Modelling Massive Star Formation Rowan Smith ZAH/ITA University of Heidelberg Ian Bonnell, Henrik Beuther, Paul Clark, Simon Glover, Ralf Klessen, Steven Longmore, Amy Stutz, Rahul Shetty

Motivation 1

Observations: Environment Massive stars usually form at the centre of dense star forming clumps. Pre-stellar massive cores either extremely short lived or don’t exist Motte et. al Star forming clumps form at the hub points of filaments. Peretto et. al. 2012, Myers 2009, Schneider et. al. 2012

Observations: Inflow Peretto et al found the mass in the central pc of a massive IRDC (SDC335) could be doubled in a million years. Kirk et al 2013 found infall gradients of ~ 30 M sol Myr -1 along the southern filament of Serpens South -radial contraction onto the filament at ~ 130 M sol Myr -1

Observations: Fragmentation Interferometry observations usually (but not always) reveal substructure on core size scales i.e. less than 0.1 pc scale. see Bontemps et al. 2012, Rodon et al. 2012, Duart-Cabral et al Girart et al mG Fragmentation with an entrained magnetic field. Palau et al & massive dense ~0.1 pc cores 5 one dominant source, 9 many (>4) sources low fragmentation = stronger magnetic field

The SPH Simulation Loosely based on Orion A M sol Smooth Particle Hydrodynamics 15.5 million particles Barytropic equation of state Sink particles for star formation Heating from sinks Self gravity Decaying turbulence No magnetic fields Equivalent to a massive star forming region. see also Bonnell et al. 2011

Massive Stars and Collapsing Gas 2

Collapsing Clumps Filament collapsing along its axis - evolves to a more compact state with less sub-structure Clump Alpha in column density blue: 0.05 gcm -2 yellow: 5 gcm x 10 5 yrs Rotating massive protostellar core at the centre. But no obvious pre-stellar core at early times.

Interferometry Observations Longmore et al observed clumps of gas where massive stars were thought to be forming. Used maser emission and chemical tracers to estimate their relative ages. YOUNGOLD 0.75 t dyn 1.0 t dyn 1.25 t dyn

Subsequent accretion A guessing game- which one of these cores forms a massive star? The positions at which accreted material passes through a shell of radius r = 0.1 pc around a sink over 20,000 yr.

Subsequent accretion Answer: thermal jeans mass 0.8 M  1.9 M  2.7 M  11.5 M  Smith et al. 2011a

The Core Mass Function Correspondence between cores and stars within the simulated massive cluster is only for the sample as a whole rather than for individual stars. Implies accretion from outside core. Smith et. al. 2009a There is a resemblance between the stellar IMF and core mass function e.g Alves et al and many others

Fate Red = p-cores Solid blue = sinks Yellow = mass which will be accreted by the most massive sink within 2.4 x 10 5 yrs. Massive star is mainly built out of gas that initially comes from the surrounding clump. See also Wang et al t= 1 t dyn Clump Alpha Smith et. al. 2009b This is in contrast to core accretion models for massive star formation e.g. McKee & Tan 2003

Massive Starless Cores Generally massive condensations exhibit some sub-structure consistent with the predictions of these simulations. Caveats: My simulations lack magnetic fields (see Myers et al. 2013) It is important to see what such regions would look like in actual observations. Tan et al Rodon et al. 2012

A Comment 2) Supersonic turbulence is not an isotropic pressure and so it cannot support a core without also inducing fragment in regions that have been compressed. Krumholz et al Competitive Accretion vs. Turbulent Cores -> Probably both wrong 1) What we see in the simulations (Smith+ 2009, Wang+ 2010) is not competitive accretion in the original Bondi-Hoyle sense. The gas and cores are well coupled. It is the global collapse of the cloud that feeds the proto- stars.

Accretion and Filaments 3

Velocity Map Large scale collapse Flow is not purely radial. Multiple filaments form a hub. (see Myers 2011, Smith et. al. 2011, Schneider et al. 2012, Kirk et al. 2013, Perretto et al. 2013/2014)

Subsequent accretion thermal jeans mass 0.8 M  1.9 M  2.7 M  11.5 M 

Irregular Shapes Cores situated in more filamentary enviroments are more massive at the end of the simulation. Low mass sinks tend to form from more spherical cores. Type Number Percentage % % % Smith et. al. 2011a

New Arepo Simulations A slight digression...

New Arepo Simulations Suite of small scale simulations: 10 4 solar mass turbulent clouds Chemistry, gas self-shielding, heating and cooling, self-gravity. Jeans length always refined by at least 16 cells.

Plummer-like Profiles For super-critical non-isothermal filaments, when we fit with a Plummer-like profile as done in Arzoumanian et al Power law profiles are flat p~2 without magnetic fields. No systematic variation in filament properties with initial turbulence type (i.e solenoidal, compressive, mix). now available on the arXiv Smith et al. 2014b submitted

Filament Comparison A11 - Arzoumanian et al J12a - Juvela et al. 2012a Filaments in Planck Cold Cores: Juvela et al. 2012a The simulated filaments are very similar to observed filaments. But no constant filament width of 0.1 pc see also Hennemann et al. 2012

Filament Formation The filaments seen in column density are actually made up of a network of sub-filaments as in Hacar et al The filament forms from smaller clumpy filaments being collected together by gravitational collapse. Sub-filament size consistent with the Jeans radius in 12K n=10 5 cm -3 gas.

Synthetic Observations 5

Observations Fuller et. al Chen et. al see talk by Chang Won Lee yesterday for the low mass case

Massive Star Line Profiles Optically thick line profiles often show a characteristic broad peak with a small red shoulder. HCN F(2-1)HCO + Smith et al Post-process the massive star forming regions with radmc-3d

Line of sight Superposition of large scale collapse motion, with smaller scale local core collapse within the massive star forming region. Supersonic infall as proposed by Motte et. al from observations of Cygnus X. See also Schneider et. al Multiple density peaks (cores) along the line of sight. Linewidths due to collapse not supportive turbulence, rotation, or outflows.

Comparision to Observations Csengeri et. al Red fit from their model. Similar wide profiles with a small shoulder in observations.

Optically Thin Profiles Multiple components in the optically thin lines. This has the potential to be diagnostic. N 2 H + (1-0) isolated hyperfine component observed over 0.06pc HWFM beam Beuther et al. 2013

Optically Thin Profiles This also becomes more apparent when observed with a narrow beam - implications for ALMA N 2 H + (1-0) isolated hyperfine component observed over 0.06pc HWFM beam black = HCO + (1-0) red = N 2 H + (1-0) *4

Future Work 5

Galactic Scale Arepo Simulations Sub-pc resolution study of the formation of molecular gas in a spiral galaxy. Smith et al. 2014a

Galactic Scale ICs These are the ideal initial conditions to revisit previous molecular clouds simulations and make observational predictions for low and high mass star- forming cores.

Conclusions 6

1.Massive stars in these simulations are fed primarily from gas from the clump rather than the core (defined in 3D using the gravitational potential). 2.Filamentary flows can feed massive protostellar cores through gravitational collapse. 3.Filament profiles in new hydro simulations with Arepo have p~2 profiles and a non-constant width. 4.Synthetic observations of the simulated massive star forming regions often have little self absorption. In optically thin dense gas tracers there are multiple line components when observed with a narrow beam. 5.Using Arepo and galactic scale simulations we are now revisiting this problem with more accurate methods and better initial conditions.

Sink Heating Basic fit to MC models Robitaille et. al a: 0.33 M < 10 - a: 1.1 M > 10 - q: -0.4 to -0.5 Overestimates feedback Spherical symmetric Spherical symmetric Isolated Isolated Underestimates column densities Underestimates column densities Ignores cluster structure, discs etc Ignores cluster structure, discs etc

CO emission Filamentary inter-arm clouds may be the observable parts of much larger structures.

Density & Accretion Accreted gas has a lower density and hence a longer free fall time. - needs a long free fall time to reach the central sink without fragmentin on the way. see also Wang et al. 2010

Is a core sufficient? Smith et. al. 2009a There is a resemblance between the stellar IMF and core mass function e.g Alves et al and many others Potential wells in in Smith et al. 2008b resemble the stellar IMF. Note that these are potential wells not observational cores.

Angular momentum Smith et al The angular momentum vector of the material accreted onto the core is not coherent. This will encourage fragmentation in the cores and may change the orientation of jets and outflows over time.

Global Collapse 3

Collapsing Clumps Clump Beta in column density blue: 0.05 gcm -2 yellow: 5 gcm -2 Region formed by converging shocks - evolves to a more compact state with enhanced densities 2.4 x 10 5 yrs

Comparison to Low Mass Cores Compared to low mass cores massive star forming region line profiles are: - Brighter and with larger linewidths - Less variable with viewing angles - Only weak self absorption signatures, broad blue peaks and small red shoulders - Have multiple line components in optically thin dense gas tracers Smith et al. 2013