Simulating galaxy formation: sub-grid models at the intermediate scale

Simulating galaxy formation:
sub-grid models at the
intermediate scale
Alexander Hobbs
ETH Zürich
Collaborators:
Sergei Nayakshin, Chris Power, Justin Read,
Seung-Hoon Cha, Andrew King

http://www.strw.leidenuniv.nl/~booth/movies/25mpc-object1-dens.avi

A pretty movie...
...but what is actually
going on here?

Simulation and movie credit: Craig Booth (Schaye et al., 2010)

M halo
~ 10 14 M sun
~ 500 kpc
Simulation and movie credit: Craig Booth (Schaye et al., 2010)

M halo
~ 10 14 M sun
~ 500 kpc
Satellites/Dwarfs?
Simulation and movie credit: Craig Booth (Schaye et al., 2010)

IGM
M halo
~ 10 14 M sun
~ 500 kpc
Satellites/Dwarfs?
Simulation and movie credit: Craig Booth (Schaye et al., 2010)

What are the
“gastrophysical”
processes?

Simulation and movie credit: Craig Booth (Schaye et al., 2010)

Star formation
and feedback
Simulation and movie credit: Craig Booth (Schaye et al., 2010)

Accretion and outflow
Simulation and movie credit: Craig Booth (Schaye et al., 2010)

Accretion and outflow
Simulation and movie credit: Craig Booth (Schaye et al., 2010)

~ 400 kpc

~ 400 kpc ~ 50 kpc
Bulge
Galactic disc

~ 400 kpc ~ 50 kpc ~ 5 kpc
Bulge
Galactic disc

Molecular
clouds
~ 500 pc
~ 400 kpc ~ 50 kpc ~ 5 kpc
Bulge
Galactic disc

Accretion
disc
Jet
Molecular
clouds
Torus
~ 50 pc
~ 500 pc
~ 400 kpc ~ 50 kpc ~ 5 kpc
Bulge
Galactic disc

Accretion
disc
Jet
Molecular
clouds
Torus
~ 10 -4 pc
~ 50 pc
~ 500 pc
~ 400 kpc ~ 50 kpc ~ 5 kpc
Bulge
Galactic disc

Accretion
disc
Jet
Molecular
clouds
Torus
~ 10 -4 pc
~ 50 pc
~ 500 pc
~ 400 kpc ~ 50 kpc ~ 5 kpc
Bulge
1 kpc
Galactic disc
1 kpc

In large-scale simulations we
require “sub-grid” models!

Sub-grid models
Star formation (SF)
- Sink particle creation
...locally enforce Schmidt law
SMBH accretion
- Bondi-Hoyle capture rate
SF feedback
- Thermal energy coupling
- “Blast-wave” feedback
AGN outflow
- Thermal energy coupling
- Momentum coupling

What improvements can
we bring to the table?

(1) Accretion models
- Current approach extremely idealised...and unphysical
...takes no account of angular momentum
...ignores presence of host dark matter halo
- Accretion from large scales ignores intermediate-scale processes
...such as feedback from star formation, e.g., supernovae

(1) Accretion models
- Current approach extremely idealised...and unphysical
...takes no account of angular momentum
...ignores presence of host dark matter halo
- Accretion from large scales ignores intermediate-scale processes
...such as feedback from star formation, e.g., supernovae

(1) Accretion models
- Current approach extremely idealised...and unphysical
...takes no account of angular momentum
...ignores presence of host dark matter halo
- Accretion from large scales ignores intermediate-scale processes
...such as feedback from star formation, e.g., supernovae
“Accretion disc particle” method
Power, Nayakshin & King (2010)

“Accretion disc particle” method (Power, Nayakshin & King 2010)
- Angular momentum
provides a natural
barrier to accretion
- Only material with the
lowest angular momentum
should accrete
- Introduce sub-grid method
to track and utilise the
angular momentum
evolution of sub-grid disc:

(1) Accretion models
- Current approach extremely idealised...and unphysical
...takes no account of angular momentum
...ignores presence of host dark matter halo
- Accretion from large scales ignores intermediate-scale processes
...such as feedback from star formation, e.g., supernovae

(1) Accretion models
- Current approach extremely idealised...and unphysical
...takes no account of angular momentum
...ignores presence of host dark matter halo
- Accretion from large scales ignores intermediate-scale processes
...such as feedback from star formation, e.g., supernovae

(1) Accretion models
- Current approach extremely idealised...and unphysical
...takes no account of angular momentum
...ignores presence of host dark matter halo
- Accretion from large scales ignores intermediate-scale processes
...such as feedback from star formation, e.g., supernovae

(1) Accretion models
- Current approach extremely idealised...and unphysical
...takes no account of angular momentum
...ignores presence of host dark matter halo
- Accretion from large scales ignores intermediate-scale processes
...such as feedback from star formation, e.g., supernovae
Supernovae-driven turbulent
feeding model
Hobbs, Nayakshin, Power & King (submitted)

Supernovae-driven turbulent feeding model
(Hobbs, Nayakshin, Power & King, submitted)
- Turbulent flows ubiquitous in astrophysics
- Evidence for supersonic turbulence in star
forming regions of molecular clouds (Larson
1981, Motte 1998, Padoan & Nordlund 2002)
- Simulations of supersonic turbulence in AGN
discs popular in the literature (Wada & Norman
2002, Chen et al. 2009, Brüggen & Scannapeico 2009)
Projection of isothermal supersonic turbulence, v turb ~ 100
km s -1 , seeded into constant density spherical shell with M ~
10 7 M sun
, v rot
= 60 km s -1 , t ~ 10 5 yrs after starburst
Star formation feedback
at kpc scales

Supernovae-driven turbulent feeding model
(Hobbs, Nayakshin, Power & King, submitted)
http://www.astro.le.ac.uk/~aph11/sweden_talk/turbulent_movie.avi

(2) AGN outflow models
- Injection of thermal energy/momentum to BH particle neighbours
simplistic and SPH-resolution dependant
- Require a feedback method that can be used regardless of
geometry or optical depth, with independent resolution

(2) AGN outflow models
- Injection of thermal energy/momentum to BH particle neighbours
simplistic and SPH-resolution dependant
- Require a feedback method that can be used regardless of
geometry or optical depth, with independent resolution
Dynamic Monte-Carlo radiation
transfer in SPH
Nayakshin, Cha & Hobbs (2009)

Dynamic Monte-Carlo radiation transfer in SPH
(Nayakshin, Cha & Hobbs 2009)
- Radiative transfer equation along a ray:
continuous
stochastic
- Explicitly follow photon's trajectory:
- Fraction of momentum absorbed (Δp γ
)
passed to SPH particles via kernel
Photons emitted isotropically & stochastically
- Re-emission of photons in random dir'n
with probability Δp γ
/ p γ0
to ensure timeaveraged
conservation of photon energy

Dynamic Monte-Carlo radiation transfer in SPH
(Nayakshin, Cha & Hobbs 2009)
- Radiative transfer equation along a ray:
continuous
stochastic
- Explicitly follow photon's trajectory:
- Fraction of momentum absorbed (Δp γ
)
passed to SPH particles via kernel
Particles found will interact if they contain the
photon within their own smoothing length
- Re-emission of photons in random dir'n
with probability Δp γ
/ p γ0
to ensure timeaveraged
conservation of photon energy

Combining sub-grid models: (1) + (2)
“Accretion disc particle” method
Power, Nayakshin & King (2010)
+
Dynamic Monte-Carlo radiation
transfer in SPH
Nayakshin, Cha & Hobbs (2009)

“Accretion disc particle” method + radiative feedback. Isothermal EQS:
http://www.astro.le.ac.uk/~cbp1/research/feedback/rot_xy.vrot0.3.isotherm.avi
Bondi-Hoyle prescription + radiative feedback. Isothermal EQS:
http://www.astro.le.ac.uk/~cbp1/research/feedback/rot_xy.vrot0.3.bondi.avi

“Accretion disc particle” method + radiative
feedback. Isothermal EQS.
Bondi-Hoyle prescription + radiative
feedback. Isothermal EQS.

The goal:
Embed sub-grid models in
galaxy formation simulations

http://www.astro.le.ac.uk/~cbp1/research/merger/mw_merger.mpeg
Simulation by Alexander Hobbs, Chris Power, Justin Read
Movie by Chris Nixon

Summary
Have developed / in the process of developing sub-grid models:
(1) For accretion (2) For outflow
“Accretion disc particle” method
Supernovae-driven turbulent
feeding model
Dynamic Monte-Carlo
radiative transfer

Conclusions
- Development of better sub-grid models for large-scale
simulations is very important!
- In particular, the development of physically motivated models
that take account of intermediate scale processes is a priority
- Preliminary tests of such models in simulations have shown
significant differences compared to previous results
- There is a clear need for physical self-consistency in simulations
of galaxy formation and evolution at large scales
- GPUs could help! More computing power means we can
explore a greater number of physical processes

Thank you

Simulating galaxy formation:
sub-grid models at the
intermediate scale
Alexander Hobbs
ETH Zürich
Collaborators:
Sergei Nayakshin, Chris Power, Justin Read,
Seung-Hoon Cha, Andrew King
1

http://www.strw.leidenuniv.nl/~booth/movies/25mpc-object1-dens.avi
2

A pretty movie...
...but what is actually
going on here?
3

Simulation and movie credit: Craig Booth (Schaye et al., 2010)
4

M halo
~ 10 14 M sun
~ 500 kpc
Simulation and movie credit: Craig Booth (Schaye et al., 2010)
5

M halo
~ 10 14 M sun
~ 500 kpc
Satellites/Dwarfs?
Simulation and movie credit: Craig Booth (Schaye et al., 2010)
6

IGM
M halo
~ 10 14 M sun
~ 500 kpc
Satellites/Dwarfs?
Simulation and movie credit: Craig Booth (Schaye et al., 2010)
7

What are the
“gastrophysical”
processes?
8

Simulation and movie credit: Craig Booth (Schaye et al., 2010)
9

Star formation
and feedback
Simulation and movie credit: Craig Booth (Schaye et al., 2010)
10

Accretion and outflow
Simulation and movie credit: Craig Booth (Schaye et al., 2010)
11

Accretion and outflow
Simulation and movie credit: Craig Booth (Schaye et al., 2010)
12

~ 400 kpc
13

~ 400 kpc ~ 50 kpc
Bulge
Galactic disc
14

~ 400 kpc ~ 50 kpc ~ 5 kpc
Bulge
Galactic disc
15

Molecular
clouds
~ 500 pc
~ 400 kpc ~ 50 kpc ~ 5 kpc
Bulge
Galactic disc
16

Accretion
disc
Jet
Molecular
clouds
Torus
~ 50 pc
~ 500 pc
~ 400 kpc ~ 50 kpc ~ 5 kpc
Bulge
Galactic disc
17

Accretion
disc
Jet
Molecular
clouds
Torus
~ 10 -4 pc
~ 50 pc
~ 500 pc
~ 400 kpc ~ 50 kpc ~ 5 kpc
Bulge
Galactic disc
18

1 kpc
Accretion
disc
Jet
Molecular
clouds
Torus
~ 10 -4 pc
~ 50 pc
~ 500 pc
~ 400 kpc ~ 50 kpc ~ 5 kpc
Bulge
Galactic disc
1 kpc
19

In large-scale simulations we
require “sub-grid” models!
20

Sub-grid models
Star formation (SF)
- Sink particle creation
...locally enforce Schmidt law
SMBH accretion
- Bondi-Hoyle capture rate
SF feedback
- Thermal energy coupling
- “Blast-wave” feedback
AGN outflow
- Thermal energy coupling
- Momentum coupling
21

What improvements can
we bring to the table?
22

(1) Accretion models
- Current approach extremely idealised...and unphysical
...takes no account of angular momentum
...ignores presence of host dark matter halo
- Accretion from large scales ignores intermediate-scale processes
...such as feedback from star formation, e.g., supernovae
23

(1) Accretion models
- Current approach extremely idealised...and unphysical
...takes no account of angular momentum
...ignores presence of host dark matter halo
- Accretion from large scales ignores intermediate-scale processes
...such as feedback from star formation, e.g., supernovae
24

(1) Accretion models
- Current approach extremely idealised...and unphysical
...takes no account of angular momentum
...ignores presence of host dark matter halo
- Accretion from large scales ignores intermediate-scale processes
...such as feedback from star formation, e.g., supernovae
“Accretion disc particle” method
Power, Nayakshin & King (2010)
25

“Accretion disc particle” method (Power, Nayakshin & King 2010)
- Angular momentum
provides a natural
barrier to accretion
- Only material with the
lowest angular momentum
should accrete
- Introduce sub-grid method
to track and utilise the
angular momentum
evolution of sub-grid disc:
26

(1) Accretion models
- Current approach extremely idealised...and unphysical
...takes no account of angular momentum
...ignores presence of host dark matter halo
- Accretion from large scales ignores intermediate-scale processes
...such as feedback from star formation, e.g., supernovae
27

(1) Accretion models
- Current approach extremely idealised...and unphysical
...takes no account of angular momentum
...ignores presence of host dark matter halo
- Accretion from large scales ignores intermediate-scale processes
...such as feedback from star formation, e.g., supernovae
28

(1) Accretion models
- Current approach extremely idealised...and unphysical
...takes no account of angular momentum
...ignores presence of host dark matter halo
- Accretion from large scales ignores intermediate-scale processes
...such as feedback from star formation, e.g., supernovae
29

(1) Accretion models
- Current approach extremely idealised...and unphysical
...takes no account of angular momentum
...ignores presence of host dark matter halo
- Accretion from large scales ignores intermediate-scale processes
...such as feedback from star formation, e.g., supernovae
Supernovae-driven turbulent
feeding model
Hobbs, Nayakshin, Power & King (submitted)
30

Supernovae-driven turbulent feeding model
(Hobbs, Nayakshin, Power & King, submitted)
- Turbulent flows ubiquitous in astrophysics
- Evidence for supersonic turbulence in star
forming regions of molecular clouds (Larson
1981, Motte 1998, Padoan & Nordlund 2002)
- Simulations of supersonic turbulence in AGN
discs popular in the literature (Wada & Norman
2002, Chen et al. 2009, Brüggen & Scannapeico 2009)
Projection of isothermal supersonic turbulence, v turb ~ 100
km s -1 , seeded into constant density spherical shell with M ~
10 7 M sun , v rot = 60 km s -1 , t ~ 10 5 yrs after starburst
Star formation feedback
at kpc scales
31

Supernovae-driven turbulent feeding model
(Hobbs, Nayakshin, Power & King, submitted)
http://www.astro.le.ac.uk/~aph11/sweden_talk/turbulent_movie.avi
32

(2) AGN outflow models
- Injection of thermal energy/momentum to BH particle neighbours
simplistic and SPH-resolution dependant
- Require a feedback method that can be used regardless of
geometry or optical depth, with independent resolution
33

(2) AGN outflow models
- Injection of thermal energy/momentum to BH particle neighbours
simplistic and SPH-resolution dependant
- Require a feedback method that can be used regardless of
geometry or optical depth, with independent resolution
Dynamic Monte-Carlo radiation
transfer in SPH
Nayakshin, Cha & Hobbs (2009)
34

Dynamic Monte-Carlo radiation transfer in SPH
(Nayakshin, Cha & Hobbs 2009)
- Radiative transfer equation along a ray:
continuous
stochastic
- Explicitly follow photon's trajectory:
- Fraction of momentum absorbed (Δp γ
)
passed to SPH particles via kernel
Photons emitted isotropically & stochastically
- Re-emission of photons in random dir'n
with probability Δp γ
/ p γ0
to ensure timeaveraged
conservation of photon energy
35

Dynamic Monte-Carlo radiation transfer in SPH
(Nayakshin, Cha & Hobbs 2009)
- Radiative transfer equation along a ray:
continuous
stochastic
- Explicitly follow photon's trajectory:
- Fraction of momentum absorbed (Δp γ
)
passed to SPH particles via kernel
Particles found will interact if they contain the
photon within their own smoothing length
- Re-emission of photons in random dir'n
with probability Δp γ
/ p γ0
to ensure timeaveraged
conservation of photon energy
36

Combining sub-grid models: (1) + (2)
“Accretion disc particle” method
Power, Nayakshin & King (2010)
+
Dynamic Monte-Carlo radiation
transfer in SPH
Nayakshin, Cha & Hobbs (2009)
37

“Accretion disc particle” method + radiative feedback. Isothermal EQS:
http://www.astro.le.ac.uk/~cbp1/research/feedback/rot_xy.vrot0.3.isotherm.avi
Bondi-Hoyle prescription + radiative feedback. Isothermal EQS:
http://www.astro.le.ac.uk/~cbp1/research/feedback/rot_xy.vrot0.3.bondi.avi
38

“Accretion disc particle” method + radiative
feedback. Isothermal EQS.
Bondi-Hoyle prescription + radiative
feedback. Isothermal EQS.
39

The goal:
Embed sub-grid models in
galaxy formation simulations
40

http://www.astro.le.ac.uk/~cbp1/research/merger/mw_merger.mpeg
41
Simulation by Alexander Hobbs, Chris Power, Justin Read
Movie by Chris Nixon

Summary
Have developed / in the process of developing sub-grid models:
(1) For accretion (2) For outflow
“Accretion disc particle” method
Supernovae-driven turbulent
feeding model
Dynamic Monte-Carlo
radiative transfer
42

Conclusions
- Development of better sub-grid models for large-scale
simulations is very important!
- In particular, the development of physically motivated models
that take account of intermediate scale processes is a priority
- Preliminary tests of such models in simulations have shown
significant differences compared to previous results
- There is a clear need for physical self-consistency in simulations
of galaxy formation and evolution at large scales
- GPUs could help! More computing power means we can
explore a greater number of physical processes
43

Thank you
44