arXiv subjects

Christine M. Simpson

Publications and source records attributed to Christine M. Simpson.

At least 19 recordsLinked to original sources

arXiv2026

Overcoming Orchestration Bottlenecks at Exascale: A Decentralized, Policy-Driven Approach for Sim-AI Ensembles

Scientific computing is increasingly shifting from monolithic applications to coupled simulation-AI workflows composed of highly heterogeneous tasks with diverse hardware, scale, and runtime requirements. As these workflows scale to leadership-class systems, the resulting extreme ensemble sizes and task variability can create orchestration bottlenecks. System-level schedulers are often configured for limited throughput, while workflow tools face scalability issues due to rigid control-plane topologies and static scheduling heuristics. We introduce EnsembleLauncher, a recursively hierarchical workflow orchestrator for exascale systems, featuring a fully decentralized control plane and a programmable scheduling policy interface. On the Aurora supercomputer, EnsembleLauncher successfully scales to the entire machine with up to eight million serial tasks, outperforming state-of-the-art tools by more than four times. Additionally, we implement a programmable scheduling interface and demonstrate a significant impact of scheduling policies on resource utilization for high-variance ensembles and active learning pipelines representative of modern coupled simulation-AI workflows.

cs.DC↗

arXiv2025

In-Transit Data Transport Strategies for Coupled AI-Simulation Workflow Patterns

Coupled AI-Simulation workflows are becoming the major workloads for HPC facilities, and their increasing complexity necessitates new tools for performance analysis and prototyping of new in-situ workflows. We present SimAI-Bench, a tool designed to both prototype and evaluate these coupled workflows. In this paper, we use SimAI-Bench to benchmark the data transport performance of two common patterns on the Aurora supercomputer: a one-to-one workflow with co-located simulation and AI training instances, and a many-to-one workflow where a single AI model is trained from an ensemble of simulations. For the one-to-one pattern, our analysis shows that node-local and DragonHPC data staging strategies provide excellent performance compared Redis and Lustre file system. For the many-to-one pattern, we find that data transport becomes a dominant bottleneck as the ensemble size grows. Our evaluation reveals that file system is the optimal solution among the tested strategies for the many-to-one pattern.

cs.DC↗

arXiv2025

Auriga Streams III: the mass-metallicity relation does not rule out tidal mass loss in Local Group satellites

The mass-metallicity relation is a fundamental galaxy scaling law that has been extended to the faintest systems in the Local Group. We show that the small scatter in this relation, which has been used to argue against tidal mass loss in Local Group satellites, is consistent with the level of disruption in the Auriga simulations. For every accreted system in Auriga, we compute stellar masses and metallicities two ways: considering the total system (bound + lost material) and only considering the progenitor. Accreted systems in Auriga have a tight relation between total stellar mass and metallicity, with scatter at a fixed stellar mass driven by age. When only considering the progenitor, the tidally evolved mass-metallicity relation has similar scatter ($\sim$0.27 dex) as observed for the Local Group satellites ($\sim$0.23 dex). Satellites that lie above the relation have experienced substantial mass loss and typically have low metallicity for their total stellar mass. Even satellites that fall exactly on the evolved relation can lose over half of their stellar mass. Only satellites substantially below the evolved relation are reliably intact. Based on their offset from the observed relation, we predict which Milky Way and M31 satellites have tidal tails waiting to be discovered.

astro-ph.GA↗

arXiv2025

Auriga Superstars: Improving the resolution and fidelity of stellar dynamics in cosmological galaxy simulations

Cosmological hydrodynamical simulations have become an indispensable tool to understand galaxies. However, computational constraints still severely limit their numerical resolution. This not only restricts the sampling of the stellar component and its direct comparison to detailed observations, but also the precision with which it is evolved. To overcome these problems we introduce the \emph{Superstars} method. This method increases the stellar mass resolution in cosmological galaxy simulations in a computationally inexpensive way for a fixed dark matter and gas resolution without altering any global properties of the simulated galaxies. We demonstrate the \emph{Superstars} method for a Milky Way-like galaxy of the Auriga project, improving the stellar mass resolution by factors of $8$ and $64$ at an additional cost of only $10\%$ and $500\%$, respectively. We show and quantify that this improves the sampling of the stellar population in the disc and halo without changing the properties of the central galaxy or its satellites, unlike simulations that change the resolution of all components (gas, dark matter, stars). Moreover, the better stellar mass resolution reduces numerical heating of the stellar disc in its outskirts and keeps substructures in the stellar disc and inner halo more coherent. It also makes lower mass and lower surface brightness structures in the stellar halo more visible. The \emph{Superstars} method is straightforward to incorporate in any cosmological galaxy simulation that does not resolve individual stars.

astro-ph.GA↗

arXiv2025

Quantifying the intrinsic variability due to randomness of the Auriga galaxy formation model

Numerical simulations have become an indispensable tool in astrophysics. To interpret their results, it is critical to understand their intrinsic variability, that is, how much the results change with numerical noise or inherent stochasticity of the physics model. We present a set of seven realisations of high-resolution cosmological zoom-in simulations of a Milky Way-like galaxy with the Auriga galaxy formation model. All realisations share the same initial conditions and code parameters, but draw different random numbers for the inherently stochastic parts of the model. We show that global galaxy properties at $z=0$, including stellar mass, star formation history, masses of stellar bulge and stellar disc, the radius and height of the stellar disk change by less than $10\%$ between the different realisations, and that magnetic field structures in the disc and the halo are very similar. In contrast, the star formation rate today can vary by a factor of two and the internal morphological structure of the stellar disc can change. The time and orbit of satellite galaxies and their galaxy properties when falling into the main halo are again very similar, but their orbits start to deviate after first pericenter passage. Finally, we show that changing the mass resolution of all matter components by a factor of $8$ in the Auriga model changes galaxy properties significantly more than the intrinsic variability of the model, and that these changes are systematic. This limits detailed comparisons between simulations at different numerical resolutions.

astro-ph.GA↗

arXiv2024

Auriga Streams II: orbital properties of tidally disrupting satellites of Milky Way-mass galaxies

Galaxies like the Milky Way are surrounded by complex populations of satellites at all stages of tidal disruption. In this paper, we present a dynamical study of the disrupting satellite galaxies in the Auriga simulations that are orbiting 28 distinct Milky Way-mass hosts across three resolutions. We find that the satellite galaxy populations are highly disrupted. The majority of satellites that remain fully intact at present day were accreted recently without experiencing more than one pericentre ($n_{\rm peri} \lesssim 1$) and have large apocentres ($r_{\rm apo} \gtrsim 200$ kpc) and pericentres ($r_{\rm peri} \gtrsim 50$ kpc). The remaining satellites have experienced significant tidal disruption and, given full knowledge of the system, would be classified as stellar streams. We find stellar streams in Auriga across the range of pericentres and apocentres of the known Milky Way dwarf galaxy streams and, interestingly, overlapping significantly with the Milky Way intact satellite population. We find no significant change in satellite orbital distributions across resolution. However, we do see substantial halo-to-halo variance of $(r_\text{peri}, r_\text{apo})$ distributions across host galaxies, as well as a dependence of satellite orbits on host halo mass - systems disrupt at larger pericentres and apocentres in more massive hosts. Our results suggest that either cosmological simulations (including, but not limited to, Auriga) are disrupting satellites far too readily, or that the Milky Way's satellites are more disrupted than current imaging surveys have revealed. Future observing facilities and careful mock observations of these systems will be key to revealing the nature of this apparent discrepancy.

astro-ph.GA↗

arXiv2024

Auriga Streams I: disrupting satellites surrounding Milky Way-mass haloes at multiple resolutions

In a hierarchically formed Universe, galaxies accrete smaller systems that tidally disrupt as they evolve in the host's potential. We present a complete catalogue of disrupting galaxies accreted onto Milky Way-mass haloes from the Auriga suite of cosmological magnetohydrodynamic zoom-in simulations. We classify accretion events as intact satellites, stellar streams, or phase-mixed systems based on automated criteria calibrated to a visually classified sample, and match accretions to their counterparts in haloes re-simulated at higher resolution. Most satellites with a bound progenitor at the present day have lost substantial amounts of stellar mass -- 67 per cent have $f_\text{bound} < 0.97$ (our threshold of lost stellar mass to no longer be considered intact), while 53 per cent satisfy a more stringent $f_\text{bound} < 0.8$. Streams typically outnumber intact systems, contribute a smaller fraction of overall accreted stars, and are substantial contributors at intermediate distances from the host centre ($\sim$0.1 to $\sim$0.7$R_\text{200m}$, or $\sim$35 to $\sim$250 kpc for the Milky Way). We also identify accretion events that disrupt to form streams around massive intact satellites instead of the main host. Streams are more likely than intact or phase-mixed systems to have experienced preprocessing, suggesting this mechanism is important for setting disruption rates around Milky Way-mass haloes. All of these results are preserved across different simulation resolutions, though we do find some hints that satellites disrupt more readily at lower resolution. The Auriga haloes suggest that disrupting satellites surrounding Milky Way-mass galaxies are the norm and that a wealth of tidal features waits to be uncovered in upcoming surveys.

astro-ph.GA↗

arXiv2023

An ever-present $Gaia$ snail shell triggered by a dark matter wake

We utilize a novel numerical technique to model star formation in cosmological simulations of galaxy formation - called Superstars - to simulate a Milky Way-like galaxy with $\gtrsim10^8$ star particles to study the formation and evolution of out-of-equilibrium stellar disc structures in a full cosmological setting. In the plane defined by the coordinate and velocity perpendicular to the mid-plane (vertical phase space, $\{Z,V_Z\}$), stars in Solar-like volumes at late times exhibit clear spirals qualitatively similar in shape and amplitude to the $Gaia$ ``Snail shell'' phase spiral. We show that the phase spiral forms at a look back time of $\sim 6$ Gyr during the pericentric passage of a $\sim10^{10}$ $\rm M_{\odot}$ satellite on a polar orbit. This satellite stimulates the formation of a resonant wake in the dark matter halo while losing mass at a rate of $\sim0.5$-$1$ dex per orbit loop. The peak magnitude of the wake-induced gravitational torque at the Solar radius is $\sim 8$ times that from the satellite, and triggers the formation of a disc warp that wraps up into a vertical phase spiral over time. As the wake decays, the phase spiral propagates several Gigayears to present-day and can be described as ``ever-present'' once stable disc evolution is established. These results suggest an alternative scenario to explain the $Gaia$ phase spiral which does not rely on a perturbation from bar buckling or a recent direct hit from a satellite.

astro-ph.GA↗

arXiv2023

Colour and infall time distributions of satellite galaxies in simulated Milky-Way analogs

We use the Auriga simulations to probe different satellite quenching mechanisms operating at different mass scales ($10^5 M_\odot \lesssim M_\star \lesssim 10^{11} M_\odot$) in Milky Way-like hosts. Our goal is to understand the origin of the satellite colour distribution and star-forming properties in both observations and simulations. We find that the satellite populations in the Auriga simulations, which was originally designed to model Milky Way-like host galaxies, resemble the populations in the Exploration of Local VolumE Satellites (ELVES) Survey and the Satellites Around Galactic Analogs (SAGA) survey in their luminosity function in the luminosity range $-12 \lesssim M_V \lesssim -15$ and resemble ELVES in their quenched fraction and colour--magnitude distribution in the luminosity range $-12 \lesssim M_g \lesssim -15$. We find that satellites transition from blue colours to red colours at the luminosity range $-15 \lesssim M_g \lesssim -12$ in both the simulations and observations and we show that this shift is driven by environmental effects in the simulations. We demonstrate also that the colour distribution in both simulations and observations can be decomposed into two statistically distinct populations based on their morphological type or star-forming status that are statistically distinct. In the simulations, these two populations also have statistically distinct infall time distributions. The comparison presented here seems to indicate that the tension between the quenched fraction in SAGA and simulations is resolved by the improved target selection of ELVES, but there are still tensions in understanding the colours of faint galaxies, of which ELVES appears to have a significant population of faint blue satellites not recovered in Auriga.

astro-ph.GA↗

arXiv2022

Machine Learning for Galactic Archaeology: A chemistry-based neural network method for identification of accreted disc stars

We develop a method ('Galactic Archaeology Neural Network', GANN) based on neural network models (NNMs) to identify accreted stars in galactic discs by only their chemical fingerprint and age, using a suite of simulated galaxies from the Auriga Project. We train the network on the target galaxy's own local environment defined by the stellar halo and the surviving satellites. We demonstrate that this approach allows the detection of accreted stars that are spatially mixed into the disc. Two performance measures are defined - recovery fraction of accreted stars, and the probability that a star with a positive (accreted) classification is a true-positive result, P(TP). As the NNM output is akin to an assigned probability, we are able to determine positivity based on flexible threshold values that can be adjusted easily to refine the selection of presumed-accreted stars. We find that GANN identifies accreted disc stars within simulated galaxies, with high recovery fraction and/or high P(TP). We also find that stars in Gaia-Enceladus-Sausage (GES) mass systems are over 50% recovered by our NNMs in the majority (18/24) of cases. Additionally, nearly every individual source of accreted stars is detected at 10% or more of its peak stellar mass in the disc. We also demonstrate that a conglomerated NNM, trained on the halo and satellite stars from all of the Auriga galaxies provides the most consistent results, and could prove to be an intriguing future approach as our observational capabilities expand.

astro-ph.GA↗

arXiv2022

Simulating radio synchrotron emission in star-forming galaxies: small-scale magnetic dynamo and the origin of the far infrared-radio correlation

In star-forming galaxies, the far-infrared (FIR) and radio-continuum luminosities obey a tight empirical relation over a large range of star-formation rates (SFR). We examine magneto-hydrodynamic galaxy simulations with cosmic rays (CRs), accounting for their advective and anisotropic diffusive transport. We show that gravitational collapse of the proto-galaxy generates a corrugated accretion shock, which injects turbulence and drives a small-scale magnetic dynamo. As the shock propagates outwards and the associated turbulence decays, the large velocity shear between the supersonically rotating cool disc with respect to the (partially) pressure-supported hot circumgalactic medium excites Kelvin-Helmholtz surface and body modes. Those inject turbulence and drive multiple small-scale dynamos, which exponentially amplify magnetic fields. They grow in scale to reach equipartition with thermal and CR energies in Milky Way-mass galaxies. In small galaxies, the magnetic energy saturates at the turbulent energy while it fails to reach equipartition with thermal and CR energies. We solve for steady-state spectra of CR protons, secondary electrons/positrons from hadronic CR-proton interactions with the interstellar medium, and primary shock-accelerated electrons at supernovae. The radio-synchrotron emission is dominated by primary electrons, irradiates the magnetised disc, bulge, and bubble-shaped magnetically-loaded outflows of our simulated Milky Way-mass galaxy. Our star-forming and star-bursting galaxies with saturated magnetic fields match the global FIR-radio correlation (FRC) across four orders of magnitude. Its intrinsic scatter arises due to (i) different magnetic saturation levels that result from different seed magnetic fields, (ii) different radio synchrotron luminosities for different specific SFRs at fixed SFR and (iii) a varying radio intensity with galactic inclination. (abridged)

astro-ph.GA↗

arXiv2022

Linking the brightest stellar streams with the accretion history of Milky Way-like galaxies

According to the current galaxy formation paradigm, mergers and interactions play an important role in shaping present-day galaxies. The remnants of this merger activity can be used to constrain galaxy formation models. In this work we use a sample of thirty hydrodynamical simulations of Milky Way-mass halos, from the AURIGA project, to generate surface brightness maps and search for the brightest stream in each halo as a function of varying limiting magnitude. We find that none of the models shows signatures of stellar streams at $μ_{r}^{lim} \leq 25$ mag arcsec$^{-2}$. The stream detection increases significantly between 27 and 28 mag arcsec$^{-2}$. Nevertheless, even at 30 mag arcsec$^{-2}$, 13 percent of our models show no detectable streams. We study the properties of the brightest streams progenitors (BSPs). We find that BSPs are accreted within a broad range of infall times, from 1.6 to 10 Gyr ago, with only 25 percent accreted within the last 5 Gyrs; thus most BSPs correspond to relatively early accretion events. We also find that 37 percent of the BSPs survive to the present day. The median infall times for surviving and disrupted BSPs are 5.6 and 6.7 Gyr, respectively. We find a clear relation between infall time and infall mass of the BSPs, such that more massive progenitors tend to be accreted at later times. However, we find that the BSPs are not, in most cases, the dominant contributor to the accreted stellar halo of each galaxy.

astro-ph.GA↗

arXiv2022

How Cosmic Rays Mediate the Evolution of the Interstellar Medium

We explore the impact of diffusive cosmic rays (CRs) on the evolution of the interstellar medium (ISM) under varying assumptions of supernova explosion environment. In practice, we systematically vary the relative fractions of supernovae (SN) occurring in star-forming high-density gas and those occurring in random locations decoupled from star-forming gas to account for SN from run-away stars or explosions in regions that have been cleared by prior SN, stellar winds, or radiation. We explore various mixed models by adjusting these fractions relative to each other. We find that in the simple system of a periodic stratified gas layer the ISM structure will evolve to one of two solutions: a "peak driving" state where warm gas is volume filling or a "thermal runaway" state where hot gas is volume filling. CR pressure and transport are important factors that strongly influence the solution state the ISM reaches and have the ability to flip the ISM between solutions. Observable signatures such as gamma ray emission and HI gas are explored. We find that gamma ray luminosity from pion decay is largely consistent with observations for a range of model parameters. The thickness of the HI gas layer may be too compact, however, this may be due to a large cold neutral fraction of midplane gas. The volume fraction of hot gas evolves to stable states in both solutions, but neither settles to a Milky Way-like configuration, suggesting that additional physics which is omitted here (e.g. a cosmological circum-galactic medium, radiation transport, or spectrally resolved and spatially varying CR transport) may be required.

astro-ph.GA↗

arXiv2022

Formation and fate of low metallicity stars in IllustrisTNG50

Low metallicity stars give rise to unique spectacular transients and are of immense interest for understanding stellar evolution. Their importance has only grown further with the recent detections of mergers of stellar mass black holes that likely originate mainly from low metallicity progenitor systems. Moreover, the formation of low metallicity stars is intricately linked to galaxy evolution, in particular to early enrichment and to later accretion and mixing of lower metallicity gas. Because low metallicity stars are difficult to observe directly, cosmological simulations are crucial for understanding their formation. Here we quantify the rates and locations of low metallicity star formation using the high-resolution TNG50 magnetohydrodynamical cosmological simulation, and we examine where low metallicity stars end up at $z=0$. We find that $20\%$ of stars with $Z_*<0.1\,\mathrm{Z_\odot}$ form after $z=2$, and that such stars are still forming in galaxies of all masses at $z=0$ today. Moreover, most low-metallicity stars at $z=0$ reside in massive galaxies. We analyse the radial distribution of low metallicity star formation, and discuss the curious case of seven galaxies in TNG50 that form stars from primordial gas even at $z=0$.

astro-ph.GA↗

arXiv2021

Determining the full satellite population of a Milky Way-mass halo in a highly resolved cosmological hydrodynamic simulation

We investigate the formation of the satellite galaxy population of a Milky Way-mass halo in a very highly resolved magneto-hydrodynamic cosmological zoom-in simulation (baryonic mass resolution $m_b =$ 800 $\rm M_{\odot}$). We show that the properties of the central star-forming galaxy, such as the radial stellar surface density profile and star formation history, are: i) robust to stochastic variations associated with the so-called ``Butterfly Effect''; and ii) well converged over 3.5 orders of magnitude in mass resolution. We find that there are approximately five times as many satellite galaxies at this high resolution compared to a standard ($m_b\sim 10^{4-5}\, \rm M_{\odot}$) resolution simulation of the same system. This is primarily because 2/3rds of the high resolution satellites do not form at standard resolution. A smaller fraction (1/6th) of the satellites present at high resolution form and disrupt at standard resolution; these objects are preferentially low-mass satellites on intermediate- to low-eccentricity orbits with impact parameters $\lesssim 30$ kpc. As a result, the radial distribution of satellites becomes substantially more centrally concentrated at higher resolution, in better agreement with recent observations of satellites around Milky Way-mass haloes. Finally, we show that our galaxy formation model successfully forms ultra-faint galaxies and reproduces the stellar velocity dispersion, half-light radii, and $V$-band luminosities of observed Milky Way and Local Group dwarf galaxies across 6 orders of magnitude in luminosity ($10^3$-$10^{9}$ $\rm L_{\odot}$).

astro-ph.GA↗

arXiv2021

Satellites Around Milky Way Analogs: Tension in the Number and Fraction of Quiescent Satellites Seen in Observations Versus Simulations

We compare the star-forming properties of satellites around Milky Way (MW) analogs from the Stage~II release of the Satellites Around Galactic Analogs Survey (SAGA-II) to those from the APOSTLE and Auriga cosmological zoom-in simulation suites. We use archival GALEX UV imaging as a star-formation indicator for the SAGA-II sample and derive star-formation rates (SFRs) to compare with those from APOSTLE and Auriga. We compare our detection rates from the NUV and FUV bands to the SAGA-II H$α$ detections and find that they are broadly consistent with over $85\%$ of observed satellites detected in all three tracers. We apply the same spatial selection criteria used around SAGA-II hosts to select satellites around the MW-like hosts in APOSTLE and Auriga. We find very good overall agreement in the derived SFRs for the star-forming satellites as well as the number of star-forming satellites per host in observed and simulated samples. However, the number and fraction of quenched satellites in the SAGA-II sample are significantly lower than those in APOSTLE and Auriga below a stellar mass of $M_*\sim10^{8}\,M_{\odot}$, even when the SAGA-II incompleteness and interloper corrections are included. This discrepancy is robust with respect to the resolution of the simulations and persists when alternative star-formation tracers are employed. We posit that this disagreement is not readily explained by vagaries in the observed or simulated samples considered here, suggesting a genuine discrepancy that may inform the physics of satellite populations around MW analogs.

astro-ph.GA↗

arXiv2020

A tale of two populations: surviving and destroyed dwarf galaxies and the build up of the Milky Way's stellar halo

We use magneto-hydrodynamical simulations of Milky Way-mass haloes from the Auriga project to examine the properties of surviving and destroyed dwarf galaxies that are accreted by these haloes over cosmic time. We show that the combined luminosity function of surviving and destroyed dwarfs at infall is similar in the various Auriga haloes, and is dominated by the destroyed dwarfs. There is, however, a strong dependence on infall time: destroyed dwarfs have typically early infall times, $t_{infall}<6$ Gyr, whereas the majority of dwarfs accreted at $t_{infall}>10$ Gyr have survived to the present day. Because of their late infall the surviving satellites today had higher metallicites at infall than their destroyed counterparts of similar infall mass; the difference is even more pronounced for the present-day metallicites of satellites, many of which continue to form stars after infall. In agreement with previous work, we find that a small number of relatively massive destroyed dwarf galaxies dominate the mass of the stellar haloes. However, there is a significant radial dependence: while 90 per cent of the mass in the inner regions ($<\,20\,$kpc) is contributed, on average, by only 3 massive progenitors, the outer regions ($>\,100\,$kpc) typically have $\sim8$ main progenitors of relatively lower mass. Finally, we show that a few massive progenitors dominate the metallicity distribution of accreted stars, even at the metal poor end. Contrary to common assumptions in the literature, dwarf galaxies of mass $M_{*}<10^7 \, M_{\odot}$ make up less than 10 per cent of the accreted, metal poor stars ([Fe/H] $<\,-3$) in the inner $50\,$kpc.

astro-ph.GA↗

arXiv2020

Magnetising the circumgalactic medium of disk galaxies

The circumgalactic medium (CGM) is one of the frontiers of galaxy formation and intimately connected to the galaxy via accretion of gas on to the galaxy and gaseous outflows from the galaxy. Here we analyse the magnetic field in the CGM of the Milky Way-like galaxies simulated as part of the \textsc{Auriga} project that constitutes a set of high resolution cosmological magnetohydrodynamical zoom simulations. We show that before $z=1$ the CGM becomes magnetised via galactic outflows that transport magnetised gas from the disk into the halo. At this time the magnetisation of the CGM closely follows its metal enrichment. We then show that at low redshift an in-situ turbulent dynamo that operates on a timescale of Gigayears further amplifies the magnetic field in the CGM and saturates before $z=0$. The magnetic field strength reaches a typical value of $0.1\,μG$ at the virial radius at $z=0$ and becomes mostly uniform within the virial radius. Its Faraday rotation signal is in excellent agreement with recent observations. For most of its evolution the magnetic field in the CGM is an unordered small scale field. Only strong coherent outflows at low redshift are able to order the magnetic field in parts of the CGM that are directly displaced by these outflows.

astro-ph.GA↗