Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “distributed inference”

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 307 records · Page 17
All sourcesNASA NTRSDOE OSTI
DOE OSTI2020

Bayesian inference of nuclear symmetry energy from measured and imagined neutron skin thickness in Sn 116 , 118 , 120 , 122 , 124 , 130 , 132 , Pb 208 , and Ca 48

The neutron skin thickness Δr np in heavy nuclei has been known as one of the most sensitive terrestrial probes of the nuclear symmetry energy E sym (ρ) around $\frac{2}{3}$ of the saturation density ρ 0 of nuclear matter. Existing neutron skin data mostly from hadronic observables suffer from large uncertainties and their extraction from experiments are often strongly model dependent. While waiting eagerly for the promised model-independent and high-precision neutron skin data for 208 Pb and 48 Ca from the parity-violating electron scattering experiments (PREX-II and CREX at JLab as well as MREX at MESA), within the Bayesian statistical framework using the Skyrme-Hartree-Fock model we infer the posterior probability distribution functions (PDFs) of the slope parameter L of the nuclear symmetry energy at ρ 0 from imagined Δr np ( 208 Pb)=0.15, 0.20, and 0.30 fm with a 1σ error bar of 0.02, 0.04, and 0.06 fm, respectively, as well as Δr np ( 48 Ca)=0.12, 0.15, and 0.25 fm, with different 1σ error bar of 0.01 and 0.02 fm, respectively. The results are compared with the PDFs of L inferred using the same approach from the available Δr np data for 116, 118, 120, 122, 124, 130, 132 Sn from hadronic probes. They are also compared with results from a recent Bayesian analysis of the radius and tidal deformability data of canonical neutron stars from GW170817 and NICER. The neutron skin data for Sn isotopes gives L = 45.5 $^{+ 26.5}_{-21.6}$ MeV surrounding its mean value or L = 53 . 4 $^{+ 18.6}_{ -29.5}$ MeV surrounding its maximum a posteriori value, respectively, with the latter smaller than but consistent with the L = 66 $^{+ 12}_{-20}$ MeV from the neutron star data within their 68% confidence intervals. We found that Δr np = 0.17 –0.18 fm in 208 Pb with an error bar of about 0.02 fm leads to a PDF of L compatible with that from analyzing the Sn data. To provide additionally useful information on L extracted from the Δr np of Sn isotopes, the experimental error bar of Δr np in 208 Pb should be at least smaller than 0.06 fm aimed by some current experiments. In addition, the Δr np ( 48 Ca) needs to be larger than 0.15 fm but smaller than 0.25 fm to be compatible with the Sn and/or neutron star results. To further improve our current knowledge about L and distinguish its PDFs in the examples considered, even higher precisions leading to significantly less than ±20 MeV error bars for L at 68% confidence level are necessary.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2024

Bayesian inference of fine features of the nuclear equation of state from future neutron star radius measurements to 0.1 km accuracy

To more precisely constrain the equation of state (EOS) of supradense neutron-rich nuclear matter, future high-precision x-ray and gravitational wave observatories are proposed to measure the radii of neutron stars (NSs) with an accuracy better than about 0.1 km. However, it remains unclear what particular aspects (other than the stiffness generally spoken of in the literature) of the EOS and to what precision they will be better constrained. In this work, within a Bayesian framework using a metamodel EOS for NSs, we infer the posterior probability distribution functions (PDFs) of incompressibility K 0 and skewness J 0 of symmetric nuclear matter (SNM) as well as the slope L, curvature K sym , and skewness J sym characterizing the density dependence of nuclear symmetry energy E sym ⁡(ρ), respectively, from mean values of NS radii consistent with existing observations and an expected accuracy Δ⁢R ranging from about 1.0 to 0.1 km. Here, we found that (1) the Δ⁢R has little effect on inferring the stiffness of SNM at suprasaturation densities, (2) smaller Δ⁢R reveals more accurately not only the PDFs but also pairwise correlations among parameters characterizing high-density E sym ⁡(ρ), (3) a double-peak feature of the PDF(K sym ) corresponding to the strong K sym – J sym and K sym – L anticorrelations is revealed when Δ⁢R is less than about 0.2 km, and the locations of the two peaks are sensitive to the maximum value of J sym reflecting the stiffness of E sym ⁡(ρ) above about 3 times the saturation density ρ 0 of SNM, and (4) the high-precision radius measurement for canonical NSs is more useful than that for massive ones for constraining the EOS of nucleonic matter around (2–3)⁢ρ 0 .

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
NASA NTRS2019

Using Deep Learning to Automate Inference of Meteoroid Pre-Entry Properties

Properly assessing the asteroid threat depends on the knowledge of asteroid pre-entry parameters, such as size, velocity, mass, density, and strength. Although a vast number of possible bodies to study exist, such characterization of asteroid populations is currently limited by substantial costs associated with space rendezvous missions and rare meteorite findings. As asteroids fragment, ablate, and decelerate in the atmosphere, they emit light detectable by ground-based and space-borne instruments. Earth’s atmosphere, thus, becomes an accessible laboratory that enables impactor risk assessments by facilitating inference of the pre-entry parameters. These asteroid pre-entry conditions are typically deduced by modeling the entry and breakup physics that best reproduce the observed light or energy deposition curve. However, this process requires extensive manual trial-and-error of uncertain modeling parameters. Automating meteor modeling and inference would improve property distributions used in risk assessments and enable population characterization as more light curves become more readily available through the presence of space assets and ground-based camera networks. We previously developed a genetic algorithm to automate meteor modeling by using the fragment-cloud model (FCM) to search for the values of the FCM input parameters (e.g., diameter) that generate energy deposition profiles that match the observed one. Now, we apply deep learning to infer asteroid diameter, velocity, and density from observed energy deposition curves. We trained and tested our neural network models with synthetic energy deposition curves modeled using the FCM rubble pile implementation. We present an application of a 1D convolutional neural network and compare its performance to other attempted regressors and machine learning techniques, such as a fully connected neural network and Random Forest regression, to demonstrate its capabilities. We validate our model weights and approach using the Chelyabinsk, Tagish Lake, Benešov, Košice, and Lost City meteors.

Tarano, Ana Maria↗
DOE OSTI2022

A Reanalysis of Public Galactic Bulge Gravitational Microlensing Events from OGLE-III and -IV

Abstract Modern surveys of gravitational microlensing events have progressed to detecting thousands per year, and surveys are capable of probing Galactic structure, stellar evolution, lens populations, black hole physics, and the nature of dark matter. One of the key avenues for doing this is the microlensing Einstein radius crossing time (t E) distribution. However, systematics in individual light curves as well as oversimplistic modeling can lead to biased results. To address this, we developed a model to simultaneously handle the microlensing parallax due to Earth's motion, systematic instrumental effects, and unlensed stellar variability with a Gaussian process model. We used light curves for nearly 10,000 OGLE-III and -IV Milky Way bulge microlensing events and fit each with our model. We also developed a forward model approach to infer the t E distribution by forward modeling from the data rather than using point estimates from individual events. We find that modeling the variability in the baseline removes a source of significant bias in individual events, and the previous analyses overestimated the number of t E > 100 day events due to their oversimplistic model ignoring parallax effects. We use our fits to identify the hundreds filling a regime in the microlensing parameter space that are 50% pure of black holes. Finally, we have released the largest-ever catalog of Markov Chain Monte Carlo parameter estimates for microlensing events.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2021

A Bimodal Diagnostic Cloud Fraction Parameterization. Part I: Motivating Analysis and Scheme Description

Cloud fraction parameterizations are beneficial to regional, convection-permitting numerical weather prediction. For its operational regional midlatitude forecasts, the Met Office uses a diagnostic cloud fraction scheme that relies on a unimodal, symmetric subgrid saturation-departure distribution. This scheme has been shown before to underestimate cloud cover and hence an empirically based bias correction is used operationally to improve performance. This first of a series of two papers proposes a new diagnostic cloud scheme as a more physically based alternative to the operational bias correction. The new cloud scheme identifies entrainment zones associated with strong temperature inversions. Additionally, for model grid boxes located in this entrainment zone, collocated moist and dry Gaussian modes are used to represent the subgrid conditions. The mean and width of the Gaussian modes, inferred from the turbulent characteristics, are then used to diagnose cloud water content and cloud fraction. It is shown that the new scheme diagnoses enhanced cloud cover for a given gridbox mean humidity, similar to the current operational approach. It does so, however, in a physically meaningful way. Using observed aircraft data and ground-based retrievals over the southern Great Plains in the United States, it is shown that the new scheme improves the relation between cloud fraction, relative humidity, and liquid water content. An emergent property of the scheme is its ability to infer skewed and bimodal distributions from the large-scale state that qualitatively compare well against observations. A detailed evaluation and resolution sensitivity study will follow in Part II.

54 ENVIRONMENTAL SCIENCES↗
NASA NTRS2006

Calculating Statistical Orbit Distributions Using GEO Optical Observations with the Michigan Orbital Debris Survey Telescope (MODEST)

NASA's Orbital Debris measurements program has a goal to characterize the small debris environment in the geosynchronous Earth-orbit (GEO) region using optical telescopes ("small" refers to objects too small to catalog and track with current systems). Traditionally, observations of GEO and near-GEO objects involve following the object with the telescope long enough to obtain an orbit suitable for tracking purposes. Telescopes operating in survey mode, however, randomly observe objects that pass through their field of view. Typically, these short-arc observation are inadequate to obtain detailed orbits, but can be used to estimate approximate circular orbit elements (semimajor axis, inclination, and ascending node). From this information, it should be possible to make statistical inferences about the orbital distributions of the GEO population bright enough to be observed by the system. The Michigan Orbital Debris Survey Telescope (MODEST) has been making such statistical surveys of the GEO region for four years. During that time, the telescope has made enough observations in enough areas of the GEO belt to have had nearly complete coverage. That means that almost all objects in all possible orbits in the GEO and near- GEO region had a non-zero chance of being observed. Some regions (such as those near zero inclination) have had good coverage, while others are poorly covered. Nevertheless, it is possible to remove these statistical biases and reconstruct the orbit populations within the limits of sampling error. In this paper, these statistical techniques and assumptions are described, and the techniques are applied to the current MODEST data set to arrive at our best estimate of the GEO orbit population distribution.

Matney, M.↗
DOE OSTI2023

Direct Bayesian inference for fault severity assessment in Digital-Twin-Based fault diagnosis

For applications in condition-based maintenance of nuclear systems, the assessment of fault severity is crucial. In this work, we developed a framework that allows for direct inference of the probability distributions of possible faults in a system. Here, we employed a model-based approach with model residuals generated from analytical redundancy relations provided by physics-based models of the system components. From real-time sensor readings, the values of the model residuals can be calculated, and the posterior probability distributions of the faults can be computed directly using the methods of Bayesian networks. From the posterior distribution of each fault, one can estimate the fault probability based on a chosen threshold and assess the severity of the fault. By eliminating the discretization and simplifications in middle steps, this approach allows us to leverage the available computational resources to provide more accurate fault probability estimates and severity assessments.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
NASA NTRS2021

3D Cloud Tomography and Droplet Size Retrieval from Multi-Angle Polarimetric Imaging of Scattered Sunlight from Above

Tomography aims to recover a three-dimensional (3D) density map of a medium or an object. In medical imaging,it is extensively used for diagnostics via X-ray computed tomography (CT). We define and derive a tomographyof cloud droplet distributions via passive remote sensing. We use multi-view polarimetric images to fit a 3Dpolarized radiative transfer (RT) forward model. Our motivation is 3D volumetric probing of vertically-developedconvectively-driven clouds that are ill-served by current methods in operational passive remote sensing. Currenttechniques are indeed based on strictly 1D RT modeling and applied to a single cloudy pixel, where cloud geometrydefaults to that of a plane-parallel slab. Incident unpolarized sunlight, once scattered by cloud droplets, changesits polarization state according to droplet size. Therefore, polarimetric measurements in the rainbow and gloryangular regions can be used to infer the droplet size distribution. This work defines and derives a framework for afull 3D tomography of cloud droplets for both their mass concentration in space and their distribution across arange of sizes. This gridded 3D retrieval of key microphysical properties is made tractable by our novel approachthat involves a restructuring and partial linearization of an open-source polarized 3D RT code to accommodate aspecial two-step iterative optimization technique. Physically-realistic synthetic clouds are used to demonstrate themethodology with rigorous uncertainty quantification, while a real-world cloud imaged by AirMSPI is processedto illustrate the new remote sensing capability

Schechner, Yoav Y.↗
NASA NTRS2004

Physical Properties of a Coronal Hole from a Coronal Diagnostics Spectrometer, Mauna Loa Coronagraph, and LASCO Observations during the Whole Sun Month

Until recently, inference of electron density distribution in the solar corona was limited by the field of view of white-light coronagraphs (typically out to 6 Rs). Now, for the first time we have a series of white- light coronagraphs (SOHO/LASCO) whose combined field of view extends from 1.1 - 30 Rs. Quantitative information on electron density distribution of coronal hole and coronal plumes/rays are estimated by using white-light, polarized brightness (pB) observations from the SOHO/LASCO/C2 and C3 and HAO/Mauna Loa Mark III coronagraphs from 1.15 to 8.0 Rs. Morphological information on the boundary of the polar coronal hole and streamer interface is determined from the white-light observations in a manner similar to the Skylab polar coronal hole boundary estimate. The average coronal hole electron density in the region 1 - 1.15 Rs is estimated from the density-sensitive EUV line ratios of Si IX 3501342 A observed by the SOHO/coronal diagnostic spectrometer (CDS). We combine these numbers with the estimate from white-light (WL) observations to obtain a density profile from 1 to 8 Rs for the plumes and the polar coronal hole. We find that white light and spectral analysis produce consistent density information. Extrapolated densities inferred from SOHO observations are compared to Ulysses in situ observations of density. Like the density inferred from the Spartan 201-03 coronagraph, the current SOHO density profiles suggest that the acceleration of the fast solar wind takes place very close to the Sun, within 10-15 Rs. The density information is used to put constraints on solar wind flow velocities and effective temperatures. Finally, these results are compared to the recent analysis of the Spartan 201-03 white-light observations.

Guhathakurta, M.↗
NASA NTRS1975

Winter bulge and diurnal variations in hydrogen inferred from AE-C composition measurements

The atomic hydrogen distribution at 250 km during December 1974 solstice was inferred, considering charge exchange equilibrium, from Atmosphere Explorer-C measurements of n(H(+)), n(O(+) and N(O). An empirical model, derived from the observations by least suqare analysis in terms of spherical harmonics, has the following characteristics: (1) n(H) increases by as much as a factor of two between the summer and winter hemispheres, (2) the n(H) diurnal variation is largest at the equator and (3) the diurnal variation is larger in the winter hemisphere than in the summer. Similar analysis of the gas temperature derived from n(N2) measurements reveals that all n(H) and Tg spherical harmonic coefficients are anticorrelated. Both the diurnal and latitudinal (annual) n(H) and Tg amplitudes are in substantial agreement with the zero flux condition, in which exospheric flow dominates the hydrogen distribution. The observed diurnal phase lag of n(H) with respect to Tg is about one hour, agreeing with theory.

Brinton, H. C.↗
NASA NTRS1989

Infrared Emission from Comets

A brief discussion of the infrared observations from 4 to 20 micrometers of seven comets is presented. The observed infrared emission from comets depends primarily on their heliocentric distance. A model based on grain populations composed of a mixture of silicate and amorphous carbon particles in the mass ratio of about 40 to 1, with a power-law size distribution similar to that inferred for comet Halley, is applied to the observations. The model provides a good match to the observed heliocentric variation of both the 10 micrometers feature and the overall thermal emission from comets West and Halley. Matches to the observations of comet IRAS-Araki-Alcock and the antitail of comet Kohoutek require slightly larger grains. While the model does not match the exact profile and position of the 3.4 micrometers feature discovered in comet Halley, it does produce a qualitative fit to the observed variation of the feature's strength as a function of heliocentric distance. The calculations predict that the continuum under the 3.4 micrometers feature is due primarily to thermal emission from the comet dust when the comet is close to the Sun and to scattered solar radiation at large heliocentric distances, as is observed. A brief discussion of the determination of cometary grain temperatures from the observed infrared emission is presented. It is found that the observed shape of the emission curve from about 4 to 8 micrometers provides the best spectral region for estimating the cometary grain temperature distribution.

Swamy, K. S. Krishna↗
DOE OSTI2023

The DESI PRObabilistic Value-added Bright Galaxy Survey (PROVABGS) Mock Challenge

The PRObabilistic Value-added Bright Galaxy Survey (PROVABGS) catalog will provide measurements of galaxy properties, such as stellar mass ($M*$), star formation rate (SFR), stellar metallicity (Z), and stellar age ($t_{age}$), for >10 million galaxies of the Dark Energy Spectroscopic Instrument (DESI) Bright Galaxy Survey. Full posterior distributions of the galaxy properties will be inferred using state-of-the-art Bayesian spectral energy distribution (SED) modeling of DESI spectroscopy and Legacy Surveys photometry. In this work, we present the SED model, the neural emulator for the model, and the Bayesian inference framework of PROVABGS. Furthermore, we apply the PROVABGS SED modeling on realistic synthetic DESI spectra and photometry, constructed using the L-Galaxies semi-analytic model. We compare the inferred galaxy properties to the true values of the simulation using a hierarchical Bayesian framework to quantify accuracy and precision. Overall, we accurately infer the true $M*$, SFR, Z, and $t_{age}$ of the simulated galaxies. However, the priors on galaxy properties induced by the SED model have a significant impact on the posteriors, which we characterize in detail. This work also demonstrates that a joint analysis of spectra and photometry significantly improves the constraints on galaxy properties over photometry alone and is necessary to mitigate the impact of the priors. With the methodology presented and validated in this work, PROVABGS will maximize information extracted from DESI observations and extend current galaxy studies to new regimes and unlock cutting-edge probabilistic analyses. https://github.com/changhoonhahn/provabgs/

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2025

Computer Vision Pipeline for Image Analysis for Freeze‐Fracture Electron Microscopy: Rosette Cellulose Synthase Complexes Case

In materials science, plant biology, agriculture, and environmental research, the automated analysis of high-magnification, complex microscopy images, such as those generated by freeze-fracture electron microscopy (FF-TEM), remains a critical challenge that limits the scalability of data interpretation. We present a deep learning computer vision pipeline for high-throughput detection and morphological characterization analysis of cellulose synthase complexes (CSCs, or rosettes) in FF-TEM images. The pipeline integrates preprocessing, detection, human-in-the-loop verification, and semantic segmentation to quantify features such as rosette diameter and inter-lobe spacing. The approach was trained and tested on a curated dataset of high-resolution FF-TEM micrographs of Physcomitrium patens, expanded via strategic tiling and augmentation to over 650 images. We compare YOLOv8 and YOLOv9 architectures and demonstrate that YOLOv9 achieves superior performance in both localization accuracy (mAP50-95 = 0.854) and inference speed. The resulting distributions revealed biological variability consistent with prior manual studies, validating the approach for high-throughput applications. Our results show that the pipeline achieves human-expert level accuracy while dramatically reducing analysis time, enabling scalable, reproducible structural characterization of intramembrane protein complexes. The pipeline is broadly applicable to other domains requiring precise interpretation of complex microscopy data and establishes a foundation for future artificial intelligence (AI)-assisted workflows in biological imaging.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2020

Modern Monte Carlo methods for efficient uncertainty quantification and propagation: A survey

Uncertainty quantification (UQ) includes the characterization, integration, and propagation of uncertainties that result from stochastic variations and a lack of knowledge or data in the natural world. Monte Carlo (MC) method is a sampling-based approach that has widely used for quantification and propagation of uncertainties. However, the standard MC method is often time-consuming if the simulation-based model is computationally intensive. This article gives an overview of modern MC methods to address the existing challenges of the standard MC in the context of UQ. Specifically, multilevel Monte Carlo (MLMC) extending the concept of control variates achieves a significant reduction of the computational cost by performing most evaluations with low accuracy and corresponding low cost, and relatively few evaluations at high accuracy and corresponding high cost. Multifidelity Monte Carlo (MFMC) accelerates the convergence of standard Monte Carlo by generalizing the control variates with different models having varying fidelities and varying computational costs. Multimodel Monte Carlo method (MMMC), having a different setting of MLMC and MFMC, aims to address the issue of UQ and propagation when data for characterizing probability distributions are limited. Multimodel inference combined with importance sampling is proposed for quantifying and efficiently propagating the uncertainties resulting from small data sets. All of these three modern MC methods achieve a significant improvement of computational efficiency for probabilistic UQ, particularly uncertainty propagation. Finally, an algorithm summary and the corresponding code implementation are provided for each of the modern MC methods. The extension and application of these methods are discussed in detail.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Enhancing approximate modular Bayesian inference by emulating the conditional posterior

In modular Bayesian analyses, complex models are composed of distinct modules, each representing different aspects of the data or prior information. In this context, fully Bayesian approaches can sometimes lead to undesirable feedback between modules, compromising the integrity of the inference. The “cut-distribution” prevents unwanted influence between modules by “cutting” feedback. The direct sampling (DS) algorithm is standard practice for approximating the cut-distribution, but it can be computationally intensive, especially when the number of imputations required is large. An enhanced method is proposed, the Emulating the Conditional Posterior (ECP) algorithm, which leverages emulation to increase the number of imputations. Through numerical experiment it is demonstrated that the ECP algorithm outperforms the traditional DS approach in terms of accuracy and computational efficiency, particularly when resources are constrained. Here, it is also shown how the DS algorithm can be improved using ideas from design of experiments. Some practical recommendations are given for algorithm choice in modular Bayesian analyses.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

O'Hare Airport roadway traffic prediction via data fusion and Gaussian process regression

This study proposes an approach of leveraging information gathered from multiple traffic data sources at different resolutions to obtain approximate inference on the traffic distribution of Chicago's O'Hare Airport area. Specifically, it proposes the ingestion of traffic datasets at different resolutions to build spatiotemporal models for predicting the distribution of traffic volume on the road network. Due to its good adaptability and flexibility for spatiotemporal data, the Gaussian process (GP) regression was employed to provide short-term forecasts using data collected by loop detectors (sensors) and supplemented by telematics data. The GP regression is used to make predictions of the distribution of the proportion of sensor data traffic volume represented by the telematics data for each location of the sensors. Consequently, the fitted GP model can be used to determine the approximate traffic distribution for a testing location outside of the training points. Policymakers in the transportation sector can find the results of this work helpful for making informed decisions relating to current and future transportation conditions in the area.

42 ENGINEERING↗
DOE OSTI2021

Joint Bayesian Inference for Near-Surface Explosion Yield and Height-of-Burst

Forensic capabilities to understand chemical and nuclear explosions are greatly aided by an accurate estimate of explosive yield with uncertainty. The relationship between explosive size and geophysical observations of seismic, acoustic, and optical waves can be exploited to provide an estimate of yield. Any near-surface yield estimate is complicated by the surface interaction, so an estimate for the explosion height-of-burst is necessarily included in the relationship. Additionally, the relationship dictates a trade-off between estimates of yield and height-of-burst. Fortunately, the surface interaction for each type of observation is different, which breaks the trade-off, and the inclusion of height-of-burst with multiple data types improves yield estimation. We define simple parametric forward models to relate seismoäcoustoöptic observations from a data set of known explosive yields and height-of-bursts. The parameters of the models and a prediction for the yield and height-of-burst of a new event can then be estimated given new observations via Bayesian inference. We report posterior distribution estimates of the parametric models using a Markov chain Monte Carlo sampling technique. These models are then used to predict the yield and height-of-burst of SUGAR, a historical near-surface nuclear explosion, using its reported historical observations. The reported yield of 1.2 ktonne Trinitrotoluene (TNT)-equivalent (Department of Energy, 2015) is within the estimated posterior. Yield uncertainty can be estimated from the spread of the posterior, which is between 0.9 and 2.1 ktonne TNT-equivalent. The posterior for height-of-burst has a wider range between 10 m below and 8 m above ground that includes the true height-of-burst of 1 m.

58 GEOSCIENCES↗
DOE OSTI2026

Generative modeling enables molecular structure retrieval from Coulomb explosion imaging

Capturing the structural changes that molecules undergo during chemical reactions in real space and time is a long-standing dream and an essential prerequisite for understanding and ultimately controlling femtochemistry. A key approach to tackle this challenging task is Coulomb explosion imaging, which benefited decisively from recently emerging high-repetition-rate X-ray free-electron laser sources. With this technique, information on the molecular structure is inferred from the momentum distributions of the ions produced by the rapid Coulomb explosion of molecules. Retrieving molecular structures from these distributions poses a highly non-linear inverse problem that remains unsolved for molecules consisting of more than a few atoms. Here, we address this challenge using a diffusion-based Transformer neural network. We show that the network reconstructs unknown molecular geometries from ion-momentum distributions with a mean absolute error below one Bohr radius, which is half the length of a typical chemical bond.

Artificial Intelligence (cs.AI)↗