SMHM - Behroozi+19 | Romanowsky & Fall+13

Main driver of galaxy formation is mass

⇒ halo-mass main driver

Galaxy formation in a 🥜 shell

1. Dark matter halo formation from cosmic fluctuations
2. Gas condenses in DM halo
3. Stars form out of gas
4. Supernova + AGNs reinject energy

increasingly non-linear

increasingly non-linear

increasingly chaotic?

⇒

Galaxy formation in a 🥜 shell

1. Dark matter halo formation from cosmic fluctuations
2. Gas condenses in DM halo
3. Stars form out of gas
4. Supernova + AGNs reinject energy

Tidal torque theory: White 84
Stochastic spin: Vitviska+ 02, Beson+ 20
Spin prediction, Porciani+ 02

Theoretical predictions of angular momentum are, at best, inaccurate

First complication

Spin prediction / measured
(in simulation)

Note: spin is \(\lambda \propto J/M_\mathrm{vir}R_\mathrm{vir}V_\mathrm{vir}\)

Second complication

Let's be optimistic and assume we can predict DM spin

Jiang+ 19

Butterfly effect: Genel+ 19, see also Thiébaut+ 08

Third complication

Resimulating the same galaxy twice yields different results ⇒ galaxies are chaotic

Why can't we predict angular momentum?

1. No good model for dark matter angular momentum,
2. Stellar AM poorly correlates with DM AM,
3. Galaxies seem chaotic.

\(z=0\)

\( z = 100\)

\(z=0\)

\( z = 100\)

[White 84]

\(z=0\)

\( z = 100\)

[Genetic modifications: Roth+16, see also Rey&Pontzen 18, Stopyra+20]

Time

Angular
momentum

\(z\rightarrow \infty\)

\(z=0\)

✅ AM of fixed DM regions responds ~linearly (so is not chaotic!)

However, particle membership hence DM halo does not

Angular
momentum

Time

Dark matter spin is hard to predict

⇒ the usual culprit are mergers

Cadiou+24
Based on the critical event theory (Cadiou +20)
itself based on Hanami 2001

We can make some reasonable predictions
of orbital spin of mergers

Given: tides + density peaks

\(z\gg 100\)

Changes to tides at \(z=100\), effect at \(z<1\)

\(\lambda_\mathrm{DM}\)

Full hydro simulations
(RAMSES, New-Horizon model):

• Resolved disk height
  \(\Delta x_\mathrm{min} = 35\ \mathrm{pc}\)
• \(M_\mathrm{200c} = 10^{12}\ \mathrm{M}_\odot\)
  at \(z=2\)
• SF, AGN, SN feedback
• 3 galaxies, 5× scenario each

Lowest tides

Low tides

Higher tides

Highest tides

\( j_0 \times 0.66\)

\( j_0 \times 0.8\)

\( j_0 \times 1.2\)

\( j_0 \times 1.5\)

See Cadiou+21a

Halo and disk evolve separately,

this allows stellar AM to be ~linear despite halo AM not being

See Cadiou+21a

Special case: no massive satellite

See Cadiou+21a

Two intermediate conclusions:
 

1. Stellar angular momentum is driven by cosmological tides
   ⇒ failure of theory, not fundamental limitation
    
2. Galaxies respond quasi-linearly to perturbations
   ⇒ morphology, bulge fraction, spin are not chaotic!

What if the galaxy had formed here instead?

What if the galaxy had formed here instead?

or here?

The “splicing” technique

1. Generate ICs
2. Integrate (\(N\)-body)
3. Select region of interest
4. Trace back to ICs
5. “Splice” the region
6. Integrate again

Splicing: equivalent of constraining density field at all points in region

\(\longleftarrow  50\,\mathrm{Mpc} \longrightarrow\)

Same halo (same initial tides + density)

forming closer and closer to filament

Repeat for 5 halos, 9 positions

Deviation from the mean

For different quantities

\(\sigma\)

\(\langle q\rangle\)

In practice, we change the potential so we're testing sensitivity to non-linear gravitational coupling to large-scale structures

Simple isothermal CGM             vs.           Complex multiphase CGM

• Well-mixed (isothermal)
• "Bathtub" modelling

• Not well-mixed (turbulence)
• How much gas reaches the galaxy?

Megatron simulations:
first simulation to combine at high-\(z\)
+ non-equilibrium chemistry
+ radiative transport
+ \(\sim 50-100\,\mathrm{pc}\) resolution
 

Reveals

• Complex, multiphase structure,
• Smallest structures at resolution limit,
• Non-equilibrium ionized structures
  • impact on thermochemical structure,
  • impact on observations.

O I

O II

O III

Megatron simulations:
first simulation to combine at high-\(z\)
+ non-equilibrium chemistry
+ radiative transport
+ \(\sim 50-100\,\mathrm{pc}\) resolution
 

Reveals

• Complex, multiphase structure,
• Smallest structures at resolution limit,
• Non-equilibrium ionized structures
  • impact on thermochemical structure,
  • impact on observations.

Megatron simulations:
first simulation to combine at high-\(z\)
+ non-equilibrium chemistry
+ radiative transport
+ \(\sim 50-100\,\mathrm{pc}\) resolution
 

Reveals

• Complex, multiphase structure,
• Smallest structures at resolution limit,
• Non-equilibrium ionized structures
  • impact on thermochemical structure,
  • impact on observations.

OIII column density
in simulation

assuming ionization equilibrium

Most powerful clusters are switching to GPUs
(following the AI hype...)

Manage grid, simple computation

Computation-heavy
(hydro, gravity, …)

CPU

GPU

[…]

wasted
time

wasted
time

Typical approach: offloading*

Dyablo's approach:

* Arepo, Gadget3, RAMSES, ...

Amdahl's law: latency kills gains of parallelisation

Manage grid, simple computation

Computation-heavy
(hydro, gravity, …)

CPU

GPU
(or CPUs)

[…]

Typical approach: offloading

Dyablo's approach: “true” GPU computing, CPU as a puppeteer

Dyablo: Implemented physics

• Adaptive mesh-refinement,
• Hydrodynamics / MHD / radiative transfer
• Gravity
• Particles,
• Gas cooling
• Star formation and feedback,
• Cosmology,
• Diffusion terms: thermal conduction, viscosity,
• Cosmic rays, thermochemistry...

Agora initial conditions, Kim+16

Dyablo: how does it scale?

weak scaling: solar convection

Image credits: Maxime Delorme

Ad Astra CPU (2 x AMD Epyc Genoa 96 cores)

Conclusions

1. Tides set merger orbital parameters & angular momentum accretion
   Mergers are not stochastic/not rerolling the dice
    
2. Trickling down to galactic scale, which drives scaling relations
   Galaxies are less stochastic than expected
   Galaxy & DM spin are partially independent at the level of individual galaxies
    
3. Non-linear effects on halo shape and alignment are comparable to population-level
   Crucial to investigate further for the success of Euclid or LSST
    
4. Knowledge/modelling frontier: Circum Galactic Medium transport
   Non-equilibrium, multiphase state in obs. (low-\(z\)) and sims (high-\(z\))
   Cosmic web → CGM → ISM: how does AM evolve in this picture?

Tracing cosmic evolution

The (not so) new frontier:
physics in the circumgalactic medium

Tracing cosmic evolution: what's next?

Conclusions

Come talk to me if you're interested in…

• galaxy formation,
• the cosmic web,
   
• numerical simulations (on GPUs, load-balancing, on-the-fly, …),
• data visualization (interactive, volume rendering, …).

🖥️ 118c — cadiou@iap.fr — https://cphyc.github.io —@cphyc.bsky.social