SEARCH · Engineering Papers

Results for “INTERPRETATION”

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.

SourceResearch connectionFrom yearThrough year

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

At least 1,153 records · Page 64

All sources NASA NTRS DOE OSTI

DOE OSTI2023

The Zonal Patterns in Late Quaternary Tropical South American Precipitation

Speleothem oxygen isotope records (δ 18 O) of tropical South American rainfall in the late Quaternary show a zonal “South American Precipitation Dipole” (SAPD). The dipole is characterized by opposing east-west precipitation anomalies compared to the present—wetter in the east and drier in the west at the mid-Holocene (~7 ka), and drier in the east and wetter in the west at the Last Glacial Maximum (~21 ka). However, the SAPD remains enigmatic because it is expressed differently in western versus eastern δ18O records and isotope-enabled climate model simulations usually misrepresent the magnitude and/or spatial pattern of δ 18 O change. Here, we address the SAPD enigma in two parts. First, we re-interpret the δ 18 O data to account for upwind rainout effects that are known to be pervasive in tropical South America, but are not always considered in Quaternary paleoclimate studies. Our revised interpretation reconciles the δ 18 O data with cave infiltration and other proxy records, and indicates that the centroid of tropical South American rainfall has migrated zonally over time. Second, using an energy balance model of tropical atmospheric circulation, we hypothesize that zonal migration of the precipitation centroid can be explained by regional energy budget shifts, such as changing Saharan albedo associated with the African Humid Period, that have not been modeled in previous SAPD studies. This hypothesis of a migrating precipitation centroid presents a new framework for interpreting δ18O records from tropical South America and may help explain the zonal rainfall anomalies that predate the late Quaternary.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2025

Experimental Soil Warming Impacts Soil Moisture and Plant Water Stress and Thereby Ecosystem Carbon Dynamics

Experimental soil heating experiments have found a consistent increase in soil-surface CO 2 emissions ( F s ), but inconsistent soil organic carbon (SOC) responses. Interpretation of heating effects is complicated by spatial heterogeneity and soil moisture, nitrogen availability, and microbial and plant responses. Here we applied a mechanistic ecosystem model to interpret heating impacts on a California forest subjected to 1 m deep, 4°C heating. The model accurately simulated control-plot CO 2 fluxes, SOC stocks, fine root biomass, soil moisture, and soil temperature, and the observed increases in F s and decreases in fine root biomass. We show that a complex suite of interactions can lead to a consistent increase in F s (~17%) over the 5-year study period, with very small changes in SOC stocks (<1%). Modeled increases in leaf water stress from soil drying reduced GPP and NPP. The resulting reduction in leaf and fine root allocation increased fine root litter inputs to the soil and reduced root exudation. Soil heating led to about a 50% larger increase in root autotrophic respiration than in heterotrophic respiration, with the heating effect on both these fluxes decreasing over the simulation period. Increased heterotrophic respiration led to increased soil N availability and plant N uptake. These heating responses are mechanistically linked, of magnitudes that can affect ecosystem dynamics, and long-term observations of them are rarely made. Therefore, we conclude that a coupled observational and mechanistic modeling framework is needed to interpret manipulation experiments, and to improve projections of climate change impacts on terrestrial ecosystem carbon dynamics.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2024

Revealing the hidden structure of disordered materials by parameterizing their local structural manifold

Abstract Durable interest in developing a framework for the detailed structure of glassy materials has produced numerous structural descriptors that trade off between general applicability and interpretability. However, none approach the combination of simplicity and wide-ranging predictive power of the lattice-grain-defect framework for crystalline materials. Working from the hypothesis that the local atomic environments of a glassy material are constrained by enthalpy minimization to a low-dimensional manifold in atomic coordinate space, we develop a generalized distance function, the Gaussian Integral Inner Product (GIIP) distance, in connection with agglomerative clustering and diffusion maps, to parameterize that manifold. Applying this approach to a two-dimensional model crystal and a three-dimensional binary model metallic glass results in parameters interpretable as coordination number, composition, volumetric strain, and local symmetry. In particular, we show that a more slowly quenched glass has a higher degree of local tetrahedral symmetry at the expense of cyclic symmetry. While these descriptors require post-hoc interpretation, they minimize bias rooted in crystalline materials science and illuminate a range of structural trends that might otherwise be missed.

36 MATERIALS SCIENCE↗

DOE OSTI2019

Analyzing machine learning models to accelerate generation of fundamental materials insights

Abstract Machine learning for materials science envisions the acceleration of basic science research through automated identification of key data relationships to augment human interpretation and gain scientific understanding. A primary role of scientists is extraction of fundamental knowledge from data, and we demonstrate that this extraction can be accelerated using neural networks via analysis of the trained data model itself rather than its application as a prediction tool. Convolutional neural networks excel at modeling complex data relationships in multi-dimensional parameter spaces, such as that mapped by a combinatorial materials science experiment. Measuring a performance metric in a given materials space provides direct information about (locally) optimal materials but not the underlying materials science that gives rise to the variation in performance. By building a model that predicts performance (in this case photoelectrochemical power generation of a solar fuels photoanode) from materials parameters (in this case composition and Raman signal), subsequent analysis of gradients in the trained model reveals key data relationships that are not readily identified by human inspection or traditional statistical analyses. Human interpretation of these key relationships produces the desired fundamental understanding, demonstrating a framework in which machine learning accelerates data interpretation by leveraging the expertize of the human scientist. We also demonstrate the use of neural network gradient analysis to automate prediction of the directions in parameter space, such as the addition of specific alloying elements, that may increase performance by moving beyond the confines of existing data.

36 MATERIALS SCIENCE↗

DOE OSTI2019

Multi-component background learning automates signal detection for spectroscopic data

Abstract Automated experimentation has yielded data acquisition rates that supersede human processing capabilities. Artificial Intelligence offers new possibilities for automating data interpretation to generate large, high-quality datasets. Background subtraction is a long-standing challenge, particularly in settings where multiple sources of the background signal coexist, and automatic extraction of signals of interest from measured signals accelerates data interpretation. Herein, we present an unsupervised probabilistic learning approach that analyzes large data collections to identify multiple background sources and establish the probability that any given data point contains a signal of interest. The approach is demonstrated on X-ray diffraction and Raman spectroscopy data and is suitable to any type of data where the signal of interest is a positive addition to the background signals. While the model can incorporate prior knowledge, it does not require knowledge of the signals since the shapes of the background signals, the noise levels, and the signal of interest are simultaneously learned via a probabilistic matrix factorization framework. Automated identification of interpretable signals by unsupervised probabilistic learning avoids the injection of human bias and expedites signal extraction in large datasets, a transformative capability with many applications in the physical sciences and beyond.

Ament, Sebastian E.↗

DOE OSTI2025

SA-GAT-SR: self-adaptable graph attention networks with symbolic regression for high-fidelity material property prediction

Recent advances in machine learning have demonstrated an enormous utility of deep learning approaches, particularly Graph Neural Networks (GNNs) for materials science. These methods have emerged as powerful tools for high-throughput prediction of material properties, offering a compelling enhancement and alternative to traditional first-principles calculations. While the community has predominantly focused on developing increasingly complex and universal models to enhance predictive accuracy, such approaches often lack physical interpretability and insights into materials behavior. Here, we introduce a novel computational paradigm—Self-Adaptable Graph Attention Networks integrated with Symbolic Regression (SA-GAT-SR)—that synergistically combines the predictive capability of GNNs with the interpretative power of symbolic regression. Our framework employs a self-adaptable encoding algorithm that automatically identifies and adjust attention weights so as to screen critical features from an expansive 180-dimensional feature space while maintaining O(n) computational scaling. The integrated SR module subsequently distills these features into compact analytical expressions that explicitly reveal quantum-mechanically meaningful relationships, achieving 23 × acceleration compared to conventional SR implementations that heavily rely on first-principle calculations-derived features as input. This work suggests a new framework in computational materials science, bridging the gap between predictive accuracy and physical interpretability, offering valuable physical insights into material behavior.

36 MATERIALS SCIENCE↗

DOE OSTI2020

Machine learning for endoleak detection after endovascular aortic repair

Diagnosis of endoleak following endovascular aortic repair (EVAR) relies on manual review of multi-slice CT angiography (CTA) by physicians which is a tedious and time-consuming process that is susceptible to error. We evaluate the use of a deep neural network for the detection of endoleak on CTA for post-EVAR patients using a novel data efficient training approach. 50 CTAs and 20 CTAs with and without endoleak respectively were identified based on gold standard interpretation by a cardiovascular subspecialty radiologist. The Endoleak Augmentor, a custom designed augmentation method, provided robust training for the machine learning (ML) model. Predicted segmentation maps underwent post-processing to determine the presence of endoleak. The model was tested against 3 blinded general radiologists and 1 blinded subspecialist using a held-out subset (10 positive endoleak CTAs, 10 control CTAs). Model accuracy, precision and recall for endoleak diagnosis were 95%, 90% and 100% relative to reference subspecialist interpretation (AUC = 0.99). Accuracy, precision and recall was 70/70/70% for generalist1, 50/50/90% for generalist2, and 90/83/100% for generalist3. The blinded subspecialist had concordant interpretations for all test cases compared with the reference. In conclusion, our ML-based approach has similar performance for endoleak diagnosis relative to subspecialists and superior performance compared with generalists.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2018

Fast and accurate view classification of echocardiograms using deep learning

Echocardiography is essential to cardiology. However, the need for human interpretation has limited echocardiography’s full potential for precision medicine. Deep learning is an emerging tool for analyzing images but has not yet been widely applied to echocardiograms, partly due to their complex multi-view format. The essential first step toward comprehensive computer-assisted echocardiographic interpretation is determining whether computers can learn to recognize these views. We trained a convolutional neural network to simultaneously classify 15 standard views (12 video, 3 still), based on labeled still images and videos from 267 transthoracic echocardiograms that captured a range of real-world clinical variation. Our model classified among 12 video views with 97.8% overall test accuracy without overfitting. Even on single low-resolution images, accuracy among 15 views was 91.7% vs. 70.2–84.0% for board-certified echocardiographers. Data visualization experiments showed that the model recognizes similarities among related views and classifies using clinically relevant image features. Our results provide a foundation for artificial intelligence-assisted echocardiographic interpretation.

Madani, Ali↗

DOE OSTI2025

A Representation Fusion Framework for Decoupling Diagnostic Information in Multimodal Learning

Modern medicine increasingly relies on multimodal data, ranging from clinical notes to imaging and genomics, to guide diagnosis and treatment. However, integrating these heterogeneous data sources in a principled and interpretable manner remains a major challenge. We present MODES (Multi-mOdal Disentangled Embedding Space), a representation fusion framework that explicitly separates shared and modality-specific factors of variation, offering a structured latent space for multimodal information that improves both prediction and interpretability. By leveraging pre-trained unimodal foundation models, MODES mitigates the dependency on extensive paired datasets, crucial in data-scarce clinical settings. We introduce a masking strategy that optimizes representation dimensionality by eliminating low-information dimensions, to achieve compact, information-rich representations. Our framework demonstrates superior performance in predicting diagnoses and phenotypes compared to unimodal and conventional fusion models. MODES also enables robust diagnostic inference in missing data scenarios, offering an opportunity toward interpretable and efficient multimodal diagnostics in personalized healthcare.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2013

A simple model of universe describing the early inflation and the late accelerated expansion in a symmetric manner

We construct a simple model of universe which 'unifies' vacuum energy and radiation on the one hand, and matter and dark energy on the other hand in the spirit of a generalized Chaplygin gas model. Specifically, the phases of early inflation and late accelerated expansion are described by a generalized equation of state p/c{sup 2} = αρ+kρ{sup 1+1/n} having a linear component p = αρc{sup 2} and a polytropic component p = kρ{sup 1+1/n}c{sup 2}. For α= 1/3, n= 1 and k=−4/(3ρ{sub P}), where ρ{sub P}= 5.1610{sup 99} g/m{sup 3} is the Planck density, this equation of state describes the transition between the vacuum energy era and the radiation era. For t≥ 0, the universe undergoes an inflationary expansion that brings it from the Planck size l{sub P}= 1.6210{sup −35} m to a size a{sub 1}= 2.6110{sup −6} m on a timescale of about 23.3 Planck times t{sub P}= 5.3910{sup −44} s (early inflation). When t > t{sub 1}= 23.3t{sub P}, the universe decelerates and enters in the radiation era. We interpret the transition from the vacuum energy era to the radiation era as a second order phase transition where the Planck constant ℏ plays the role of finite size effects (the standard Big Bang theory is recovered for ℏ= 0). For α= 0, n=−1 and k=−ρ{sub Λ}, where ρ{sub Λ}= 7.0210{sup −24} g/m{sup 3} is the cosmological density, the equation of state p/c{sup 2} = αρ+kρ{sup 1+1/n} describes the transition from a decelerating universe dominated by pressureless matter (baryonic and dark matter) to an accelerating universe dominated by dark energy (late inflation). This transition takes place at a size a{sub 2}= 0.204l{sub Λ}. corresponding to a time t{sub 2}= 0.203t{sub Λ} where l{sub Λ}= 4.38 10{sup 26} m is the cosmological length and t{sub Λ}= 1.46 10{sup 18} s the cosmological time. The present universe turns out to be just at the transition between these two periods (t{sub 0}∼t{sub 2}). Our model gives the same results as the standard ΛCDM model for t≫t{sub P} and completes it by incorporating a phase of early inflation for t < 23.3t{sub P} in a very natural manner. Furthermore, it reveals a nice 'symmetry' between the early and the late evolution of the universe. The early universe is modeled by a polytrope n=+ 1 and the late universe by a polytrope n=−1. Furthermore, the cosmological constant Λ in the late universe plays a role similar to the Planck constant ℏ in the early universe. The mathematical formulae in the early and in the late universe are then strikingly symmetric. We interpret the cosmological constant as a fundamental constant of Nature describing the 'cosmophysics' just like the Planck constant describes the 'microphysics'. The Planck density and the cosmological density represent fundamental upper and lower bounds differing by 122 orders of magnitude. The cosmological constant 'problem' may be a false problem. Finally, we show that our model admits a scalar field interpretation based on a quintessence field or a tachyon field.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

DOE OSTI2013

Experimental validation of Mueller-Stokes theory and investigation of the influence of the Cotton-Mouton effect on polarimetry in a magnetized fusion plasma

Mueller-Stokes theory can be used to calculate the polarization evolution of an electromagnetic (EM) wave as it propagates through a magnetized plasma. Historically, the theory has been used to interpret polarimeter signals from systems operating on fusion plasmas. These interpretations have mostly employed approximations of Mueller-Stokes theory in regimes where either the Faraday rotation (FR) or the Cotton-Mouton (CM) effect is dominant. The current paper presents the first systematic comparison of polarimeter measurements with the predictions of full Mueller-Stokes theory where conditions transition smoothly from a FR-dominant (i.e., weak CM effect) plasma to one where the CM effect plays a significant role. A synthetic diagnostic code, based on Mueller-Stokes theory accurately reproduces the trends evident in the experimentally measured polarimeter phase over this entire operating range, thereby validating Mueller-Stokes theory. The synthetic diagnostic code is then used to investigate the influence of the CM effect on polarimetry measurements. As expected, the measurements are well approximated by the FR effect when the CM effect is predicted to be weak. However, the code shows that as the CM effect increases, it can compete with the FR effect in rotating the polarization of the EM-wave. This results in a reduced polarimeter response to the FR effect, just as observed in the experiment. The code also shows if sufficiently large, the CM effect can even reverse the handedness of a wave launched with circular polarization. This helps to understand the surprising experimental observations that the sensitivity to the FR effect can be nearly eliminated at high enough B{sub T} (2.0 T). The results also suggest that the CM effect on the plasma midplane can be exploited to potentially measure magnetic shear in tokamak plasmas. These results establish increased confidence in the use of such a synthetic diagnostic code to guide future polarimetry design and interpret the resultant experimental data.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

DOE OSTI2016

Combined theoretical and experimental study of the valence, Rydberg, and ionic states of chlorobenzene

New photoelectron (PE) and ultra violet (UV) and vacuum UV (VUV) spectra have been obtained for chlorobenzene by synchrotron study with higher sensitivity and resolution than previous work and are subjected to detailed analysis. In addition, we report on the mass-resolved (2 + 1) resonance enhanced multiphoton ionization (REMPI) spectra of a jet-cooled sample. Both the VUV and REMPI spectra have enabled identification of a considerable number of Rydberg states for the first time. The use of ab initio calculations, which include both multi-reference multi-root doubles and singles configuration interaction (MRD-CI) and time dependent density functional theoretical (TDDFT) methods, has led to major advances in interpretation of the vibrational structure of the ionic and electronically excited states. Franck-Condon (FC) analyses of the PE spectra, including both hot and cold bands, indicate much more complex envelopes than previously thought. The sequence of ionic states can be best interpreted by our multi-configuration self-consistent field computations and also by comparison of the calculated vibrational structure of the B and C ionic states with experiment; these conclusions suggest that the leading sequence is the same as that of iodobenzene and bromobenzene, namely: X{sup 2}B{sub 1}(3b{sub 1}{sup −1}) < A{sup 2}A{sub 2}(1a{sub 2}{sup −1}) < B{sup 2}B{sub 2}(6b{sub 2}{sup −1}) < C{sup 2}B{sub 1}(2b{sub 1}{sup −1}). The absorption onset near 4.6 eV has been investigated using MRD-CI and TDDFT calculations; the principal component of this band is {sup 1}B{sub 2} and an interpretation based on the superposition of FC and Herzberg-Teller contributions has been performed. The other low-lying absorption band near 5.8 eV is dominated by a {sup 1}A{sub 1} state, but an underlying weak {sup 1}B{sub 1} state (πσ{sup ∗}) is also found. The strongest band in the VUV spectrum near 6.7 eV is poorly resolved and is analyzed in terms of two ππ{sup ∗} states of {sup 1}A{sub 1} (higher oscillator strength) and {sup 1}B{sub 2} (lower oscillator strength) symmetries, respectively. The calculated vertical excitation energies of these two states are critically dependent upon the presence of Rydberg functions in the basis set, since both manifolds are strongly perturbed by the Rydberg states in this energy range. A number of equilibrium structures of the ionic and singlet excited states show that the molecular structure is less subject to variation than corresponding studies for iodobenzene and bromobenzene.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

DOE OSTI2016

Light leptonic new physics at the precision frontier

Precision probes of new physics are often interpreted through their indirect sensitivity to short-distance scales. In this proceedings contribution, we focus on the question of which precision observables, at current sensitivity levels, allow for an interpretation via either short-distance new physics or consistent models of long-distance new physics, weakly coupled to the Standard Model. The electroweak scale is chosen to set the dividing line between these scenarios. In particular, we find that inverse see-saw models of neutrino mass allow for light new physics interpretations of most precision leptonic observables, such as lepton universality, lepton flavor violation, but not for the electron EDM.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

DOE OSTI2017

Metastable defect response in CZTSSe from admittance spectroscopy

Admittance spectroscopy is a useful tool used to study defects in semiconductor materials. However, metastable defect responses in non-ideal semiconductors can greatly impact the measurement and therefore the interpretation of results. Here, admittance spectroscopy was performed on Cu2ZnSn(S,Se) 4 where metastable defect response is illustrated due to the trapping of injected carriers into a deep defect state. To investigate the metastable response, admittance measurements were performed under electrically and optically relaxed conditions in comparison to a device following a low level carrier-injection pretreatment. The relaxed measurement demonstrates a single capacitance signature while two capacitance signatures are observed for the device measured following carrier-injection. The deeper level signature, typically reported for kesterites, is activated by charge trapping following carrier injection. Both signatures are attributed to bulk level defects. The significant metastable response observed on kesterites due to charge trapping obscures accurate interpretation of defect levels from admittance spectroscopy and indicates that great care must be taken when performing and interpreting this measurement on non-ideal devices.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

DOE OSTI2020

Characterizing magnetized plasmas with dynamic mode decomposition

Accurate and efficient plasma models are essential to understand and control experimental devices. Existing magnetohydrodynamic or kinetic models are nonlinear and computationally intensive and can be difficult to interpret, while often only approximating the true dynamics. In this work, data-driven techniques recently developed in the field of fluid dynamics are leveraged to develop interpretable reduced-order models of plasmas that strike a balance between accuracy and efficiency. In particular, dynamic mode decomposition (DMD) is used to extract spatio-temporal magnetic coherent structures from the experimental and simulation datasets of the helicity injected torus with steady inductive (HIT-SI) experiment. Three-dimensional magnetic surface probes from the HIT-SI experiment are analyzed, along with companion simulations with synthetic internal magnetic probes. A number of leading variants of the DMD algorithm are compared, including the sparsity-promoting and optimized DMD. Optimized DMD results in the highest overall prediction accuracy, while sparsity-promoting DMD yields physically interpretable models that avoid overfitting. These DMD algorithms uncover several coherent magnetic modes that provide new physical insights into the inner plasma structure. These modes were subsequently used to discover a previously unobserved three-dimensional structure in the simulation, rotating at the second injector harmonic. Finally, using data from probes at experimentally accessible locations, DMD identifies a resistive kink mode, a ubiquitous instability seen in magnetized plasmas.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

DOE OSTI2024

Theory and modeling of molecular modes in the NMR relaxation of fluids

Traditional theories of the nuclear magnetic resonance (NMR) autocorrelation function for intra-molecular dipole pairs assume a single-exponential decay, yet the calculated autocorrelation of realistic systems displays a rich, multi-exponential behavior, resulting in anomalous NMR relaxation dispersion (i.e., frequency dependence). We develop an approach to model and interpret the multi-exponential intra-molecular autocorrelation using simple, physical models within a rigorous statistical mechanical development that encompasses both rotational diffusion and translational diffusion in the same framework. Here, we recast the problem of evaluating the autocorrelation in terms of averaging over a diffusion propagator whose evolution is described by a Fokker–Planck equation. The time-independent part admits an eigenfunction expansion, allowing us to write the propagator as a sum over modes. Each mode has a spatial part that depends on the specified eigenfunction and a temporal part that depends on the corresponding eigenvalue (i.e., correlation time) with a simple, exponential decay. The spatial part is a probability distribution of the dipole pair, analogous to the stationary states of a quantum harmonic oscillator. Drawing inspiration from the idea of inherent structures in liquids, we interpret each of the spatial contributions as a specific molecular mode. These modes can be used to model and predict the NMR dipole–dipole relaxation dispersion of fluids by incorporating phenomena on the molecular level. We validate our statistical mechanical description of the distribution in molecular modes with molecular dynamics simulations interpreted without any relaxation models or adjustable parameters: the most important poles in the Padé–Laplace transform of the simulated autocorrelation agree with the eigenvalues predicted by the theory

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

DOE OSTI2025

VERITAS : A density-functional theory-based multiband kinetic model for understanding x-ray spectroscopy of dense plasmas

X-ray spectroscopy has long been a powerful diagnostic tool for hot, dilute plasmas, providing insights into plasma conditions by measuring line shifts and broadenings of atomic transitions. The technique critically depends on the accuracy of atomic physics models used to interpret spectroscopic measurements for inferring plasma properties such as free-electron density and temperature. Over the past decades, the atomic and plasma physics communities have developed robust atomic physics models to account for various processes in hot, dilute classical plasmas. While these models have been successful in that regime, their applicability becomes uncertain when interpreting x-ray spectroscopy experiments of above-solid-density plasmas. Given that finite-temperature density-functional theory (DFT) offers a more accurate description of dense plasma environments, we present the development of a DFT-based multi-band kinetic model, VERITAS, designed to improve the interpretation of x-ray spectroscopic measurements in high-density plasmas produced by laser-driven spherical implosions. This work details the VERITAS model and its application to both time-integrated and time-resolved x-ray spectra from implosion experiments on OMEGA. The advantages and limitations of the VERITAS model will also be discussed, along with potential directions for advancing x-ray spectroscopy of dense and superdense plasmas.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

DOE OSTI2026

Attention-based explainability for structure–property relationships

Machine learning methods are emerging as a universal paradigm for constructing correlative structure–property relationships in materials science based on multimodal characterization. However, this necessitates the development of methods for the physical interpretability of the resulting correlative models. Here, we demonstrate the potential of attention-based neural networks for revealing structure–property relationships and the underlying physical mechanisms, using the ferroelectric properties of PbTiO3 thin films as a case study. Through the analysis of attention scores, we disentangle the influence of distinct domain patterns on the polarization switching process. The attention-based Transformer model is explored both as a direct interpretability tool and as a surrogate for explaining representations learned via unsupervised machine learning, enabling the identification of physically grounded correlations. We compare attention-derived interpretability scores with classical SHapley Additive exPlanations analysis and show that, in contrast to applications in natural language processing, attention mechanisms in materials science exhibit high efficiency in highlighting meaningful structural features.

Slautin, Boris [Independent Researcher]↗