Curt Struck

Iowa State Univ.

The Recurrent Nature of Central Starbursts

*** Starburst ‘04 - Cambridge ***

Intro.: Some Questions.

• Are central bursts (r ≤ 1 kpc) inherently recurrent?

This fundamental question is closely related to others…

• What turns bursts off?

• What turns bursts on in isolated, non-barred galaxies?

Is it always inflow or compression?

• When does turning them off induce a recovery?

• Are bursts part of a deterministic process or stochastic?

Caveat: there are many different kinds of bursts: ULIRG bursts, bursts in dwarfs, bar and interaction driven bursts, bursts + AGN, etc.

Core of Arp 284, HST image from Lancon et al. 2001

Evidence for Recurring Bursts

…on timescales of a few times 107 - 108 yrs. comes from several directions.

• Mixed populations in galaxy cores, especially in interacting galaxies. E.g., Bendo & Joseph (2004, ISO), Boeker et al. (2004, Sp survey), Mayya et al. (2004, Op/IR)

• Detailed models for specific nearby galaxies suggest multiple, discrete bursts in some cases, e.g. Arp 284, Lancon et al. (2001), a largely unobscured burst (Brandl et al. 2004, Spitzer IRS).

• Discrete star cluster ages, e.g.,

NGC 3077 (Harris et al. 2004).

1. They run out of gas.

30 yrs. of color and spectral models suggest that most burst populations are much younger than the gas consumption times. Though eventually the gas is consumed or lost.

2. They blow the gas away in a wind.

• Wind mass fluxes comparable to the SFR; so time scale ≈ consumption time.

• Not to mention post-starburst galaxies with gas-rich cores, including:

- N5195 (M51 comp.), Kohno et al. (2002), PASJ 54, 541

- N4736, N5055 (but not N7331) Tosaki & Shioya (1998), IAU Symp. 184, 247

- N7715 (Arp 284b) Smith, Struck, & Pogge 1997

How do Starbursts End? Global Conceptual Ideas

3. Maybe burstiness is an illusion, e.g., noisy but continuous SF..

Unlikely in closest, best-studied systems.

4. The gas is all converted to warm/hot phases.

and the molecular clouds are all fried. Except we see many phases including the molecular, so a less extreme version of this is needed.

5. The central disk inflates. (As in dwarf galaxy models.)

- Much gas doesn’t escape, nor go into hot phase.

- Making it much harder for cloud collisions or grav. instabilities to regrow the clouds, though eventually cools and settles back down.

- Should be able to detect, and probably have, more later.

From S. Galactic Plane Survey, Gaensler, et al., (2001)

What Do Models Tell Us?(about evolution of bursts & winds)

• A few influential models for parts of the phenomena:

- For molecular obs. - PDR models have proven useful, e.g., Wolfire, Tielens, & Hollenbach (1990). Bursts make nucleus a PDR.

- For SB winds - steady, adiabatic outflows as per Chevalier & Clegg (1984). Or like fountain flows of Field & Shapiro 1976.

• Silich, Tenorio-Tagle, & Rodriguez-Gonzalez (2004) argue that cooling is very important in winds, can quench less energetic ones.

• Also semi-analytic models of Tutukov and collaborators on various outcomes in the battle between injected turbulence and gravity in nuclei (e.g., Firmani & Tutukov 1994, Tutukov & Kruegel 1995).

• …and much work on dynamic models of dwarf galaxies…

Numerical Hydro. Models of Galaxy Cores

• To test these ideas - exploratory, 3-d, SPH+N-body models, with Couchman, Pierce & Thomas (1995) Hydra 3.0 code, including optically thin cooling, and SF feedbacks (T threshold, finite heating time to max T).

(Also see Ph.D. thesis (2001) of D. C. Smith for similar studies of disk SF.)

Brief Results:

• With good gas supply and time delays models are intrinsically bursty at a low level (like the old cloud fluid models of Scalo & Struck-

Marcell 1986). Big bursts occur only occasionally (also Bergvall, this conference).

See Fig.

• Feedbacks introduce much turbulence and the central disk inflates in the vertical direction (like option 5 above). Figs &, Movie show effects of (double) burst.

These models are very simple, but they provide important clues.

Some Technical Notes

Number of Particles: 10,000 dark matter, 9550 stars, 9550gas particles. Gas particle mass: 6 x 106 Mo, Total galaxy mass: 1.75 x 1011 Mo. Cooling: Standard diffuse ISM cooling curve (Sutherland & Dopita).

Feedback Heating: Temperature (< 10,000K) and density (> 0.14 cm-3) thresholds. Once these thresholds are exceeded they must continue to be exceeded for a time of 107 yrs. before feedback heating is turned on. Then the heating is maintained for a time of also about 107 yr. Once heating is initiated the internal energy of the particle is increased by 10% each timestep (of about 105 yr.) until a maximum temperature of about 106 K is reached. With the given particle mass, this corresponds to and energy of 1.1 x 1054 ergs per star-forming particle.

If we assume a typical supernova injects about 1051 ergs into the gas, then we require about 1100 SN to generate the needed energy. If we assume that the mass of a typical SN is about 10 Mo, and adopt a Salpeter mass function for a mass range 0.2 - 100 Mo, so that about 10% of the stars are SN progenitors, then we need about to form a star cluster of mass ≈ 105 Mo to provide the needed energy. This is about 2% of the gas particle mass. Note: there is gas is not consumed in these models; losses are assumed to be resupplied (slow feeding).

These estimates do not include energy losses due to cooling. In the numerical model, the mean gas density is so low, that these losses are not important. In reality, the cooling time would be shorter than the adopted heating time at gas densities greater than about 0.1 cm-3, so cooling could be large, and the number of stars required to produce the adopted heating would be much greater than the above estimate. However, the feedback heating used here should not be equated with thermal energy. It is more an algorithm for inducing both increased thermal energy and particle kinetic energy and mass motions via local pressure effects. In the simulations, once the heating is turned off the affected particles cool rapidly due to both adiabatic expansion and radiative cooling.The pressure generated by the feedback and its affects on surrounding particles seems generally realistic, even if the thermal details are not. These ‘details’, however, are needed determine quantiites like the mass fractions and scale heights of different thermal phases.

QuickTime™ and aGIF decompressor

are needed to see this picture.

Green particles are hot (generally heated by recent SF. Note the 2 burst peaks, flocculent spirals, and large holes.

Can We Assemble the Pieces? (Conceptual & numerical)

• Simple onion skin models (hot bubble inside warm PDR inside…) probably not too good.

• Models & observations suggest that there are many ways feedback from young clusters generates turbulence in the surrounding ISM:

- magneto-acoustic shocks

- photo-heating, both near and farther

- accleration of natal cloud shreds

- partial breakout within disk, along ‘soft’ paths

- fountain fallback on many scales

Turbulence is strong in the gas that is not blown away.

• There are a number of lines of evidence that the gas in SB cores is stirred and excited, but not totally fried and ejected (though it’s often hard to tell how many layers we’re looking through!)

- Large HI scale heights (M82, NGC 2403)

- Hot and large scale height dust (e.g., NGC 891, Howk & Savage)

- Declining HCN/CO intensities with SB age (Gao & Solomon 2004)

- In M82 (and others) H emission seems to require interaction of hot wind with ambient clouds (e.g., Chevalier & Clegg 1985).

- Hot, X-ray gas filling factor is small, < 20%,

(see Strickland 2004 (I.A.U. Symp. 222).

NOAO/WIYN image, Westmoquette et al.

A mosaic model incorporating many of the ideas & numerical results above, can explain these facts. Some features include:

- Clusters form in dense regions, not throughout. They heat their immediate surroundings, create ‘hot spots’ and often break out as wind gusts.

- Hot spots also drive turbulence over a wider area.

- Turbulent pressure propagates SF to new hot spots --> more gusts that can add up to a wind (or not in less gas rich nuclei).

- Later, central regions are so turbulently stirred, shredded, and puffed up that SF crashes.

- If not too much gas is consumed or blown out, clouds eventually reform --> recurrent bursts.

This mosaic picture suggests answers to many of the questions posed in the Intro. E.g., what turns off non-driven SBs?

Turbulent shredding and pressure-driven vertical inflation.

Two Winds

• In a mosaic model, hot winds and superwinds are the sum of hot spot gusts. Maybe still describable as Chevalier/Clegg adiabatic outflows (if sound speed > escape speed).

• In the limit of pervasive injected turbulence, the puffing up of the multiphase (cool/warm) central disk, may be viewed as a dense, slow wind. A transient wind of atomic clumps, molecular shreds, dust cirrus, etc. (Like the fast/slow winds of pre-planetary nebulae.)

Burst’s End

• When the burst ends, cooling will be rapid in the slow, dense ‘wind.’

• Recurrence time will be determined by fallback and cloud re-assembly times. For fully developed bursts the fallback time is of order 108 yr.

• With cooling, fallback and deflation, gravitational instabilities are very likely. Transient spirals or bars appear to be common in the numerical models.

• These determine the location of a new generation of hot spots, and a new round of activity. (If the turbulence is not pervasive the hot spot activity may continue sporatically and stochastically, and the wind may never grow beyond gusts.)

Conclusions/Conjectures• Starbursts are naturally recurrent. Internally driven by

feedback/relaxation, but…

- recurrence is very ‘noisy’ (turbulence).

- many bursts ignite prematurely and fizzle as hot spots. There are often several recurrence times between strong internally driven bursts.

- Bursts in isolated nuclei damped by consumption and loss.

• Small scale spirals and bars are common & transient. Acoustic as much as gravitational. A symptom of the burst cycle, like vertical inflation.

• Force fed nuclei can be ‘on’ continuously, but this should be seen in spectral synthesis. Driven SBs can be very strong. Interaction between driving & internal dynamics is complex.