DISRUPTION OF MOLECULAR CLOUDS BY EXPANSION OF H II REGIONS

Jeong-Gyu Kim (Seoul Nat’l Univ.) Woong-Tae Kim (Seoul Nat’l Univ.)

Eve Ostriker (Princeton Univ.)

Sep 29, 2016

EAMA10 MEETING

Stellar Feedback in Action30 Dor in LMC

Energy output from massive stars removes residual gas around star clusters, leading to cloud dispersal.

X-ray(0.5-8keV)

Hα

IR(8μm)

Destructive Effects of UV Radiation from Hot Stars

• Photoionization (e.g., Matzner 02)

• Loss of fuel for star formation

• Thermal-pressure-driven expansion of H II regions

• Radiation pressure on dust grains

• Important in dense environment (e.g., Krumholz+09)

• Both from ionizing/non-ionizing UV

Why Radiative Feedback?• Protostellar outflows:

- Too weak to destroy massive GMCs (e.g., Matzner02; Nakamura+10)

• Hot gas created by shocked stellar winds: - Leaks out of local environment

(Harper-Clark+09; Rosen+14)

• Supernova explosions: - Begins ~3-4 Myr after the formation of massive stars

(Krumholz+09; Fall+10)

No. 2, 2009 DYNAMICAL MODELS OF THE CARINA NEBULA BUBBLE 1697

Table 1Properties of Tr 16 in The Carina Nebula Using All Stars of Spectral Type B2or Earlier Assuming Standard Mass-Loss Rates (Repolust et al. 2004) (From

Smith 2006a).

Number of stars ! B2 72Total bolometric luminosity 1.7 × 107 L⊙Total ionizing flux (Q) 5.9 × 1050 s−1

Total mass-loss rate Mw 1.1 × 10−3 M⊙ yr−1

Total stellar wind luminosity 6.7 × 104 L⊙

the appendices), we describe the Chevalier & Clegg model, theCastor et al. model, and our modifications to the Castor et al.models. We also describe our numerical methods. In Section 4,we present the results of our modeling, including a parameterstudy of the (modified) Castor et al. bubbles. In Section 5, wediscuss the X-ray emission. In Section 6, we discuss the resultsand compare to previous work. We give our conclusions in thefinal section.

2. OBSERVATIONS OF CARINA

The distance to Carina is 2.3 kpc (Allen & Hillier 1993; Smith2006b). The complex contains about 70 O stars (Smith 2006a),of which 47 (including evolved massive stars) reside in Trumpler16 (Tr 16), a star cluster with a radius r ∼ 3 pc. This makesCarina one of the most massive (and best studied) star formationregions in the Milky Way. The bolometric luminosity emergingfrom the stars is 9.6 × 1040 erg s−1. This is about twice the far-infrared luminosity of the complex (Smith & Brooks 2007a),indicating that half the stellar flux is intercepted by dust grainswithin ∼ 10–20 pc of the stars.

We note that this rather low ratio of infrared to bolometricluminosity is consistent with the fact that we can see Tr 16in the optical, and with turbulent models of the ISM, but it isinconsistent with the assumption of the Castor et al. model thatthe ISM and swept-up bubble shell surrounding the cluster ishomogeneous. It suggests that hot gas can escape the bubble, apoint we return to later.

The number of ionizing photons emitted per second isQ = 9 × 1050 s−1. The kinetic energy of the stellar winds isLw ∼ 3.5 × 1038 erg s−1, assuming standard estimates of themass-loss rates; we argue below that this is likely to be anoverestimate by a factor of 3–10. The massive stars of Tr 16 areresponsible for about 70% of the bolometric and stellar windluminosity and of the ionizing flux of the entire nebula; seeTable 1.

A reasonably reliable estimate for the age of Tr 16 is given bythe age of η Carina, currently the most massive star in the region.η Carina is ∼ 3.6 million years old, since it has clearly evolvedoff the main sequence, while estimates of its initial mass basedon its current luminosity and stellar models (Bressan et al. 1993)are in the range of 120M⊙. We thus assume that the cluster isapproximately 3.6 million years old. It is believed that no SNehave occurred in Tr 16 (Retallack 1983; Lopez & Meaburn 1984;Whiteoak 1994), but in any case the integrated energy depositedby the winds, 4 × 1052 erg, substantially exceeds the energyinput of an SN, of order 1051 erg.

For the 17th anniversary of the Hubble Space Telescope,Smith & Brooks (2007b) took a 50 light-year-wide mosaic ofthe center of the Carina nebula. This image has unprecedenteddetail of globules within the bubble and star birth at the edgeof the bubble. Triggered star formation around the edge of

Figure 1. Cartoons of the Castor et al. bubble (i) and the modified Castor et al.bubble (ii). The Castor et al. bubbles have a four-zone structure. Zone a ishypersonic stellar wind; zone b is shocked stellar wind and evaporated materialfrom the shell; zone c is the bubble shell; and zone d is the ambient ISM. (ii) Thestructure of the modified Castor et al. bubble, showing leaking from the shell.Note the larger size of region a, smaller size of region b, and material escapinginto region d.

the bubble in Carina is also observed in X-ray and near-IRstudies of Carina (Sanchawala et al. 2007). There are manydirectional markers, such as pillars, showing that the bubbleedge is dominated by the effects of Tr 16. The images showthat the main bubble shell is markedly nonspherical. Smallerbubble structures within the main bubble can be seen. Smith& Brooks (2007a) compiled a multiwavelength analysis of theCarina nebula enabling approximate edges to the bubble to belocated. We estimate radii in the range 10–20 pc.

There have been many observations of Carina in X-rays,which establish the existence of diffuse emission from thenebula itself (Seward et al. 1979; Flaccomio 2005; Hamaguchiet al. 2007a, 2007b). The largest area coverage of the Carinaregion is provided by Seward et al. (1979), who used theEinstein Observatory to find a diffuse X-ray emission of 4.8 ×1034erg s−1 for the 0.5–3.0 keV band (scaled from their distanceestimate of 2.6–2.3 kpc).

The pressure inside the bubble surrounding Tr 16 contributesdirectly to the outward force on the surrounding material. Weestimate the pressure in several ways.

First, the pressure can be calculated from the free–free radioemission, e.g., as reported in Huchtmeier & Day (1975), whomeasured a flux of Sν = 1.43×103 Jy. From calculations of theexpected bremsstrahlung radio flux at the observer,

nH ii = 3.6!

R

pc

"−3/2 !D

kpc

"!T

K

"1/4 !Sν

Jy

"1/2

ϵ−1/2H ii cm−3,

(1)where ϵH ii is the volume filling factor of the H ii gas inside thebubble. For the radius of the bubble, R = 10 pc, and distance,D = 2.3 kpc, we find nH ii = 100 cm−3 for TH ii = 104 K. Thus,P = 1.4 × 10−10ϵ

−1/2H ii dynes cm−2.

As a consistency check, we note that the pressure of an Hii region powered by a cluster emitting Q ionizing photons persecond is given by

PH ii =

#3Q

4πr3αrϵH iikbTH ii, (2)

where αr ≈ 4 × 10−13 cm3 s−1 is the recombination coefficient,and TH ii ≈ 104 K. Using Q = 9 × 1050 s−1 and r ≈ 10 pc, wefind PH ii ≈ 2 × 10−10ϵ

−1/2H ii dynes cm−2.

Second, we can use the observed X-ray flux to estimatethe pressure in the hot (Tx ∼ 107 K) gas. From Figure 1 in

Environments of Star Formation

more compact deeper grav. potential well

(see also Tan et al. 2014)

Environments of Star Formation

Data from Roman-Duval+10 Heyer09

GMCs

in the MW

Environments of Star Formation

super star clusters

GMCs

in the MW

Importance of Radiation Pressure/Photoionization Feedback (1D Spherical Model)

Kim, Kim, & Ostriker (2016)

radiation pressure dominated

photoionization dominated

Key Questions• Can H II region feedback still be effective in realistic

turbulent clouds? • What is the net efficiency and duration of star formation? • In what environment radiation pressure/photoionization more

important?

Dale et al. 2012, 2013240 J. E. Dale, B. Ercolano and I. A. Bonnell

Figure 7. Gallery of final states of clusters, as shown by column-density maps observed down the z-axis. White dots represent sink particles (individual starsin Runs UP, UQ and UF, clusters otherwise) and are not to scale. Note the different physical sizes and the different column-density scales.

densities are much less marked in the simulations presented here,implying that there must be some other effect at work. The aboveargument is simple and strictly only applies when the backgroundcloud is stationary.

In the turbulent clouds studied in this series of papers, the turbu-lent velocity field also likely plays a role. As was remarked earlier,

the bound clouds from Paper I are all quite close to virial equilib-rium and, if anything, are slowly contracting. At any given pointin these clouds, the turbulent flows are travelling in random di-rections but, averaged over a sufficiently large volume, the flowshave divergences close to zero or slightly negative. The unboundclouds presented here, by contrast, are all expanding, so that their

at Seoul National U

niversity on June 25, 2015http://m

nras.oxfordjournals.org/D

ownloaded from

Photoionisation only

– 12 –

Fig. 2.— Snapshots from the fiducial model evolution. We show column densities in the y-z

plane and all star particles projected onto the y-z plane, calculated at times t2, t10, t50 and t90,

when the total stellar mass is 2%, 10%, 50% and 90% of the final value. This corresponds to

t/t↵,0 = 0.43, 0.59, 1.06 and 1.57 as shown, where t↵,0 = 4.29 Myr is the initial free-fall time in the

cloud. The color scale for the gas column density (top) is in units of M� pc�2 and the color scale

for the particle mass (bottom) is in units of M�. The box size is 60 pc, 4 times the initial cloud

radius.

Radiation pressure onlyRaskutti et al. 2016

Numerical Simulations• Disruption of turbulent molecular clouds by UV radiation

feedback from a young star cluster

- Run models with differing cloud mass M, surface density Σ, and feedback

- No feedback - Photoionization - Radiation pressure - Photoionization + radiation

pressure

- Grid-based code Athena (Stone et al. 2008) - Initially uniform density sphere - Turbulent velocity fields with power

spectrum |δv|∝k-4, with EK=|EG|

- Star formation via sink particle method - Sink particles also becomes radiation

sources - Point source radiative transfer using

adaptive ray tracing - Constant light-to-mass ratio Ψ=943L⦿/M⦿, Li/Ln=1.25

- Gas temperature: Tneu=10 K, Tion=8000 K

Long Characteristics Methods for Radiative Transfer

• Solve the transfer equation along rays cast from multiple point sources: n ·rI⌫ = �I⌫

Abel & Wandelt (2002)

• Adaptive ray trace • Rays are split when angular resolution becomes coarse.

•

Fiducial model: ⌃cl ⇠ 70M� pc�2

t = 1.73 t↵

30pc

• Striking resemblance to observed nebula! • UV radiation sculpting pillars or elephant trunks

Evolution of key characteristics

• Net star formation efficiency of ~10% • Cloud disruption occurs within ~1-2 free-fall times!

Mstar

Moutflow,ion

Moutflow

Mgas

⟨xe⟩

Mion

Relative Importance of Feedback Mechanisms

• Radiation-pressure-only feedback model requires more star formation for disruption

• Photoionization (photoevaporation + thermal pressure) is an efficient feedback mechanism for destroying typical GMCs in the Milky Way.

Fiducial model: Mcl = 5⇥ 104 M� Rcl = 15pc⌃cl ⇠ 70M� pc�2

• Disruption requires higher SF efficiency than the fiducial model.

• Radiation pressure more efficient in driving outflows, consistent with the results of 1D models.

Relative Importance of Feedback MechanismsMassive, high-Σ cloud: Mcl = 1⇥ 106 M� Rcl = 45pc⌃cl ⇠ 160M� pc�2

Comparison with 1D Model

• Dividing line roughly consistent with the results of the 1D model

rad pressure dominated

photoionization dominated

Summary• Disruption of turbulent molecular clouds by UV

radiation feedback from massive stars • Expanding H II regions can be a primary feedback

mechanism for cloud disruption • Typical GMCs ( )

• photoionization dominated • Net SF efficiency of ~10%

• Dense, massive clouds • radiation pressure dominated • Tends to have higher SF efficiency

• Complete disruption occurs within 1-2 free-fall time

M . 105 M� ⌃ . 102 M� pc�2