arXiv subjects

Prateek Sharma

Publications and source records attributed to Prateek Sharma.

At least 19 recordsLinked to original sources

arXiv2026

Simulating megaparsec-scale jets of radio galaxies: Magneto-hydrodynamics of jets reaching 5 Mpc

Extragalactic jets have long prompted the question of how far relativistic outflows can extend, with some radio sources reaching 5 - 7 Mpc in length. These great extents motivate investigations into their ages, propagation dynamics, stability, and impact on the environment. We perform 3D high-resolution numerical simulations of two jet configurations involving continuous injection at different powers propagating in low-density regions of the cosmos (static and laminar), investigating the conditions for jet collimation versus disruption at extreme scales. We show that the combined effects of higher jet thrust (enhanced kinetic power), improved collimation (suppression of transverse distortions), and magnetic stabilization (strengthened poloidal field) can sustain a laterally confined flow, enabling such a jet to reach 5 Mpc in just 15 Myr (injecting a total energy of $2.3 \times 10^{61}$ erg into the environment). In contrast, a jet lacking these conditions dissipates more rapidly, forming lobe-like morphologies and reaching only $\sim 3$ Mpc over $\sim35$ Myr (injecting total energy of $8.1 \times 10^{60}$ erg). Pinch and kink MHD instabilities are identified as the primary drivers of transverse distortions; their suppression allows the persistence of a fast spine alongside a slower, dissipative head (location of maximum environmental interaction). We find that the jet-head propagation shows two regimes: one with speed $\sim0.5 c$; the other with speed from $\sim 0.2 c$ to $\sim 0.05 c$. We consider a proxy of synchrotron emission and find that radiation is concentrated in regions of enhanced compression and magnetic amplification, primarily near the first recollimation shock (producing a bright radio spot) and at the jet-head interaction zone (producing the radio termination lobe). Such jets facilitate the transport of substantial energy and magnetic flux into underdense cosmic regions.

astro-ph.GA↗

arXiv2025

Impact of Cosmic Ray Distribution on the Growth and Saturation of Bell Instability

Cosmic rays (CRs) streaming in weakly magnetized plasmas can drive large-amplitude magnetic fluctuations via nonresonant streaming instability (NRSI), or Bell instability. Using one-dimensional kinetic simulations, we investigate how mono-energetic and power-law CR momentum distributions influence the growth and saturation of NRSI. The linear growth is governed solely by the CR current and is largely insensitive to the CR distribution. However, the saturation depends strongly on the CR distribution and is achieved through CR isotropization, which quenches the driving current. Mono-energetic CRs effectively amplify the magnetic field and isotropize. For power-law distributions, the lowest-energy CRs dominate current relaxation and magnetic growth, while the highest-energy CRs remain weakly scattered, limiting their contribution to saturation. In the absence of low-energy CRs, high-energy particles amplify magnetic fields effectively and isotropize. Accounting for these effects, we provide a modified saturation prescription valid for both relativistic and nonrelativistic CRs. We propose a layered CR-confinement scenario upstream of astrophysical shocks, relevant to particle acceleration to high energies.

astro-ph.HE↗

arXiv2025

Multiphase gas in Circumgalactic cloud complexes: Insights from kiloparsec-scale Magnetohydrodynamic Turbulence Simulations

The circumgalactic medium (CGM) is the diffuse gas surrounding a galaxy's halo, and it plays a vital role in the galactic baryon cycle. However, its mass distribution across the virial phase and the cooler, denser atomic phase, remains uncertain. To investigate this, we perform high-resolution magnetohydrodynamic simulations of 0.125--8 kpc-scale representative patches of the CGM, with parameters informed by quasar absorption line observations. Our simulations resolve the cooling length (the minimum across all temperatures of $c_s t_{\rm cool}$, where $c_s$ is the sound speed and $t_{\rm cool}$ is the cooling time in isobaric conditions), allowing us to track the evolution of cold gas more accurately. We find that low-density CGM gas ($3\times10^{-4}$ cm$^{-3}$) cannot sustain cold gas below $10^4$ K for long, due to a large value of the ratio between the cooling to mixing time ($t_{\rm cool}/t_{\rm mix}$). In contrast, higher-density environments ($3\times10^{-3}~{\rm cm}^{-3}$) reach a turbulent multiphase steady state, with up to $50\%$ of the mass in the cold phase, occupying only about $1\%$ of the volume. To connect with large-volume cosmological simulations and small ${\rm pc}$-scale idealized simulations, we explore different box sizes (0.125--8 kpc) and identify a key scaling relation: simulations with similar $t_{\rm cool}/t_{\rm mix}$ exhibit comparable cold gas mass fractions and lifetimes. Importantly, we find that simply sub-sampling (reducing box-size) a small region from a large-volume simulation while maintaining a constant turbulent energy density injection rate from larger to smaller scales artificially shortens $t_\mathrm{mix}$, leading to inaccurate predictions for cold gas survival. This means that cold gas at small $\lesssim 10$ kpc scales arises in relatively dense, quiescent regions of the CGM rather than the turbulent ones undergoing cascade from large scales.

astro-ph.GA↗

arXiv2025

Woven by the Whirls: The growth and entrainment of cold clouds in turbulent hot winds

Galactic and intergalactic flows often exhibit relative motion between the cold dense gas and the hot diffuse medium. Such multiphase flows -- involving gas at different temperatures, densities, and ionization states -- for instance, galactic winds, are frequently turbulent. However, idealized simulations typically model the winds and driven turbulence separately, despite their intertwined roles in galaxy evolution. To address this, we investigate the survival of a dense cloud in a hot wind subject to continuous external turbulent forcing. We perform 3D hydrodynamic simulations across a range of turbulent Mach numbers in the hot phase $\mathcal{M}_{\rm turb}=v_{\rm turb}/c_{\rm s, wind}$ from 0.1 to 0.7 ($c_{\rm s, wind}$ and $v_{\rm turb}$ being the sound speed and the turbulent velocity in the hot phase, respectively). We find that in spite of the additional subsonic turbulence, cold clouds can survive if the cooling time of the mixed gas $t_{\rm cool, mix}$ is shorter than a modified destruction time $\tilde{t}_{\rm cc}$, i.e., $t_{\rm cool,mix}/\tilde{t}_{\rm cc}<1$ where $\tilde{t}_{\rm cc}=t_{\rm cc}/(1+\left(\mathcal{M}_{\rm turb}/\left(f_{\rm mix}\mathcal{M}_{\rm wind}\right)\right)^2)^{1/2}$, where $f_{\rm mix}\sim0.6$ is a fudge factor. Moreover, in the `survival regime', turbulence can enhance the growth of cold clouds by up to an order of magnitude because of more efficient stretching and an associated increase in the surface area. This increase in mass transfer between the phases leads to significantly faster entrainment of cold material in turbulent winds. In contrast to the narrow filamentary tails formed in laminar winds, turbulence stretches the cold gas orthogonally, dispersing it over a larger area and changing absorption line signatures.

astro-ph.GA↗

arXiv2025

Universal Structure of Turbulent Radiative Mixing Layers

Turbulent radiative mixing layers (TRMLs), where shear-driven turbulence between dense and diffuse gas produces rapidly cooling intermediate-temperature gas, are ubiquitous in the interstellar and circumgalactic media. Using a quasi-steady Reynolds decomposition, we separate mean and turbulent components. In quasi-isobaric TRMLs, upstream gas cools and compresses before streamwise momentum is fully mixed, yielding a negative shear stress (R_{xz}) and a positive compressive stress (R_{zz}) that together sustain a steady radiative conversion of hot to cold gas. A pronounced thermal-pressure dip develops within the TRML, while radiative losses are offset by the divergences of enthalpy flux and (subdominant) turbulent heat flux (Q_t). The volume-averaged temperature follows a tanh profile, resulting in universal emissivity distributions that are consistent with simulations. Contrary to previous claims, the cooling-rate surface density saturates and becomes independent of box size in the strong-cooling limit, establishing the universal structure of TRMLs.

astro-ph.GA↗

arXiv2025

MQFQ-Sticky: Fair Queueing For Serverless GPU Functions

Hardware accelerators like GPUs are now ubiquitous in data centers, but are not fully supported by common cloud abstractions such as Functions as a Service (FaaS). Many popular and emerging FaaS applications such as machine learning and scientific computing can benefit from GPU acceleration. However, FaaS frameworks (such as OpenWhisk) are not capable of providing this acceleration because of the impedance mismatch between GPUs and the FaaS programming model, which requires virtualization and sandboxing of each function. The challenges are amplified due to the highly dynamic and heterogeneous FaaS workloads. This paper presents the design and implementation of a FaaS system for providing GPU acceleration in a black-box manner (without modifying function code). Running small functions in containerized sandboxes is challenging due to limited GPU concurrency and high cold-start overheads, resulting in heavy queueing of function invocations. We show how principles from I/O scheduling, such as fair queuing and anticipatory scheduling, can be translated to function scheduling on GPUs. We develop MQFQ-Sticky, an integrated fair queueing and GPU memory management approach, which balances the tradeoffs between locality, fairness, and latency. Empirical evaluation on a range of workloads shows that it reduces function latency by 2x to 20x compared to existing GPU and CPU queueing policies.

cs.DC↗

arXiv2025

TeV spectral bump of cosmic-ray protons and helium nuclei: the role of nearby supernova remnants

The existence of nearby discrete cosmic-ray sources can lead to many interesting effects on the observed properties of cosmic rays. Recent measurements of cosmic rays with the CALET and the DAMPE experiments have revealed a bump-like new feature in the proton and helium spectra in the energy range of ~ (1-100) TeV/nucleon. The origin of the feature is not clearly understood. In this paper, considering an improved and more detailed analysis than previous works, and using an updated age and distance estimates of nearby supernova remnants along with an energy-dependent escape process for cosmic rays from the remnants, we show that the spectral bump can be explained by the contribution of cosmic rays from the nearby supernova remnants, in particular, the Vela remnant. We also show that the contribution from the nearby remnants agrees well with the observed spectra of the heavier cosmic-ray elements from carbon to iron as well as with the measured all-particle cosmic-ray spectrum beyond the knee region when combined with a background flux of cosmic rays originating from distant supernova remnants.

astro-ph.HE↗

arXiv2025

Fading in the Flow: Suppression of cold gas growth in expanding galactic outflows

Multiphase outflows, revealed by multi-wavelength observations, are crucial in redistributing gas and metals within and around galaxies. These outflows are often modelled theoretically using wind tunnel simulations of a cold ($\sim 10^4$ K) cloud interacting with a uniform hot ($\sim 10^6$ K) wind. However, real outflows expand downstream, a feature overlooked in most idealised simulations. We study how an expanding wind affects the survival, morphology, and dynamics of a cloud. We conduct idealised hydrodynamic simulations with optically thin radiative cooling of a cloud in an expanding wind, modelled using the adiabatic Chevalier & Clegg (1985) solution. We find that clouds remain locally isobaric with the wind, leading to a steep decline in their density contrast and eventual dissolution downstream. Compared to a plane-parallel wind, this suppresses cold gas mass growth because as clouds travel downstream, the surrounding mixed boundary layer becomes diffuse and less radiative. Our analytical scaling arguments show that cloud expansion and local pressure equilibrium are the key regulators of cold mass growth. Unlike traditional simulations, our model accounts for the differential expansion experienced by the long cometary tails of clouds in wind tunnels. This creates a strong head-to-tail emission gradient in the filamentary cold gas, which is more consistent with observations. We also demonstrate that the dynamics of individual clouds can substantially alter the radial properties of their host multiphase outflows.

astro-ph.GA↗

arXiv2025

OmniLearn: A Framework for Distributed Deep Learning over Heterogeneous Clusters

Deep learning systems are optimized for clusters with homogeneous resources. However, heterogeneity is prevalent in computing infrastructure across edge, cloud and HPC. When training neural networks using stochastic gradient descent techniques on heterogeneous resources, performance degrades due to stragglers and stale updates. In this work, we develop an adaptive batch-scaling framework called OmniLearn to mitigate the effects of heterogeneity in distributed training. Our approach is inspired by proportional controllers to balance computation across heterogeneous servers, and works under varying resource availability. By dynamically adjusting worker mini-batches at runtime, OmniLearn reduces training time by 14-85%. We also investigate asynchronous training, where our techniques improve accuracy by up to 6.9%.

cs.LG↗

arXiv2025

Computational Astrophysics, Data Science & AI/ML in Astronomy: A Perspective from Indian Community

In contemporary astronomy and astrophysics (A&A), the integration of high-performance computing (HPC), big data analytics, and artificial intelligence/machine learning (AI/ML) has become essential for advancing research across a wide range of scientific domains. These tools are playing an increasingly pivotal role in accelerating discoveries, simulating complex astrophysical phenomena, and analyzing vast amounts of observational data. For India to maintain and enhance its competitive edge in the global landscape of computational astrophysics and data science, it is crucial for the Indian A&A community to fully embrace these transformative technologies. Despite limited resources, the expanding Indian community has already made significant scientific contributions. However, to remain globally competitive in the coming years, it is vital to establish a robust national framework that provides researchers with reliable access to state-of-the-art computational resources. This system should involve the regular solicitation of computational proposals, which can be assessed by domain experts and HPC specialists, ensuring that high-impact research receives the necessary support. By building such a system, India can cultivate the talent, infrastructure, and collaborative environment necessary to foster world-class research in computational astrophysics and data science.

astro-ph.IM↗

arXiv2024

Misty, patchy, and turbulent: constraining the cool circumgalactic medium with mCC

The circumgalactic medium (CGM) is the largest baryon reservoir around galaxies, but its extent, mass, and temperature distribution remain uncertain. We propose that cool gas ($\sim 10^4$ K) in the CGM resides in clumpy structures referred to as cloud complexes (CCs) rather than uniformly filling the entire CGM volume. Each CC contains a mist of tiny cool cloudlets dispersed in a warm/hot medium ($\sim 10^5$ - $10^6$~K). Modeling CCs in the mist limit (unit area covering fraction within a CC) simplifies the calculation of observables like ion absorption columns, equivalent widths, compared to modeling individual cloudlets from first principles. Through Monte Carlo realizations of CCs, we explore how CC properties affect the observed variation in observables. We find that a power-law distribution of CCs ($dN_{\rm CC}/dR \propto R^{-1}$) with a total of $\sim 10^3$ CCs each with a radius of $\sim 10$ kpc and total cool gas mass of $\sim 10^{10} M_\odot$ reproduces MgII column density and equivalent width distribution trends with impact parameter for the COS-Halos sample (Werk+ 2013). We further show that the area-averaged MgII column density, combined with the area covering fraction, provides a robust proxy for estimating the cool CGM mass, independent of other model parameters. Modeling a larger number of (smaller size) cloudlets within a CC shows that line blending from individual cloudlets results in turbulent broadening on the CC scale. This work presents a practical framework for linking CGM models with observations of a multiphase CGM, illuminating the distribution of cool gas in galaxy halos.

astro-ph.GA↗

arXiv2024

The Jevons Paradox In Cloud Computing: A Thermodynamics Perspective

How do we explain the simultaneous growth in energy efficiency of cloud computing and its energy consumption? The Jevons paradox provides one perspective of this phenomenon. However, it is not clear or obvious \emph{why} the Jevons paradox exists, and \emph{when} is it applicable. To answer these questions, we seek inspiration from thermodynamics, and model the cloud as a thermodynamic system. We find that system growth, due to the revenue generation of cloud platforms, is a key driver behind energy consumption. This thermodynamic model provides energy consumption insights into modern hyperscale clouds, and we validate it using data from Meta and Google. Our investigation points to the necessity of future work in new and meaningful efficiency metrics, implications for future applications and edge clouds, and the need for studying system-wide energy and sustainability.

cs.DC↗

arXiv2024

FaasMeter: Energy-First Serverless Computing

Functions as a Service has emerged as a popular abstraction for a wide range of cloud applications and an important cloud workload. We present the design and implementation of FaasMeter, a FaaS control plane which provides energy monitoring, accounting, control, and pricing as first-class operations. The highly diverse and dynamic workloads of FaaS create additional complexity to measuring and controlling energy usage which FaasMeter can mitigate. We develop a new statistical energy disaggregation approach to provide accurate and complete energy footprints for functions, despite using noisy and coarse-grained system-level power (not just CPU power readings). Our accurate and robust footprints are achieved by combining conventional power models with Kalman filters and Shapley values. FaasMeter is a full-spectrum energy profiler, and fairly attributes energy of shared resources to functions (such as energy used by the control plane itself). We develop new energy profiling validation metrics, and show that FaasMeter's energy footprints are accurate to within 1\% of carefully obtained marginal energy ground truth measurements.

cs.DC↗

arXiv2024

Breaking into the window of primordial black hole dark matter with x-ray microlensing

Primordial black holes (PBHs) in the mass range $10^{-16}-10^{-11}~M_\odot$ may constitute all the dark matter. We show that gravitational microlensing of bright x-ray pulsars provide the most robust and immediately implementable opportunity to uncover PBH dark matter in this mass window. As proofs of concept, we show that the currently operational NICER telescope can probe this window near $10^{-14}~M_\odot$ with just two months of exposure on the x-ray pulsar SMC-X1, and that the forthcoming STROBE-X telescope can probe complementary regions in only a few weeks. These times are comparable to the week-long exposures obtained by NICER on various individual sources. We take into account the effects of wave optics and the finite extent of the source, which become important for the mass range of our PBHs. We also provide a spectral diagnostic to distinguish microlensing from transient background events and to broadly mark the PBH mass if true microlensing events are observed. In light of the powerful science case, i.e., the imminent discovery of dark matter searchable over multiple decades of PBH masses with achievable exposures, we strongly urge the commission of a dedicated large broadband telescope for x-ray microlensing. We derive the microlensing reach of such a telescope by assuming sensitivities of detector components of proposed missions, and find that with hard x-ray pulsar sources PBH masses down to a few $10^{-17}~M_\odot$ can be probed.

astro-ph.HE↗

arXiv2024

Ram pressure stripping in clusters: Gravity can bind the ISM but not the CGM

We explore the survival of a galaxy's circumgalactic medium (CGM) as it experiences ram pressure stripping (RPS) moving through the intracluster medium (ICM). For a satellite galaxy, the CGM is often assumed to be entirely stripped/evaporated, an assumption that may not always be justified. We carry out 3D-hydrodynamic simulations of the interstellar and circumgalactic media (ISM+CGM) of a galaxy like JO201 moving through the ICM. The CGM can survive long at cluster outskirts ($\gtrsim2 \rm \ Gyr$) but at smaller cluster-centric distances, 90\% of the CGM mass is lost within $\sim 500$ Myr. The gravitational restoring force on the CGM is mostly negligible and the CGM-ICM interaction is analogous to \textit{`cloud-wind interaction'}. The CGM stripping timescale does not depend on the ram pressure but on the CGM to ICM density contrast $\chi$. Two distinct regimes emerge for CGM stripping: the $\chi >1$ regime, which is the well-known \textit{`cloud crushing'} problem, and the $\chi <1$ regime, which we refer to as the (relatively unexplored) \textit{`bubble drag'} problem. The first pericentric passage near the cluster core can rapidly -- over a crossing time $t_{\rm drag} \sim R/v_{\rm rel}$ -- strip the CGM in the \textit{bubble drag} regime. The ISM stripping criterion unlike the CGM criterion, still depends on the ram pressure $\rho_{\rm ICM} v_{\rm rel}^2$. The stripped tails of satellites contain contributions from both the disk and the CGM. The X-ray plume in M89 in the Virgo cluster and a lack of it in the nearby M90 might be attributed to their orbital histories. M90 has likely undergone stripping in the bubble drag regime due to a pericentric passage close to the cluster center.

astro-ph.GA↗

arXiv2023

Dissipation of AGN jets in a clumpy interstellar medium

Accreting supermassive black holes (SMBHs) frequently power jets that interact with the interstellar/circumgalactic medium (ISM/CGM), regulating star-formation in the galaxy. Highly supersonic jets launched by active galactic nuclei (AGN) power a cocoon that confines them and shocks the ambient medium. We build upon the models of narrow conical jets interacting with a smooth ambient medium, to include the effect of dense clouds that are an essential ingredient of a multiphase ISM. The key physical ingredient of this model is that the clouds along the supersonic jet-beam strongly decelerate the jet-head, but the subsonic cocoon easily moves around the clouds without much resistance. We propose scalings for important physical quantities -- cocoon pressure, head & cocoon speed, and jet radius. We obtain, for the first time, the analytic condition on clumpiness of the ambient medium for the jet to dissipate within the cocoon and verify it with numerical simulations of conical jets interacting with a uniform ISM with embedded spherical clouds. A jet is defined to be dissipated when the cocoon speed exceeds the speed of the jet-head. We compare our models to more sophisticated numerical simulations, direct observations of jet-ISM interaction (e.g., quasar J1316+1753), and discuss implications for the Fermi/eROSITA bubbles. Our work also motivates effective subgrid models for AGN jet feedback in a clumpy ISM unresolved by the present generation of cosmological galaxy formation simulations.

astro-ph.GA↗

arXiv2023

Between the cosmic-ray `knee' and the `ankle': Contribution from star clusters

We show that massive young star clusters may be possible candidates that can accelerate Galactic cosmic rays (CRs) in the range of $10^7\hbox{--}10^9$ GeV (between the `knee' and `ankle'). Various plausible scenarios such as acceleration at the wind termination shock (WTS), supernova shocks inside these young star clusters, etc. have been proposed,since it is difficult to accelerate particles up to the $10^7\hbox{--}10^9$ GeV range in the standard paradigm of CR acceleration in supernova remnants. We consider a model for the production of different nuclei in CRs from massive stellar winds using the observed distribution of young star clusters in the Galactic plane. We present a detailed calculation of CR transport in the Galaxy, taking into account the effect of diffusion, interaction losses during propagation, and particle re-acceleration by old supernova remnants to determine the all-particle CR spectrum. Using the maximum energy estimate from the Hillas criterion, we argue that a young massive star cluster can accelerate protons up to a few tens of PeV. Upon comparison with the observed data, our model requires a CR source spectrum with an exponential cutoff of $5\times 10^7 Z$ GeV ($50\,Z$~PeV) from these clusters together with a cosmic-ray injection fraction of $\sim 5\%$ of the wind kinetic energy. We discuss the possibility of achieving these requirements in star clusters, and the associated uncertainties, in the context of considering star clusters as the natural accelerator of the `second component' of Galactic cosmic rays.

astro-ph.HE↗

arXiv2023

Multiphase Neutral Interstellar Medium: Analyzing Simulation with H I 21cm Observational Data Analysis Techniques

Several different methods are regularly used to infer the properties of the neutral interstellar medium (ISM) using atomic hydrogen (H I) 21cm absorption and emission spectra. In this work, we study various techniques used for inferring ISM gas phase properties, namely the correlation between brightness temperature and optical depth $(T_B(v)$, $τ(v))$ at each channel velocity ($v$), and decomposition into Gaussian components, by creating mock spectra from a 3D magnetohydrodynamic simulation of a two-phase, turbulent ISM. We propose a physically motivated model to explain the $T_B(v)-τ(v)$ distribution and relate the model parameters to properties like warm gas spin temperature and cold cloud length scales. Two methods based on Gaussian decomposition -- using only absorption spectra and both absorption and emission spectra -- are used to infer the column density distribution as a function of temperature. In observations, such analysis reveals the puzzle of large amounts (significantly higher than in simulations) of gas with temperature in the thermally unstable range of $\sim$200 K to $\sim$2000 K and a lack of the expected bimodal (two-phase) temperature distribution. We show that, in simulation, both methods are able to recover the actual gas distribution in the simulation till temperatures $\lesssim2500$~K (and the two-phase distribution in general) reasonably well. We find our results to be robust to a range of effects such as noise, varying emission beam size, and simulation resolution. This shows that the observational inferences are unlikely to be artifacts, thus highlighting a tension between observations and simulations. We discuss possible reasons for this tension and ways to resolve it.

astro-ph.GA↗