first linear polarization measurements from the 2015 long-duration balloon flight of Spider, which is an experiment that is designed to map the polarization of the cosmic microwave background (CMB) on degree angular scales. The results from these measurements include maps and angular power spectra from observations of 4.8% of the sky at 95 and 150 GHz, along with the results of internal consistency tests on these data. While the polarized CMB anisotropy from primordial density perturbations is the dominant signal in this region of sky, Galactic dust emission is also detected with high significance. Galactic synchrotron emission is found to be negligible in the Spider bands. We employ two independent foreground-removal techniques to explore the sensitivity of the cosmological result to the assumptions made by each. The primary method uses a dust template derived from Planck data to subtract the Galactic dust signal. A second approach, which constitutes a joint analysis of Spider and Planck data in the harmonic domain, assumes a modified-blackbody model for the spectral energy distribution of the dust with no constraint on its spatial morphology. Using a likelihood that jointly samples the template amplitude and r parameter space, we derive 95% upper limits on the primordial tensor-to-scalar ratio from Feldman–Cousins and Bayesian constructions, finding r < 0.11 and r < 0.19, respectively. Roughly half the uncertainty in r derives from noise associated with the template subtraction. New data at 280 GHz from Spider's second flight will complement the Planck polarization maps, providing powerful measurements of the polarized Galactic dust emission.

79 ASTRONOMY AND ASTROPHYSICS↗

Electrostatics of Granular Material (EGM): Space Station Experiment

Aggregates were observed to form very suddenly in a lab-contained dust cloud, transforming (within seconds) an opaque monodispersed cloud into a clear volume containing rapidly-settling, long hair-like aggregates. The implications of such a "phase change" led to a series of experiments progressing from the lab, to KC-135, followed by micro-g flights on USML-1 and USML-2, and now EGM slated for Space Station. We attribute the sudden "collapse" of a cloud to the effect of dipoles. This has significant ramifications for all types of cloud systems, and additionally implicates dipoles in the processes of cohesion and adhesion of granular matter. Notably, there is the inference that like-charged grains need not necessarily repel if they are close enough together: attraction or repulsion depends on intergranular distance (the dipole being more powerful at short range), and the D/M ratio for each grain, where D is the dipole moment and M is the net charge. We discovered that these ideas about dipoles, the likely pervasiveness of them in granular material, the significance of the D/M ratio, and the idea of mixed charges on individual grains resulting from tribological processes --are not universally recognized in electrostatics, granular material studies, and aerosol science, despite some early seminal work in the literature, and despite commercial applications of dipoles in such modern uses as "Krazy Glue", housecleaning dust cloths, and photocopying. The overarching goal of EGM is to empirically prove that (triboelectrically) charged dielectric grains of material have dipole moments that provide an "always attractive" intergranular force as a result of both positive and negative charges residing on the surfaces of individual grains. Microgravity is required for this experiment because sand grains can be suspended as a cloud for protracted periods, the grains are free to rotate to express their electrostatic character, and Coulombic forces are unmasked. Suspended grains will be "interrogated" by applied electrical fields. In one module, grains will be immersed in an inhomogeneous electric field and allowed to be attracted towards or repelled from the central electrode of the module: part of the grain's speed will be a function of its net charge (monopole), part will be a function of the dipole. Observed grain position vs. time will provide a curve that can be deconvolved into the dipole and monopole forces responsible, since both have distinctive radial dependencies. In a second approach, the inhomogeneous field will be alternated at low frequency (e.g., every 5-10 seconds) so that the grains are alternately attracted and repelled from the center of the field. The resulting "zigzag" grain motion will gradually drift inwards, then suddenly change to a unidirectional inward path when a critical radial distance is encountered (a sort of "Coulombic event horizon") at which the dipole strength supersedes the monopole strength --thus proving the presence of a dipole, while also quantifying the D/M ratio. In a second module, an homogeneous electric field eliminates dipole effects (both Coulombic and induced) to provide calibration of the monopole and to more readily evaluate net charge statistical variance. In both modules, the e-fields will be exponentially step-ramped in voltage during the experiment, so that the field "nominalizes" grain speed while spreading the response time --effectively forcing each grain to "wait its turn" to be measured. In addition to rigorously quantifying M, D, and the D/M ratio for many hundreds of grains, the experiment will also observe gross electrometric and RF discharge phenomena associated with grain activity. The parameter space will encompass grain charging levels (via intentional triboelectrification), grain size, cloud density, and material type. Results will prove or disprove the dipole hypothesis. In either case, light will be shed on the role of electrostatic forces in governing granular systems. Knowledge so gained can be applied to natural clouds such as protostellar and protoplanetary dust and debris systems, planetary rings, planetary dust palls and aerosols created by volcanic, impact, aeolian, firestorm, or nuclear winter processes. The data are also directly applicable to adhesion, cohesion, transport, dispersion, and collection of granular materials in industrial, agricultural, pharmaceutical applications, and in fields as diverse as dust contamination of space suits on Mars and crop spraying on Earth.

Marshall, J.↗

Predicting electromagnetic ion cyclotron wave amplitude from unstable ring current plasma conditions

Electromagnetic ion cyclotron (EMIC) waves in the Earth's inner magnetosphere are enhanced fluctuations driven unstable by ring current ion temperature anisotropy. EMIC waves can resonate with relativistic electrons and play an important role in precipitation of MeV radiation belt electrons. In this study, we investigate the excitation and saturation of EMIC instability in a homogeneous plasma using both linear theory and nonlinear hybrid simulations. We have explored a four-dimensional parameter space, carried out a large number of simulations, and derived a scaling formula that relates the saturation EMIC wave amplitude to initial plasma conditions. Finally, such scaling can be used in conjunction with ring current models like ring current-atmosphere interactions model with self-consistent magnetic field to provide global dynamic EMIC wave maps that will be more accurate inputs for radiation belt modeling than statistical models.

79 ASTRONOMY AND ASTROPHYSICS↗

Single–Column Model Simulations of Subtropical Marine Boundary–Layer Cloud Transitions Under Weakening Inversions

Results are presented of the GASS/EUCLIPSE single-column model inter-comparison study on the subtropical marine low-level cloud transition. A central goal is to establish the performance of state-of-the-art boundary-layer schemes for weather and climate models for this cloud regime, using large-eddy simulations of the same scenes as a reference. A novelty is that the comparison covers four different cases instead of one, in order to broaden the covered parameter space. Three cases are situated in the North-Eastern Pacific, while one reflects conditions in the North-Eastern Atlantic. A set of variables is considered that reflects key aspects of the transition process, making use of simple metrics to establish the model performance. Using this method some longstanding problems in low level cloud representation are identified. Considerable spread exists among models concerning the cloud amount, its vertical structure and the associated impact on radiative transfer. Here, the sign and amplitude of these biases differ somewhat per case, depending on how far the transition has progressed. After cloud breakup the ensemble median exhibits the well-known “too few too bright” problem. The boundary layer deepening rate and its state of decoupling are both underestimated, while the representation of the thin capping cloud layer appears complicated by a lack of vertical resolution. Encouragingly, some models are successful in representing the full set of variables, in particular the vertical structure and diurnal cycle of the cloud layer in transition. An intriguing result is that the median of the model ensemble performs best, inspiring a new approach in subgrid parameterization.

54 ENVIRONMENTAL SCIENCES↗

Inferring subsurface heterogeneity from push-drift tracer tests

In this study, we consider the late-time tailing in a tracer test performed with a push-drift methodology (i.e., quasi-radial injection followed by drift under natural gradient). Numerical simulations of such tests are performed on 1000 multi-Gaussian 2-D log-hydraulic conductivity field realizations of varying heterogeneity, each under eight distinct mean flow directions. The ensemble pdfs of solute return times are found to exhibit power law tails for each considered variance of the log-hydraulic conductivity field, σ$2\atop{ln K}$. The tail exponent is found to relate straightforwardly to σ$2\atop{ln K}$ and, within the parameter space we explored, to be independent of push-phase pumping rate, pumping duration, and local-scale dispersivity. We conjecture that individual push-drift tracer tests in wells with screened intervals much greater than the vertical correlation length of the aquifer will exhibit quasi-ergodicity and that their tail exponent may be used to infer σ$2\atop{ln K}$. Lastly, we calibrate a predictive relationship of this sort from our Monte Carlo study, and apply it to data from a push-drift test performed at a site of approximately known heterogeneity—closely matching the existing best estimate of heterogeneity.

58 GEOSCIENCES↗

Interfacial Atomistic Mechanisms of Lithium Metal Stripping and Plating in Solid–State Batteries

All-solid-state batteries based on a Li metal anode represent a promising next-generation energy storage system, but are currently limited by low current density and short cycle life. Further research to improve the Li metal anode is impeded by the lack of understanding in its failure mechanisms at lithium–solid interfaces, in particular, the fundamental atomistic processes responsible for interface failure. Here, using large-scale molecular dynamics simulations, the first atomistic modeling study of lithium stripping and plating on a solid electrolyte is performed by explicitly considering key fundamental atomistic processes and interface atomistic structures. In the simulations, the interface failure initiated with the formation of nano-sized pores, and how interface structures, lithium diffusion, adhesion energy, and applied pressure affect interface failure during Li cycling are observed. By systematically varying the parameters of solid-state lithium cells in the simulations, the parameter space of applied pressures and interfacial adhesion energies that inhibit interface failure during cycling are mapped to guide selection of solid-state cells. Furthermore, this study establishes the atomistic modeling for Li stripping and plating, and predicts optimal solid interfaces and new strategies for the future research and development of solid-state Li-metal batteries.

25 ENERGY STORAGE↗

Rational Design of A Chemical Bath Deposition Based Tin Oxide Electron Transport Layer for Perovskite Photovoltaics

Chemical bath deposition is widely used to deposit SnO x as an electron transport layer in perovskite solar cells (PSCs). The conventional recipe uses thioglycolic acid (TGA) to facilitate attachments of SnO x particles onto the substrate. However, nonvolatile TGA has been reported to harm the operational stability of PSCs. In this work, we introduced a volatile oxalic acid (OA) as an alternative to TGA. OA, a dicarboxylic acid, functions as a chemical linker for the nucleation and attachment of particles to the substrate in the chemical bath. Moreover, OA can be readily removed through thermal annealing followed by a mild H 2 O 2 treatment, as shown by FTIR measurements. Synergistically, the mild H 2 O 2 treatment selectively oxidizes the surface of the SnOx layer, minimizing nonradiative interface carrier recombination. EELS (electron-energy-loss-spectroscopy) confirms that the SnO x surface is dominated by Sn 4+ , while the bulk is a mixture of Sn 2+ and Sn 4+ . This rational design of a CBD SnO x layer leads to devices with T85~1,500h, a significant improvement over the TGA-based device with T80~250h. Our champion device reached a power conversion efficiency of 24.6%. This work offers a rationale for optimizing the complex parameter space of CBD SnO x to achieve efficient and stable PSCs.

14 SOLAR ENERGY↗

Sieving Hydrogen Isotopes via Machine Learning Assisted Chemical Vapor Deposition (CVD) of High‐Quality Monolayer Hexagonal Boron Nitride (h‐BN) on Iron Foils

Atomically thin two-dimensional (2D) ceramics, such as monolayer hexagonal boron nitride (h-BN), present potential for disruptive advances in separations. However, sub-atomic scale separation of hydrogen isotopes (H + /D + ) require near pristine 2D material membranes, and scalable synthesis of such high-quality h-BN comparable to mechanically exfoliated crystals remains a significant challenge. Here, we report a scalable Fe-catalyzed chemical vapor deposition (CVD) process for bottom-up synthesis of large-area, high-quality monolayer h-BN films, overcoming key limitations of conventional ammonia-based routes. By leveraging mechanistic insights and higher CVD temperatures, we suppress multilayer formation and achieve uniform monolayer h-BN coverage on commercially available Fe foils. Machine learning enables systematic exploration of the complex, multi-dimensional CVD parameter space (growth time, temperature, precursor temperature, multilayer faction, coverage), providing data-driven approaches to visualize and identify process regimes facilitating predominantly monolayer h-BN growth with minimal secondary nuclei/ad-layers. The optimized Fe-catalyzed CVD h-BN membranes show high-quality as observed by proton/deuteron (H + /D + ) selectivity ≈8.45, approaching the highest quality benchmark of mechanically exfoliated h-BN (H + /D + selectivity ≈10) as well as significantly outperforming Cu-catalyzed CVD h-BN membranes (H + /D + selectivity ≈3.62, control selectivity ≈1.7). Our work provides a scalable cost-effective route for high-quality monolayer h-BN synthesis for sub-atomic scale separations (H + /D + ) and demonstrates the broader potential of machine learning-guided optimization of CVD for advancing synthesis of 2D materials.

36 MATERIALS SCIENCE↗

Oxidation Driven Thin‐Film Solid‐State Metal Dealloying Forming Bicontinuous Nanostructures

Abstract Solid‐state metal dealloying (SSMD) is a promising method for fabricating nanoscale metallic composites and nanoporous metals across a range of materials. Thin‐film SSMD is particularly attractive due to its ability to create fine features via solid‐state interfacial reactions within a thin‐film geometry, which can be integrated into devices for various applications. This work examines a new dealloying couple, namely the Nb–Al alloy with the dealloying agent Sc, as previously predicted in the machine‐learning (ML) models. Prior ML predictions aimed to guide the design of nanoarchitectured materials through dealloying, relying on intuition‐driven discovery within a large parameter space. However, this work reveals that at the nanoscale, the involvement of oxygen in thin film processing may instead drive the dealloying process, resulting in the formation of bicontinuous nanostructures similar to those formed by metal‐agent dealloying. The phase evolution, as well as chemical and morphological changes, are closely analyzed using a combination of X‐ray absorption spectroscopy, diffraction, and scanning transmission electron microscopy to understand the mechanisms behind nanostructure formation. The findings suggest a potential pathway for utilizing oxygen to drive the formation of bicontinuous metal–metal oxide nanocomposites, paving the way for further development of functional nanoporous materials in diverse fields.

36 MATERIALS SCIENCE↗

When Machine Learning Meets 2D Materials: A Review

The availability of an ever-expanding portfolio of 2D materials with rich internal degrees of freedom (spin, excitonic, valley, sublattice, and layer pseudospin) together with the unique ability to tailor heterostructures made layer by layer in a precisely chosen stacking sequence and relative crystallographic alignments, offers an unprecedented platform for realizing materials by design. However, the breadth of multi-dimensional parameter space and massive data sets involved is emblematic of complex, resource-intensive experimentation, which not only challenges the current state of the art but also renders exhaustive sampling untenable. To this end, machine learning, a very powerful data-driven approach and subset of artificial intelligence, is a potential game-changer, enabling a cheaper – yet more efficient – alternative to traditional computational strategies. It is also a new paradigm for autonomous experimentation for accelerated discovery and machine-assisted design of functional 2D materials and heterostructures. Here, the study reviews the recent progress and challenges of such endeavors, and highlight various emerging opportunities in this frontier research area.

2D materials↗

Inverse Design of Photonic Surfaces via High throughput Femtosecond Laser Processing and Tandem Neural Networks

Abstract This work demonstrates a method to design photonic surfaces by combining femtosecond laser processing with the inverse design capabilities of tandem neural networks that directly link laser fabrication parameters to their resulting textured substrate optical properties. High throughput fabrication and characterization platforms are developed that generate a dataset comprising 35280 unique microtextured surfaces on stainless steel with corresponding measured spectral emissivities. The trained model utilizes the nonlinear one‐to‐many mapping between spectral emissivity and laser parameters. Consequently, it generates predominantly novel designs, which reproduce the full range of spectral emissivities (average root‐mean‐squared‐error < 2.5%) using only a compact region of laser parameter space 25 times smaller than what is represented in the training data. Finally, the inverse design model is experimentally validated on a thermophotovoltaic emitter design application. By synergizing laser‐matter interactions with neural network capabilities, the approach offers insights into accelerating the discovery of photonic surfaces, advancing energy harvesting technologies.

36 MATERIALS SCIENCE↗

The Origin of Improved Performance in Boron‐Alloyed Silicon Nanoparticle‐Based Anodes for Lithium‐Ion Batteries

Stabilizing the solid electrolyte interphase (SEI) remains a key challenge for silicon‐based lithium‐ion battery anodes. Alloying silicon with secondary elements like boron has emerged as a promising strategy to improve the cycle life of silicon anodes, yet the underlying mechanism remains unclear. To address this knowledge gap, how boron concentration influences battery performance is systematically investigated. These results show a near‐monotonic increase in cycle lifetime with higher boron content, with boron‐rich electrodes significantly outperforming pure silicon. Additionally, silicon‐boron alloy anodes exhibit nearly three times longer calendar life than pure silicon. Through detailed mechanistic analysis, alternative contributing factors are systematically ruled out, and it is proposed that improved passivation arises from a strong permanent dipole at the nanoparticle surface. This dipole, formed by undercoordinated and highly Lewis acidic boron, creates a static, ion‐dense layer that stabilizes the electrochemical interface, reducing parasitic electrolyte decomposition and enhancing long‐term stability. These findings suggest that, within the SEI framework, the electric double layer is an important consideration in surface passivation. This insight provides an underexplored parameter space for optimizing silicon anodes in next‐generation lithium‐ion batteries.

25 ENERGY STORAGE↗

Advanced Solution Methods for Microkinetic Models of Catalytic Reactions: A Methanol Synthesis Case Study

Microkinetic models, combined with experimentally measured reaction rates and orders, play a key role in elucidating detailed reaction mechanisms in heterogeneous catalysis and have typically been solved as systems of ordinary differential equations. In this work, we demonstrate a new approach to fitting those models to experimental data. For the specific example treated here, by reformulating a typical microkinetic model for a continuous stirred tank reactor to a system of nonlinear equations, we achieved a 1000-fold increase in solution speed. The reduced computational cost allows a more systematic search of the parameter space, leading to better fits to the available experimental data. We applied this approach to the problem of methanol synthesis by CO/CO2 hydrogenation over a supported-Cu catalyst, an important catalytic reaction with a large industrial interest and potential for large-scale CO2 chemical fixation.

Environmental Molecular Sciences Laboratory↗

Expeditious modeling of vapor transport and reactions in polymeric materials

We present a methodology for approximating dynamic adsorption of vapor coupled with diffusion in polymeric materials. In previous publications, the dynamic adsorption was represented by ordinary differential equations (ODEs) and solved in concentration and parameter space. To accelerate the calculation, we have developed a statistical approximation method using computationally cheap surrogate models (e.g., algebraic polynomials) that replace the ODE solutions of adsorption and are coupled with the diffusion equations. Since the polynomial presentation of the adsorption term is obtained in a standard format prior to modeling coupled sorption‐diffusion, the adsorption operator can be expressed as input data in the transport code. Compared to conventional operator‐splitting methods, the polynomial approximation of adsorption offers better computational efficiency. The methodology is demonstrated and validated using a dynamic Langmuir adsorption model that is coupled to diffusion and absorption models and applied to a water vapor sorption‐diffusion process in polydimethylsiloxane polymers. © 2017 American Institute of Chemical Engineers AIChE J , 63: 4079–4089, 2017

Sun, Yunwei↗

ytopt: Autotuning Scientific Applications for Energy Efficiency at Large Scales

As we enter the exascale computing era, efficiently utilizing power and optimizing the performance of scientific applications under power and energy constraints has become critical and challenging. We propose a low-overhead autotuning framework to autotune performance and energy for various hybrid MPI/OpenMP scientific applications at large scales and to explore the tradeoffs between application runtime and power/energy for energy efficient application execution, then use this framework to autotune four ECP proxy applications—XSBench, AMG, SWFFT, and SW4lite. Our approach uses Bayesian optimization with a Random Forest surrogate model to effectively search parameter spaces with up to 6 million different configurations on two large-scale HPC production systems, Theta at Argonne National Laboratory and Summit at Oak Ridge National Laboratory. The experimental results show that our autotuning framework at large scales has low overhead and achieves good scalability. Using the proposed autotuning framework to identify the best configurations, we achieve up to 91.59% performance improvement, up to 21.2% energy savings, and up to 37.84% EDP (energy delay product) improvement on up to 4096 nodes.

Autotuning↗

Iterative sampling of expensive simulations for faster deep surrogate training

Deep neural network (DNN) surrogates of expensive physics simulations are enabling a rapid change in the way that common experimental design and analysis tasks are approached. Surrogate models allow simulations to be performed in parallel and separately from downstream tasks, thereby enabling analyses that would be impossible with the simulation in-the-loop; surrogates based on DNNs can effectively emulate diverse non-scalar data of the types collected in fusion and laboratory-astrophysics experiments. The challenge is in training the surrogate model, for which large ensembles of physics simulations must be run, preferably without wasting computational effort on uninteresting simulations. Here, in this paper, we present an iterative sampling scheme that can preferentially propose simulations in interesting regions of parameter space without neglecting unexplored regions, allowing high-quality and wide-ranging surrogate models to be trained using 2–3 times fewer simulations compare to space-filling designs. Our approach uses an explicit importance function defined on the simulation output space, balanced against a measure of simulation density which serves as a proxy for surrogate accuracy. It is easy to implement and can be tuned to find interesting simulations early in the study, allowing surrogates to be trained quickly and refined as new simulations become available; this represents an important step towards the routine generation of deep surrogate models quickly enough to be truly relevant to experimental work.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗