Note: I did not attend the Monday and Tuesday talks. Here is the program with all of the videos and slides (if provided).

Wednesday, April 16th

“Star formation in self-gravitating turbulent fluids” – Murray (CITA)

Collapse models: originally, the turbulent motions were time-independent; however the turbulent flows should be dynamical!
Turbulent pressure is always significant during the collapse.
(Robertson & Goldreich 2012) Controlled turbulent collapse in cosmology.
Solving the original equations from Shu but with a dynamical turbulent velocity. Infall occurs but at a longer timescale than Keplerian because of pressure support.
The density profile asymptotes to a stable attractor inside the sphere of influence of the star.
The general result from this model is that the stellar mass scales as t².
Outside the sphere of influence, pressure gradient must balance gravity: p ~ 0.2 in Larson’s equations, i.e. v_r ~ r^p; slow dependence on radius. Density is nearly proportional to 1/r^1.5.
The application of this 1-D model is a sub-grid model. If a simulation resolves a GMC, then the infall velocities and SFR for the GMC can be calculated.
Self-gravity in turbulent simulations is crucial to calculate the SFR (obvious)

“The life-cycle of molecular clouds in Milky Way-like disk galaxies” – Walch (Köln)

Series of papers: Peters+ (2010); Walch+ (2011, 2012). Multi-scale simulations of GMCs + RT in isolated disk galaxies.
Considering a metallicity-dependent cooling (collaborating with S. Glover).
Including many different heating processes (photo-electric dust emission; LW radiation; interstellar radiation field; X-ray ionization heating; Cosmic ray ionization heating; kinematic effects)
Uses HEALPix Level 1 (48 pixels) to calculate the column density for each cell for self-shielding. See TreeCol (Clark+ 2012).
Switching off cooling (even for a short time!) results in an over-injection of momentum by a factor of 6-8.
Ionizing radiation enhancing the momentum input by a factor of 2.
Developed a method of a mixed input of thermal energy and momentum for SN explosions, depending on the ambient density.
Studied the SN input positions inside turbulent box simulations: at random positions (migration) and at peak densities. The former results in a hot ISM, where little rarefied gas exists in the peak placement SNe.
Stratified disk simulations with this model: random position model drives cold outflows from the disk; peak position model drives no outflows, and there is a large GMC forming in the simulation because the SN energy cannot halt the collapse.

“Feedback and accretion in star cluster formation” – Matzner (Toronto)

What are the relative roles of feedback and accretion in star cluster formation?
Focus on the Serpens South protocluster that has filamentary structures and a central star cluster.
Self-similar virialized growth? SFE, SFR_ff, f_gas vary slowly.
Accretion-driven turbulence. Can also drive turbulence by expanding and colliding HII regions.
Feedback vs. inflow: Outflows precede and outnumber massive stellar feedback, coincide in time with star cluster formation, emit far more momentum than starlight, highly collimated by B-fields and exert forces at wide angles (Matzner & McKee 1999), couple well with the dense ISM at r < 0.1 pc.
Critical parameter: mean outflow momentum per stellar mass.
Accretion-driven turbulence should be strongly affected by the presence of accretion.
Outflow driving may not be strong enough on its own to sustain turbulence.

“Formation and Evolution of Star Clusters” – Fall (STScI)

Focusing on cluster mass functions in different galaxies (MW, LMC, SMC, M83, M51, Antennae). All have dN/dM ~ M^-2, regardless of the cluster ages.
The clump mass function has a power-law slope of approximately -1.7.
Why do the MFs of old (globular) clusters have such different (non-power-law) shapes than those of young clusters?
Why do the MFs of young clusters of different ages have nearly the same shape?
Stellar dynamical distruption: stellar mass loss through tidal limitation (10-100 Myr), tidal disturbances by MCs (>100 Myr), stellar dynamical evaporation (>1 Gyr).
For evaporation models, see Baumgardt & Makino (2003). There should be newer references, though.

Thursday, April 17th

“Star formation self-regulation: concepts and computations” – Ostriker (Princeton)

From some simple arguments on momentum input, the SFR is inversely proportional to the amount of feedback (p_*/m_*).
(Ostriker & Shetty 2011, 2012) SFR surface density is independent of turbulent and SF efficiency (ε_ff) on small scales. ε_ff ~ 0.5-1%. Used p_*/m_* ~ 3000 km/s.
SN-driven turbulence is insensitive to the density, which makes the velocity dispersion weakly dependent on SFR.
Self-regulation when the dynamical crossing (vertical) time of the disk is much greater than free-fall time.
(Kim, Ostriker, & Kim 2013) Simulation results. Interesting to inspect SFR surface density vs. equilibrium pressure.
Radiative feedback: direct radiation, reprocessed radiation (overall ISM & individual cluster-forming cloud), where the IR optical depth maxes out theoretically around 10-20 in the most intense starburst galaxies, e.g. Arp 220. Most likely a factor of a few.
(Skinner & Ostriker 2014) Cluster-forming simulation with IR radiation. Athena-RT. Inspecting whether the cloud is supported by radiation pressure and the radiation trapping fraction. For kappa = 4, 20, 40, SF efficiency is large at 0.8, 0.6, 0.5, respectively. Eddington factors (kappa = 20) is roughly 0.7-0.8 for all radii. The optical depth is around 5, but the trapping factor is a factor of 2-3 lower.

“Regulating star formation with turbulence at high redshift” – Slyz (Oxford)

“NUT” re-simulations of a 5 x 10^11 solar mass halo at z=0. DM particle mass of 50,000 solar masses with star particle masses of ~10^4 solar masses. Analyzing the data at z = 3.
How to choose the SF regions in simulations? Overdense (as usual) or self-gravitating systems (Hopkins+ 2013). The latter recipe results in lower SFRs by a factor of 10 with feedback.
The density distributions with the self-grav. SF recipe is very unstructured, even without feedback, whereas the density threshold method is very disk-like.
Higher gas surface densities in the self-gravitating method.
Rotation curves: Density threshold with momentum-driven feedback results in a flat curve, compared to peaked ones with energy-driven feedback and no feedback. The self-grav. method results in a very clumpy and turbulent ISM, and correspondingly, the rotation curve is not smooth.
JHW comment: The clumps will probably form super star clusters, which might evolve into globular clusters.

“Star formation, stellar feedback, and their effects on the circum-galactic medium” – Faucher-Giguère (NWU)

FIRE Simulations. On what scale is SF regulated? Star-forming clouds (ε_ff ~ 0.01) with supersonic turbulence or galaxy-scale balance with feedback-regulated models?
Galaxy scale: SF regulation through global hydrostatic balance.
In addition to the model presented by Ostriker, they imposed the restriction of Q ~ 1, i.e. ε_ff scales with f_gas. Finding that ε_ff is higher than the supersonic models.
Observation: 1/3 of the MW’s SF occurs in 33 GMCs (perhaps see Rahman & Murray 2010). For gas content in z ~ 2 galaxies, see Genzel+ (2011).
Proposing that GMC formation is the main limiting factor for SF.
FIRE simulations track 11 elements and r-process tracer.
Directly constrain the baryon cycle (inflow/outflow) regulating galaxy growth through absorption systems. Studying a sample of 12 LBGs at z = 2-3, finding time-variability in the absorption systems because of the burstiness of the SF.
Metallicity distribution of absorption systems show a bimodality (Lehner+ 2013; Wotta+ 2014), suggesting that these trace inflows (metal-poor) and outflows (metal-rich).
M. Fall comment: Careful about comparing with QSO absorption systems because they are very clumpy.

“The impact of feedback on the ISM” – Dobbs (Exeter)

Isolated galaxy simulations – investigating different resolutions and feedback prescriptions.
ε = 1%: feedback doesn’t disrupt clouds; 5%: clear spiral arms; 10%: feedback dominates structure
Cloud-cloud collisions occur on the order of 10 Myr. However, this could depend on the definition of a cloud (see Tasker & Tan 2009).
Shear is important for the largest complexes, but feedback is more effective on regulating SF in smaller clouds.
Zoom-in simulations of a molecular complex inside a global disk simulation. Again, varying formation and feedback prescriptions. The feedback has a larger impact on the vertical gas distribution and the amount of hot gas.

Lunch break

“Feedback Effects in High-z Galaxies” – Dekel (Hebrew)

Cold flows to feed z ~ 2 galaxies.
(Danovich+ 2012, 2014) Angular momentum build-up in 4 phases: (1) cosmic web, (2) AM transport to small radii, (3) inner halo – outer tilted ring, (4) inner bulge.
Finding disk with a tilted outer ring in their simulations.
(Ceverino+ 2014) Including photo-heating and photo-ionization and radiation pressure, but the latter is only input locally in dense gas. Why isn’t this coupled with the radiative transfer?
Stronger feedback allows for the cold flows to penetrate farther into the galaxy.
(Dekel & Mandelker 2014) Bathtub Toy Model: sSFR is independent of SF efficiency and mass loading factor, if ignoring any stellar accretion.
At z ~2, the toy model cannot match observations because there is stellar accretion. Can there be some bias in the observations because they only probe the strongly star-forming galaxies?
Violent disk instabilities: caused by high accretion rates and higher mean densities. Creates giant clumps and transient perturbations with mass scales ~10^9 solar masses. The clumps migrate inwards because of torques, encounters, outflows, and dynamical friction. Migration timescale is around 8 dynamical times.
(Dekel & Krumholz 2013) Most relevant feedback processes considered. Ψ_w = 2.5, η = 1-2. For η < 4, the gas fraction is constant.
Radiative feedback (approximately modeled) results in the disruption of small clumps, and the disk expands.
Inflow down a potential gradient provides the energy for driving the dispersion to Q ~ 1, compensating for the turbulent decay (e.g. Krumholz & Burkert 2010; Bournaud+ 2011). Compact stellar systems (aka compactification) expected in gas-rich or low-spin systems.
In a merger, SFR timescale is constant if DM-dominated and is M_gas^-1/2 (? … double-check – slide change happened before I typed this) if gas-dominated.
Two modes of evolution: First the gas-rich inflow makes the system more dense and at a constant high sSFR (~10), and then the gas is ejected and sSFR decrease, leaving the stellar mass nearly constant.

“Stellar feedback experiments in a galactic disk with HD and RHD” – Rosdahl (Leiden)

Note: Overcooling problem (Navarro & Benz 1991). I’ve been using Katz’s paper.
Feedback implementation experiments with isolated disk galaxies, using RAMSES. Stellar particle mass of 600 solar masses. Schmidt law for SFR.
Stochastic thermal feedback (Dalla Vecchia & Schaye 2012): Thermal dump: too much SF when compared to KS relation. With the stochastic feedback, their results cross the KS relation but the slope is too steep. Can match relation with low SF efficiency of 0.2%, but this is also accomplished without feedback.
Kinetic feedback (Dubois+ 2008): bubble radius of 150pc, experiment with different mass loading factors between 1 and 10. With η=1, Matches the KS relation at ~30 Msun/pc^2, but the slope is still steep. With η=10, better match with a turnover below ~10 Msun/pc^2.
Delayed cooling: Matches at >20 Msun/pc^2. Implemented with a tracer field that represented an unresolved turbulent medium that expands. Once it drops below 100 or 200 km/s, cooling is turned back on.
Ramses-RT: IR can indefinitely scatter, so multiply the radiation pressure by the IR optical depth.
With the inclusion of RT, spiral arms are created. No more clumps! SN + RT feedback creates a constant SFR instead of a suppression, then burst (from the clump formation). Almost matches the KS relation at >50 Msun/pc^2.
Without IR radiation pressure, increases SFR by 10-20%. Without any radiation pressure, increases SFR by ~30-50%. No effects on KS relation.
Main effect of RT is heating, resulting in lower density peaks.

“Cosmic-ray driven winds from high-redshift disk galaxies” – Hanasz (Copernicus)

Cosmic rays are in average equipartition with turbulent and magnetic energy densities. 10% of SN energy goes into the acceleration of CRs.
Cosmic rays escape from the galaxy or star-forming regions from buoyant flows (Longair 1966, 1967).
CR-driven wind models: e.g. Hanasz+ (2013), Salem & Bryan (2013)
Utilizing PIERNIK MHD code. Diffusive transport of CRs. Need the magnetic field at cell faces. The CR diffusion-advection timestep is determined by the parallel and perpendicular components of the diffusion tensor.
Injecting B-field dipoles at small scales. Results in a uniform magnetic field after ~4 Gyr.
Cosmic ray distribution is very smooth and perpendicularly propagates out of the disk. In some SF regions, vertical outflows up to 2300 km/s (!) with mass loading factors ~ 1 are created. Results in a total outflow flux of ~10 Msun/yr at a height of 30-40 kpc.

“Ray-Tracing and Flux-Limited-Diffusion for simulating Stellar Radiation Feedback” – Kuiper (MPIA)

On Tuesday, S. Davis compared the driving of outflows between FLD and variable Eddington tensor methods. In this talk, FLD will be compared to ray tracing.
(Kuiper & Klessen 2013) Calculating radiation field from stars with ray tracing, and thermal dust re-emission with FLD. Comparing RADMC (reference code), Hybrid method, and FLD.
In the optically thin case, the hybrid scheme is accurate to 3%. FLD gives the incorrect temperature slope, which is mainly caused by the grey approximation.
In the optically thick case, the hybrid scheme is only accurate up to 46%, which is similar to the MC calculation, and FLD misses the shadowing effects.
Radiation pressure cavities in massive star formation simulations are stable. Massive SF do not form through Rayleigh-Taylor instabilities. With the RT or hybrid scheme, the radiation pressures are 1-2 orders of magnitude higher than FLD.
(Kuiper+ 2011) Science application: massive star formation. Eddington ratio of ~75. Simulation stopped when M_star = 26 Msun. In their 2010 paper, they experimented with more massive cores (up to 480 Msun within 0.1 pc; JHW comment: for Pop III stars, typical values are around 1000 Msun), and they form stars up to 140 solar masses.

Friday, April 18th

Note: I missed the first half of the first talk, “Why does the IMF appear so universal?” by Hennebelle (CEA-Sacley).

“Radiation Feedback and the Origin of the IMF” – Krumholz (UCSC)

Historical blurb about the arguments between Eddington & Jeans in 1926 about Eddington not giving credit to Jeans. Pretty interesting.
Don’t use H-alpha / FUV and H-alpha equivalent width as proxies for the upper IMF. Dwarfs are H-alpha deficient, but it’s not a good measure of the IMF.
(Dobbs+ 2005) Fragmentation simulation with a Jeans mass of 1 Msun, but it fragmented into 100 Msun clumps. The reason is that the gas isn’t isothermal. See Krumholz+ (2006) for a protostellar model. Useful for our simulations? Also see Stacy+ (2012), Safranek-Shrader (2014) and references therein.
Understanding the IMF peak: What density to use in the Jeans mass? Furthermore, isothermal MHD equations give a scale-free system, which cannot explain a peak.
(Martel+ 2006) Problem of non-convergence of the IMF in simulations. Isothermality at fault. EOSs are a bad fit to the ISM (Offner+ 2009; Glover & Clark 2012).
(Krumholz 2011) Simple model of a collapsing cloud with a central radiation source. From using a polytrope, he finds that the peak is 0.15 [ (P/k_b) / 10^6 ] Msun.
(Krumholz+ 2012) Comparing this model to a simulation. Recovers the IMF and binarity fraction as a function of stellar mass.
(Myers+ 2012 or 2013) SF simulations at different metallicities. All the same above 5% solar, but things get interesting at lower Z.

“Towards Self-consistent Modeling of Star Formation and the Initial Mass Function” – Nordlund (Copenhagen)

Resolve all the things!
Philosophy: Use larger scales for the ICs / BCs of small-scale simulations. Basically, zoom-in simulations.
Zooming into solar-mass stars (arxiv/1309.2278) Proof of concept (RAMSES). 40 pc simulation box with 0.01 AU resolution, forming 4 solar-mass stars. First, drive the turbulence. Then turn off driving and let massive stars form and feedback that drives turbulence. Finally inspect how solar-mass stars form.
Outer parts: disk wind speeds ~ 10 km/s, driven by inclined B-fields. Inner parts: highly colimated jet with outflow velocities ~100 km/s. Not necessary to resolve the stellar surface.
Rapid accretion (angular momentum transport) relies on large scale B-fields instead of MRI. Magnetic braking isn’t catastrophic but is chaotic.
For a realistic IMF, need MHD supersonic turbulence, Respecting Larson’s relations. But what’s not needed is Non-isothermal EOS, radiative transfer, and local feedback.

“Towards constraining the IMF of Pop III stars” – Stacy (Berkeley)

Note: Beginning with a primer on Pop III stars (not writing this down because I’m up-to-date unlike present-day SF!).
Pop III SF without feedback: Stacy+ (2010). Fragmentation.
With radiative feedback (Stacy+ 2012): Ionizing and LW radiation. Binary of 20 and 10 Msun companions.
Inspecting ten different Pop III star-forming halos to get a bigger statistically sample with the same physics as their 2012 paper. They find that the total stellar mass in the halos range between 20 and 120 Msun. 36% binarity fraction. Find an IMF slope of -0.17 after 5 kyr of protostellar evolution. Around half of the stars are ejected from the accretion disk.
(Stacy & Bromm 2014) Low-mass Pop III SF mode. Central SF’ing gas has a relatively high angular momentum, only resulting in a dM_*/dt ~ 2 x 10^-3 Msun/yr. Projected stellar masses between 1-5 solar masses.
Indirect evidence and signatures for Pop III: low or zero metallicity Lyman-limit systems and DLAs at z > 3 (Simcoe+ 2012; Fumagalli+ 2011).