Kentaro Nagamine

We present the extension of the differentiable hydrodynamics code, diffhydro , enabling scalable partial differential equation-constrained inference and integrated hybrid physics–machine learning (ML) models for a wide range of astrophysical applications. New physics additions include radiative heating/cooling, Ornstein–Uhlenbeck–driven turbulence, and self-gravity via multigrid Poisson. We demonstrate good agreement with the Athena++ code on standard validation tests like Sedov–Taylor, Kelvin–Helmholtz, and driven/decaying turbulence. We further introduce a solver-in-the-loop neural corrector that reduces coarse-grid errors during time integration while preserving stability. The addition of custom adjoints facilitates efficient end-to-end gradients and multidevice scaling. We present simulations up to 1024 ³ elements, run on distributed GPU systems, and we show gradient-based reconstructions of complex initial conditions in turbulent, and self-gravitating, radiatively cooling flows. The code is written in JAX, and the solver’s modular finite-volume components are compiled by Accelerated Linear Algebra into fused accelerator kernels, delivering high-throughput forward runs and tractable differentiation through long integrations.

Context. Numerical efforts increasingly resolve the small-scale structure of the circumgalactic medium (CGM), but the dynamical impact of ionising radiation on its cold 10 ⁴ K component remains poorly understood. Aims. We investigated the evolution of static cold gas structures exposed to the extreme-ultraviolet (EUV) radiation of quasars. Methods. We developed an analytical framework to describe the evolution of such clouds, introducing a new threshold that defines when a cloud becomes radiation-shielded. The framework was validated using radiation-hydrodynamic simulations of single static clouds. Results. The framework predicts three evolutionary regimes: (i) an optically thin regime, in which radiation uniformly ionises the cloud; (ii) a radiation-shielded regime, where the cloud remains largely unaffected; and (iii) a “rocket-effect” regime, in which an ionisation front propagates through the cloud, ionising the illuminated side while compressing the opposite side and accelerating the surviving cold clump. In this latter regime, the cloud’s Lyman- α (Ly α ) luminosity can increase by up to one order of magnitude compared to the optically thin case. Such luminosities are as high as 70% of the values obtained from a fluorescent regime without considering hydrodynamical response. Unless the cloud is self-shielded, at least ∼50 − 60% of Ly α emission arises from recombination. Applying this framework to both a population of clouds along a line of sight and a ray propagating inside a single cold stream, we find that the cold inter-galactic medium (IGM) and CGM around bright quasars ( L _ν, LL ∼ 10 ^31.6 erg s ⁻¹ Hz ⁻¹ ) is likely fully ionised, whereas the one around faint quasars ( L _ν, LL ∼ 10 ^28.6 erg s ⁻¹ Hz ⁻¹ ) predominantly experiences a rocket-effect regime. Conclusions. These results imply that the hydrodynamical response of cold CGM structures to quasar radiation must be considered when deriving CGM and IGM physical properties from observations, particularly for faint quasars.

Quasars are among the most luminous objects. They are powered by accretion onto supermassive black holes. They are thought to impact cosmological evolution primarily through energetic winds, known as quasar-mode feedback, yet the efficiency and spatial extent of this process remain poorly constrained. Here we present X-Ray Imaging and Spectroscopy Mission (XRISM) observations of H1821+643—the nearest galaxy cluster with a central quasar (redshift z = 0.297)—which was a rare opportunity to directly probe quasar-mode feedback in the intracluster medium. High-resolution spectroscopy reveals exceptionally broadened Fe XXV emission lines from the intracluster medium, with a velocity dispersion of approximately 300 km s ⁻¹ , far exceeding values observed in nearby cluster cores. These lines originate predominantly at radii of 20–100 kpc from the centre. Assuming that turbulence from a quasar-driven shock led to the broadening of the lines, the energy injected by the quasar beyond galactic scales (≳20 kpc) is estimated to be ≳1–10% of its radiative energy. Notably, this feedback efficiency exceeds previous multiwavelength estimates by orders of magnitude (≲0.01%) and reaches the levels required by the latest cosmological hydrodynamical simulations. This finding of vigorous turbulence indicates that quasar-mode feedback plays a central role in regulating galaxy and cluster evolution at high redshift.

Fast radio bursts (FRBs) provide a sensitive probe of ionized baryons through their dispersion measure (DM). In addition to slowly evolving cosmological terms, at least three repeaters now show clear secular DM-decrease episodes, FRB 20190520B, FRB 20220529A, and FRB 20121102 (the last with a two-stage trend: early mild rise and late decline), supporting a dense, dynamically evolving local environment. We adopt a forward-modeling approach and use time-dependent 1D supernova remnant (SNR) simulations for a young magnetar embedded in SN ejecta, combining single-star (SS) and binary-stripped (BS) progenitors with hydrodynamic and nonequilibrium ionization calculations to follow shock structure, ionization, and electron density. The shocked region contributes only limited DM (≲10 pc cm ⁻³ ), while the dominant time-varying component is the unshocked ejecta, whose early behavior follows DM ∝ t ^{− α} with α ≃ 1.8–1.9. Although shocked-region DM is small, shock-amplified magnetic fields can still generate substantial rotation measure (RM); in our shock-only RM framework, only the 11 M _⊙ SS model reproduces the FRB 20121102 RM evolution. BS progenitors generally yield smaller DM than SS models at fixed M _ZAMS , with composition-dependent mean molecular weights introducing nonmonotonic mass trends. Matching the observed d DM/ d t of FRB 20190520B, FRB 20220529A, and the late-stage slope of FRB 20121102, we infer local SNR DM contributions of ∼10 to a few × 10 ² pc cm ⁻³ . We also find that GHz escape is allowed in most models, with τ _ff = 1 typically reached by t _esc ≲ 70 yr; for weakly ionized ejecta, the source can be nearly transparent from very early times. These results support a young core-collapse supernova/SNR origin for a substantial fraction of DM _source and highlight that physically consistent local-environment modeling is essential for robust FRB cosmological DM inferences.

We present a new constraint on the primordial helium abundance, Y _P , based on Subaru observations. A major source of uncertainty in previous Y _P determinations is the lack of extremely metal-poor galaxies (EMPGs; 0.01–0.1 Z _⊙ ), which have metallicities a few to ten times lower than the metal-poor galaxies (MPGs; 0.1–0.4 Z _⊙ ) predominantly used in earlier studies, requiring substantial extrapolation to zero metallicity. Here, we perform Subaru near-infrared spectroscopy of 29 galaxies, including 14 EMPGs. By incorporating existing optical spectra, we derive He/H for each galaxy using photoionization modeling of helium and hydrogen emission lines, including the He i 10830 Å line to break the density–temperature degeneracy. After carefully selecting galaxies with robust He/H determinations, and adding 58 galaxies from previous studies, we obtain Y P = 0.240 2 − 0.0040 + 0.0040 . This Y _P value is ∼1 σ lower than most of the previous estimates, but agrees with recent determinations using EMPGs and the cosmic microwave background constraint from the Atacama Cosmology Telescope experiment. Our result indicates N eff = 2.5 4 − 0.25 + 0.20 , showing a mild (∼2 σ ) tension with the Standard Model and Planck results. These tensions may suggest a nonzero lepton asymmetry ( ξ _e ≠ 0), which would alleviate the tension with ξ e = 0.0 5 − 0.03 + 0.02 . More observations of EMPGs and further assessments of systematic uncertainties are essential to test the potential tension more rigorously.

Extragalactic fast radio bursts (FRBs) have emerged as powerful probes of turbulence within the intergalactic medium (IGM), a phenomenon that plays a crucial role in various cosmological and astrophysical processes. In this study, we employ the structure function analysis on the dispersion measures of over 3000 FRBs, leveraging the recently released CHIME/FRB Catalog 2 alongside previously observed sources. By comparing our results with mock datasets generated from cosmological simulations, we find excellent agreement at large angular separations. At small angular scales, our findings reveal a potential scaling behavior consistent with a two-dimensional Kolmogorov power spectrum. From this scaling, we constrain the turbulence outer scale to be on the order of several Mpc, which aligns with theoretical expectations, independent observations of the low-redshift IGM, and cosmological simulations. Ultimately, to conclusively confirm this Kolmogorov-like turbulent cascade and overcome current small-sample statistical limitations, a larger sample of FRBs with subarcsecond localization is required.

Active galactic nuclei (AGN) jets are powerful drivers of galaxy evolution, depositing energy and momentum into the circumgalactic and intracluster medium (CGM/ICM) and regulating gas cooling and star formation. We investigate the dynamics of jet evolution in the self-similar regime using the smoothed particle hydrodynamics (SPH) code GADGET4-Osaka, systematically vary jet-launching schemes, artificial-viscosity prescriptions, mass resolution, and jet lifetimes and compare the results with grid-based simulation. Our analysis combines quantitative diagnostics of jet size and energetics with detailed morphological and thermodynamic characterizations from slice maps and phase diagrams. We find that jet lobe growth follows analytic self-similar scaling relations and converges with resolution, but is highly sensitive to the choice of artificial viscosity. While the overall jet size tracks self-similar predictions, the partitioning of thermal and kinetic energy departs significantly from the idealized picture, reflecting enhanced dissipation and mixing, which is consistent with the jet propagation in grid-based simulations. These results establish robust benchmarks for SPH-based jet modeling, provide insight into the physical and numerical factors shaping jet–medium interactions, and lay the groundwork for future studies of AGN feedback in realistic galactic and cluster environments.

We present results from CROCODILE-DWARF, a new suite of cosmological zoom-in hydrodynamic simulations of isolated field dwarf galaxies with halo masses of ∼10 ¹⁰ M _⊙ at z = 0, performed with the gadget 4 -osaka code. The simulations include detailed modeling of star formation, chemical enrichment, and supernova feedback using the CELib and grackle libraries, achieving baryonic resolutions of ∼2 × 10 ³ M _⊙ . Our study focuses on how assembly history governs the structural and kinematic diversity of dwarf galaxies within the ΛCDM framework. The simulated galaxies reproduce the observed stellar-to-halo mass, mass–metallicity, and size–mass relations for nearby dwarf galaxies, including those of the Local Group, yielding stellar masses of ∼10 ⁷ M _⊙ . The galaxies display a broad range of rotational support, where gas is generally more rotationally supported than stars. Differences in morphology and kinematics primarily reflect variations in halo assembly timescales and merger activity. Early-assembling, high-concentration halos form stars efficiently and become gas-poor by z = 0, while late-assembling, low-concentration halos remain gas-rich due to delayed star formation and rejuvenated gas accretion. We find a trend between rotational support and the cumulative merger mass fraction, providing tentative evidence that dynamical heating induced by mergers plays a role in shaping the kinematic diversity. In some cases, late-time mergers induce the formation of extended gas disks by delivering fresh gas and angular momentum. These results demonstrate that it is assembly history, rather than halo mass alone, that shapes the present-day kinematic and morphological diversity of dwarf galaxies.

We present the first z = 0 H I column density distribution function, f ( N _HI ), extending down to log ( N HI / cm − 2 ) = 17.8 . This was derived from high-sensitivity 21 cm emission-line imaging at ∼1 kpc resolution. At high column densities (19.8 < log ( N HI / cm − 2 ) < 21.3), our results align with earlier z = 0 studies but benefit from 100 times greater sensitivity. Comparisons with z ∼ 3 quasar absorption-line studies reveal that f ( N _HI ) at z = 0 is systematically lower by 0.1–0.4 dex for 19.2 < log ( N HI / cm − 2 ) < 21 . However, the distributions become comparable at 17.8 < log ( N HI / cm − 2 ) < 19.2 , suggesting weak evolution in this regime. Extrapolating the length incidence ( d N / d X ) for log ( N HI / cm − 2 ) > 17.5 implies a covering fraction ( f _cov ) of ∼0.7 within 1 kpc-scale H I -detected pixels at z = 0. Notably, for 17.8 < log ( N HI / cm − 2 ) < 20 , impact parameters at a given N _HI are significantly lower than previous z ∼ 0 absorption-line results and TNG50 simulation predictions. This discrepancy indicates challenges in identifying galaxy counterparts for absorbers and in recovering low–column density H I within cosmological simulations. Finally, we derive a covering fraction of 0.006 for log ( N HI / cm − 2 ) > 17.8 gas within the virial radius around Milky Way–like galaxies. These findings provide new constraints on the baryonic flows and gaseous dynamics governing galaxy evolution.

Recent observations from the James Webb Space Telescope have revealed unexpectedly luminous galaxies, exhibiting stellar masses and luminosities significantly higher than predicted by theoretical models at Cosmic Dawn. In this study, we present a suite of cosmological zoomed-in simulations targeting high-redshift ( z ≥ 10) galaxies with dark matter halo masses in the range 10 ¹⁰ –10 ¹¹ M _⊙ at z = 10, using state-of-the-art galaxy formation simulation codes ( Enzo , Ramses , Changa , Gadget-3 , Gadget-4 , and Gizmo ). This study aims to evaluate the convergence of the participating codes and their reproducibility of high-redshift galaxies with the galaxy formation model calibrated at relatively low redshift, without additional physics for high-redshift environments. The subgrid physics follows the AGORA CosmoRun framework, with adjustments to resolution and initial conditions to emulate similar physical environments in the early Universe. The participating codes show consistent results for key galaxy properties (e.g., stellar mass), but also reveal notable differences (e.g., metallicity), indicating that galaxy properties at high redshifts are highly sensitive to the feedback implementation of the simulation. Massive halos ( M _halo ≥5 × 10 ¹⁰ M _⊙ at z = 10) succeed in reproducing observed stellar masses, metallicities, and UV luminosities at 10 ≤ z ≤ 12 without requiring additional subgrid physics, but tend to underpredict those properties at higher redshift. We also find that varying the dust-to-metal ratio modestly affects UV luminosity of simulated galaxies, whereas the absence of dust significantly enhances it. In future work, higher-resolution simulations will be conducted to better understand the formation and evolution of galaxies at Cosmic Dawn.