Search NASA⌕ Search

SEARCH · Search NASA

Results for “Edge Modeling”

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 1,189 records · Page 66

Characterizing the negative triangularity reactor core operating space with integrated modeling

Negative triangularity (NT) has received renewed interest as a fusion reactor regime due to its beneficial power-handling properties, including low scrape-off layer power and a larger divertor wetted area that facilitates simple divertor integration. NT experiments have also demonstrated core performance on par with positive triangularity (PT) high confinement mode (H-mode) without edge-localized modes (ELMs), encouraging further study of an NT reactor core. In this work, we use integrated modeling to scope the operating space around two NT reactor strategies. The first is the high-field, compact fusion pilot plant concept Modular, Adjustable, NT ARC (MANTA) and the second is a low field, high aspect ratio concept based on work by Medvedev et al. By integrating equilibrium, core transport, and edge ballooning instability models, we establish a range of operating points with less than 50 MW scrape-off layer power and fusion power comparable to PT H-mode reactor concepts. Heating and seeded impurities are leveraged to accomplish the same fusion performance and scrape-off layer exhaust power for various pressure edge boundary conditions. Scans over these pressure edge conditions accommodate any current uncertainty of the properties of the NT edge and show that the performance of an NT reactor will be extremely dependent on the edge pressure. The high-field case is found to enable lower scrape-off layer power because it is capable of reaching high fusion powers at a relatively compact size, which allows increased separatrix density without exceeding the Greenwald density limit. Adjustments in NT shaping exhibit small changes in fusion power, with an increase in fusion power density seen at weaker NT. Infinite-n ballooning instability models indicate that an NT reactor core can reach fusion powers comparable to leading PT H-mode reactor concepts while remaining ballooning-stable. Seeded krypton is leveraged to further lower scrape-off layer power since NT does not have a requirement to remain in H-mode while still maintaining high confinement. We contextualize the NT reactor operating space by comparing to popular PT H-mode reactor concepts, and find that NT exhibits competitive ELM-free performance with these concepts for a variety of edge conditions while maintaining relatively low scrape-off layer power.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Phased Array Characterization of Slat Noise Radiation from a High-Lift Common Research Model

A test campaign was conducted in the open-jet test section of the NASA Langley 14- by 22-Foot Subsonic Tunnel on a 10%-scale, semispan version of a High-Lift Common Research Model (CRM-HL) incorporating a leading-edge slat, trailing-edge flap and removable high-fidelity main landing gear. Far field acoustic measurements were obtained on baseline and acoustically treated model configurations using a traversing 97-microphone phased array that viewed the pressure side of the airframe. A modified version of the DAMAS deconvolution method incorporating corrections for shear layer acoustic wave decorrelation was employed to determine the locations and strengths of relevant noise sources along the span of the slat and in the vicinity of the flap edges and landing gear. The main goal of the test campaign was to use the array data to evaluate the effectiveness of various slat noise reduction concepts. It was found that slat-gap fillers produced significant broadband noise reduction for all Mach numbers and airfoil angles of attack that were investigated. Mach number scaling revealed an approximate velocity to the sixth power relationship for the slat noise at higher frequencies. Directivity measurements obtained from integration of DAMAS pressure-squared values over defined geometric zones around the model showed that baseline slat noise radiation approximated a dipole as a function of polar emission angle, becoming approximately omnidirectional with the introduction of the slat-gap filler. The test campaign clearly demonstrated the noise reduction benefits that can be obtained by applying appropriate treatments to leading edge slats on commercial transport-class aircraft.

William M Humphreys, Jr.↗

Quasiconical flowfield structure of the three-dimensional single fin interaction

A series of conical and three-dimensional computations have been performed for the swept oblique shock wave/turbulent boundary-layer interaction generated by a 20-deg sharp fin at Mach 4 and freestream Reynolds number of 2.18 x 10 exp 5 based on the incoming boundary-layer thickness. The Reynolds-averaged compressible Navier-Stokes equations are employed with turbulence incorporated using the Baldwin-Lomax and Jones-Launder models. The computed results are basically similar for both turbulence models and display general agreement with experimental data for surface pressure and surface flow direction, although underestimating the size of the primary vortex. The computed three-dimensional flowfield displays quasiconical behavior of the surface pressure, surface flow direction, and flowfield contours of static pressure, density, and Mach number over the extent of the computational domain except for an inception region near the fin leading edge. Certain features of the flowfield model are not observed in the computations, namely, a 'normal' shock near the attachment line, transonic shocklets in the expansion region, and secondary separation. The absence of these features in the computation is believed to be indirectly attributable to limitations in the turbulence models.

Knight, Doyle D.↗

Analysis of pedestal plasma transport

An H-mode edge pedestal plasma transport benchmarking exercise was undertaken for a single DIII-D pedestal. Transport modelling codes used include 1.5D interpretive (ONETWO, GTEDGE), 1.5D predictive (ASTRA) and 2D ones (SOLPS, UEDGE). The particular DIII-D discharge considered is 98889, which has a typical low density pedestal. Profiles for the edge plasma are obtained from Thomson and charge-exchange recombination data averaged over the last 20% of the average 33.53 ms repetition time between type I edge localized modes. The modelled density of recycled neutrals is largest in the divertor X-point region and causes the edge plasma source rate to vary by a factor similar to 10(2) on the separatrix. Modelled poloidal variations in the densities and temperatures on flux surfaces are small on all flux surfaces up to within about 2.6 mm (rho(N) > 0.99) of the mid-plane separatrix. For the assumed Fick's-diffusion-type laws, the radial heat and density fluxes vary poloidally by factors of 2-3 in the pedestal region; they are largest on the outboard mid-plane where flux surfaces are compressed and local radial gradients are largest. Convective heat flows are found to be small fractions of the electron (less than or similar to 10%) and ion (less than or similar to 25%) heat flows in this pedestal. Appropriately averaging the transport fluxes yields interpretive 1.5D effective diffusivities that are smallest near the mid-point of the pedestal. Their 'transport barrier' minima are about 0.3 (electron heat), 0.15 (ion heat) and 0.035 (density) m(2) s(-1). Electron heat transport is found to be best characterized by electron-temperature-gradient-induced transport at the pedestal top and paleoclassical transport throughout the pedestal. The effective ion heat diffusivity in the pedestal has a different profile from the neoclassical prediction and may be smaller than it. The very small effective density diffusivity may be the result of an inward pinch flow nearly balancing a diffusive outward radial density flux. The inward ion pinch velocity and density diffusion coefficient are determined by a new interpretive analysis technique that uses information from the force balance (momentum conservation) equations; the paleoclassical transport model provides a plausible explanation of these new results. Finally, the measurements and additional modelling needed to facilitate better pedestal plasma transport modelling are discussed.

ASDEX-UPGRADE↗

Constraint-dependent CTOA determination for stable ductile crack growth

The crack tip opening angle (CTOA) has been used as a reliable fracture toughness to characterize stable crack growth for thin-wall structures in low-constraint conditions. Recently, it is found that CTOA can also be utilized as a robust fracture parameter to describe the arrest fracture toughness for gas transmission pipelines in ductile steels. For better understanding the concept of CTOA, this paper presents a brief review on the CTOA definitions and its test methods. And then, four CTOA models are developed for determining the CTOA toughness using conventional single edge notched bend (SENB) specimens, including the load–displacement linear fit model, the logarithmic load–displacement linear fit model, the stable tearing energy model, and the J-differentiation model. Here fracture test data from SENB specimens in A285 carbon steel are employed to evaluate the proposed CTOA models. The results show that these CTOA models can determine either identical or comparable critical CTOA values for stable ductile crack growth using the SENB specimens. A constraint-corrected CTOA resistance curve and a constraint-corrected critical CTOA equation are also obtained for A285 carbon steel.

42 ENGINEERING↗

Bifurcation theory of a one-dimensional transport model for the L-H transition

Transitions between low and high-confinement (L-H transitions) in magnetically confined plasmas can appear as three qualitatively different types: sharp, smooth, and oscillatory. Bifurcation analysis unravels these possible transition types and how they are situated in parameter space. In this paper the bifurcation analysis is applied to a 1-dimensional model for the radial transport of energy and density near the edge of magnetically confined plasmas. This phenomenological L-H transition model describes the reduction of the turbulent transport by E×B-flow shear self-consistently with the evolution of the radial electric field. Therewith, the exact parameter space, including the threshold values of the control parameters, of the possible L-H transitions in the model is determined. Furthermore, a generalised equal area rule is derived to describe the evolution of the transport barrier in space and time self-consistently. Applying this newly developed rule to the model analysed in this paper reveals a naturally occurring transition to an extra wide transport barrier that may correspond to the improved confinement known as the very-high-confinement mode.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

The X-Ray Warm Absorber in NGC 3516

The Seyfert 1 galaxy, NGC 3516 has been the subject of many absorption line studies at both ultra-violet and X-ray wavelengths. In the UV, strong, broad, variable associated metal line absorption with velocity width approximately 2000 km s(exp -1) is thought to originate in gas with N(sub H)greater than or approximately 10(exp 10)cm (exp -2) lying between 0.01 and 9 pc from the central active nucleus. The Ginga X-ray data are consistent with several possibilities: a warm absorber and a cold absorber combined either with partial covering or an unusually strong reflection spectrum. We present ROSAT observations of NGC 3516 which show a strong detection of a warm absorber dominated by a blend of 0 VII/O VIII edges at approximately 0.8 keV with N(sub H) approximately 7x10(exp 21)cm(exp -2), and U:8-12. We argue that NGC 3516 contains an outflowing 'XUV' absorber showing the presence of X-ray absorption edges which are consistent with the presence of broad absorption lines in the old IUE spectra and their disappearance in the new UV observations. Our dynamical model suggests that the O VII absorption edge will continue to weaken compared to the O VIII edge, an easily testable prediction with future missions like AXAF. Eventually the source would be transparent to the X-rays unless a new absorption system is produced.

Mathur, Smita↗

SERS Theory: The Chemical Effect of Rhodamine 6G Adsorption on Silver Surfaces on Its Raman Spectrum

The SERS spectra of small organic molecules can be very sensitive to the physisorption or chemisorption of the molecules onto metallic surfaces. However it is challenging to model the results, as typical SERS substrates have a broad distribution of possible surface sites, and the populations of these sites and the SERS enhancement factors associated with them are very difficult to determine, especially when electromagnetic hot spots dominate the observed spectra. However it is important to account for these effects if one is to make a quantitative interpretation of measured SERS spectra. Indeed, the influence of different sites is important to both the electromagnetic and chemical mechanisms in SERS, where the former mechanism is concerned with local field enhancement effects arising from plasmon excitation, while the latter arises from the effect of chemisorption induced charged transfer on the Raman intensities, as can be modeled using static limit Raman intensity calculations[1]. Extensive studies have been performed on simple organics like pyridine bound to gold and silver clusters, and it was found that both the chemical and electrodynamic enhancement factors associated with binding pyridine in various orientations result in noticeable changes in the SERS spectra compared to the gas phase spectrum of pyridine.[2] One of the most important benchmark molecules for SERS and TERS studies is Rhodamine 6G (R6G). This is a strongly fluorescent dye molecule in solution, but when adsorbed onto SERS substrates, its fluorescence is quenched, and the resulting SERS intensities are among the largest observed for any molecule. As a result R6G has played a crucial role in the development of single molecule SERS, and in the development of SERS substrates. [3, 4] Previously, the SERS spectrum of R6G has been modeled as if it were a gas phase molecule. Jensen and Schatz[5] incorporated resonant effects as well as vibronic effects and found that resonance excitation to the S1 excited state of R6G resulted in good agreement with spectra obtained from SERS experiments. In another study, intermolecular charge transfer was studied with R6G bound to a Ag2 cluster.[6] It is unclear if a cluster model as small as two silver atoms can accurately capture the surface effects of a nanoparticle with thousands of atoms. Recent single molecule, ultra high vacuum, tip enhanced Raman spectroscopy (SM-UHV-TERS) measurements for R6G[3] were moderately reproduced by gas phase resonance Raman calculations. Of particular importance in this work was the comparison of TERS spectra at room temperature with those at a low temperature of 10K. It was found that lowering the temperature led to both sharpening of the vibrational spectra, and shifts in the spectra (up to 20 cm-1 with some lines shifting red, some blue and some not at all), suggesting that the molecule transitions from a largely gas phase molecule at room temperature to being strongly chemisorbed at 10K. A potential energy distribution analysis performed by Chulhai and Jensen[3] showed that the modes that were shifted from the gas phase had vibrational activity associated with the ethylamine component of R6G, which is the likely moiety associated with Raman calculations. What remains unresolved concerning the SM-UHV-TERS studies of R6G is direct evidence that the chemical binding of R6G to a silver nanoparticle surfaces causes frequency shifts and/or intensity changes in the Raman scattering intensities of specific modes, especially those involving vibrations of the ethylamine moiety. The theoretical work described in this chapter directly addresses this issue by calculating Raman spectra for R6G interacting with a 20-atom silver cluster, and with a flat Ag(111) surface. Tetrahedral Ag20 is chosen for part of these studies as it provides a model nanoparticle structure that has a well-defined fragment of a (111) surface, as well as having edges and vertices that can be used to model coordinately unsaturated sites on a particle. In addition, Ag20 has been found to provide a reasonable reference structure for determining the chemical contribution to SERS enhancements.[1] Two different geometries of the Ag20/R6G system are considered in the present study, corresponding to two different binding configurations of R6G to a pristine Ag(111) surface. In addition, we study adsorption of R6G on a flat Ag(111) surface of using plane wave DFT calculations to determine optimized geometries and normal mode properties.

77 NANOSCIENCE AND NANOTECHNOLOGY↗

Thin-layer approximation and algebraic model for separated turbulent flows

An algebraic turbulence model for two- and three-dimensional separated flows is specified that avoids the necessity for finding the edge of the boundary layer. Properties of the model are determined and comparisons made with experiment for an incident shock on a flat plate, separated flow over a compression corner, and transonic flow over an airfoil. Separation and reattachment points from numerical Navier-Stokes solutions agree with experiment within one boundary-layer thickness. Use of law-of-the-wall boundary conditions does not alter the predictions significantly. Applications of the model to other cases are contained in companion papers.

Baldwin, B.↗

Stalled-Flow and Head-Loss Model for Diffuser Pumps

Modeling procedure approximates inlet transition zone (blade leading edge to blade throat) of diffuser pump as two-dimensional cascade, properties of which are well known. Model applied to stators as well as rotors. Procedure much faster than previous methods.

Meng, S. Y.↗

Fate and transport of uranium (VI) in weathered saprolite

We conducted batch and column experiments to investigate sorption and transport of uranium (U) in the presence of saprolite derived from interbedded shale, limestone, and sandstone sequences. Sorption kinetics were measured at two initial concentrations (C0; 1, 10 mM) and three soil:solution ratios (Rs/w; 0.005, 0.25, 2 kg/L) at pH 4.5 (pH of the saprolite). The rate of U loss from solution (mmole/L/h) increased with increasing Rs/w. Uranium sorption exhibited a fast phase with 80% sorption in the first eight hours for all C0 and Rs/w values and a slow phase during which the reaction slowly approached (pseudo) equilibrium over the next seven days. The pH-dependency of U sorption was apparent in pH sorption edges. U(VI) sorption increased over the pH range 4e6, then decreased sharply at pH > 7.5. U(VI) sorption edges were well described by a surface complexation model using calibrated parameters and the reaction network proposed by Waite et al. (1994). Sorption isotherms measured using the same Rs/w and pH values showed a solids concentration effect where U(VI) sorption capacity and affinity decreased with increasing solids concentration. Moreover, this effect may have been due to either particle aggregation or competition between U(VI) and exchangeable cations for sorption sites. The surface complexation model with calibrated parameters was able to predict the general sorption behavior relatively well, but failed to reproduce solid concentration effects, implying the importance of appropriate design if batch experiments are to be utilized for dynamic systems. Transport of U(VI) through the packed column was significantly retarded. We also conducted transport simulations using the reactive transport model HydroGeoChem (HGC) v5.0 that incorporated the surface complexation reaction network used to model the batch data. Model parameters reported by Waite et al. (1994) provided a better prediction of U transport than optimized parameters derived from our sorption edges. The results presented in this study highlight the challenges in defining appropriate conditions for batch-type experiments used to extrapolate parameters for transport models, and also underline a gap in our ability to transfer batch results to transport simulations.

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

Simulations of the edge plasma: the role of atomic, molecular and surface physics

Atomic, molecular and surface physics plays an important role in simulations of the edge plasma in present day tokamaks, and in the predictive simulations of new devices. The edge plasma - in this context, the Scrape-Off Layer (SOL), the Private Flux Region (PFR) and core region close to the separatrix (or Last Closed Flux Surface, LCFS) - provides the boundary conditions for the main plasma, and is the region where much of the power and all of the particle exhaust occurs. It is also the region where the plasma interacts with solid surfaces, puffed gases and gas arising from recycling. The results of plasma edge simulations can depend strongly on the availability and quality of the atomic, molecular and surface data (the peak plasma temperature at the divertor was found to vary by a factor of five dependent on the choice of atomic physics data in a recent sensitivity analysis). The current material choice for ITER with Plasma Facing Components (PFCs) consisting of C, Be and W also presents challenges, both in the availability of the necessary data for W, and in the plethora of charge states for W. Another challenge presented by the material choice is the likely presence of mixed materials formed by the migration of material from one surface to another. These introduce effects like alloying and preferential sputtering as well as new (much longer) time-scales in the problem.Efforts to incorporate a bundled charge state model within one of the present edge simulation codes, SOLPS, will be described, as well as efforts to address some of the questions raised by mixed materials. Some issues related to data consistency and traceability within the context of the European effort on Integrated Tokamak Modelling will also be addressed.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Magnetic Polarization Measurements of the Multi-modal Plasma Response to 3D fields in the EAST Tokamak

A multi-modal plasma response to applied non-axisymmetric fields has been found in EAST tokamak plasmas. Here, multi-modal means the radial and poloidal structure of an individually driven toroidal harmonic is not fixed. The signature of such a multi-modal response is the magnetic polarization (ratio of radial and poloidal components) of the plasma response field measured on the low field side device mid-plane. A difference in the 3D coil phasing (the relative phase of two coil arrays) dependencies between the two responses is observed in response to n=2 fields in the same plasma for which the n=1 responses are well synchronized. Neither the maximum radial nor the maximum poloidal field response to n=2 fields agrees with the best applied phasing for mitigating edge localized modes, suggesting that the edge plasma response is not a dominant component of either polarization. GPEC modeling reproduces the discrepant phasing dependences of the experimental measurements, and confirms the edge resonances are maximized by the coil phasing that mitigates ELMs in the experiments. The model confirms the measured plasma response is not dominated by resonant current drive from the external field. Instead, non-resonant contributions play a large role in the diagnostic signal for both toroidal harmonics n=1 and n=2. The analysis in this paper demonstrates the ability of 3D modeling to connect external magnetic sensor measurements to the internal plasma physics and accurately predict optimal applied 3D field configurations in multi-modal plasmas.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

EdgeAI: Machine learning via direct attached accelerator for streaming data processing at high shot rate x-ray free-electron lasers

We present a case for low batch-size inference with the potential for adaptive training of a lean encoder model. We do so in the context of a paradigmatic example of machine learning as applied in data acquisition at high data velocity scientific user facilities such as the Linac Coherent Light Source-II x-ray Free-Electron Laser. We discuss how a low-latency inference model operating at the data acquisition edge can capitalize on the naturally stochastic nature of such sources. We simulate the method of attosecond angular streaking to produce representative results whereby simulated input data reproduce high-resolution ground truth probability distributions. By minimizing the mean-squared error between the decoded output of the latent representation and the ground truth distributions, we ensure that the encoding layers and resulting latent representation maintains full fidelity for any downstream task, be it classification or regression. We present throughput results for data-parallel inference of various batch sizes, some with throughput exceeding 100 k images per second. We also show in situ training below 10 s per epoch for the full encoder–decoder model as would be relevant for streaming and adaptive real-time data production at our nation’s scientific light sources.

97 MATHEMATICS AND COMPUTING↗

A multiscale anisotropic polymer network model coupled with phase field fracture

The study of polymers has continued to gain substantial attention due to their expanding range of applications, spanning essential engineering fields to emerging domains like stretchable electronics, soft robotics, and implantable sensors. These materials exhibit remarkable properties, primarily stemming from their intricate polymer chain network, which, in turn, increases the complexity of precisely modeling their behavior. Especially for modeling elastomers and their fracture behavior, accurately accounting for the deformations of the polymer chains is vital for predicting the rupture in highly stretched chains. Despite the importance, many robust multiscale continuum frameworks for modeling elastomer fracture tend to simplify network deformations by assuming uniform behavior among chains in all directions. Recognizing this limitation, our study proposes a multiscale fracture model that accounts for the anisotropic nature of elastomer network responses. At the microscale, damage in the chains is assumed to be driven by both the chain's entropy and the internal energy due to molecular bond distortions. In order to bridge the stretching in the chains to the macroscale deformation, we employ the maximal advance path constraint network model, inherently accommodating anisotropic network responses. As a result, chains oriented differently can be predicted to exhibit varying stretch and, consequently, different damage levels. To drive macroscale fracture based on damages in these chains, we utilize the micromorphic regularization theory, which involves the introduction of dual local-global damage variables at the macroscale. The macroscale local damage variable is obtained through the homogenization of the chain damage values, resulting in the prediction of an isotropic material response. The macroscale global damage variable is subjected to nonlocal effects and boundary conditions in a thermodynamically consistent phase field continuum formulation. Moreover, the total dissipation in the system is considered to be mainly due to the breaking of the molecular bonds at the microscale. To validate our model, we employ the double-edge notched tensile test as a benchmark, comparing simulation predictions with existing experimental data. Additionally, to enhance our understanding of the fracturing process, we conduct uniaxial tensile experiments on a square film made up of polydimethylsiloxane (PDMS) rubber embedded with a hole and notches and then compare our simulation predictions with the experimental observations. Furthermore, we visualize the evolution of stretch and damage values in chains oriented along different directions to assess the predictive capacity of the model. In conclusion, the results are also compared with another existing model to evaluate the utility of our model in accurately simulating the fracture behavior of rubber-like materials.

42 ENGINEERING↗

Deciphering the Solvation Structure of Aqueous ZnCl 2 Solutions from X-ray Absorption Spectra Using the Interpretable Graph Neural Network

Machine learning (ML) provides powerful pathways for predicting spectroscopic observables from atomic structures, but its broader impact depends on making model predictions interpretable in terms of physical and chemical principles. Here, we introduce a physics-guided graph neural network (GNN) model that predicts Zn K-edge X-ray spectroscopy (XAS) spectra of aqueous ZnCl 2 solutions. Training data are generated from ab initio XAS calculations on molecular dynamics snapshots obtained using a machine learning interatomic potential. The GNN reproduces experimental spectra across concentrations from dilute (<0.1 m) to highly concentrated (30 m, “water-in-salt”) regimes and scales efficiently to large, disordered liquid systems beyond the reach of conventional ab initio approaches. Gradient-based attribution analysis reveals that the model learns physically meaningful structure-spectrum relationships. Ligand-specific attributions reflect orbital hybridization patterns and the origin of the excitations derived from the density functional theory. Bond-length attributions recover spectral shifts consistent with multiple-scattering theory. Finally, this work bridges data-driven prediction with electronic-structure theory, establishing a general paradigm for interpretable ML that links atomic structure, electronic structure, and spectroscopic observables.

25 ENERGY STORAGE↗

Description of the Electronic Structure of Oxyhemoglobin Using Fe L-Edge X-ray Absorption Spectroscopy

The electronic structure of oxyhemoglobin has been controversial since the discovery of the compound’s diamagnetism in 1936. Here, this study uses partial fluorescence yield Fe L-edge X-ray absorption spectroscopy (XAS) in the 3s→2p fluorescence on oxyhemoglobin solutions, measured using a transition-edge sensor detector, to obtain a quantitative experimental description of the electronic structure of the O 2 -bound iron site. The spectrum is very different from typical low-spin Fe II and Fe III heme spectra, and multiplet simulations indicate a mixed ground configuration with ∼57% low-spin Fe III and ∼43% low-spin Fe II character. This is also very different from the Fe II character found for the picket-fence porphyrin model complex. The oxyhemoglobin L-edge XAS data further show that the O 2 ligand engages in a weak σ- but strong π-bond with the iron ion, leading to the overall strong Fe–O 2 bond required for O 2 transport.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Do graph neural networks learn traditional jet substructure?

At the CERN LHC, the task of jet tagging, whose goal is to infer the origin of a jet given a set of final-state particles, is dominated by machine learning methods. Graph neural networks have been used to address this task by treating jets as point clouds with underlying, learnable, edge connections between the particles inside. We explore the decision-making process for one such state-of-the-art network, ParticleNet, by looking for relevant edge connections identified using the layerwise-relevance propagation technique. As the model is trained, we observe changes in the distribution of relevant edges connecting different intermediate clusters of particles, known as subjets. The resulting distribution of subjet connections is different for signal jets originating from top quarks, whose subjets typically correspond to its three decay products, and background jets originating from lighter quarks and gluons. This behavior indicates that the model is using traditional jet substructure observables, such as the number of prongs -- energetic particle clusters -- within a jet, when identifying jets.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗