arXiv subjects

Jun-Young Lee

Publications and source records attributed to Jun-Young Lee.

14 recordsLinked to original sources

arXiv2026

RAMSES-GPU: Cell-by-Cell Adaptive Mesh Refinement with Magneto-Hydrodynamics and Self-Gravity on Graphics Processing Units

We present the implementation and optimization of the cosmological simulation code RAMSES on Graphics Processing Units (GPUs) using CUDA Fortran. This accelerated version ports the main computational routines, including hydrodynamics, particle dynamics, and self-gravity, to multi-GPU architectures. We detail our strategy for managing cell-by-cell Adaptive Mesh Refinement (AMR) on the GPU, utilizing bucket sort with prefix sums for AMR level sorting, radix sort via the CUB library for Hilbert key ordering, and an fnv64 hash table with linear probing for fast spatial indexing. Portability across diverse hardware architectures is achieved via a dispatcher and C-Fortran wrappers, calling CUDA, HIP, and Metal kernels directly translated from the CUDA Fortran framework. Hydrodynamics updates are executed via a Godunov MUSCL-Hancock HLLC Riemann solver managed through a three-tier shared-memory kernel architecture (named rock, paper, and scissor). Particle mass deposition uses Cloud-in-Cell (CIC) interpolation optimized with atomic additions or prefix sums, combined with a kick-drift-kick time integration pusher. Self-gravity is handled via a Multigrid (MG) Poisson solver performing hierarchical V-cycles on individual levels. Performance benchmarks conducted on NVIDIA A100 and H200 GPUs demonstrate substantial accelerations compared to multi-core CPUs, yielding 10x up to a 100x speedup for standard test problems such as the Sedov blast wave, molecular core collapse, and cosmological simulations. Finally, we briefly discuss additional accelerated physics modules, including equilibrium cooling, polytropic equations of state, ideal and non-ideal magneto-hydrodynamics (MHD), and stellar feedback.

astro-ph.IM↗

arXiv2026

A Scalable Fast Multipole Method Poisson Solver for the RAMSES code: II. Adaptive Mesh Refinement and Adaptive Time Stepping

We present an extended implementation of a scalable, O(N) Poisson solver based on the fast multipole method (FMM), fully compatible with adaptive mesh refinement (AMR) and adaptive time stepping (ATS) within the RAMSES framework. Building on the unigrid algorithm in Lee & Teyssier (2026), we introduce several novel elements, including the use of multiple FMM trees, one for each AMR level, a merged FMM tree for all active levels to optimize neighbor searches, and the introduction of the concept of "nearest-field" to enforce force symmetry across refinement levels. Across a broad set of test problems, we find excellent, percent-level agreement with our reference traditional multigrid (MG) solver. We show, however, that FMM exhibits better momentum conservation properties across coarse-fine AMR interfaces. Finally, despite the overhead introduced by the spatio-temporal adaptivity, FMM shows better scalability than MG across various AMR configurations, with the largest gains obtained for the largest configurations.

astro-ph.IM↗

arXiv2026

A Scalable Fast Multipole Method Poisson Solver for the RAMSES code: I. Unigrid Algorithm

We present a scalable Poisson solver with $O(N)$ complexity based on the fast multipole method (FMM) implemented in RAMSES. Our FMM constructs a hierarchy of FMM grids on top of the pre-existing Cartesian grid which is used to compute the force for hydrodynamics or particle-mesh simulations. In contrast to the $O(N)$ multigrid solver (MG) - an iterative method that requires multiple V-cycles through a multi-resolution hierarchy of Cartesian grids - the FMM algorithm performs just one upward pass through the same hierarchy, during which multipole expansions are accumulated and shifted, followed by a single downward pass, in which local expansions are propagated. Numerical tests indicate that FMM attains accuracy comparable to that of MG for smooth potentials and is particularly well-suited for problems with isolated boundary conditions, since it avoids the approximate Dirichlet boundary conditions required by MG schemes. Although in theory FMM requires around 30 times more floating-point operations than MG, its higher arithmetic intensity leads to comparable performance and better scalability relative to MG.

astro-ph.IM↗

arXiv2026

Learning the Universe with the 2nd Generation of CAMELS: Varying 35 parameters of the IllustrisTNG model in (50Mpc/h)^3 boxes

We present a new set of 1,192 cosmological simulations as part of the CAMELS project, in which a space of 35 cosmological, astrophysical, and numerical parameters is explored around the fiducial IllustrisTNG model. The volume of each of these simulations is (50Mpc/h)^3, eight times larger than that of previous CAMELS simulations. This provides lower sample variance as well as access to more massive halos and more diverse environments. We focus this work on exploring the advantages these differences provide for parameter inference powered by neural networks. We generate training sets based on the matter power spectra, projected maps of the volumes, graphs representing galaxy spatial distributions, and thermodynamical properties of massive halos. We employ multilayer perceptrons, convolutional neural networks, graph neural networks, and Gaussian processes, respectively, to extract information on the simulation parameters from these inputs while comparing systematically to analogous results from our previous generation of (25Mpc/h)^3 simulations. We generally find that the new, larger volumes produce tighter marginal constraints on the parameters, to degrees that vary between the different inputs. The improvements, however, scale more weakly than with the square root of the increase in the amount of data (i.e., physical volume). We interpret this as originating either from information loss due to mode coupling or from complex degeneracies in parameter space. We also discuss the effects on statistics of the intergalactic medium temperature from four new parameters that are varied in these simulations, which control the amplitude and timing of the ionizing background radiation. We publicly release the simulation outputs and ancillary data at https://camels.readthedocs.io.

astro-ph.CO↗

arXiv2025

The Denario project: Deep knowledge AI agents for scientific discovery

We present Denario, an AI multi-agent system designed to serve as a scientific research assistant. Denario can perform many different tasks, such as generating ideas, checking the literature, developing research plans, writing and executing code, making plots, and drafting and reviewing a scientific paper. The system has a modular architecture, allowing it to handle specific tasks, such as generating an idea, or carrying out end-to-end scientific analysis using Cmbagent as a deep-research backend. In this work, we describe in detail Denario and its modules, and illustrate its capabilities by presenting multiple AI-generated papers generated by it in many different scientific disciplines such as astrophysics, biology, biophysics, biomedical informatics, chemistry, material science, mathematical physics, medicine, neuroscience and planetary science. Denario also excels at combining ideas from different disciplines, and we illustrate this by showing a paper that applies methods from quantum physics and machine learning to astrophysical data. We report the evaluations performed on these papers by domain experts, who provided both numerical scores and review-like feedback. We then highlight the strengths, weaknesses, and limitations of the current system. Finally, we discuss the ethical implications of AI-driven research and reflect on how such technology relates to the philosophy of science. We publicly release the code at https://github.com/AstroPilot-AI/Denario. A Denario demo can also be run directly on the web at https://huggingface.co/spaces/astropilot-ai/Denario, and the full app will be deployed on the cloud.

cs.AI↗

arXiv2025

CosmoBench: A Multiscale, Multiview, Multitask Cosmology Benchmark for Geometric Deep Learning

Cosmological simulations provide a wealth of data in the form of point clouds and directed trees. A crucial goal is to extract insights from this data that shed light on the nature and composition of the Universe. In this paper we introduce CosmoBench, a benchmark dataset curated from state-of-the-art cosmological simulations whose runs required more than 41 million core-hours and generated over two petabytes of data. CosmoBench is the largest dataset of its kind: it contains 34 thousand point clouds from simulations of dark matter halos and galaxies at three different length scales, as well as 25 thousand directed trees that record the formation history of halos on two different time scales. The data in CosmoBench can be used for multiple tasks -- to predict cosmological parameters from point clouds and merger trees, to predict the velocities of individual halos and galaxies from their collective positions, and to reconstruct merger trees on finer time scales from those on coarser time scales. We provide several baselines on these tasks, some based on established approaches from cosmological modeling and others rooted in machine learning. For the latter, we study different approaches -- from simple linear models that are minimally constrained by symmetries to much larger and more computationally-demanding models in deep learning, such as graph neural networks. We find that least-squares fits with a handful of invariant features sometimes outperform deep architectures with many more parameters and far longer training time. Still there remains tremendous potential to improve these baselines by combining machine learning and cosmology to fully exploit the data. CosmoBench sets the stage for bridging cosmology and geometric deep learning at scale. We invite the community to push the frontier of scientific discovery by engaging with this dataset, available at https://cosmobench.streamlit.app

cs.LG↗

arXiv2025

Cosmology with Topological Deep Learning

The standard cosmological model with cold dark matter posits a hierarchical formation of structures. We introduce topological neural networks (TNNs), implemented as message-passing neural networks on higher-order structures, to effectively capture the topological information inherent in these hierarchies that traditional graph neural networks (GNNs) fail to account for. Our approach not only considers the vertices and edges that comprise a graph but also extends to higher-order cells such as tetrahedra, clusters, and hyperedges. This enables message-passing between these heterogeneous structures within a combinatorial complex. Furthermore, our TNNs are designed to conserve the $E(3)$-invariance, which refers to the symmetry arising from invariance against translations, reflections, and rotations. When applied to the Quijote suite, our TNNs achieve a significant reduction in the mean squared error. Compared to our GNNs, which lack higher-order message-passing, ClusterTNNs show improvements of up to 22% in $Ω_{\rm m}$ and 34% in $σ_8$ jointly, while the best FullTNN achieves an improvement of up to 60% in $σ_8$. In the context of the CAMELS suite, our models yield results comparable to the current GNN benchmark, albeit with a slight decrease in performance. We emphasize that our topology and symmetry-aware neural networks provide enhanced expressive power in modeling the large-scale structures of our universe.

astro-ph.CO↗

arXiv2024

Inferring Cosmological Parameters on SDSS via Domain-Generalized Neural Networks and Lightcone Simulations

We present a proof-of-concept simulation-based inference on $Ω_{\rm m}$ and $σ_{8}$ from the SDSS BOSS LOWZ NGC catalog using neural networks and domain generalization techniques without the need of summary statistics. Using rapid lightcone simulations, ${\rm L{\scriptsize -PICOLA}}$, mock galaxy catalogs are produced that fully incorporate the observational effects. The collection of galaxies is fed as input to a point cloud-based network, ${\texttt{Minkowski-PointNet}}$. We also add relatively more accurate ${\rm G{\scriptsize ADGET}}$ mocks to obtain robust and generalizable neural networks. By explicitly learning the representations which reduces the discrepancies between the two different datasets via the semantic alignment loss term, we show that the latent space configuration aligns into a single plane in which the two cosmological parameters form clear axes. Consequently, during inference, the SDSS BOSS LOWZ NGC catalog maps onto the plane, demonstrating effective generalization and improving prediction accuracy compared to non-generalized models. Results from the ensemble of 25 independently trained machines find $Ω_{\rm m}=0.339 \pm 0.056$ and $σ_{8}=0.801 \pm 0.061$, inferred only from the distribution of galaxies in the lightcone slices without relying on any indirect summary statistics. A single machine that best adapts to the ${\rm G{\scriptsize ADGET}}$ mocks yields a tighter prediction of $Ω_{\rm m}=0.282 \pm 0.014$ and $σ_{8}=0.786 \pm 0.036$. We emphasize that adaptation across multiple domains can enhance the robustness of the neural networks in observational data.

astro-ph.CO↗

arXiv2009

Magnetic domain-wall motion by propagating spin waves

We found by micromagnetic simulations that the motion of a transverse wall (TW) type domain wall in magnetic thin-film nanostripes can be manipulated via interaction with spin waves (SWs) propagating through the TW. The velocity of the TW motion can be controlled by changes of the frequency and amplitude of the propagating SWs. Moreover, the TW motion is efficiently driven by specific SW frequencies that coincide with the resonant frequencies of the local modes existing inside the TW structure. The use of propagating SWs, whose frequencies are tuned to those of the intrinsic TW modes, is an alternative approach for controlling TW motion in nanostripes.

cond-mat.mtrl-sci↗

arXiv2008

Criterion for transformation of transverse domain wall to vortex or antivortex wall in soft magnetic thin-film nanostripes

We report on the criterion for the dynamic transformation of the internal structure of moving domain walls (DWs) in soft magnetic thin-film nanostripes above the Walker threshold field, Hw. In order for the process of transformation from transverse wall (TW) to vortex wall (VW) or antivortex wall (AVW) occurs, the edge-soliton core of the TW-type DW should grow sufficiently to the full width at half maximum of the out-of-plane magnetizations of the core area of the stabilized vortex (or antivortex) by moving inward along the transverse (width) direction. Upon completion of the nucleation of the vortex (antivortex) core, the VW (AVW) is stabilized, and then its core accompanies the gyrotropic motion in a potential well (hill) of a given nanostripe. Field strengths exceeding the Hw, which is the onset field of DW velocity breakdown, are not sufficient but necessary conditions for dynamic DW transformation.

cond-mat.mtrl-sci↗

arXiv2008

Origin of the increased velocities of domain wall motions in soft magnetic thin-film nanostripes beyond the velocity-breakdown regime

It is known that oscillatory domain-wall (DW) motions in soft magnetic thin-film nanostripes above the Walker critical field lead to a remarkable reduction in the average DW velocities. In a much-higher-field region beyond the velocity-breakdown regime, however, the DW velocities have been found to increase in response to a further increase of the applied field. We report on the physical origin and detailed mechanism of this unexpected behavior. We associate the mechanism with the serial dynamic processes of the nucleation of vortex-antivortex (V-AV) pairs inside the stripe or at its edges, the non-linear gyrotropic motions of Vs and AVs, and their annihilation process. The present results imply that a two-dimensional soliton model is required for adequate interpretation of DW motions in the linear- and oscillatory-DW-motion regimes as well as in the beyond-velocity-breakdown regime.

cond-mat.mtrl-sci↗

arXiv2007

Dynamics of magnetic topological solitons in soft magnetic nanostripes

The motions of domain walls driven by magnetic field in soft magnetic nanostripes were calculated. The domain walls reveal steady motions in the low fields and oscillations of their internal structure above a critical field. A developed soliton model of the walls explains its dynamics by the motions of a limited number of magnetic topological solitons such as vortex and antivortex. We predict the reduced wall velocity and critical field in the low-field regime, and increased wall-oscillation frequency in nanostripes, compared to 1D Walker solution for bulk magnets. The critical field and velocity are determined by the nanostripe sizes, whereas the oscillation frequency depends only on the field strength.

nlin.PS↗

arXiv2007

Understanding of complex periodic transformations of moving domain walls in magnetic nanostripes

The magnetic field (or electric current) driven domain-wall motion in magnetic nanostripes is of considerable interest because it is essential to the performance of information storage and logic devices. One of the currently key problems is to unveil the complex behaviors of oscillatory domain-wall motions under applied magnetic fields stronger than the so-called Walker field, beyond which the velocity of domain walls markedly drops. Here, we provide not only considerably better understandings but also new details of the complex domain-wall motions. In a certain range just above the Walker field, the motions are not chaotic but rather periodic with different unique periodicities of dynamic transformations of a moving domain wall between the different types of its internal structure. Three unique periodicities found, which consist of different types of domain wall that are transformed from type one to another. The transformation periods vary with the field strength and the nanostripe width. This novel phenomenon can be described by the dynamic motion of a limited number of magnetic topological solitons such as vortex and antivortex in nanostripes.

cond-mat.other↗

arXiv2007

Ultrafast Vortex-Core Reversal Dynamics in Ferromagnetic Nanodots

To verify the exact underlying mechanism of ultrafast vortex-core reversal as well as the vortex state stability we conducted numerical calculations of the dynamic evolution of magnetic vortices in Permalloy cylindrical nanodots under an oscillating in-plane magnetic field over a wide range of the frequency and amplitude. The calculated results reveal different kinds of the non-trivial dynamic responses of vortices to the driving external field. In particular, the results offer insight into the 10 ps scale underlying physics of the ulrafast vortex-core reversals driven by small amplitude (~10 Oe) in-plane fields. This work also provides fundamentals of how to manipulate effectively the dynamical switching of the vortex-core orientation.

cond-mat.mtrl-sci↗