Resolution Dependence of Cloud-Wind Simulations Cholla ( Schneider & Robertson 2015 ) , numpy ( Van Der Walt et al. 2011 ) , matplotlib ( Hunter 2007 ) , hdf5 ( The HDF Group 1997-2019 ) seaborn ( Waskom 2021 )
Hannah J. Leary, Evan Schneider, Helena M. Richie · 2025
Read Resolution Dependence of Cloud-Wind Simulations Cholla ( Schneider & Robertson 2015 ) , numpy ( Van Der Walt et al. 2011 ) , matplotlib ( Hunter 2007 ) , hdf5 ( The HDF Group 1997-2019 ) seaborn ( Waskom 2021 ) by Hannah J. Leary, Evan Schneider, Helena M. Richie (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-13186.txt.
Introduction
Hannah Leary, Evan Schneider, and Helena Richie test how grid detail affects simulations of gas clouds hit by wind in a 2025 paper. Such cloud-wind setups model how interstellar clouds are shredded or pushed by fast flows. The team ran wind-tunnel simulations at two wind speeds, one below and one above the speed of sound, each at five levels of detail.
In the slower wind, the cloud's fate changed sharply at a threshold detail of sixteen cells per cloud radius. The trend was not smooth, which surprised the authors. In the faster wind, results changed more steadily with detail, matching the different balance of forces early on. The study warns that simulation detail can change conclusions, not just sharpen them.
Catalog record
| CREATOR | Hannah J. Leary, Evan Schneider, Helena M. Richie |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 8 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-13186
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Resolution Dependence of Cloud-Wind Simulations Cholla ( Schneider & Robertson 2015 ) , numpy ( Van Der Walt et al. 2011 ) , matplotlib ( Hunter 2007 ) , hdf5 ( The HDF Group 1997-2019 ) seaborn ( Waskom 2021 )
Hannah J. Leary, Evan Schneider, Helena M. Richie
Abstract
wind-tunneand accelerationalso in bothsubsonic and ingtions run lla code, we investigate the effect that numerical resolution has on clFord evolsetup in sexplore imulWnumerical e consider upersonic and subsonic regimes. We begin with two groups of adiabatic wind tunnel simulations with wind speeds of 100 km/s and In the subsonic case, we find that a0 km/s, respectively. s per cloud radiusoisthl at five vary in the cloud evolutions:In the supersonic case the trend is more monotonic, consistent with the difference in timescales for which ram pressure acceleration dominates over mixing in the early acceleration.
I
Introduction
Galactic winds play a crucial role in galaxy formation, as they regulate star formation and enrich the circumgalactic medium (CGM) with metals. These winds are observed to be highly multiphase, with cool clouds of gas entrained in a fast moving hot wind
(
Veilleux et al. 2005
;
Rupke 2018
, see)
. On small scales in idealized wind tunnel simulations, mixing has been shown to play a key role in cloud evolution. In the adiabatic limit, Kelvin-Helmholtz (KH) and other hydrodynamic instabilities destroy clouds in several characteristic “cloud crushing times”
(
Klein et al. 1994
)
. For a cloud with radius
R_{cl}
in a wind of speed
v_{w}
, and a density contrast
\chi=\frac{\rho_{cl}}{\rho_{w}}
, this timescale is given by
t_{\mathrm{cc}}=\chi^{1/2}\frac{R_{\mathrm{cl}}}{v_{\mathrm{w}}}.
(1)
Because the growth rate of KH instabilities in simulations is sensitive to the smallest resolved scale
(
Chandrasekhar 1961
)
, the rate of mixing will be impacted by the resolution of the simulation. Indeed, previous work has demonstrated that as simulation resolution increases, finer structures appear and the number of cloudlets increases
(
Cooper et al. 2009
;
Schneider & Robertson 2017
)
. Despite the important role of mixing, the relationship between cloud evolution and resolution is an under-explored area, particularly in the low resolution limit most relevant in cosmological simulations. In this Note, we explore the effects of numerical resolution on cloud evolution in the simple case of a spherical, adiabatic cloud. We also investigate the role of wind speed by investigating both the supersonic and the less commonly studied subsonic wind regime.
Figure 1:
The evolution of clouds in two wind speeds: 100 km/s (top) and 1000 km/s (bottom). (a) and (c) show snapshots of column density at resolutions
R_{cl}/\Delta x
= 4, 8, 16, 32, and 48, as well as at three time-steps: 0, 4, and 7
t_{cc}
. (b) and (d) show the change in cloud mass over time normalized to the initial cloud mass, and the average cloud velocity normalized by the wind velocity over time.
II
Methods and Analysis
To investigate the effects of resolution on cloud evolution, we ran two sets of simulations at five different resolutions:
R_{cl}/\Delta x
= 4, 8, 16, 32, and 48 (hereafter
R_{4}
,
R_{8}
,
R_{16}
,
R_{32}
,
R_{48}
). Simulations were run with the Cholla hydrodynamics code
(
Schneider & Robertson 2015
)
using PPM reconstruction, an HLLC Riemann solver, and the Van Leer integrator. We employ a standard wind tunnel setup consisting of a long box with a constant wind entering from the left boundary and outflow boundaries on all other sides. All simulations have box dimensions of 32
R_{cl}
x 16
R_{cl}
x 16
R_{cl}
.
In order to explore potential differences in mixing for the subsonic versus supersonic regime, we use wind speeds of 100 km/s (M = 0.852) and 1000 km/s (M = 8.52), respectively. The wind density and temperature are the same for all simulations:
n_{w}=0.01
cm
-3
and
T_{w}=10^{6}
K. Clouds are initialized as spheres with a radius of 50 pc and zero initial velocity, and are positioned 3.2
R_{cl}
away from the
-x
boundary (Figure
1
, first column). The cloud-wind density contrast is
\chi
=
10^{2}
, and the cloud temperature is
T_{cl}=10^{4}
K . All simulations run for a total of 10
t_{cc}
.
In the following analysis, we consider cells with a density above
\frac{1}{3}
of the initial cloud density to be “cloud” material (Our results are not sensitive to the precise density threshold). We measure cloud mass
M_{cl}=\sum\rho_{i}(>\frac{1}{3}\rho_{\mathrm{cl,init}})\mathrm{dV},
(2)
and mass-weighted average cloud velocity,
\bar{v}_{\mathrm{x}}
,
\bar{v}_{\mathrm{x}}=\frac{\sum\rho_{i}(>\frac{1}{3}\rho_{\mathrm{cl,init}})v_{x,i}\mathrm{dV}}{M_{cl}},
(3)
every 0.2
t_{cc}
, where
\rho_{i}
is the mass density of the
i
th cell and
\mathrm{dV}
is the cell volume.
III
Results & Discussion
Figure
1
illustrates the difference in cloud evolution at each resolution. Panels (a) and (c) show column density snapshots of the cloud in its initial state, in a partially disrupted state at 4
t_{cc}
, and late in its evolution at 7
t_{cc}
. These snapshots clearly show how hydrodynamic instabilities destroy the cloud within several
t_{cc}
in all cases. However, in both the subsonic and supersonic regimes, the cloud’s acceleration and destruction is affected by the resolution of the simulation. Somewhat surprisingly, this trend does not appear to be monotonic. In the subsonic case in particular (panel a) the destruction follows an almost parabolic trend, with a turning point at
R_{16}
.
In panels (b) and (d) we further investigate this behavior by plotting the normalized cloud mass
M_{cl}/M_{cl_{i}}
and the normalized average cloud velocity
\bar{v}_{x}/v_{w}
over time. In (b) we see in the subsonic wind case that the
R_{16}
cloud survives the longest and is accelerated to significantly lower terminal velocities than the other clouds, reaching a terminal velocity of only 23% of the wind speed. In the supersonic wind, (d) shows that while the
R_{16}
cloud is again the slowest to be completely destroyed, the velocity trend is more monotonic, with the lowest resolutions accelerating the most quickly and the highest resolutions the most slowly. In both subsonic and supersonic regimes (b) and (d) show that the highest speeds are achieved in the highest resolution simulations, with
R_{32}
reaching 35% of the wind speed in both the subsonic and supersonic cases. Results appear to be converging at
R_{32}
in both cases.
The early-time acceleration at all resolutions is similar, while the late-time behavior of the velocities diverge. This is likely a result of a change in acceleration mechanism throughout the cloud evolution. Ram pressure is the dominant acceleration mechanism in the beginning of the interaction, before mixing becomes efficient and destroys the cloud. While the efficiency of mixing is resolution-dependent, the ram pressure only depends on the surface area of the cloud, which is the same for all simulations. Because the ram pressure of the wind scales as
\rho v^{2}
, we expect the acceleration due to ram pressure to be 100 times greater in the supersonic models, while the mixing rate scales only linearly with the shear velocity. This may explain why the resolution trend seen in the subsonic versus supersonic simulations is different.