arXiv subjects

Thibault Lechien

Publications and source records attributed to Thibault Lechien.

9 recordsLinked to original sources

arXiv2026

You're Gonna Need a Bigger Core: Calibrating Massive Star Models against Galactic OB-type Stars

The evolution of massive stars above 8 M$_\odot$ depends critically on the amount of mixing above the convective core during the main sequence. However, current models typically extrapolate results from lower-mass stars, where constraints from asteroseismology and eclipsing binary systems are more readily available. A new opportunity to study the evolution of massive stars and their distribution in the Hertzsprung--Russell diagram arises by combining the IACOB spectroscopic sample of over 900 Galactic OB-type stars with Gaia distances. We use this homogeneously analyzed sample to place population-level constraints on main-sequence evolution. We analyze the data by forward modeling stellar evolution tracks with MESA and applying Bayesian inference. This enables us to 1) determine a physically-motivated, data-driven location of the terminal-age main sequence, 2) constrain convective boundary mixing and resulting core masses, and 3) provide a set of massive star models calibrated against modern data. We explore how boundary mixing depends on mass and find that it is well described by a constant overshooting parameter in the mass range of 12 to 40 M$_\odot$, with $\alpha_{\mathrm{ov}} = 0.33 \pm 0.02$, or $f_{\mathrm{ov}} = 0.028 \pm 0.003$ in the step and exponential overshooting schemes respectively. We find evidence against a continuation of the trend to increase with mass that is found at lower masses. Instead, the data does not exclude a decreasing trend at the high mass end. We find that the resulting helium core masses are 10 to 40% larger than other commonly used overshooting prescriptions. Combining our findings with existing observational constraints for low- and intermediate-mass stars, we propose a new mass-dependent overshooting prescription for a wide range of masses. Our calibration and model set are particularly useful for population and spectral synthesis applications.

astro-ph.SR↗

arXiv2026

Multipolar Magnetic-Field Inference for PSR J0740+6620 with Neural-Network-Accelerated NICER Pulse-Profile Modeling

We investigate the multipolar surface magnetic-field structure of the high-mass millisecond pulsar PSR J0740+6620 using the 32-bin bolometric NICER pulse profile of Dittmann et al. (2024). Building on the neural-network surrogate framework of Olmschenk et al. (2025), we model the emitting regions as open-field-line footpoints of an offset dipole plus axisymmetric quadrupole static vacuum field, rather than as prescribed geometric hotspots. We fix the stellar mass, radius, observer inclination, and hotspot temperature ratio to the Dittmann et al. (2024) maximum-likelihood values and explore the resulting 11-dimensional magnetic-field space. To make this feasible, we train convolutional neural-network surrogates on $5.12\times10^7$ synthetic bolometric light curves and use them in a parallel ensemble Markov Chain Monte Carlo calculation on 4000 CPU cores, accelerating likelihood evaluations by a factor of $\gtrsim 400$. We perform independent inferences for two calibrated temperature-weight prescriptions, Tw=1.31 and Tw=1.41, encoding the relative bolometric weight associated with the hotspot temperature difference. The posteriors, posterior-predictive light curves, and maximum-likelihood values are very similar, indicating weak sensitivity to this choice. The offset model reproduces the observed double-peaked profile and yields broad, multimodal posteriors, reflecting both the background-dominated data and degeneracies of the multipolar parameterization. The hotspot-density map shows that pulse phases constrain the approximate azimuthal placement of the emission, while latitude, surface extent, and morphology remain weakly constrained. A restricted zero offset run is disfavored within the adopted field basis. This work extends neural-network-accelerated magnetic-field inference to PSR J0740+6620 and motivates future energy-dependent, force-free, and joint X-ray/$\gamma$-ray extensions.

astro-ph.HE↗

arXiv2026

Disentangling the Galactic binary zoo: Machine learning classification of stellar remnant binaries in LISA data

The Laser Interferometer Space Antenna (LISA) will open a new observational window in the millihertz gravitational-wave band, enabling the detection of tens of thousands of compact stellar remnant binaries across the Milky Way. Most of LISA's sources will be double white dwarf (WDWD) systems, while neutron star-white dwarf (NSWD) binaries and higher-mass systems will be orders of magnitude rarer but of significant astrophysical interest. Disentangling these populations is challenging due to the strong overlap in their gravitational-wave features. In this work, we investigate the use of machine-learning techniques to classify LISA-detectable binaries based solely on LISA observables. Using mock catalogues of Galactic binaries constructed from population-synthesis studies, we evaluate a range of machine-learning classifiers. We find that ensemble-based methods-particularly gradient-boosting algorithms such as XGBoost-deliver the best performance on our highly imbalanced dataset. WDWD systems are identified with a recall of $\sim 99\%$, reflecting their dominant presence, and high-mass binaries are also classified with high recall ($\ge 85\%$). In contrast, NSWD systems remain the most challenging population to distinguish: their features overlap strongly with those of WDWD binaries, making them particularly prone to misclassification. Despite this, XGBoost correctly identifies 85.6% of NSWD systems in our simulated LISA detections, outperforming simple statistical approaches based on kernel density estimation. We further demonstrate that machine-learning classification can effectively support the interpretation of LISA data, enabling the identification of eccentric binaries and extremely rare subclasses.

astro-ph.HE↗

arXiv2025

Binary stars take what they get: Evidence for Efficient Mass Transfer from Stripped Stars with Rapidly Rotating Companions

Binary stars and their interactions shape the formation of compact binaries, supernovae, and gravitational wave sources. The efficiency of mass transfer - the fraction of mass retained by the accretor during binary interaction - is a critical parameter that significantly impacts the final fate of these systems. However, this parameter is observationally poorly constrained due to a scarcity of well-characterized post-mass-transfer binaries. Be+sdOB binaries, consisting of a rapidly rotating Be star and a stripped hot subdwarf companion, are particularly valuable for studying mass transfer since they represent clear examples of past binary interaction. Recently, a significantly expanded observational sample of 16 Be+sdOB binaries with well-constrained masses was obtained through combined spectroscopic and interferometric observations. In this work, we compile and analyze this sample to provide robust constraints on the mass transfer efficiency in binaries that underwent stable mass transfer during the donor's hydrogen-shell burning phase. Our analysis reveals that mass transfer was predominantly conservative: half of the systems require mass transfer efficiencies above 50%. This challenges commonly adopted assumptions of highly non-conservative mass transfer in binary evolution modeling. Our findings are inconsistent with models that account for spin-up and limit accretion due to a centrifugal barrier. We also find tension with a commonly used mass transfer model in rapid population synthesis that limits accretion based on the thermal timescale of the accretor. These results have strong implications for almost all products of binary evolution including the variety of supernovae, white dwarfs, blue stragglers, runaway stars, X-ray binaries, and gravitational-wave sources.

astro-ph.SR↗

arXiv2025

Pioneering High-Speed Pulsar Parameter Estimation Using Convolutional Neural Networks

Accurate thermal emission models of neutron stars are essential for constraining the dense matter equation of state. However, incorporating realistic magnetic field structures is computationally prohibitive, severely constraining feasible parameter space exploration. In this work, we develop a neural network (NN) emulator to generate model thermal bolometric X-ray light curves of millisecond pulsars with multipolar magnetic fields. We assess the NN's predictive and computational performance across a broad parameter space. We find that for a static vacuum field model, the NN provides a >400 times speedup. We integrate this NN emulator into a Monte Carlo Markov Chain (MCMC) framework to replace the computationally expensive physical model during parameter exploration. Applied to PSR J0030+0451, this approach allows the MCMC to reach equilibrium in ~1 day on 4000 cores, where with the original physical model alone it would have taken more than a year on the same hardware. We compare posterior distributions by running equivalent MCMC iterations with both the NN and the physical model, evaluate differences in distributions when continuing the physical model MCMC from the NN MCMC equilibrium state, and assess variations in posterior distributions resulting from NNs trained on datasets of different sizes. Our NN architecture is agnostic to the underlying physics of the physical model and can be trained for any other physical model, opening many previously intractable avenues of analysis. The NN speed remains the same regardless of the complexity of the physical model it was trained to emulate, allowing greater speedups for more complex physical models.

astro-ph.HE↗

arXiv2023

Automated Semiconductor Defect Inspection in Scanning Electron Microscope Images: a Systematic Review

A growing need exists for efficient and accurate methods for detecting defects in semiconductor materials and devices. These defects can have a detrimental impact on the efficiency of the manufacturing process, because they cause critical failures and wafer-yield limitations. As nodes and patterns get smaller, even high-resolution imaging techniques such as Scanning Electron Microscopy (SEM) produce noisy images due to operating close to sensitivity levels and due to varying physical properties of different underlayers or resist materials. This inherent noise is one of the main challenges for defect inspection. One promising approach is the use of machine learning algorithms, which can be trained to accurately classify and locate defects in semiconductor samples. Recently, convolutional neural networks have proved to be particularly useful in this regard. This systematic review provides a comprehensive overview of the state of automated semiconductor defect inspection on SEM images, including the most recent innovations and developments. 38 publications were selected on this topic, indexed in IEEE Xplore and SPIE databases. For each of these, the application, methodology, dataset, results, limitations and future work were summarized. A comprehensive overview and analysis of their methods is provided. Finally, promising avenues for future work in the field of SEM-based defect inspection are suggested.

cs.CV↗

arXiv2023

Dark Matter reconstruction from stellar orbits in the Galactic Centre

Context. Current constraints on distributed matter in the innermost Galactic Centre (such as a cluster of faint stars and stellar remnants, Dark Matter or a combination thereof) based on the orbital dynamics of the visible stars closest to the central black hole, typically assume simple functional forms for the distributions. Aims. We take instead a general model agnostic approach in which the form of the distribution is not constrained by prior assumptions on the physical composition of the matter. This approach yields unbiased - entirely observation driven - fits for the matter distribution and places constraints on our ability to discriminate between different density profiles (and consequently between physical compositions) of the distributed matter. Methods. We construct a spherical shell model with the flexibility to fit a wide variety of physically reasonable density profiles by modelling the distribution as a series of concentric mass shells. We test this approach in an analysis of mock observations of the star S2. Results. For a sufficiently large and precise data set, we find that it is possible to discriminate between several physically motivated density profiles. However, for data coming from current and expected next generation observational instruments, the potential for profile distinction will remain limited by the precision of the instruments. Future observations will still be able to constrain the overall enclosed distributed mass within the apocentre of the probing orbit in an unbiased manner. We interpret this in the theoretical context of constraining the secular versus non-secular orbital dynamics.

astro-ph.GA↗

arXiv2022

Neural network approach to reconstructing spectral functions and complex poles of confined particles

Reconstructing spectral functions from propagator data is difficult as solving the analytic continuation problem or applying an inverse integral transformation are ill-conditioned problems. Recent work has proposed using neural networks to solve this problem and has shown promising results, either matching or improving upon the performance of other methods. We generalize this approach by not only reconstructing spectral functions, but also (possible) pairs of complex poles or an infrared (IR) cutoff. We train our network on physically motivated toy functions, examine the reconstruction accuracy and check its robustness to noise. Encouraging results are found on both toy functions and genuine lattice QCD data for the gluon propagator, suggesting that this approach may lead to significant improvements over current state-of-the-art methods.

hep-lat↗

arXiv2022

Evolving test instances of the Hamiltonian completion problem

Predicting and comparing algorithm performance on graph instances is challenging for multiple reasons. First, there is usually no standard set of instances to benchmark performance. Second, using existing graph generators results in a restricted spectrum of difficulty and the resulting graphs are usually not diverse enough to draw sound conclusions. That is why recent work proposes a new methodology to generate a diverse set of instances by using an evolutionary algorithm. We can then analyze the resulting graphs and get key insights into which attributes are most related to algorithm performance. We can also fill observed gaps in the instance space in order to generate graphs with previously unseen combinations of features. This methodology is applied to the instance space of the Hamiltonian completion problem using two different solvers, namely the Concorde TSP Solver and a multi-start local search algorithm.

cs.AI↗