1. Feedback in Starburst Galaxies
Todd Thompson
Princeton University
with Eliot Quataert, Norm Murray, & Eli Waxman

2. Outline Goal: A model for the global structure of starbursts.
Why starbursts? The physical conditions.
Radiation pressure feedback.
Magnetic fields, cosmic rays, & -rays.

3. Systematics of Star Formation
Schmidt Law:
``Star-forming” galaxies:
Extended, few-kpc scales.
~ billion year timescales.
``Starburst” galaxies:
Compact, 100’s pc scales.
1-100 million year timescales.
Pressure: P ~  G g2
Starbursts
Star-forming galaxies
Kennicutt (1998)

4. Regulation & Feedback in Galaxies
Low star formation efficiency:
Suggests feedback and/or regulation over a broad range of conditions.
Q~1 observed in disks.
(Martin & Kennicutt 2001)
Stellar processes (?): Stellar winds, radiation, supernovae, HII regions, etc.
Non-stellar processes (?): MRI. (Sellwood & Balbus 99;Piontek & Ostriker 04)
Starbursts
Star-forming galaxies
Kennicutt (1998)

5. Why Starbursts?

6. M82
IRAS
M51
NOAO
NGC 253
Arp 220

7. Backgrounds & Starbursts
Dole et al. (2006)

8. Why Starbursts? Starbursts & U/LIRGs What do we want to know?
lie on the same scaling relations with normal galaxies.
constitute a large fraction of the IR background, the star-formation rate density at high z (also, -ray & MeV/TeV  backgrounds).
may be a key phase in the growth of super-massive black holes & spheroids.
are connected physically to super-star clusters, starburst cores.
have turbulent velocities v > 10 km/s.
What do we want to know?
Constituents: radiation, gas/dust, magnetic fields, and cosmic rays.
The origin and systematics of the scaling relations of galaxies.

9. The Physical Conditions
Arp 220 (d ~ 80 Mpc):
Two counter-rotating cores, ~100pc.
Circumbinary disk R~300pc.
gas ~ 5 g cm-2
n ~ cm-3
Mgas ~ M
v ~ 100 km s-1
LFIR ~ 21012 L
LX ~ 3109 L
tdyn ~ 106 n4-1/2 yr
300 pc
Solomon, Sakamoto
Armus et al 2006; Brandl et al 2006; Higdon et al 2006
Beswick 2006; Mundell et al; Lonsdale et al

10. Pressures
Accounting:

11. What processes regulate Star Formation in ULIRGs?
The standard lore: Energy injection by supernovae, stellar winds, HII regions (e.g., McKee & Ostriker ‘77). However, in a dense ISM, radiative losses are large: E  n-1/4.
Another Option: Radiation Pressure:
Starburst photons absorbed & scattered by dust: UV ~ 100’s cm2/g.
Dust is collisionally coupled to gas:  ~ 0.01 pc a0.1 n3-1.
Starbursts: optically thick to re-radiated IR : IR ~ gasIR > 1.
Radiative diffusion: efficient coupling to cold, dusty component, most of the mass.
Scoville (2003) Thompson, Quataert, & Murray (2005)

12. Radiation Pressure Supported Starbursts
Radiative flux:
Radiative diffusion:
Radiation pressure:
Obtain Eddington-limited starbursts:

13. Some Predictions The “Schmidt”-law for optically-thick starbursts:
Higher  implies more pressure support, which implies a lower star formation rate & efficiency.
Kennicutt (1998)

14. The Rosseland Mean Opacity
Sublimation: Tsub ~ 1000 K.
Dust dominates T < 1000 K.
At T < 200 K — in the Rayleigh limit — = 0T2.
Overall normalization is dependent on metallicity and the dust-to-gas ratio.
Semenov et al. (2003)

15. Some Predictions The “Schmidt”-law: When = 0T2:
no dependence on anything, but 0.

16. A Characteristic Flux? ULIRGs are compact. Intrinsic size?
Appeal to radio size, hoping that the radio reliably traces the star formation.
Data from Condon et al. (1991)

17. Evidence for a Characteristic Flux?
Davies et al. (2006)

18. Why Radiation Always Wins
Schmidt law:
Flux:
Radiation pressure:
Hydrostatic pressure:
Critical surface density:

19. Magnetic Fields & Cosmic Rays

20. The FIR-Radio Correlation
How do CR electrons cool?
Radio synchrotron from CR e-’s accelerated by SNe.
FIR traces star formation, massive stars, SNe.
“Calorimeter” theory: synchrotron cooling timescale shorter than the escape time:
tsynch < < tescape
(Völk‘89; generally unaccepted)
galaxy = CR beam dump
Starbursts
Star-Forming Galaxies
Yun et al. (2001)

21. Magnetic Fields & Cosmic Rays
In the Milky Way, B~5-10G and
In starburst galaxies, how do we estimate B?
“Minimum energy” (UB~UCR; Burbidge 1956): (~5-10G in MW). Depends on the ratio [p/e] and on the injected CR spectral index.
Magnetic energy density in equipartition with total hydrostatic pressure: (~5-10G in MW)

22. Magnetic Fields Conclusion:
Magnetic fields in star-forming galaxies are both minimum energy & equipartition.
and

23. Magnetic Fields Conclusion: Either
the minimum energy estimate is wrong, or
magnetic fields are dynamically weak in starburst galaxies.
Thompson et al. (2006)

24. Bmin Must Underestimate the True Field
UBmin/Uph measures the importance of synchrotron relative to IC cooling.
If Bmin is correct, IC dominates for starbursts.
This contradicts the linearity of the FIR-radio correlation.
UBmin /Uph

25. Magnetic Fields & FIR-Radio Correlation
In the limit of very strong cooling (the “calorimeter” limit):
The observed Schmidt Law says that
Therefore, in the limit of strong cooling:

26. Magnetic Fields Conclusion:
If a fraction ~1% of 1051 ergs per SN goes to CR electrons, and they cool rapidly, the observed trend is reproduced.
Implies that B is in fact larger than Bmin.
Thompson et al. (2006)

27. Magnetic Fields in Starbursts
Observations thus imply rapid electron cooling.
Strong evidence for the calorimeter theory for the FIR-radio correlation: tcool< < tescape.
So, how big is B?
Well, B is big enough that the synchrotron cooling timescale is << tesc. But, what is tesc?
Very uncertain:
Diffusion in MW tesc ~107.5 yrs Maybe advection (winds!) in starbursts tesc ~105.5 yrs (?).

28. Magnetic Fields in Starbursts
Argument/Problem:
The strongest objection to the calorimeter theory for FIR-radio correlation: if synchrotron dominates cooling and tcool< < tesc, the radio spectral indices of starbursts at GHz should be steep “cooled” : F ~ - , with  ~
This is not observed. Spectral indices at GHz are ~constant & not steep: F ~ - , with  ~ 0.7.
Solution:
If CRs interact with matter at mean density & B~Beq, then
Ionization losses dominate for low-energy CRs, not high. This effect changes the expected slope of the radio spectrum at a characteristic frequency ~GHz.

29. Magnetic Fields in Starbursts
Ionization losses flatten the radio spectra
Ionization is important only if CRs interact with ISM of ~mean density.
Prediction: spectral break ubiquitous at GHz ’s for all galaxies obeying FIR-radio.
Because this only works if B~Beq, this is the best argument for B >> Bmin in starbursts.
Steeper
p=2.5
p=2.0
Flatter

30. Summary Observations indicate Radiation pressure
feedback is important, SF is inefficient, starbursts are dusty, disks have Q~1.
Radiation pressure
can dominate feedback in the optically thick regions of starbursts.
yields qualitative change to Schmidt Law.
couples to the cold dusty component, most of the mass.
predict starburst structure: T, Teff, F, , , v, SFR/area, efficiency
are in good agreement with observations (local & high-z ULIRGs).
Magnetic Fields in Starbursts
are larger than Bmin and probably ~ Beq.
are large enough that the “calorimeter” theory for FIR-radio is preferred.
are consistent with starburst radio spectral indices only if CRs interact with ISM of mean density so that ionization/bremsstrahlung losses are important.
-Ray Observations of Starbursts
will constrain the ISM density seen by CR protons.
will constrain the energetics of CR acceleration. - Lastly, (CRp/CRe) ~ 10.
Thompson et al. (2005), (2006ab)

31. The Present & The Future
Radiation pressure feedback:
Embedded sources, porosity, transport, multi-phase ISM.
The gravitational instability in radiation pressure dominated backgrounds.
Starburst winds, scaling relations: Faber-Jackson, M-.
Other mechanisms for feedback:
HII regions, stellar winds, supernovae, gravity.
The starburst-AGN fueling connection.
The FIR-radio correlation:
Test prediction of spectral breaks at GHz.
Electron calorimetry in normal star-forming galaxies (?).
Starbursts: what is the role of the secondary electron/positrons?
Backgrounds: neutrino (MeV to >TeV), -ray, FIR, & radio.
What is the energy density of cosmic rays in starburst galaxies?

32. The End

33. Constraining the Average Density “Seen” by Cosmic Rays

34. -Rays from Starbursts
Assume SNe accelerate both CR protons & electrons.
The GeV protons collide with ambient gas:
Proton-proton collisions produce
If pp<< esc, then the starburst is a “proton calorimeter,” and all of the proton energy goes into ’s (1/3), e+,-’s (1/6), and ’s (1/2).
What is esc? As for CR electrons, very uncertain.
Thompson et al. (2006)

35. -Rays from Starbursts
Massive star formation  IR emission  Supernovae:
where  is the fraction of 1051 ergs per supernova to CRp’s. This is a FIR--Ray correlation analogous to FIR-radio.
How do we constrain ? Assume the e+,-’s from p-p cool via only synchrotron in the starburst:
Observed FIR-radio correlation:
Thompson et al. (2006)

36. -Rays from Starbursts
NGC 253
Arp 220

37. -Rays from Starbursts
If GLAST sees a larger flux from NGC 253:
Then  >  more energy per SN to CR protons.
Because from secondary electrons/positrons, another process (not synchrotron) must dominate CR electron cooling.
If GLAST sees a smaller flux from NGC 253:
Either the CRs interact ISM below mean density, rapid escape,
or,  < 0.05  less energy per SN to CR protons.
These options can in principle be distinguished by modeling the IC and relativistic bremsstrahlung emission at -ray energies since the latter also depends on density.

38. The Diffuse -Ray Background
Massive star formation  IR emission  Supernovae.
+ star formation rate history of the universe.
+ the fraction of all star formation at high-z that occurs in “proton calorimeters” (high density).
For an individual galaxy:
For the history of star formation:
Thompson et al. (2006)

39. The -Ray Background