On the consistency of jet feedback modelling across different astrophysics hydrodynamical codes

TL;DR

This work tests a code-agnostic subgrid AGN jet launcher by implementing it in SWIFT (SPH), AREPO (moving-mesh), and PLUTO (AMR grid) to isolate hydrodynamical solver effects. It compares jets in uniform and stratified media across resolutions and jet speeds, revealing that lobe morphology is code-dependent (SWIFT: short/wide/hot; AREPO: long/thin/cool; PLUTO: intermediate) while the ambient energy deposition remains broadly similar (~60% of injected energy). Stratified environments amplify deviations from self-similar evolution and produce longer, thinner lobes with different mixing, yet remnants and ambient heating converge in behavior across codes. The study concludes that solver differences are subdominant to subgrid-model uncertainties in cosmological simulations, though high-resolution, code-comparative tests are essential to calibrate jet subgrid prescriptions for robust, cross-code predictions.

Abstract

Active Galactic Nuclei (AGN) feedback is essential in cosmological simulations of galaxy formation, yet its implementation has to rely on subgrid models due to limited resolution. We present a novel subgrid jet-launching method for galaxy formation simulations and implement it in three hydrodynamical codes: the smoothed particle hydrodynamics (SPH) code SWIFT, the moving-mesh code AREPO, and the Eulerian grid code PLUTO. To isolate the impact of hydrodynamical solvers on jet evolution, we compare idealised jets and their remnants in uniform and stratified media across resolutions and jet parameters. In uniform media, all jets drive bow shocks, inflate hot lobes, exhibit backflows, and evolve self-similarly. For the parameters explored, SWIFT lobes are shorter, wider, and hotter; AREPO lobes are longer, thinner, and cooler; while PLUTO lobes display complex flows with intermediate characteristics. In stratified media, jets deviate from self-similar evolution, inflating longer and thinner lobes due to lower external ram pressure. After switch-off, SWIFT jets evolve into smooth cylindrical bubbles, AREPO jets produce long filamentary remnants, and PLUTO jets yield intermediate-length remnants with varying degrees of mixing. Despite such differences, all jets and remnants have a similar impact on the ambient medium. We conclude that variations in lobe properties between codes emerge even for identical subgrid prescriptions, since the coupling of jet feedback to resolvable scales and the effective resolution depend on the hydrodynamical method. In structure formation simulations, these solver differences are likely subdominant to uncertainties in subgrid modelling and calibration, while averaging over galaxy populations may lessen their impact.

Figures (13)

• Figure 1: Sketch of our virtual-particle model for a single injection event. Two virtual particles are launched in opposite directions, inside two cones, and travel for $10 \, \mathrm{kpc}$ without interacting with the medium, before they inject mass, momentum, and energy into their gas neighbours. Denoted are the different components of the resulting jets and lobes they inflate.
• Figure 2: Comparison of swift, arepo and pluto jets. Top: Overview slices in the $y=0$ plane in the uniform medium after $98 \, \mathrm{Myr}$ at our highest resolution ($N = 10^8$, $M_\mathrm{gas} = 2\times 10^5 \, \mathrm{M}_\odot$) simulations. Mapped quantities are temperature (top left), density (top right), pressure (bottom left), and velocity magnitude (bottom right). All jets drive a bow shock in the ambient medium, inflate lobes, and develop backflows. swift jets inflate shorter, wider, and hotter lobes, arepo jets produce longer, thinner, and cooler lobes, and pluto jets yield lobes that are of intermediate length and width and relatively hot. Bottom: Zoomed-in (white box in the top panel) temperature map displaying the underlying discretisation method for each code. We can distinguish differences between codes in the spatial resolution of the lobes and the medium, as well as in the development of Kelvin-Helmholtz instabilities.
• Figure 3: Radial velocity slices of jets simulated with the three codes in the uniform medium after $98 \, \mathrm{Myr}$. All jets display a central sheath of high-velocity material moving outwards along the jet axis, surrounded by backflows. swift jets show the least coherent backflows, arepo jets display stronger and more streamlined backflows, and pluto jets present fast but irregular backflows.
• Figure 4: Temperature (top) - density (bottom) slices of the time evolution of the jets simulated with the three codes in the uniform medium. Different rows showcase the different jets at progressively later times. The swift lobes retain their shape and evolve self-similarly from early on, while arepo and pluto lobes are becoming more elongated, for at least the first $50 \, \mathrm{Myr}$. The top and bottom lobes progressively fuse in all codes due to backflows and the growth of KH instabilities.
• Figure 5: Plots of lobe length and width with respect to time for jets simulated with the three codes in the uniform medium. Black lines represent the lobes simulated with swift, blue lines the arepo lobes, and red lines the pluto lobes. Solid lines represent estimates of length, and dotted lines represent estimates of width. The dashed grey lines represent the analytic, self-similar solution. All jets produce lobes that expand self-similarly by the end of the simulations. swift under-predicts, arepo over-predicts, and pluto agrees with the normalised analytic solution for lobe length.