The Earliest Stages of Star and Planet Formation:Core Collapse, and the Formation of Disks and Outflows

Zhi-Yun LiUniversity of Virginia

Robi BanerjeeUniversitat Hamburg

Ralph E. PudritzMcMaster University

Jes K. JørgensenCopenhagen University

Hsien Shang and Ruben KrasnopolskyAcademia Sinica

Anaelle MauryHarvard-Smithsonian Center for Astrophysics

The formation of stars and planets are connected through disks. Our theoretical understand-ing of disk formation has undergone drastic changes in recent years,and we are on the brink ofa revolution in disk observation enabled by ALMA. Large rotationally supported circumstellardisks, although common around more evolved young stellar objects, arerarely detected duringthe earliest, “Class 0” phase; a few excellent candidates have been discovered recently aroundboth low- and high-mass protostars though. In this early phase, prominent outflows areubiquitously observed; they are expected to be associated with at least small magnetized disks.Whether the paucity of large Keplerian disks is due to observational challenges or intrinsicallydifferent properties of the youngest disks is unclear. In this review wefocus on the observationsand theory of the formation of early disks and outflows, and their connections with the firstphases of planet formation. Disk formation — once thought to be a simple consequence ofthe conservation of angular momentum during hydrodynamic core collapse — is far moresubtle in magnetized gas. In this case, the rotation can be strongly magnetically braked.Indeed, both analytic arguments and numerical simulations have shown that disk formation issuppressed in the strict ideal magnetohydrodynamic (MHD) limit for the observed level of coremagnetization. We review what is known about this “magnetic braking catastrophe”, possibleways to resolve it, and the current status of early disk observations. Possible resolutions includenon-ideal MHD effects (ambipolar diffusion, Ohmic dissipation and Hall effect), magneticinterchange instability in the inner part of protostellar accretion flow, turbulence, misalignmentbetween the magnetic field and rotation axis, and depletion of the slowly rotatingenvelope byoutflow stripping or accretion. Outflows are also intimately linked to disk formation; they area natural product of magnetic fields and rotation and are important signposts of star formation.We review new developments on early outflow generation since PPV. The properties of earlydisks and outflows are a key component of planet formation in its early stages and we reviewthese major connections.

1. Overview

This review focuses on the earliest stages of star andplanet formation, with an emphasis on the origins of earlydisks and outflows, and conditions that characterize the ear-liest stages of planet formation.

The importance of disks is obvious. They are the

birthplace of planets, including those in our solar sys-tem. Nearly 1000 exoplanets have been discovered to date(http://exoplanet.eu; see the chapters by Chabrier et al. andHelled et al. in this volume). The prevalence of planets in-dicates that disks must be common at least at some point intime around Sun-like stars. Observations show that this is

1

indeed the case.Direct evidence for circumstellar disks around young

stellar objects first came from the HST observations ofthe so-called Orion “proplyds” (O’Dell & Wen 1992; Mc-Caughrean & O’Dell 1996), where the disks are seen in sil-houette against the bright background. More recently, withthe advent of millimeter and sub-millimeter arrays, there isnow clear evidence from molecular line observations thatsome protoplanetary disks have Keplerian velocity fields,indicating rotational support (e.g., ALMA observations ofTW Hydra; see§ 2 on observations of early disks). Indirectevidence, such as protostellar outflows and infrared excessin spectral energy distribution, indicates that the majority, ifnot all, low-mass, Sun-like stars pass through a stage withdisks, in agreement with the common occurrence of exo-planets.

Theoretically, disk formation — once thought to be atrivial consequence of the conservation of angular momen-tum during hydrodynamic core collapse — is far more sub-tle in magnetized gas. In the latter case, the rotation canbe strongly magnetically braked. Indeed, disk formation issuppressed in the strict ideal MHD limit for the observedlevel of core magnetization; the angular momentum of theidealized collapsing core is nearly completely removed bymagnetic braking close to the central object. How is thisresolved?

We review what is known about this so-called “magneticbraking catastrophe” and its possible resolutions (§ 3). Im-portant processes to be discussed include non-ideal MHDeffects (ambipolar diffusion, Ohmic dissipation and Halleffect), magnetic interchange instability in the inner partof protostellar accretion flow, turbulence, misalignment be-tween the magnetic field and rotation axis, and depletion ofthe slowly rotating envelope by outflow stripping or accre-tion. We then turn to a discussion of the launch of the ear-liest outflows, and show that two aspects of such outflows— the outer magnetic “tower” and the inner centrifugallydriven disk wind that have dominated much of the discus-sion and theory of early outflows, are actually two regimesof the same unified MHD theory (§ 4). In § 5, we discussthe earliest phases of planet formation in such disks whichare likely quite massive and affected by the angular mo-mentum transport via both strong spiral waves and powerfuloutflows. We synthesize the results in§ 6.

2. Observations

2.1. Dense Cores

Dense cores are the basic units for the formation of atleast low-mass stars. Their properties determine the char-acteristics of the disk, outflow and planets—the byproductsof the star formation process. Particularly important for theformation of the outflow and disk (and its embedded plan-ets) are the core rotation rate and magnetic field strength.We begin our discussion of observations with these two keyquantities, before moving on to early outflows and disks.

Rotation rate is typically inferred from the velocity gra-

dient measured across a dense core (Goodman et al. 1993;Caselli et al. 2002). Whether the gradient traces true ro-tation or not is still under debate (Bergin & Tafalla 2007;Dib et al. 2010). For example, infall along a filament canmimic rotation signature (Tobin et al. 2012a; for syntheticline emission maps from filament accretion and their inter-pretation, see also Smith et al. 2012, 2013). If the gradientdoes trace rotation, then the rotational energy of the corewould typically be a few percent of the gravitational en-ergy. Such a rotation would not be fast enough to preventthe dense core from gravitational collapse. It is, however,more than enough to form a large,102 AU-sized circum-stellar disk, if angular momentum is conserved during thecore collapse.

Magnetic fields are observed in the interstellar mediumon a wide range of scales (see chapter by H.-B. Li et al. inthis volume). Their dynamical importance relative to grav-ity is usually measured by the ratio of the mass of a regionto the magnetic flux threading the region. On the core scale,the field strength was characterized by Troland & Crutcher(2008), who carried out an OH Zeeman survey of a sampleof dense cores in nearby dark clouds. They inferred a me-dian valueλlos ≈ 4.8 for the dimensionless mass-to-fluxratio (in units of the critical value[2πG1/2]−1, Nakano &Nakamura 1978). Geometric corrections should reduce theratio to a typical value ofλ ≈ 2 (Crutcher 2012). It corre-sponds to a ratio of magnetic to gravitational energy of tensof percent, since the ratio is given roughly byλ−2. Sucha field is not strong enough to prevent the core from col-lapsing into stars. It is, however, strong enough to dominatethe rotation in terms of energy and, therefore, is expected tostrongly affect disk formation (§ 3).

2.2. Early Outflows

Jets and outflows are observed during the formation ofstars over the whole stellar spectrum, from brown dwarfs(e.g., Whelan et al. 2005, 2012) to high-mass stars (e.g.,Motogi et al. 2013; Carrasco-Gonzalez et al. 2010; Qiu etal. 2011, 2008, 2007; Zhang et al. 2007), strongly indicatinga universal launching mechanism at work (see§ 4 and alsochapter by Frank et al. in this volume).

Young brown dwarfs have optical forbidden line spectrasimilar to those of low-mass young stars that are indica-tive of outflows (Whelan et al. 2005, 2006). The inferredoutflow speeds of order40–80 km s−1 are somewhat lowerthan those of young stars (Whelan et al. 2007; Joergens etal. 2013). Whelan et al. (2012) suggest that brown dwarfoutflows can be collimated and episodic, just as their low-mass star counterparts. There is some indication that the ra-tio of outflow and accretion rates,Mout/Maccr, is higher foryoung brown dwarf and very-low mass stars (Comeron etal. 2003; Whelan et al. 2009; Bacciotti et al. 2011) than forthe classical T-Tauri stars (e.g., Hartigan et al. 1995; Sicilia-Aguilar et al. 2010; Fang et al. 2009; Ray et al. 2007).Whether this is generally true remains to be established.

Since planetary systems like the Jovian system with its

2

Galilean moons are thought to be built up from planetarysub-disks (Mohanty et al. 2007), one would naturally expectthem to launch outflows as well (see e.g., Machida et al.2006; Liu & Schneider 2009). There is, however, no directobservational evidence yet for such circum-planetary disk-driven outflows.

At the other end of the mass spectrum, there is nowevidence that outflows around young massive stars can behighly collimated, even at relative late evolutionary stages(e.g., Carrasco-Gonzalez et al. 2010; Rodrıguez et al. 2012;Chibueze et al. 2012; Palau et al. 2013). For example, in-terferometric observations reveal that the young, luminous( ∼ 105L⊙), object IRAS 19520+2759 drives a well colli-mated CO outflow, with a collimation factor of 5.7 (Palauet al. 2013). It appears to have evolved beyond a centralB-type object, but still drives a collimated outflow, in con-trast with the expectation that massive YSO outflows decol-limate as they evolve in time (Beuther & Shepherd 2005).Interestingly, an HII region has yet to develop in this source.It may be quenched by protostellar accretion flow (Keto2002, 2003; Peters et al. 2010, 2011) or absent because thecentral stellar object is puffed up by rapid accretion (andthus not hot enough at surface to produce ionizing radia-tion; Hosokawa et al. 2010).

2.3. Early Disks

From an observational point of view, the key question toaddress is when rotationally supported circumstellar disksare first established and become observable. It is clear that,after approximately 0.5 Myr (Evans et al. 2009), gaseousKeplerian (protoplanetary) disks are present on the scalesof ∼100–500 AU around both low- and intermediate-massstars (“T Tauri” and “Herbig Ae” stars, respectively —or Class II young stellar objects; see Dutrey et al. 2007).Whether they are present at earlier times requires high-resolution studies of the youngest protostars, e.g., Class0and Class I objects.

One way to constrain the process of disk formation is tostudy the rotation rates on difference scales (see the caveatsin inferring rotation rate in§ 2.1). From larger to smallerscales, a clear progression in kinematics is evident (Fig. 1;reproduced from Belloche 2013): at large distances fromthe central protostar and in prestellar cores, the specificangular momentum decreases rapidly toward smaller radii,implying that the angular velocity is roughly constant (e.g.,Goodman et al. 1993; Belloche et al. 2002). Observationsof objects in relatively late stages of evolution suggest thatthe specific angular momentum tends to a constant value(vrot ∝ r−1) between∼ 102 and 104 AU, as expectedfrom conservation of angular momentum under infall. Arotationally supported disk is expected to show increasingspecific angular momentum as function of radius (Keple-rian rotation,vrot ∝ r−0.5). To characterize the propertiesof disks being formed, the task at hand is to search for thelocation where the latter two regimes, Keplerian disk withvrot ∝ r−0.5 and the infalling envelope withvrot ∝ r−1,

separate.

2.3.1. Techniques

Observationally the main challenge in revealing theearliest stages of the circumstellar disks is the pres-ence of the larger scale protostellar envelopes during theClass 0 and I stages, which reprocess most of the emis-sion from the central protostellar object itself at shorterwavelengths and easily dominate the total flux at longerwavelengths. The key observational tools at different wave-length regimes are (for convenience we define the near-infrared asλ<3 µm, mid-infrared as 3µm<λ<50 µm,far-infrared as 50µm<λ<250 µm, (sub)millimeter as250µm<λ<4 mm):

Mid-infrared: at mid-infrared wavelengths the observa-tional signatures of young stars are dominated by thebalance between the presence of warm dust and de-gree of extinction on small-scales — and in partic-ular, highly sensitive observations with theSpitzerSpace Telescopein 2003–2009 have been instrumen-tal in this field. One of the key results relevant todisk formation during the embedded stages is thatsimple infalling envelope profiles (e.g.ρ ∝ r−2

or r−1.5) cannot extend unmodified to. few hun-dred AU scales (Jørgensen et al. 2005b; Enoch et al.2009): the embedded protostars are brighter in themid-infrared than expected from such profiles, whichcan for example be explained if the envelope is flat-tened or has a cavity on small scales. Indeed, extinc-tion maps reveal that protostellar environments arecomplex on103 AU scales (Tobin et al. 2010) and insome cases show asymmetric and filamentary struc-tures that complicate the canonical picture of forma-tion of stars from the collapse of relatively sphericaldense cores.

Far-infrared: at far-infrared wavelengths the continuumemission comes mainly from thermal dust grains withtemperatures of a few tens of K. This wavelengthrange is accessible almost exclusively from spaceonly. Consequently, observatories such as the In-frared Space Observatory (ISO) and Herschel SpaceObservatory (Pilbratt et al. 2010) have been the maintools for characterizing protostars there. At the timeof writing, surveys from the Herschel Space Obser-vatory are starting to produce large samples of deeplyembedded protostellar cores that can be followed upby other instruments. The observations at far-infraredwavelengths provide important information about thepeak of the luminosity of the embedded protostarsand the distribution of low surface brightness dust— but due to the limitation in angular resolution lessinformation on the few hundred AU scales of disksaround more evolved YSOs.

(Sub)millimeter: the (sub)millimeter wavelengths providea unique window on the thermal radiation from the

3

cooler dust grains on small scales (Andre et al. 1993;Chandler & Richer 2000). Aperture synthesis obser-vations at these wavelengths resolve scales down to∼100 AU or better in nearby star forming regions.The flux from the thermal dust continuum emissionis strongly increasing with frequencyν asFν ∝ ν2

or steeper, making these wavelengths ideally suitedfor detecting dense structures while discriminatingfrom possible free-free emission. Likewise the highspectral resolution for a wide range of molecular rota-tional transitions can be tailored to study the structure(e.g., temperature and kinematics) of the differentcomponents in the protostellar systems (Jørgensen etal. 2005a).

2.3.2. Millimeter Continuum Surveys

The main observational tool for understanding disk for-mation is millimeter surveys using aperture synthesis tech-nique that probe how the matter is distributed on the fewhundred AU scales. The use of such a technique to ad-dress this question goes back to Keene & Masson (1990)and Terebey et al. (1993) with the first larger arcsecondscale surveys appearing in the early 2000s (e.g., Looney etal. 2000) and detailed radiative transfer modeling appearingabout the same time (e.g. Hogerheijde & Sandell 2000; Har-vey et al. 2003; Looney et al. 2003; Jørgensen et al. 2004,2005a).

The general conclusion from these studies is that, inmost cases, both the large- and small-scale continuum emis-sion cannot be reproduced by a single analytical model of asimple, axisymmetric envelope. In some cases, e.g., Brownet al. (2000); Jørgensen et al. (2005a); Enoch et al. (2009),these structures are well resolved on a few hundred AUscales. Some noteworthy counter-examples where no ad-ditional dust continuum components are required includeL483 (Jørgensen 2004), L723 (Girart et al. 2009) and threeof the nine Class 0 protostars in Serpens surveyed by Enochet al. (2011). Typically the masses for individual objects de-rived in these studies agree well with each other once sim-ilar dust opacities and temperatures are adopted. Jørgensenet al. (2009) compared the dust components for a sampleof 18 embedded Class 0 and I protostars and did not findan increase in mass of the modeled compact componentwith bolometric temperature as one might expect from thegrowth of Keplerian disks. Jørgensen et al. (2009) sug-gested that this could reflect the presence of the rapid for-mation of disk-like structure around the most deeply em-bedded protostars, although the exact kinematics of thosearound the most deeply embedded (Class 0) sources wereunclear. An unbiased survey of embedded protostars inSerpens with envelope masses larger than 0.25M⊙ andluminosities larger than 0.05L⊙ by Enoch et al. (2011)finds similar masses for the compact structures around thesources in that sample. Generally these masses are found tobe small relative to the larger scale envelopes in the Class 0stage on 10,000 AU scales — but still typically one to twoorders of magnitude larger than the mass on similar, few

hundred AU, scales extrapolated from the envelope.Still, the continuum observations do not provide an un-

ambiguous answer to what these compact components rep-resent. By the Class I stage some become the Kepleriandisks surrounding the protostars. The compact componentsaround the Class 0 protostars could be the precursors tothese Keplerian disks. However, it is unlikely that suchmassive rotationally supported disks could be stably sup-ported given the expected low stellar mass for the Class 0protostars: they should be prone to fragmentation (see§ 5).

An alternative explanation for the compact dust emissiondetected in interferometric continuum observations may bethe presence of “pseudo-disks”. In the presence of magneticfields, torsional Alfven waves in twisted field lines carryaway angular momentum, preventing the otherwise naturalformation of large rotationally supported disks. However,strong magnetic pinching forces deflect infalling gas to-ward the equatorial plane to form a flattened structure—the“pseudo-disk” (Galli & Shu 1993; Allen et al. 2003; Fro-mang et al. 2006). Unlike Keplerian disks observed at laterstages, this flattened inner envelope is not supported by ro-tational motions, but can be partially supported by magneticfields. Observationally disentangling disks and pseudo-disks is of paramount importance since accretion onto theprotostar proceeds very differently through a rotationally-supported disk or a magnetically-induced pseudo-disk.

Maury et al. (2010) compared the results from an IRAMPlateau de Bure study of five Class 0 to synthetic modelimages from three numerical simulations — in particu-lar focusing on binarity and structure down to scales of∼50 AU. The comparison shows that magnetized modelsof protostar formation including pseudo-disks agree bet-ter with the observations than, e.g., pure hydrodynamicalsimulation in the case of no initial perturbation or turbu-lence. With turbulence, magnetized models can producesmall disks (∼ 100AU) (Seifried et al. 2013), which mightstill be compatible with Maury et al. (2010)’s observations.The compact continuum components could also representthe “magnetic walls” modeled by Tassis & Mouschovias(2005), although in some sources excess unresolved emis-sion remains unaccounted for in this model (Chiang et al.2008).

2.3.3. Kinematics

A number of more evolved Class I young stellar ob-jects show velocity gradients that are well fitted by Keple-rian profile (e.g., Brinch et al. 2007; Lommen et al. 2008;Jørgensen et al. 2009) (see also Harsono et al. 2013, andLee 2011 for evidence of Keplerian disks around Class Iobject TMC1A and HH 111, respectively). Generally theproblem in studying the kinematics on disk scales in em-bedded protostars is that many of the traditional line tracersare optically thick in the larger scale envelope. Jørgensenetal. (2009), for example, showed that the emission from the(sub)millimeter transitions of HCO+ would be opticallythick on scales of∼100 AU for envelopes with masses

4

larger than 0.1M⊙. An alternative is to trace less abundantisotopologues. Recently one embedded protostar, L1527,was found to show Keplerian rotation in13CO/C18O 2–1 (Tobin et al. 2012b, see also Murillo et al. 2013 forVLA1623A); this result was subsequently strengthened byALMA observations (N. Ohashi, priv. comm., see alsohttp://www.almasc.org/upload/presentations/DC-05.pdf).Another example of Keplerian motions is found in the pro-tostellar binary L1551-NE (a borderline Class 0/I source),in 13CO/C18O 2–1 (Takakuwa et al. 2012). Large circum-stellar disks are inferred around a number of high-massyoung stellar objects, including G31.41+0.31 (Cesaroniet al. 1994), IRAS 20126+4104 (Cesaroni et al. 1997),IRAS 18089-1732 (Beuther et al. 2004), IRAS 16547-4247 (Franco-Hernandez et al. 2009) and IRAS 18162-2048 (Fernandez-Lopez et al. 2011). Whether they arerotationally supported remains uncertain.

Brinch et al. (2009) investigated the dynamics of thedeeply embedded protostar NGC 1333-IRAS2A in subarc-second resolution observations of HCN (4–3) and the sameline of its isotopologue H13CN. Through detailed line ra-diative transfer modeling they showed that the∼ 300 AUcompact structure seen in dust continuum was in fact dom-inated by infall rather than rotation.

Pineda et al. (2012) presented some of the first ALMAobservations of the deeply embedded protostellar binaryIRAS 16293-2422. These sensitive observations revealeda velocity gradient across one component in the binary,IRAS16293A, in lines of the complex organic moleculemethyl formate. However, this velocity gradient does notreflect Keplerian rotation and does not require a centralmass beyond the envelope mass enclosed on the same scale.This is also true for the less dense gas traced by lines of therare C17O and C34S isotopologues in extended Submillime-ter Array (eSMA) observations (0.5′′ resolution; Favre et al.2014).

The above observations paint a picture of complex struc-ture of the material on small scales around low-mass proto-stellar systems. They raise a number of important potentialimplications, which we discuss next.

2.4. Implications and Outlook

2.4.1. Protostellar Mass Evolution

An important constraint on the evolution of protostarsis how the bulk mass is transported and accreted from thelarger scales through the circumstellar disks onto the cen-tral stars (§ 5). An important diagnostic from the aboveobservations is to compare the disk masses — either fromdust continuum or line observations (taking into accountthe caveats about the dust properties and/or chemistry) tostellar masses inferred, e.g., from the disk dynamical pro-files. Jørgensen et al. (2009) compared these quantities forasample of predominantly Class I young stellar objects withwell-established disks: an updated version of this figure in-cluding recent measurements of dynamical masses for addi-tional sources (Tobin et al. 2012b; Takakuwa et al. 2012) is

shown in Fig. 2. These measurements are compared to stan-dard semi-analytic models for collapsing rotating protostars(Visser et al. 2009): these models typically underestimatethe stellar masses relative to the disk masses. These simplemodels of collapse of largely spherical cores are likely in-applicable on larger scales where filamentary structures aresometimes observed (Tobin et al. 2010; Lee et al. 2012).Still, this comparison illustrates a potential avenue to ex-plore when observations of a large sample of embedded pro-tostars become available in the ALMA era; it provides di-rect measures of the accretion rates that more sophisticatednumerical simulations need to reproduce.

2.4.2. Grain growth

Multi-wavelength continuum observations in the mil-limeter and submillimeter are also interesting for studyingthe grain properties close to the newly formed protostars.In more evolved circumstellar disks a flattened slope of thespectral energy distribution at millimeter or longer wave-lengths is taken as evidence that significant grain growthhas taken place (see, e.g., Natta et al. 2007 for a review). Onlarger scales of protostellar envelope where the continuumemission is optically thin, ISM-like dust would result insubmillimeter spectral slopes of 3.5–4. The more compactdust components observed at few hundred AU scales havelower spectral indices of 2.5–3.0 (Jørgensen et al. 2007;Kwon et al. 2009; Chiang et al. 2012), either indicating thatsome growth of dust grains to millimeter sizes has occurredor that the compact components are optically thick. At leastin a few cases where the dust emission is clearly resolvedthe inferred spectral slopes are in agreement with those ob-served for more evolved T Tauri stars, indicating that dustrapidly grows to millimeter sizes (Ricci et al. 2010). As-sessing the occurrence of grain growth during the embeddedstages is important for not only understanding the forma-tion of the seeds for planetesimals (§ 5), but also evaluatingthe non-ideal MHD effects in magnetized core collapse anddisk formation, since they depend on the ionization level,which in turn is strongly affected by the grain size distribu-tion (§ 3).

2.4.3. Chemistry

The presence of disk-like structures on hundred AUscales may also have important implications for the chem-istry in those regions. The presence of a disk may changethe temperature, allowing molecules to freeze-out againwhich would otherwise stay in the gas-phase. Water and itsisotopologues are a particularly clear example of these ef-fects: for example in observations of the H18

2 O isotopologuetoward the centers of a small sample of Class 0 protostars,Jørgensen & van Dishoeck (2010) and Persson et al. (2012)found lower H2O abundances than expected in the typicalgas-phase, consistent with a picture in which a significantfraction of the material at scales. 100 AU has tempera-tures lower than∼ 100 K. This complicates the interpreta-tion of the chemistry throughout the envelope. For example,

5

it makes extrapolations of envelope physical and abundancestructures from larger scales invalid (Visser et al. 2013).Anongoing challenge is to better constrain the physical struc-ture of the protostellar envelopes and disks on these scales.This must be done before significant progress can be madeon understanding the initial conditions for chemical evo-lution in protoplanetary disks. The challenge is even moreformidable for massive stars, although Isokoski et al. (2013)did not find any chemical differentiation between massivestars with and without disk-like structures.

2.4.4. Linking Observations and Theory

Observations suggest that, in principle, there is typicallymore than enough angular momentum on the core scale toform large,102 AU scale rotationally supported disks. Thecommon presence of fast jets around deeply embedded pro-tostars implies that the formation of such disks has begunearly in the process of star formation. There is, however,currently little direct evidence that large, well-developed,Keplerian disks are prevalent around Class 0 protostars, asone may naively expect based on angular momentum con-servation during hydrodynamic core collapse. The paucityof large, early disks indicates that disk formation is notas straightforward as generally expected. The most likelyreason is that star-forming cores are observed to be signifi-cantly magnetized, and magnetic fields are known to inter-act strongly with rotation. They greatly affect, perhaps evencontrol, disk formation, as we show next.

3. Theory of Magnetized Disk Formation

How disks form has been a long-standing theoreticalproblem in star formation. Early work on this topic was re-viewed by Bodenheimer (1995) and Boss (1998). Both re-views listed a number of unsolved problems. Topping bothlists was the effect of the magnetic field, which turns out topresent a formidable obstacle to disk formation. Substan-tial progress has been made in recent years in overcomingthis obstacle, especially through Ohmic dissipation, field-rotation misalignment, and turbulence. This progress willbe summarized below.

3.1. Magnetic Braking Catastrophe in Ideal MHDLimit

The basic difficulty with disk formation in magnetizeddense cores can be illustrated using analytic arguments inthe strict ideal MHD limit where the magnetic field linesare perfectly frozen into the core material. In this limit,as a finite amount of mass is accreted onto the central ob-ject (the protostar), a finite amount of magnetic flux will bedragged into the object as well. The magnetic flux accumu-lated at the center forms a magnetic split monopole, withthe field lines fanning out radially (Galli et al. 2006; seethe sketch in their Fig. 1). As a result, the magnetic fieldstrength increases rapidly with decreasing distance to thecenter, asB ∝ r−2. The magnetic energy density, whichis proportional to the field strength squared, increases with

Fig. 1.— Progression of specific angular momentum asfunction of scale and/or evolutionary stage of young stel-lar objects. Figure from Belloche (2013).

Fig. 2.— Updated version of Fig. 18 from Jørgensen et al.(2009). Predicted stellar mass,Mstar, vs. disk mass,Mdisk,both measured relative to the envelope massMenv in themodels of Visser et al. (2009) withΩ0 = 10−14 s−1 andcs of (a) 0.19 km s−1 and (b) 0.26 km s−1 (solid lines).The sources for which stellar, disk and envelope massesare measured are shown with blue dots (Jørgensen et al.2009; Takakuwa et al. 2012; Tobin et al. 2012b). Finally,the dashed lines indicateMdisk/Mstar ratios of 1% (upper)and 10% (lower).

6

decreasing radius even more rapidly, asEB ∝ r−4. This in-crease is faster than, for example, the energy density of theaccretion flow, which can be estimated approximately fromspherical free-fall collapse asEff ∝ r−5/2. As the infallingmaterial approaches the central object, it will become com-pletely dominated by the magnetic field sooner or later. Thestrong magnetic field at small radii is able to remove all ofthe angular momentum in the collapsing flow, leading tothe so-called “magnetic braking catastrophe” for disk for-mation (Galli et al. 2006). The braking occurs naturallyin a magnetized collapsing core, because the faster rotatingmatter that has already collapsed closer to the rotation axisremains connected to the more slowly rotating material atlarger (cylindrical) distances through field lines. The dif-ferential rotation generates a fieldline twist that brakes therotation of the inner, faster rotating part and transports itsangular momentum outward (see Mouschovias & Paleolo-gou 1979, 1980 for analytic illustrations of magnetic brak-ing).

The catastrophic braking of disks in magnetized densecores in the ideal MHD limit was also found in many nu-merical as well as semi-analytic calculations. Krasnopolsky& K onigl (2002) were the first to show semi-analytically,using the so-called “thin-disk” approximation, that the for-mation of rotationally supported disks (RSDs hereafter) canbe suppressed if the efficiency of magnetic braking is largeenough. However, the braking efficiency was parametrizedrather than computed self-consistently. Similarly, Dapp &Basu (2010) and Dapp et al. (2012) demonstrated that, inthe absence of any magnetic diffusivity, a magnetic split-monopole is produced at the center and RSD formation issuppressed, again under the thin-disk approximation.

Indirect evidence for potential difficulty with disk forma-tion in ideal MHD simulations came from Tomisaka (2000),who studied the (2D) collapse of a rotating, magnetizeddense core using a nested grid under the assumption of ax-isymmetry. He found that, while there is little magneticbraking during the phase of runaway core collapse leadingup to the formation of a central object, once an outflow islaunched, the specific angular momentum of the materialat the highest densities is reduced by a large factor (up to∼ 104) from the initial value. The severity of the magneticbraking and its deleterious effect on disk formation were notfully appreciated until Allen et al. (2003) explicitly demon-strated that the formation of a large, numerically resolvable,rotationally supported disk was completely suppressed in2D by a moderately strong magnetic field (correspondingto a dimensionless mass-to-flux ratioλ of several) in an ini-tially self-similar, rotating, magnetized, singular isothermaltoroid (Li & Shu 1996). They identified two key ingredi-ents behind the efficient braking during the accretion phase:(1) concentration of the field lines at small radii by the col-lapsing flow, which increases the field strength, and (2) thefanning out of field lines due to equatorial pinching, whichincreases the lever arm for magnetic braking.

The catastrophic braking that prevents the formation ofRSDs during the protostellar accretion phase has been con-

firmed in several subsequent 2D and 3D ideal MHD simu-lations (Mellon & Li 2008; Hennebelle & Fromang 2008;Duffin & Pudritz 2009; Seifried et al. 2012; Santos-Lima etal. 2012). Mellon & Li (2008), in particular, formulatedthe disk formation problem in the same way as Allen etal. (2003), by adopting a self-similar rotating, magnetized,singular isothermal toroid as the initial configuration. Al-though idealized, the adopted initial configuration has theadvantage that the subsequent core collapse should remainself-similar. The self-similarity provides a useful checkonthe correctness of the numerically obtained solution. Theyfound that the disk formation was suppressed by a field asweak asλ = 13.3.

Hennebelle & Fromang (2008) carried out 3D simula-tions of the collapse of a rotating dense core of uniformdensity and magnetic field into the protostellar accretionphase using an ideal MHD AMR code (RAMSES). Theyfound that the formation of a RSD is suppressed as longasλ is of order 5 or less. However, theλ = 5 case inPrice & Bate’s (2007) SPMHD simulations appears to haveformed a small disk (judging from the column density dis-tribution). It is unclear whether the disk is rotationally sup-ported or not, since the disk rotation rate was not given inthe paper. Furthermore, contrary to the grid-based simula-tions, there appears little, if any, outflow driven by twistedfield lines in the SPH simulations, indicating that the ef-ficiency of magnetic braking is underestimated (prominentoutflows are produced, however, in other SPMHD simula-tions, e.g., Burzle et al. 2011 and Price et al. 2012 ). An-other apparently discrepant result is that of Machida et al.(2011). They managed to form a large,102-AU scale, rota-tionally supported disk for a very strongly magnetized coreof λ = 1 even in the ideal MHD limit (their Model 4) usinga nested grid and sink particle. This contradicts the resultsfrom other simulations and semi-analytic calculations.

To summarize, both numerical simulations and analyticarguments support the notion that, in the ideal MHD limit,catastrophic braking makes it difficult to form rotationallysupported disks in (laminar) dense cores magnetized to arealistic level (with a typicalλ of a few). In what follows,we will explore the potential resolutions that have been pro-posed in the literature to date.

3.2. Non-ideal MHD Effects

Dense cores of molecular clouds are lightly ionized(with a typical electron fractional abundance of order10−7;Bergin & Tafalla 2007). As such, the magnetic field is notexpected to be perfectly frozen into the bulk neutral mate-rial. There are three well known non-ideal MHD effectsthat can in principle break the flux-freezing condition thatlies at the heart of the magnetic braking catastrophe in thestrict ideal MHD limit. They are ambipolar diffusion, theHall effect, and Ohmic dissipation (see Armitage 2011 fora review). Roughly speaking, in the simplest case of anelectron-ion-neutral medium, both ions and electrons arewell tied to the magnetic field in the ambipolar diffusion

7

regime. In the Hall regime, electrons remain well tied to thefield, but not ions. At the highest densities, both electronsand ions are knocked off the field lines by collisions be-fore they finish a complete gyration; in such a case, Ohmicdissipation dominates. This simple picture is complicatedby dust grains, whose size distribution in dense cores isrelatively uncertain (see§ 2.4), but which can become thedominant charge carriers. Under typical cloud conditions,ambipolar diffusion dominates over the other two effects atdensities typical of cores (e.g., Nakano et al. 2002, Kunz &Mouschovias 2010). It is the most widely studied non-idealMHD effect in the context of core formation and evolutionin the so-called “standard” picture of low-mass star forma-tion out of magnetically supported clouds (Nakano 1984;Shu et al. 1987; Mouschovias & Ciolek 1999). It is the ef-fect that we will first concentrate on.

Ambipolar diffusion enables the magnetic field lines thatare tied to the ions to drift relative to the bulk neutral ma-terial. In the context of disk formation, its most impor-tant effect is to redistribute the magnetic flux that wouldhave been dragged into the central object in the ideal MHDlimit to a circumstellar region where the magnetic fieldstrength is greatly enhanced (Li & McKee 1996). Indeed,the enhanced circumstellar magnetic field is strong enoughto drive a hydromagnetic shock into the protostellar accre-tion flow (Li & McKee 1996; Ciolek & Konigl 1998; Con-topoulos et al. 1998; Krasnopolsky & Konigl 2002; Tassis& Mouschovias 2007; Dapp et al. 2012). Krasnopolsky &Konigl (2002) showed semi-analytically, using the 1D thin-disk approximation, that disk formation may be suppressedin the strongly magnetized post-shock region if the mag-netic braking is efficient enough. The braking efficiency,parametrized in Krasnopolsky & Konigl (2002), was com-puted self-consistently in the 2D (axisymmetric) simula-tions of Mellon & Li (2009), which were performed un-der the usual assumption of ion density proportional to thesquare root of neutral density. 3D simulations of AD wereperformed by Duffin & Pudritz (2009) using a specially de-veloped, single fluid AMR code (Duffin & Pudritz 2008) aswell as by a two fluid SPH code (Hosking & Whitworth2004). Mellon & Li (2009) found that ambipolar diffu-sion does not weaken the magnetic braking enough to allowrotationally supported disks to form for realistic levels ofcloud core magnetization and cosmic ray ionization rate. Inmany cases, the magnetic braking is even enhanced. Thesefindings were strengthened by Li et al. (2011), who com-puted the ion density self-consistently using the simplifiedchemical network of Nakano and collaborators (Nakano etal. 2002; Nishi et al. 1991) that includes dust grains. Anexample of their simulations is shown in Fig. 3. It showsclearly the rapid slowdown of the infalling material nearthe ambipolar diffusion-induced shock (located at a radius∼ 1015 cm in this particular example) and the nearly com-plete braking of the rotation in the post-shock region, whichprevents the formation of a RSD. The suppression of RSDis also evident from the fast, supersonic infall close to thecentral object. We should note that RSD formation may still

be possible if the cosmic ray ionization rate can be reducedwell below the canonical value of10−17 s−1 (Mellon & Li2009), through for example the magnetic mirroring effect,which may turn a large fraction of the incoming cosmic raysback before they reach the disk-forming region (Padovani etal. 2013).

1014 1015 1016 1017

Radius along equator (cm)

-1•105

-5•104

0

5•104

1•105

Infa

ll an

d ro

tatio

n sp

eed

(cm

/s)

Model REF+AHO

Model REF-AHO

Model REF

Rotation

Infall

Fig. 3.— Infall and rotation speeds along the equatorialplane of a collapsing rotating, magnetized dense core dur-ing the protostellar mass accretion phase for three represen-tative models of Li et al. (2011). Model REF (solid lines)includes only ambipolar diffusion, whereas the other twoinclude all three non-ideal MHD effects, especially the Halleffect. The initial magnetic field and rotation axis are in thesame direction in one model (REF+, dashed lines) and inopposite directions in the other (REF−, dotted).

As the density increases, the Hall effect tends to be-come more important (the exact density for this to hap-pen depends on the grain size distribution). It is less ex-plored than ambipolar diffusion in the star formation lit-erature. A unique feature of this effect is that it can ac-tively increase the angular momentum of a collapsing, mag-netized flow through the so-called “Hall spin-up” (Wardle& Ng 1999). In the simplest case of electron-ion-neutralfluid, the spin-up is caused by the current carriers (theelectrons) moving in the azimuthal direction, generatinga magnetic torque through field twisting; the toroidal cur-rent is produced by gravitational collapse, which drags thepoloidal field into a pinched, hourglass-like configuration.The Hall spin-up was studied numerically by Krasnopol-sky et al. (2011) and semi-analytically by Braiding & War-dle (2012a,b). Krasnopolsky et al. (2011) showed that arotationally supported disk can form even in an initiallynon-rotatingcore, provided that the Hall coefficient is largeenough. Interestingly, when the direction of the initial mag-

8

netic field in the core is flipped, the disk rotation is reversed.This reversal of rotation is also evident in Fig. 3, where theHall effect spins up the nearly non-rotating material in thepost-AD shock region to highly supersonic speeds, but indifferent directions depending on the field orientation. TheHall effect, although dynamically significant, does not ap-pear capable of forming a rotationally supported disk undertypical dense core conditions according to Li et al. (2011).This inability is illustrated in Fig. 3, where the equatorialmaterial collapses supersonically on the102-AU scale evenwhen the Hall effect is present.

Ohmic dissipation becomes the dominant nonidealMHD effects at high densities (e.g., Nakano et al. 2002).It has been investigated by different groups in connec-tion with disk formation. Shu et al. (2006) studied semi-analytically the effects of a spatially uniform resistivity onthe magnetic field structure during the protostellar mass ac-cretion phase. They found that, close to the central object,the magnetic field decouples from the collapsing materialand becomes more or less uniform. They suggested thata rotationally supported disk may form in the decoupledregion, especially if the resistivity is higher than the clas-sic (microscopic) value. This suggestion was confirmedby Krasnopolsky et al. (2010; see also Santos-Lima et al.2012), who found numerically that a large,102AU-scale,Keplerian disk can form around a0.5 M⊙ star, providedthat the resistivity is of order1019 cm2 s−1 or more; sucha resistivity is significantly higher than the classic (micro-scopic) value over most of the density range relevant to diskformation.

Machida & Matsumoto (2011) and Machida et al. (2011)studied disk formation in magnetized cores including onlythe classic value of resistivity estimated from Nakano etal.’s (2002) numerical results. The former study found thata relatively small, 10 AU-scale, rotationally supported diskformed within a few years after the formation of the stellarcore. Inside the disk, the density is high enough for mag-netic decoupling to occur due to Ohmic dissipation. Thiswork was extended to much later times by Machida et al.(2011), who included a central sink region in the simula-tions. They concluded that the small RSD can grow to large,102-AU size at later times, especially after the most of theenvelope material has fallen onto the disk and the centralobject. A caveat, pointed out by Tomida et al. (2013; seealso Dapp & Basu 2010), is that they used a form of in-duction equation that is, strictly speaking, inappropriate forthe non-constant resistivity adopted in their models; it maygenerate magnetic monopoles that are subsequently cleanedaway using the Dedner’s method (Dedner et al. 2002). Thisdeficiency was corrected in Tomida et al. (2013), who car-ried out radiative MHD simulations of magnetized core col-lapse to a time shortly (∼ 1 year) after the formation ofthe second (protostellar) core. They found that the forma-tion of a (small, AU-scale) rotationally supported disk wassuppressed by magnetic braking in the ideal MHD limitbut was enabled by Ohmic dissipation at this early time;the latter result is in qualitative agreement with Machida

& Matsumoto (2011) and Machida et al. (2011), althoughit remains to be seen how the small disks in Tomida et al.(2013)’s simulations evolve further in time.

Dapp & Basu (2010) studied the effects of Ohmic dissi-pation on disk formation semi-analytically, using the “thin-disk” approximation for the mass distribution and an ap-proximate treatment of magnetic braking. The approxima-tions enabled them to follow the formation of both the firstand second core. They found that a small, sub-AU, rota-tionally supported disk was able to form soon after the for-mation of the second core in the presence of Ohmic dissi-pation; it was suppressed in the ideal MHD limit, in agree-ment with the later 3D simulations of Tomida et al. (2013).This work was extended by Dapp et al. (2012) to includea set of self-consistently computed charge densities from asimplified chemical network and ambipolar diffusion. Theyshowed that their earlier conclusion that a small, sub-AUscale, RSD is formed through Ohmic dissipation holds evenin the presence of a realistic level of ambipolar diffusion.This conclusion appears reasonably secure in view of thebroad agreement between the semi-analytic work and nu-merical simulations. When and how such disks grow to themuch larger,102AU-scale, size deserve to be explored morefully.

3.3. Magnetic Interchange Instabilities

The formation of a large-scale RSD in a magnetized coreis made difficult by the accumulation of magnetic flux nearthe accreting protostar. As discussed earlier, this is espe-cially true in the presence of a realistic level of ambipolardiffusion, which redistributes the magnetic flux that wouldhave been dragged into the central object to the circumstel-lar region (Ohmic dissipation has a similar effect, see Li etal. 2011 and Dapp et al. 2012). The result of the flux re-distribution is the creation of a strongly magnetized regionclose to the protostar where the infall speed of the accret-ing flow is slowed down to well below the free-fall value(i.e., it is effectively held up by magnetic tension againstthegravity of the central object), at least in 2D (assuming ax-isymmetry). It has long been suspected that such a magnet-ically supported structure would become unstable to inter-change instabilities in 3D (Li & McKee 1996; Krasnopol-sky & Konigl 2002). Recent 3D simulations have shownthat this is indeed the case.

Magnetic interchange instability in a protostellar accre-tion flow driven by flux redistribution was first studied indetail by Zhao et al. (2011). They treated the flux redis-tribution through a sink particle treatment: when the massin a cell is accreted onto a sink particle, the magnetic fieldis left behind in the cell (see also Seifried et al. 2011 andCunningham et al. 2012); it is a crude representation ofthe matter-field decoupling expected at high densities (oforder1012 cm−3 or higher; Nakano et al. 2002; Kunz &Mouschovias 2010). The decoupled flux piles up near thesink particle, leading to a high magnetic pressure that is re-leased through the escape of field lines along the directions

9

of least resistance. As a result, the magnetic flux draggedinto the decoupling region near the protostar along someazimuthal directions is advected back out along other di-rections in highly magnetized, low-density, expanding re-gions. Such regions are termed DEMS (decoupling-enabledmagnetic structure) by Zhao et al. (2011); they appear to bepresent in the formally ideal MHD simulations of Seifried etal. (2011), Cunningham et al. (2012) and Joos et al. (2012)as well.

Krasnopolsky et al. (2012) improved upon the work ofZhao et al. (2011) by including two of the physical pro-cesses that can lead to magnetic decoupling: ambipolar dif-fusion and Ohmic dissipation. They found that the basicconclusion of Zhao et al. (2011) that the inner part of theprotostellar accretion flow is driven unstable by magneticflux redistribution continues to hold in the presence of real-istic levels of non-ideal MHD effects (see Fig. 4 for an illus-trative example). The magnetic flux accumulated near thecenter is transported outward not only diffusively by the mi-croscopic non-ideal effects, but also advectively throughthebulk motions of the strongly magnetized expanding regions(the DEMS) generated by the instability. The advective fluxredistribution in 3D lowers the field strength at small radiicompared to the 2D (axisymmetric) case where the insta-bility is suppressed. It makes the magnetic braking less ef-ficient and the formation of a RSD easier in principle. Inpractice, the magnetic interchange instability does not ap-pear to enable the formation of rotationally supported disksby itself, because the highly magnetized DEMS that it cre-ates remain trapped at relatively small distances from theprotostar by the protostellar accretion flow (see Fig. 4); thestrong magnetic field inside the DEMS blocks the accretionflow from rotating freely around the center object to form acomplete disk.

3.4. Magnetic Field-Rotation Misalignment

Misalignment between the magnetic field and rotationaxis as a way to form large RSDs has been explored ex-tensively by Hennebelle and collaborators (Hennebelle &Ciardi 2009; Ciardi & Hennebelle 2010; Joos et al. 2012;see also Machida et al. 2006; Price & Bate 2007, and Boss& Keiser 2013). The misalignment is expected if the angu-lar momenta of dense cores are generated through turbulentmotions (e.g., Burkert & Bodenheimer 2000; Seifried et al.2012b; Myers et al. 2013; Joos et al. 2013). Plausible obser-vational evidence for it was recently uncovered by Hull etal. (2013) using CARMA, who found that the distributionof the angle between the magnetic field on the103AU-scaleand the bipolar outflow axis (taken as a proxy for the rota-tion axis) is consistent with being random. If true, it wouldimply that in half of the sources the two axes are misalignedby an angle greater than60. Joos et al. (2012) found thatsuch a large misalignment enables the formation of RSDsin moderately magnetized dense cores with a dimension-less mass-to-flux ratioλ of ∼ 3–5; RSD formation is sup-pressed in such cores if the magnetic field and rotation axis

-4•1015 -2•1015 0 2•1015 4•1015-4•1015

-2•1015

0

2•1015

4•1015

-14.

8

-14.8

-14.8

-14.8

-14.

4

-14.4

-14.

4

-14.

4

-14.

0

-14.0

-14.0

-14.0

-13.

6

-16.0 -15.3 -14.7 -14.0 -13.3 -12.7 -12.0

Fig. 4.— An example of the inner protostellar accretionflow driven unstable by magnetic flux redistribution (takenfrom Krasnopolsky et al. 2012). Plotted are the distribu-tion of logarithm of density (in units of g cm−3) and ve-locity field on the equatorial plane (the length is in unitsof cm). The expanding, low-density, regions near the cen-ter are the so-called “decoupling-enabled magnetic struc-ture” (DEMS) that are strongly magnetized. They present aformidable barrier to disk formation.

are less misaligned (see Fig. 5). They attributed the diskformation to a reduction in the magnetic braking efficiencyinduced by large misalignment. In more strongly magne-tized cores withλ . 2, RSD formation is suppressed inde-pendent of the misalignment angle, whereas in very weaklymagnetized cores RSDs are formed for all misalignment an-gles.

Based on the work of Hull et al. (2013) and Joos etal. (2012), Krumholz et al. (2013) estimated that the field-rotation misalignment may enable the formation of largeRSDs in∼ 10–50% of dense cores. If the upper rangeis correct, the misalignment would go a long way towardsolving the problem of excessive magnetic braking in pro-tostellar disk formation.

Li et al. (2013) carried out simulations similar to thoseof Joos et al. (2012), except for the initial conditions. Theyconfirmed the qualitative result of Joos et al. (2012) thatthe field-rotation misalignment is conducive to disk forma-tion. In particular, large misalignment weakens the strongoutflow in the aligned case and is a key reason behind theformation of RSDs in relatively weakly magnetized cores.For more strongly magnetized cores withλ . 4, RSDformation is suppressed independent of the degree of mis-alignment. This threshold value for the mass-to-flux ratio isabout a factor of 2 higher than that obtained by Joos et al.

10

(2012). The difference may come, at least in part, from thedifferent initial conditions adopted: uniform density with auniform magnetic field for Li et al. (2013) and a centrallycondensed density profile with a nonuniform but unidirec-tional field for Joos et al. (2012); the magnetic braking isexpected to be more efficient at a given (high) central den-sity for the former initial configuration, because its fieldlines would become more pinched, with a longer lever armfor braking. Whether there are other factors that contributesignificantly to the above discrepancy remains to be deter-mined.

If the result of Li et al. (2013) is correct, then a densecore must have both a large field-rotation misalignmentanda rather weak magnetic field in order to form a RSD. Thisdual requirement would make it difficult for the misalign-ment alone to enable disk formation in the majority of densecores, which are typically rather strongly magnetized ac-cording to Troland & Crutcher (2008, with a median mass-to-flux ratio ofλ ∼ 2). In a more recent study, Crutcheret al. (2010) argued, based on Bayesian analysis, that afraction of dense cores could be very weakly magnetized,with a dimensionless mass-to-flux ratioλ well above unity(see Bertram et al. 2012 for additional arguments for weakfield, including field reversal). However, since the medianmass-to-flux ratio remains unchanged for the different dis-tributions of the total field strength assumed in Crutcher’s(2012) Bayesian analysis, it is unlikely for the majority ofdense cores to haveλ much greater than the median valueof 2. For example, Li et al. (2013) estimated the fraction ofdense cores withλ > 4 at ∼ 25%. There is also concernthat the random distribution of the field-rotation misalign-ment angle found by Hull et al. (2013) on the103 AU scalemay not be representative of the distribution on the largercore scale. Indeed, Chapman et al. (2013) found that thefield orientation on the core scale (measured using a singledish telescope) is within∼ 30 of the outflow axis for 3 ofthe 4 sources in their sample (see also Davidson et al. 2011);the larger angle measured in the remaining source may bedue to projection effects because its outflow axis lies closeto the line of sight. If the result of Chapman et al. (2013)is robust and if the outflow axis reflects the rotation axis,dense cores with large misalignment between the magneticand rotation axes would be rare. In such a case, it wouldbe even less likely for the misalignment to be the dominantmechanism for disk formation.

3.5. Turbulence

Turbulence is a major ingredient for star formation (seereviews by, e.g., Mac Low & Klessen 2004 and McKee &Ostriker 2007). It can generate local angular momentum byshear flows and form highly asymmetric dense cores (seeresults from Herschel observations, e.g., Men’shchikov etal. 2010 and Molinari et al. 2010). There is increasing evi-dence that it also promotes RSD formation. Santos-Lima etal. (2012) contrasted the accretion of turbulent and laminarmagnetized gas onto a pre-existing central star, and found

that a nearly Keplerian disk was formed in the turbulent butnot laminar case (see Fig. 6). The simulations were carriedout at a relatively low resolution (with a rather large cell sizeof 15.6 AU; this was halved, however, in Santos-Lima et al.2013, who found similar results), and turbulence was drivento an rms Mach number of∼ 4, which may be too large forlow-mass cores. Nevertheless, the beneficial effect of tur-bulence on disk formation is clearly demonstrated. Theyattributed the disk formation to the turbulence-induced out-ward diffusion of magnetic flux, which reduces the strengthof the magnetic field in the inner, disk-forming, part of theaccretion flow. Similar results of disk formation in turbu-lent cloud cores are presented by Seifried et al. (2012b)and Seifried et al. (2013), although these authors attributetheir findings to different mechanisms. They argued thatthe turbulence-induced magnetic flux loss is limited welloutside their disks, based on the near constancy of an ap-proximate mass-to-flux ratio computed on a sphere severaltimes the disk size (with a500 AU radius). They proposedinstead that the turbulence-induced tangling of field linesand strong local shear are mainly responsible for the diskformation: the disordered magnetic field weakens the brak-ing and the shear enhances rotation. Similarly, Myers et al.(2013) also observed formation of a nearly Keplerian diskin their radiative MHD simulation of a turbulent massive(300M⊙) core, although they refrained from discussing theorigin of the disk in detail since it was not the focus of theirinvestigation.

Seifried et al. (2013) extended their previous work to in-clude both low-mass and high-mass cores and both subsonicand supersonic turbulence. They found disk formation inall cases. Particularly intriguing is the formation of rota-tionally dominated disks in the low-mass, subsonically tur-bulent cores. They argued that, as in the case of massivecore with supersonic turbulence of Seifried et al. (2012b),such disks are not the consequence of turbulence-inducedmagnetic flux loss, although such loss appears quite severeon the disk scale, which may have contributed to the long-term survival of the formed disk. While disks appear toform only at sufficiently high mass-to-flux ratiosλ ≥ 10 inordered magnetic fields, disks in the turbulent MHD simu-lations form at much lower and more realistic values ofλ.Seifried et al. (2012b) found thatλ increases gradually inthe vicinity of the forming disk, which may have more todo with the growing accreting mass relative to the magneticflux than with dissipative effects of the magnetic field byturbulent diffusion or reconnection.

Joos et al. (2013) investigated the effects of turbulence ofvarious strengths on disk formation in a core of intermediatemass (5M⊙). They found that an initially imposed turbu-lence has two major effects. It produces an effective diffu-sivity that enables magnetic flux to diffuse outward, broadlyconsistent with the picture envisioned in Santos-Lima etal. (2012, 2013). It also generates a substantial misalign-ment between the rotation axis and magnetic field direction(an effect also seen in Seifried et al. 2012b and Myers etal. 2013). Both of these effects tend to weaken magnetic

11

µ

0

20

45

70

80

90

α

2 3 5 17

Keplerian diskdiskno disk

Fig. 5.— Parameter space for disk formation according toJoos et al. (2012). The parameterα is the angle between themagnetic field and rotation axis, andµ is the dimensionlessmass-to-flux ratio of the dense core (denoted byλ in the restof the article). The diamond denotes disks with Keplerianrotation profile, the square those with flat rotation curve,and circle the cases with no significant disk.

ideal MHD turbulent

Fig. 6.— Accretion of rotating, magnetized material ontoa pre-existing central object with (right panel) and without(left) turbulence (adapted from Santos-Lima et al. 2012).The formation of a nearly Keplerian disk is clearly sup-pressed in the laminar case (left panel), by excessive mag-netic braking, but is enabled by turbulence (right).

braking and make disk formation easier. If the turbulence-induced magnetic diffusion is responsible, at least in part,for the disk formation, then numerical effects would be aconcern. In the ideal MHD limit, the diffusion presumablycomes from turbulence-enhanced reconnections due to fi-nite grid resolution. Indeed, Joos et al. (2013) reported thattheir simulations did not appear to be fully converged, withdisk masses differing by a factor up to∼ 2 in higher resolu-tion simulations. The situation is further complicated by nu-merical algorithms for treating magnetic field evolution, es-pecially those relying on divergence cleaning, which couldintroduce additional artificial magnetic diffusion. To makefurther progress, it would be useful to determine when andhow the reconnections occur and exactly how they lead tothe magnetic diffusion that are apparent in the simulationsof Joos et al. (2013), Santos-Lima et al. (2012), Li et al. (inpreparation) and perhaps Seifried et al. (2012b, 2013).

3.6. Other Mechanisms

The magnetic braking catastrophe in disk formationwould disappear if the majority of dense cores are non-magnetic or only weakly magnetized (λ ∼> 5 or evenλ ∼> 10, see e.g., Hennebelle & Ciardi 2009; Ciardi & Hen-nebelle 2010; Seifried et al. 2011). However, such weaklymagnetized cloud cores are rather unlikely. Although, therecent study by Crutcher et al. (2010) indicates that somecloud cores might be highly supercritical, they are certainlynot the majority. Furthermore, consider, for example, a typ-ical core of 1 M⊙ in mass and104 cm−3 in H2 numberdensity. To have a dimensionless mass-to-flux ratioλ & 5,its field strength must beB . 4.4 µG, less than the medianfield strength inferred for the atomic CNM (∼ 6 µG, Heiles& Troland 2005), which is unlikely. We therefore expectthe majority of dense cores to have magnetic fields corre-sponding toλ . 5 (in agreement with Troland & Crutcher2008), which are strong enough to make RSD formationdifficult.

Another proposed solution is the depletion of the pro-tostellar envelope. The slowly rotating envelope acts as abrake on the more rapidly rotating material closer to thecentral object that is magnetically connected to it. Its de-pletion should promote RSD formation. Indeed, Machidaet al. (2010) found that envelope depletion is conducive tothe formation of large RSDs toward the end of the mainaccretion phase. This is in line with the expectation of Mel-lon & Li (2008), who envisioned that most of the envelopedepletion is achieved through wind stripping rather than ac-cretion, as would be the case if the star formation efficiencyof individual cores is relatively low (say,∼ 1/3, Alves etal. 2007). Given the ubiquity of fast outflows, their effectson envelope depletion and disk formation should be inves-tigated in more detail.

3.7. Summary and Outlook

The formation of rotationally supported disks turns outto be much more complicated than envisioned just a decade

12

ago. This is because star-forming dense cores are observedto be rather strongly magnetized in general (although themagnetization in a fraction of them can be rather weak, seeCrutcher et al. 2010 and discussion above), with a mag-netic energy typically much higher than the rotational en-ergy. The field strength is further amplified by core col-lapse, which tends to concentrate the field lines in the re-gion of disk formation close to the protostar. Both analyticcalculations and numerical simulations have shown that thecollapse-enhanced (ordered) magnetic field can prevent theRSD formation through catastrophic magnetic braking inthe simplest case of ideal MHD limit, aligned magnetic fieldand rotation axis, and no turbulence. Ambipolar diffusion,the Hall effect, and magnetic interchange instabilities haveprofound effects on the dynamics of the inner protostellaraccretion flow, but they do not appear capable of formingRSDs by themselves under typical conditions. Ohmic dis-sipation, on the other hand, can enable the formation of atleast small, AU-scale disks at early times. How such disksevolve in the presence of the instabilities and other non-ideal MHD effects remains to be quantified. Magnetic field-rotation misalignment is conducive to disk formation, but itis unlikely to enable the formation of RSDs in the majorityof dense cores, because of the dual requirement of both arelatively weak magnetic field and a relatively large tilt an-gle that may be uncommon on the core scale. Turbulenceappears to facilitate RSD formation in a number of numer-ical simulations. It is possible that the turbulence-enhancednumerical reconnection plays a role in the appearance ofRSDs in these formally ideal MHD simulations, althoughturbulence by itself could reduce braking efficiency. Thepossible role of reconnection needs to be better understoodand quantified.

4. Early Outflows

4.1. Introduction

Generally, low mass young stellar objects are accompa-nied by highly collimated optical jets (see e.g., Cabrit et al.1997; Reipurth & Bally 2001) whereas high mass stars areoften obscured, hard to observe, and until recently thoughtto drive much less collimated outflows (e.g., Shepherd &Churchwell 1996a,b; Beuther & Shepherd 2005, and dis-cussion below). Nevertheless, there is strong evidence thatthe underlying launching process is based on the same phys-ical mechanism, namely the magneto-rotational coupling:magnetic fields anchored to an underlying rotor (e.g., anaccretion disk) will carry along gas which will be flung out-wards (the same mechanism could also apply to galacticjets, e.g., Pudritz et al. 2007, 2009). For example, Guzmanet al. (2012) concluded that collimated thermal radio jets areassociated with high-mass young stellar objects, althoughfor a relatively short time (∼ 4× 104 yr).

With the seminal theoretical work by Blandford & Payne(1982) and Pudritz & Norman (1983) the idea of magneto-centrifugally driven jets was first established. It was shownthat magnetic fields anchored to the disk around a central

object can lift off gas from the disk surface. A magneto-centrifugally driven jet will be launched if the poloidal com-ponent of magnetic field is inclined with respect to the ro-tation axis by more than 30. Numerical simulations of Ke-plerian accretion disks threaded by such a magnetic fieldhave shown that these jets are self-collimated and accel-erated to high velocities (e.g., Fendt & Camenzind 1996;Ouyed & Pudritz 1997). This driving mechanism, where thelaunching of the jet is connected to the underlying accretiondisk, predicts that jets rotate and carry angular momentumoff the disk. The first plausible observational confirmationcame from Hubble Space Telescope (HST) detection of ro-tational signatures in the optical jet of DG Tauri (Bacciottiet al. 2002; see, however, Soker 2005, Cerqueira et al. 2006and Fendt 2011 for different interpretations of the observa-tion). Further evidence is provided by UV (Coffey et al.2007) and IR observations (Chrysostomou et al. 2008).

The most viable mechanism to launch jets and wider an-gle outflows from accretion disks around YSOs is the cou-pling through magnetic fields where the gas from the disksurface is accelerated by the Lorentz force. Generally, thisforce can be divided into a magnetic tension and a mag-netic pressure term. In an axisymmetric setup, the magnetictension term is responsible for the magneto-centrifugal ac-celeration and jet collimation via hoop stress. The mag-netic pressure can also accelerate gas off the underlyingdisk. These magnetic pressure driven winds are sometimesknown asmagnetic twist(Shibata & Uchida 1985),plasmagun (Contopoulos 1995), ormagnetic tower(Lynden-Bell2003).

Protostellar disks around young stellar objects them-selves are the result of gravitational collapse of molecularcloud cores (see Sec. 3). Since the molecular clouds arepermeated by magnetic fields of varying strength and mor-phology (see e.g., Crutcher et al. 1999; Beck 2001; Alveset al. 2008), there should be a profound link between thecollapse and magneto-rotationally driven outflows.

There are still many unresolved problems concerningjets and outflows. These include the details of the jetlaunching, the driving of molecular outflows, the efficiencyof outflows around massive protostars, the influence of out-flow feedback on star formation, and how efficient they arein clearing off the envelope material around the young stel-lar object. Shedding light on the last problem will also helpdetermine whether there is a clear physical link betweenthe core mass function (CMF) and IMF (see the chapter byOffner et al. in this volume).

In this section we summarize our knowledge of early jetsand outflows and discuss open questions.

4.2. Jet Launching and Theoretical Modeling

Self-consistent modeling of jet launching is a chal-lenging task, especially during the earliest phases of starformation, when the core collapse has to be modeled atthe same time. The most practical approach to study theself-consistent jet launching during the collapse of self-

13

gravitating gas is through direct numerical simulation. Eventhen, the large dynamical range of length scales (from the104 AU molecular core to the sub AU protostellar disk) andtime scales (from the initial free fall time of105 years tothe orbital time of one year or less) require expensive adap-tive mesh refinement (AMR) or SPH (smoothed particlehydrodynamics) simulations that include magnetic fields.Furthermore, non-ideal MHD effects such as Ohmic dissi-pation and ambipolar diffusion complicate the calculations(see§ 3).

One of the first collapse simulations in which outflowsare observed was done more than a decade ago by Tomisaka(1998) with an axisymmetric nested grid technique. Thesesimulations of magnetized, rotating, cylindrical cloud coresshowed that a strong toroidal field component builds up,which eventually drives a bipolar outflow. Subsequently, anumber of collapse simulations from different groups anddifferent levels of sophistication were performed (amongthese are work by Tomisaka 2002; Boss 2002; Allen et al.2003; Matsumoto & Tomisaka 2004; Hosking & Whitworth2004; Machida et al. 2004, 2005a,b; Ziegler 2005; Machidaet al. 2006; Banerjee & Pudritz 2006, 2007; Machida et al.2007; Price & Bate 2007; Machida et al. 2008; Hennebelle& Fromang 2008; Duffin & Pudritz 2009; Mellon & Li2009; Commercon et al. 2010; Burzle et al. 2011; Seifriedet al. 2012). Despite the diversity in numerical approach(e.g. AMR vs. SPH simulations) and initial problem setup,all simulations enforce the same general picture, that mag-netically launched outflows are a natural outcome of mag-netized core collapse. The details of the outflows generateddepend, of course, on the initial parameters such as the de-gree of core magnetization and the core rotation rate.

4.2.1. Outflow Driving

Traditionally, there is a clear distinction between out-flows driven by centrifugal acceleration (Blandford &Payne 1982; Pudritz & Norman 1986; Pelletier & Pudritz1992) or the magnetic pressure gradient (Lynden-Bell 1996,2003). A frequently used quantity to make the distinction isthe ratio of the toroidal to poloidal magnetic field,Bφ/Bpol

(e.g. Hennebelle & Fromang 2008). If this ratio is signif-icantly above 1, the outflow is often believed to be drivenby the magnetic pressure. However, the consideration ofBφ/Bpol alone can be misleading as in centrifugally drivenflows this value can be as high as 10 (Blandford & Payne1982). Close to the disk surface one can check the inclina-tion of the magnetic field lines with respect to the verticalaxis. The field lines have to be inclined by more than 30

for centrifugal acceleration to work (Blandford & Payne1982). Although this criterion is an exact solution of theideal, stationary and axisymmetric MHD equations for anoutflow from a Keplerian disk, its applicability is limited tothe surface of the disk. A criterion to determine the drivingmechanism above the disk was used by Tomisaka (2002)comparing the centrifugal forceFc and the magnetic forceFmag. By projecting both forces on the poloidal magnetic

field lines it can be determined which force dominates theacceleration. For the outflow to be driven centrifugally,Fc

has to be larger thanFmag. However, for this criterion to beself-consistent the gravitational force and the fact that anytoroidal magnetic field would reduce the effect ofFc haveto be taken into account.

In Seifried et al. (2012), a general criterion was derivedto identify centrifugally driven regions of the outflows andto differentiate those from magnetic pressure driven out-flows. The derivation assumed a stationary axisymmetricflow, which leads to a set of constraint equations basedon conservation laws along magnetic field lines (see alsoBlandford & Payne 1982; Pudritz & Norman 1986; Pelletier& Pudritz 1992). This criterion is applicable throughout theentire outflow.

The general condition for outflow acceleration is (Seifriedet al. 2012)

∂pol

(

1

2v2φ +Φ−

vφvpol

1

4π

BφBpol

ρ+

1

4π

B2φ

ρ

)

< 0 (1)

where∂pol denotes the derivative along the poloidal mag-netic field. It describes all regions of gas acceleration in-cluding those dominated by the effect ofBφ. This generaloutflow criterion should be compared to the case of cen-trifugal acceleration whereBφ = 0, i.e. in the case with noresulting Lorentz force along the poloidal field line

r

z

1

GM

(

v2φr2

(r2 + z2)3/2 −GM

)

/(

Bz

Br

)

> 1 , (2)

whereM , r andz are the mass of the central object, thecylindrical radius and distance alongz-axis, respectively.Using both equations, one can distinguish between regionsdominated by centrifugal acceleration and those by thetoroidal magnetic pressure. Note that Eq. (2) does not as-sume a Keplerian disk, hence it is also applicable to early-type sub-Keplerian configurations.

An example of those outflow criteria is shown in Fig. 7where one can see that the centrifugally launched regionis narrower and closer to the rotation axis but faster thanthe outer part of the outflow. Generally, such early typeoutflows are driven by both mechanisms, i.e. by magneticpressure and magneto-centrifugal forces, but the centrifugallaunching should become more dominant while the under-lying disk evolves towards a more stable Keplerian config-uration.

Another, indirect, support for the outflow generationmechanism involving magnetic driving comes from a re-cent numerical study by Peters et al. (2012). Their simula-tions, which include feedback from ionizing radiation frommassive protostars, show pressure driven bipolar outflowsreminiscent of those observed around massive stars. Butdetailed analysis through synthetic CO maps show that thepressure driven outflows are typically too weak to explainthe observed ones (e.g., outflows in G5.89 Puga et al. 2006;Su et al. 2012). The failure suggests that a mechanism other

14

Fig. 7.— Application of the outflow criteria derived in Seifried et al. (2012). The left panels show that centrifugal accel-eration works mainly close to thez-axis up to a height of about 800 AU which agrees very well withthe region where thehighest velocities are found (see Eq.(2). The general criterion (see Eq. 1) is more volume filling and traces also regionsinthe outer parts [from Seifried et al. (2012)].

15

than ionizing radiation must be found to drive the massiveoutflows. Since massive star forming regions are observedto be significantly magnetized (e.g., Girart et al. 2009 andTang et al. 2009), the magnetic field is a natural candidatefor outflow driving: the observed massive outflows mightbe driven magnetically by the massive stars themselves, orby the collection of lower and intermediate-mass stars in theyoung massive cluster.

4.2.2. Outflow Collimation

A general finding of self-consistent numerical simula-tions is that the degree of outflow collimation is time de-pendent and depends on the initial field strength. At veryearly stages (103 – 104 yr) outflows in typically magnetized,massive cores (with mass-to-flux ratios ofλ . 5) are foundto be poorly collimated with collimation factors of 1 – 2 in-stead of 5 – 10, still in agreement with observations of out-flows around most young massive protostellar objects (e.g.Ridge & Moore 2001; Torrelles et al. 2003; Wu et al. 2004;Sollins et al. 2004; Surcis et al. 2011). It is suggestive thatduring the earliest stage, i.e. before the B1–B2-type phaseof the scenario described by Beuther & Shepherd (2005),the outflows are rather poorly collimated except in case ofan unusually weak magnetic field. In their further evolu-tion, however, the collimation will increase quickly due tothe development of a fast, central jet coupled to the build-up of a Keplerian disk. Therefore it might be problematicto directly link the evolutionary stage of the massive youngstellar object to the collimation of the observed outflow assuggested by Beuther & Shepherd (2005). Additional dif-ficulties to correlate ages of YSOs and the collimation ofoutflows arise from the fragmentation of massive disks. Forinstance, circum-system outflows (from around binaries orhigher multiples) from large sub-Keplerian disks at earlystages are possible. But those outflows are often uncolli-mated and might even show spherical morphologies due tofragmentation of the highly unstable accretion disk (e.g.,Peters et al. 2011; Seifried et al. 2012). Although, it seemslikely that the subsequent outflows from around single mas-sive protostars should be collimated, the evidence from nu-merical simulations of clustered star formation showing theself-consistent launching of such outflows is still missingdue to the lack of resolution.

Direct confirmation of those evolutionary scenarios isdifficult as one would need independent information of theage of the YSOs and details of the magnetization of theenvironment (see also the discussion of Ray 2009). Suchobservations are hard to obtain and therefore rather rare.However, there is an interesting observation that supportsthe picture of very early stage, poorly collimated outflowssuccessively collimating over time. Observing two spa-tially adjacent, massive protostars in the star forming regionW75N, Torrelles et al. (2003) and Surcis et al. (2011) findthe younger of the two having a spherical outflow whereasthe more evolved protostar has a well collimated outflow.Due to their close proximity to each other, they should have

similar environmental conditions. Therefore the differenceshould rather be a consequence of different evolutionarystages, where the younger, poorly collimated outflow is pos-sibly only a transient.

4.3. Feedback by Jets and Outflows

As mentioned earlier, jets and outflows are alreadypresent at very early stages of star formation. Hence, theirinfluence on the subsequent evolution within star form-ing regions may not be neglected. In particular, in clusterforming regions, outflows are believed to influence or evenregulate star formation (see also chapters by Krumholz etal. and Frank et al. in this volume) as originally proposed byNorman & Silk (1980). Since then a number of numericalsimulations tried to address this issue (e.g., Li & Nakamura2006; Banerjee et al. 2007; Nakamura & Li 2007; Baner-jee et al. 2009; Wang et al. 2010; Li et al. 2010; Hansenet al. 2012), but with different outcomes. Detailed sin-gle jet simulations demonstrated that the jet power doesnot couple efficiently to the ambient medium and is notable to drive volume-filling supersonic turbulence (Baner-jee et al. 2007). This is because the bow shock of a highlycollimated jet and developed jet instabilities mainly excitesub-sonic velocity fluctuations. Similarly, the simulationsby Hansen et al. (2012) showed that protostellar outflowsdo not significantly affect the overall cloud dynamics, atleast in the absence of magnetic fields. Otherwise, the re-sults from simulations of star cluster forming regions (e.g.,Li & Nakamura 2006; Nakamura & Li 2007; Wang et al.2010; Li et al. 2010) clearly show an impact of outflowson the cloud dynamics, the accretion rates and the star for-mation efficiency. But this seems to be only effective ifrather strong magnetic fields and a high amount of initialturbulence are present in those cloud cores. The jet energyand momentum are better coupled to a turbulent ambientmedium than a laminar one (Cunningham et al. 2009).

4.4. Future Research on Outflows around Protostars

Undoubtedly, jets and outflows from YSOs are stronglylinked through the magnetic field to both the disk and sur-rounding envelope. Deciphering the strength and mor-phological structure of magnetic fields of jets and out-flows will be a key to gain a better understanding of theseexciting phenomena. Unfortunately, there are very fewdirect measurements of magnetic field strengths in YSOjets to date. The situation should improve with the ad-vent of new radio instruments (see also Ray 2009). Forexample, theMagnetism Key Science Projectof LOFAR(http://www.lofar.org) plans to spatially resolvethe polarized structure of protostellar jets to examine theirmagnetic field structure and to investigate the impact of thefield on the launching and evolution of protostellar jets.Prime targets would be the star forming regions of Tau-rus, Perseus & Cepheus Flare molecular clouds with sub-arc-second resolution. The forthcomingSquare KilometerArray (SKA; http://www.skatelescope.org) will

16

also offer unprecedented sensitivity to probe the small scalestructure of outflows around protostars (Aharonian et al.2013).

5. Connecting Early Disks to Planet Formation

As we have seen, the first105 years in the life of a pro-tostellar disk are witness to the accretion of the bulk of thedisk mass, the rapid evolution of its basic dynamics, as wellas the most vigorous phase of its outflow activity. Thisis also the period when the basic foundations of the star’splanetary system are laid down. Giant planet formationstarts either by rapid gravitationally driven fragmentationin the more distant regions of massive disks (e.g., Mayer etal. 2002; Rafikov 2009), or by the formation of rocky plan-etary cores that over longer (Myr) time scales will accretemassive gaseous envelopes (e.g., Pollack et al. 1996, chap-ter by Helled et al.). Terrestrial planet formation is believedto occur as a consequence of the oligarchic collisional phasethat is excited by perturbations caused by the appearance ofthe giant planets (see chapter by Raymond et al.). Thereare several important connections between the first phasesof planet formation and the properties of early protostellardisks. We first give an overview of some of the essentialpoints, before focusing on two key issues.

At the most fundamental level, the disk mass is centralto the character of both star and planet formation. Mostof a star’s mass is accreted through its disk, while at thesame time, giant planets must compete for gas from thesame gas reservoir. Sufficiently massive disks, roughly atenth of the stellar mass, can in these early stages gener-ate strong spiral waves which drive rapid accretion onto thecentral star. Such disks are also prone to fragmentation.Early disk masses exceeding 0.01M⊙ provide a sufficientgas supply to quickly form Jovian planets (Weidenschilling1977). The lifetime of protostellar disks, known to be in therange 3–10 Myr, provides another of the most demandingconstraints on massive planet formation. Detailed studiesusing the Spitzer Space Telescope indicate that 80 % of gasdisks around stars of less than 2M⊙ have dissipated by 5Myr after their formation (Carpenter et al. 2006; Hernandezet al. 2008, 2010).

As already discussed, protostellar outflows are one of theearliest manifestations of star formation. Class 0 sourcesare defined as having vigorous outflows and this implies thatmagnetized disks are present at the earliest times. Magneticfields that thread such disks are required for the outflowlaunching. These fields also have a strong quenching ef-fect on the fragmentation of disks which has consequencesfor the gravitational instability picture of planet formation.How early does Keplerian behavior set in? Some simula-tions (e.g., Seifried et al. 2012b) suggest that, even withinthe first few104 years, the disk has already become Kep-lerian (§ 3) making the launch of centrifugally driven jetsall the more efficient. Before this, it is possible that angularmomentum transport by spiral waves is significant.

Stars form as members of star clusters and this may have

an effect on disk properties, and therefore, upon aspects ofplanet formation. Observations show that as much as 90% of the stars in the galactic disk originated in embeddedyoung clusters (Gutermuth et al. 2009). The disk fraction ofyoung stars in clusters such asλ Orionis (Hernandez et al.2010) and other clusters such as Upper Scorpius (Carpenteret al. 2006) is similar to that of more dispersed groups indi-cating that the dissipative time scale for disks is not stronglyaffected by how clustered the star formation process is. Wenote in passing that the late time dissipation of disks is con-trolled both by photo-evaporation, which is dominated bythe FUV and X-ray radiation fields of the whole cluster andnot of the host star, as well as by the clearing of holes indisks by multiple giant planets (see chapter by Espaillat etal.). Calculations show that FUV radiation fields, producedmainly by massive stars, would inhibit giant planet forma-tion in 1/3 to 2/3 of planetary systems, depending on thedust attenuation. However, this photo-evaporation affectsmainly the outer regions of disks, leaving radii out to 35AU relatively unscathed (Holden et al. 2011).

Rocky planetary cores with of order 10 Earth masses areessential for the core accretion picture of giant planets. Thisprocess must take place during the early disk phase — thefirst 105 years or so (see chapter by Johansen et al.) in orderto allow enough time for the accretion of a gas envelope.Therefore the appearance of planetesimals out of whichsuch giant cores are built, must also be quite rapid and isanother important part of the first phases of giant planet for-mation in early protostellar disks. It is important to realizetherefore, that the various aspects of non-ideal MHD dis-cussed in the context of disk formation in section 3, beingdependent on grain properties, take place in a rapidly evolv-ing situation wherein larger grains settle, agglomerate, andgo on to pebble formation.

Finally, a major factor in the development of planetarysystems in early disks arises from the rapid migration ofthe forming planetary cores. As is well known, the effi-cient exchange of planetary orbital angular momentum witha gaseous disk by means of Lindblad and co-rotation reso-nances leads to very rapid inward migration of small coreson 105 year time scales (Ida & Lin 2008; see chapter byBenz et al.). One way of drastically slowing such migrationis by means of planet traps — which are regions of zero nettorque on the planet that occur at disk (Masset et al. 2006;Matsumura et al. 2009; Hasegawa & Pudritz 2011, 2012).Inhomogeneities in disks, such as dead zones, ice lines, andheat transitions regions (from viscous disk heating to stellarirradiation domination) form special narrow zones wheregrowing planets can be trapped. The early appearance ofdisk inhomogeneities and such planet traps encodes the ba-sic initial architecture of forming planetary systems. Wenow turn to a couple of these major issues in more detail.

5.1. Fragmentation in Early Massive Disks

Early disks are highly time-dependent, with infall of thecore continuously delivering mass and angular momentum

17

to the forming disk. Moreover, the central star is still form-ing by rapid accretion of material through the disk. De-pending on the infall rate, the disk may or may not be self-gravitating (see below). Disk properties and masses canbe measured towards the end of this accretion phase in theClass I sources. A CARMA (Combined Array for Researchin Millimeter-wave Astronomy) survey of 10 Class I diskscarried out by Eisner (2012) as an example showed that onlya few Class I disks exceed0.1 M⊙, and the range of massesfrom< 0.01 to> 0.1 M⊙ exceeds that of disks in the ClassII phase. This is a tight constraint on the formation of mas-sive planets and already suggests that the process must havebeen well on its way before even the Class I state has beenreached.

Two quantities that control the gravitational stability ofa hydrodynamic disk are the ratio of the disk mass tothe total mass of the systemµ = Md/(Md + M∗) andthe Toomre instability parameterQ = csκ/πGΣd, whereΣd = Md/(2πR

2d) is the surface mass density of the disk.

Heating and cooling of the disk as well as its general evolu-tion alter Q and infall onto the disk changesµ (Kratter et al.2008). The mass balance in an early disk will depend uponthe efficiency of angular momentum transport, which iswidely believed to be governed by two mechanisms: gravi-tational torque as well as the magneto-rotational instability(MRI). Angular momentum transport through the disk bythese agents can be treated as “effective” viscosities:αGI

andαMRI ≃ 10−2. These drive the total disk viscosityα = αGI(Q,µ) + αMRI . The data produced by numeri-cal simulations can be used to estimateαGI in terms of Qandµ (e.g. simulations of Vorobyov & Basu 2005, 2006).The resulting accretion rate onto the central star can be writ-ten in dimensionless form asM∗/MdΩ, where the epicyclicfrequency has been replaced by the angular frequencyΩ ofthe disk. The accompanying Fig. 8 shows the accretion ratethrough the disks in this phase space. The figure shows thatthe greatest part of this disk Q-µ phase space is dominatedby MRI transport rather than by gravitational torques fromspiral waves. Evolutionary tracks for accreting stars can becomputed in this diagram, which can be used to trace theevolution of the disks. Low mass stars will have lower val-ues ofµ, MRI dominated evolution, masses of order 30 %of the system mass, and have typical outer radii of order 50AU. High mass stars by comparison are predicted to havehigh values ofµ ≃ 0.35 and an extended period of localfragmentation as the accretion rates peak, as well as a diskouter edge at 200 AU.

The fragmentation of disks is markedly affected by thepresence of significant magnetic fields. One of the maineffects of a field is to modify the Toomre criterion. Be-cause part of the action of the threading field in a disk isto contribute a supportive magnetic pressure, the ToomreQ parameter is modified with the Alfven velocityvA (thetypical propagation speed of a transverse wave in a mag-netic field);QM = (c2s + v2A)

1/2κ/πGΣd. For typical val-ues of the mass to flux ratio, magnetic energy densities indisk are comparable to thermal or turbulent energy densi-

Fig. 8.— Contours of the dimensionless accretion rateM∗/(MdΩ) from the disk onto the star from both trans-port components of the model. The lowest contour level is10−4.8, and subsequent contours increase by 0.3 dex. Theeffect of each transport mechanism is apparent in the cur-vature of the contours. The MRI causes a mild kink in thecontours across the Q=2 boundary and is more dominant athigher disk masses due to the assumption of a constant diskturbulence parameter. Adapted from Kratter et al. (2008)

ties. Seifried et al. (2011) find that the magnetic suppres-sion of disk fragmentation occurs in most of their models.Even the presence of ambipolar diffusion of the disk fielddoes not significantly enhance the prospects for fragmenta-tion (Duffin & Pudritz 2009).

Gravitational fragmentation into planets or low masscompanions requires thatQ ∼ 1, however this is not suffi-cient. Fragments must also cool sufficiently rapidly as wasfirst derived by Gammie (2001), and generalized by Kratter& Murray-Clay (2011). The condition for sufficiently rapidcooling depends, in turn, upon how the disk is heated. Theinner regions of disks are dominated by viscous heating,which changes into dominant radiative heating by irradia-tion of the disk by the central star, in the outer regions (Chi-ang & Goldreich 1997). The transition zone between thesetwo regions, which has been called a “heat transition radius”(Hasegawa & Pudritz 2011) is of importance both from theview of gravitation fragmentation (Kratter & Murray-Clay2011) as well as for the theory of planet traps. This radiusoccurs where the heating of the surface of the disk by irra-diation by the central star balances the heating of the disk atthe midplane by viscous heating. Stellar irradiation domi-nates viscous heating if the temperatureT exceeds a criticalvalue: T > [(9/8)(αΣ/σ)(k/µ)τRΩ]

1/3, whereτR is theRosseland mean opacity,α is the viscosity parameter,σ isthe Stefan-Boltzmann constant,k is Boltzmann’s constant,andΩ is the orbital angular frequency.

The long term survival of fragments depends upon threedifferent forces including gas pressure, shearing in the disk,

18

and mutual interaction and collisions (Kratter & Murray-Clay 2011). The role of pressure in turn depends upon howquickly the gas can cool and upon its primary source ofenergy. Fragmentation in the viscously heated regime —which is where giant planets may typically form — can oc-cur if the ratio of the cooling time to the dynamical timeβ is sufficiently small for a gas with adiabatic indexγ:β < ([4/9γ(γ − 1)]α−1

sat, where the saturated value of theviscousα parameter refers to the turbulent amplitude thatcan be driven by gravitational instability. Infall plays animportant role in controlling this fragmentation. Generally,a higher infall rateM drives the value ofQ downwardsas seen in the Figure. Rapid infall will tend to drive diskscloser to instability therefore, and perhaps even on to frag-mentation. Irradiated disks, by contrast, have a harder timeto fragment. The basic point here is that while gravitation-ally driven turbulence can be dissipated to maintainQ = 1,irradiated disks do not have this property. Once an irra-diated disk moves into a critical Q regime, they are moreliable to fragment since there is no intrinsic self-regulatorymechanism for maintaining disk temperature near a criti-cal value. The results indicate that irradiated disks can bedriven by infall to fragment at lower accretion rates onto thedisk.

Spiral arms compress the disk gas and are the most likelysites for fragmentation. A key question is what are thetypical masses of surviving fragments. Recent numericalsimulations of self-gravitating disks, without infall, havegone much farther into the non-linear regimes to follow thefragmentation into planet scale objects (e.g., Boley et al.2010; Rogers & Wadsley 2012). Using realistic coolingfunctions for disks, Rogers & Wadsley (2012) have simu-lated self-gravitating disks and have found a new criterionfor the formation of bound fragments. Consider a patchof a disk that has been compressed into a spiral arm ofthicknessl1 (see Figure 9). This arm will form gravita-tionally bound fragments ifl1 lies within the Hill radiusHHill = [GΣll

21/3Ω

2]1/3 of the arm, or:(l1/2HHill) < 1.This criterion also addresses the ability of shear to pre-

vent the fragmentation of the arm. Numerical simulationsverify that this criterion describes the survival of gravita-tionally induced fragments in the spiral arms. The resultshave been applied to disks around A stars and show thatfragments of masses15MJup can form and survive at largedistances of the order of 95 AU from the central star, inthis radiation heating dominated regime. Brown dwarf scalemasses seem to be preferred.

Finally we note that clumps formed in the outer parts ofdisks may collapse very efficiently, perhaps on as little asthousand year time scales. This suggests that clumps lead-ing to the formation of giant planets could collapse quicklyand survive transport to the interior regions of the disk (Gal-vagni et al. 2012). We turn now to planetary transportthrough disks.

Fig. 9.— The Hill criterion for spiral arm fragmentation: ifa piece of the spiral arm of widthl1 lies within its own Hillthickness, then that section of the arm is free to collapse andfragmentation takes place. If such a section of the spiral armlies outside of its own Hill thickness, then shear stabilizesthe arm and fragmentation does not take place. Adaptedfrom Rogers & Wadsley (2012)

5.2. Planet Traps and the Growth and Radial Migra-tion of Planets

The survival of planetary cores in early disks faces an-other classic problem arising from the exchange of orbitalangular momentum between the low mass planetary coreand its surrounding gaseous disk. In these early phases,protoplanetary cores can raise significant wakes or spiralwaves to their interior and exterior radial regions of the disk.These waves in turn exert torques back on the planet result-ing in its migration. For homogeneous disks with smooth,decreasing density and temperature profiles, the inner wake(which transfer angular momentum to the protoplanet —driving it outwards) is slightly overcome by the outer wake(which extracts orbital angular momentum from the proto-planet driving it inwards). This results in a net torque whichresults in the rapid inward (Type I) migration of the planet.As is well known, Monte Carlo population synthesis cal-culations on protoplanets in evolving accreting disks findthat protoplanetary cores migrate into the center of the diskwithin 105 years (Ida & Lin 2005, 2008) — the timescalecharacterizing early disks.

Rapid Type I migration must be slowed down by at leasta factor of 10 to make it more compatible with the lifetimeof the disk (see chapter by Benz et al.). As noted, thiscan be achieved in disks at density and temperature inho-mogeneities at whose boundaries, migration can be rapidlyslowed, or even stopped. We have already encountered twoof these in other contexts, namely dead zones and heat tran-sitions wherein there is a change in density and in disk tem-perature gradient respectively. Heat transitions were shownto be planet traps by (Hasegawa & Pudritz 2011). A thirdtype of trap is the traditional ice line, wherein the temper-ature transition gives rise to a change in disk opacity with

19

a concomitant fluctuation in the density (Ida & Lin 2008).Hasegawa & Pudritz (2011, 2012) showed that protoplane-tary cores can become trapped at these radii.

Dead zone edges act to stop planetary migration as hasbeen shown in theory (Masset et al. 2006) and numeri-cal simulations (Matsumura et al. 2009). A region that isstarved of ionization, which occurs where disk column den-sities are high enough to screen out ionizing cosmic rays,has large Ohmic dissipation which prevents the operationof the MRI instability. This region is known as a deadzone because it is unable to generate the MHD turbulencenecessary to sustain a reasonable viscosityαMRI (Gammie1996). A dead zone is most likely to be present in the in-ner regions of disks in their later stages. During the earlieststages of disk evolution, when the disks are the most mas-sive, these would extend out to 10 AU or so (Matsumura &Pudritz 2006). It is necessary to transport away disk angu-lar momentum to allow an ongoing large accretion flow thatis measured for young stars. Older models supposed thata thin surface layer is sufficiently well ionized to supportMRI turbulence. Recent simulations show, however, thatsuch MRI active surface layers may not occur (Bai & Stone2013). The inclusion of all three of the non-ideal effects dis-cussed in§ 3 (Ohmic resistivity, Hall effect, and ambipolardiffusion) strongly changes the nature of MHD instability invertically stratified disks. Going from the dense mid planeto the surface of the disk, the dominant dissipation mecha-nism changes: from Ohmic resistivity at the mid plane, toHall effect at mid-scale heights, to ambipolar diffusion inthe surface layers. The results show that ambipolar diffu-sion shuts down the MRI even in the surface layers in thepresence of a net (non-zero) magnetic flux through the disk.Instead of the operation of an MRI in the dead zone, a strongMHD disk wind is launched and this carries off the requi-site disk angular momentum very efficiently (e.g. review;Pudritz et al. 2007,§ 4 of this paper).

Planetary cores accrete most of their mass while mov-ing along with such traps. As the disk accretion rate fallsin an evolving disk, the traps move inward at differentrates. At early times, with high accretion rates, the trapsare widely separated. As the disk accrete rate falls froma high of10−6M⊙ yr−1 to lower values at later times, thetraps slowly converge at small disk radii, which likely initi-ates planet-planet interactions.

In summary then, this section has shown that the prop-erties of early magnetized disks as characterized by highinfall rates and disk masses, as well as powerful outflows,can strongly influence the early phases of planet formationand migration.

6. Synthesis

Both observational and theoretical studies of disk for-mation are poised for rapid development. Observation-ally, existing dust continuum surveys of deeply embedded“Class 0” objects indicate that a compact emission com-ponent apparently distinct from the protostellar envelope

is often present. Whether this component is a rotationallysupported disk (RSD) or not is currently unclear in gen-eral. With unprecedented sensitivity and resolution, ALMAshould settle this question in the near future.

On the theoretical side, recent development has beenspurred largely by the finding that magnetic braking is soefficient as to prevent the formation of a RSD in laminardense cores magnetized to realistic levels in the ideal MHDlimit — the so-called “magnetic braking catastrophe.” Al-though how exactly this catastrophe can be avoided remainsunclear, two ingredients emerge as the leading candidatesfor circumventing the excessive braking — Ohmic dissipa-tion and complex flow pattern (including turbulence, mis-alignment of magnetic field and rotation axis, and possi-bly irregular core shapes): the former through the decou-pling of magnetic field lines from the bulk neutral mat-ter at high densities and the latter through, at least inpart, turbulence/misalignment-induced magnetic reconnec-tion. It is likely that both ingredients play a role, withOhmic dissipation enabling a dense, small (perhaps AU-scale) RSD to form early in the protostellar accretion phase,and flow complexity facilitating the growth of the disk atlater times by weakening the magnetic braking of the lowerdensity protostellar accretion flow.

The above hybrid picture, although probably not unique,has the virtue of being at least qualitatively consistent withthe available observational and theoretical results. Thesmall Ohmic dissipation-enabled disk can in principle drivepowerful outflows that are a defining characteristic of Class0 sources from close to the central object where the mag-netic field and matter are well coupled due to thermal ion-ization of alkali metals. During the Class 0 phase, it mayhave grown sufficiently in mass to account for the compactcomponent often detected in interferometric continuum ob-servations, but not so much in size as to violate the con-straint that the majority of the compact components remainunresolved to date. The relatively small size of early diskscould result from magnetic braking.

A relatively small early disk could also result from asmall specific angular momentum of the core material tobegin with. There is a strong need to determine more sys-tematically the magnitude and distribution of angular mo-mentum in prestellar cores through detailed observations.Another need is to determine the structure of the magneticfields on the102–103 AU scale that is crucial to disk for-mation. For example, detection of magnetic field twistingwould be direct evidence for magnetic braking. With thepolarization capability coming online soon, ALMA is ex-pected to make progress on this observational front.

On the theory front, there is a strong need to carry outsimulations that combine non-ideal MHD effects with tur-bulence and complex initial conditions on the core scale, in-cluding magnetic field-rotation misalignment. This will betechnically challenging to do, but is required to firm up diskformation scenarios such as the hybrid one outlined above.

All of the evidence and theory shows that the formationof outflows is deeply connected with the birth of magne-

20

tized disks during gravitational collapse. Early outflows andlater higher speed jets may be two aspects of a common un-derlying physical picture in which acceleration is promotedboth by toroidal magnetic field pressure on larger scales aswell as centrifugal “fling” from smaller scales. Simulations,theory, and observations are converging on the idea that thecollapse and outflow phenomenon is universal covering thefull range of stellar mass scales from brown dwarfs to mas-sive stars. Finally, the earliest stages of planet formationtake place on the very same time scales as disks are formedand outflows are first launched. While little is yet knownabout this connection, it is evident that this must be im-portant. Many aspects of planet formation are tied to theproperties of early disks.

In summary, firm knowledge of disk formation will pro-vide a solid foundation for understanding the links betweenearly disks, outflows, and planets, opening the way to dis-covering the deep connections between star and planet for-mation.

Acknowledgments.ZYL is supported in part by NASANNX10AH30G, NNX14AB38G, and NSF AST1313083,RB by the Deutsche Forschungsgemeinschaft (DFG) viathe grant BA 3706/1-1, REP by a Discovery Grant fromthe National Science and Engineering Research Council(NSERC) of Canada, and JKJ by a Lundbeck FoundationJunior Group Leader Fellowship as well as Centre for Starand Planet Formation, funded by the Danish National Re-search Foundation and the University of Copenhagen’s pro-gramme of excellence.

REFERENCES

Aharonian, F. et al. 2013, arXiv:1301.4124Allen, A. et al. 2003,Astrophys. J., 599, 363Alves, F. O. et al. 2008,Astron. Astrophys., 486, L13Alves, J. et al. 2007,Astron. Astrophys., 462, L17Andre, P. et al. 1993,Astrophys. J., 406, 122Armitage, P. J. 2011,Annu. Rev. Astron. Astrophys., 49, 195Bacciotti, F. et al. 2002,Astrophys. J., 576, 222Bacciotti, F. et al. 2011,Astrophys. J. Lett., 737, L26Bai, X.-N., & Stone, J. M. 2013,Astrophys. J., 769, 76Banerjee, R. et al. 2009, in: Protostellar Jets in Context, (eds. K.

Tsinganos, T. Ray, & M. Stute), 421Banerjee, R. et al. 2007,Astrophys. J., 668, 1028Banerjee R., & Pudritz R. E. 2006,Astrophys. J., 641, 949Banerjee, R., & Pudritz, R. E. 2007,Astrophys. J., 660, 479Beck, R. 2001, Space Science Reviews, 99, 243Belloche, A. 2013 in: Role and Mechanisms of Angular Momen-

tum Transport During the Formation and Early Evolution ofStars, (eds. P. Hennebelle & C. Charbonnel), 25

Belloche, A. et al. 2002,Astron. Astrophys., 393, 927Bergin, E. A., & Tafalla, M. 2007,Annu. Rev. Astron. Astrophys.,

45, 339Bertram, E. et al. 2012,Mon. Not. R. Astron. Soc., 420, 3163Beuther, H. et al. 2004,Astrophys. J. Lett., 616, L23Beuther, H., & Shepherd, D. 2005, in: Cores to Clusters: Star

Formation with Next Generation Telescopes, (eds. M. S. N.Kumar, M. Tafalla, & P. Caselli), 105

Blandford, R. D., & Payne, D. G. 1982,Mon. Not. R. Astron. Soc.,199, 883

Bodenheimer, P. 1995,Annu. Rev. Astron. Astrophys., 33, 199Boley, A. C. et al. 2010, Icarus, 207, 509Boss, A. P. 1998 in: Origins, (eds. C. E. Woodward, J. M. Shull,

& H. A. Thronson, Jr.), 314Boss, A. P. 2002,Astrophys. J., 568, 743Boss, A. P., & Keiser, S. A. 2013,Astrophys. J., 764, 136Braiding, C. R., & Wardle, M. 2012a,Mon. Not. R. Astron. Soc.,

422, 261Braiding, C. R., & Wardle, M. 2012b,Mon. Not. R. Astron. Soc.,

427, 3188Brinch, C. et al. 2007,Astron. Astrophys., 475, 915Brinch, C. et al. 2009,Astron. Astrophys., 502, 199Brown, D. W. et al. 2000,Mon. Not. R. Astron. Soc., 319, 154Burkert, A., & Bodenheimer, P. 2000,Astrophys. J., 543, 822Burzle, F. et al. 2011,Mon. Not. R. Astron. Soc., 417, L61Cabrit, S. et al. 1997, in: Herbig-Haro Flows and the Birth of Stars,

(eds. B. Reipurth & C. Bertout), 163Carpenter, J. M. et al. 2006,Astrophys. J. Lett., 651, L49Carrasco-Gonzalez, C. et al. 2010, Science, 330, 1209Caselli, P. et al. 2002,Astrophys. J., 572, 238Cerqueira, A. H. et al. 2006,Astron. Astrophys., 448, 231Cesaroni, R. et al. 1997,Astron. Astrophys., 325, 725Cesaroni, R. et al. 1994,Astrophys. J. Lett., 435, L137Chandler, C. J. et al. 2000,Astrophys. J., 530, 851Chapman, N. L. et al. 2013,Astrophys. J., 770, 151Chiang, E. I., & Goldreich, P. 1997,Astrophys. J., 490, 368Chiang, H.-F. et al. 2008,Astrophys. J., 680, 474Chiang, H.-F. et al. 2012,Astrophys. J., 756, 168Chibueze, J. O. et al. 2012,Astrophys. J., 748, 146Chrysostomou, A. et al. 2008,Astron. Astrophys., 482, 575Ciardi, A., & Hennebelle, P. 2010,Mon. Not. R. Astron. Soc., 409,

L39Ciolek, G. E., & Konigl, A. 1998,Astrophys. J., 504, 257Coffey, D. et al. 2007,Astrophys. J., 663, 350Comeron, F. et al. 2003,Astron. Astrophys., 406, 1001Commercon, B. et al. 2010,Astron. Astrophys., 510, L3Contopoulos, J. 1995,Astrophys. J., 450, 616Contopoulos, I. et al. 1998,Astrophys. J., 504, 247Crutcher, R. M. 2012,Annu. Rev. Astron. Astrophys., 50, 29Crutcher, R. M. et al. 1999,Astrophys. J. Lett., 514, L121Crutcher, R. M. et al. 2010,Astrophys. J., 725, 466Cunningham, A. J. et al. 2009,Astrophys. J., 692, 816Cunningham, A. J. et al. 2012,Astrophys. J., 744, 185Dapp, W. B., & Basu, S. 2010,Astron. Astrophys., 521, L56Dapp, W. B. et al. 2012,Astron. Astrophys., 541, A35Davidson, J. A. et al. 2011,Astrophys. J., 732, 97Dedner, A. et al. 2002, Journal of Computational Physics, 175,

645Dib, S. et al. 2010,Astrophys. J., 723, 425Duffin, D. F., & Pudritz, R. E. 2008,Mon. Not. R. Astron. Soc.,

391, 1659Duffin, D. F., & Pudritz, R. E. 2009,Astrophys. J. Lett., 706, L46Dutrey, A. et al. 2007, in: Protostars and Planets V, (eds. B.

Reipurth, D. Jewitt, & K. Keil), 495Eisner, J. A. 2012,Astrophys. J., 755, 23Enoch, M. L. et al. 2011,Astrophys. J. Suppl., 195, 21Enoch, M. L. et al. 2009,Astrophys. J., 707, 103Evans, N. J., II et al. 2009,Astrophys. J. Suppl., 181, 321Fang, M. et al. 2009,Astron. Astrophys., 504, 461Favre, C. et al. 2014,Astrophys. J., submittedFendt, C. 2011,Astrophys. J., 737, 43Fendt, C., & Camenzind, M. 1996, Astrophysical Letters and

21

Communications, 34, 289Fernandez-Lopez, M. et al. 2011,Astron. J., 142, 97Franco-Hernandez, R. et al. 2009,Astrophys. J., 701, 974Fromang, S. et al. 2006,Astron. Astrophys., 457, 371Galli, D. et al. 2006,Astrophys. J., 647, 374Galli, D., & Shu, F. H. 1993,Astrophys. J., 417, 220Galvagni, M. et al. 2012,Mon. Not. R. Astron. Soc., 427, 1725Gammie, C. F. 1996,Astrophys. J., 457, 355Gammie, C. F. 2001,Astrophys. J., 553, 174Girart, J. M. et al. 2009,Astrophys. J., 694, 56Goodman, A. A. et al. 1993,Astrophys. J., 406, 528Gutermuth, R. A. et al. 2009,Astrophys. J. Suppl., 184, 18Guzman, A. E. et al. 2012,Astrophys. J., 753, 51Hansen, C. E. et al. & Fisher, R. T. 2012,Astrophys. J., 747, 22Harsono, D. et al. 2013,Astron. Astrophys., submittedHartigan, P. et al. 1995,Astrophys. J., 452, 736Harvey, D. W. A. et al. 2003,Astrophys. J., 583, 809Hasegawa, Y., & Pudritz, R. E. 2011,Mon. Not. R. Astron. Soc.,

417, 1236Hasegawa, Y., & Pudritz, R. E. 2012,Astrophys. J., 760, 117Heiles, C., & Troland, T. H. 2005,Astrophys. J., 624, 773Hennebelle, P., & Ciardi, A. 2009,Astron. Astrophys., 506, L29Hennebelle, P., & Fromang, S. 2008,Astron. Astrophys., 477, 9Hernandez, J. et al. 2008,Astrophys. J., 686, 1195Hernandez, J. et al. 2010,Astrophys. J., 722, 1226Hogerheijde, M. R., & Sandell, G. 2000,Astrophys. J., 534, 880Holden, L. et al. 2011, PASP, 123, 14Hosking, J. G., & Whitworth, A. P. 2004,Mon. Not. R. Astron.

Soc., 347, 1001Hosokawa, T. et al. 2010,Astrophys. J., 721, 478Hull, C. L. H. et al. 2013,Astrophys. J., 768, 159Ida, S., & Lin, D. N. C. 2005,Astrophys. J., 626, 1045Ida, S., & Lin, D. N. C. 2008,Astrophys. J., 685, 584Isokoski, K. et al. 2013,Astron. Astrophys., 554, 100Joergens, V. et al. 2013, Astronomische Nachrichten, 334, 159Joos, M. et al. 2012,Astron. Astrophys., 543, A128Joos, M. et al. 2013,Astron. Astrophys., 554, A17Jørgensen, J. K. 2004,Astron. Astrophys., 424, 589Jørgensen, J. K. et al. 2007,Astrophys. J., 659, 479Jørgensen, J. K. et al. 2005a,Astrophys. J., 632, 973Jørgensen, J. K. et al. 2004,Astron. Astrophys., 413, 993Jørgensen, J. K. et al. 2005b,Astrophys. J. Lett., 631, L77Jørgensen, J. K., & van Dishoeck, E. F. 2010,Astrophys. J. Lett.,

710, L72Jørgensen, J. K. et al. 2009,Astron. Astrophys., 507, 861Keene, J., & Masson, C. R. 1990,Astrophys. J., 355, 635Keto, E. 2002,Astrophys. J., 580, 980Keto, E. 2003,Astrophys. J., 599, 1196Krasnopolsky, R., & Konigl, A. 2002,Astrophys. J., 580, 987Krasnopolsky, R. et al. 2010,Astrophys. J., 716, 1541Krasnopolsky, R. et al. 2011,Astrophys. J., 733, 54Krasnopolsky, R. et al. & Zhao, B. 2012,Astrophys. J., 757, 77Kratter, K. M. et al. 2008,Astrophys. J., 681, 375Kratter, K. M., & Murray-Clay, R. A. 2011,Astrophys. J., 740, 1Krumholz, M. R. et al. 2013,Astrophys. J. Lett., 767, L11Kunz, M. W., & Mouschovias, T. C. 2010,Mon. Not. R. Astron.

Soc., 408, 322Kwon, W. et al. 2009,Astrophys. J., 696, 841Lee, C.-F. 2011,Astrophys. J., 741, 62Lee, K. et al. & Tobin, J. 2012,Astrophys. J., 761, 171Li, Z.-Y. et al. 2011,Astrophys. J., 738, 180Li, Z.-Y. et al. 2013,Astrophys. J., 774, 82

Li, Z.-Y., & McKee, C. F. 1996,Astrophys. J., 464, 373Li, Z.-Y., & Nakamura, F. 2006,Astrophys. J. Lett., 640, L187Li, Z.-Y., & Shu, F. H. 1996,Astrophys. J., 472, 211Li, Z.-Y. et al. & Nakamura, F. 2010,Astrophys. J. Lett., 720, L26Liu, J., & Schneider, T. 2009, AGU Fall Meeting Abstracts, P32C-

04Lommen, D. et al. & Crapsi, A. 2008,Astron. Astrophys., 481, 141Looney, L. W. et al. 2000,Astrophys. J., 529, 477Looney, L. W. et al. 2003,Astrophys. J., 592, 255Lynden-Bell, D. 1996,Mon. Not. R. Astron. Soc., 279, 389Lynden-Bell, D. 2003,Mon. Not. R. Astron. Soc., 341, 1360Machida, M. N. et al. 2006,Astrophys. J. Lett., 649, L129Machida, M. N. et al. 2007,Astrophys. J., 670, 1198Machida, M. N. et al. 2008,Astrophys. J., 676, 1088Machida, M. N. et al. 2010,Astrophys. J., 724, 1006Machida, M. N. et al. 2011, PASJ, 63, 555Machida, M. N., & Matsumoto, T. 2011,Mon. Not. R. Astron.

Soc., 413, 2767Machida, M. N. et al. & Tomisaka, K. 2005b,Mon. Not. R. Astron.

Soc., 362, 382Machida, M. N. et al. & Hanawa, T. 2005a,Mon. Not. R. Astron.

Soc., 362, 369Machida, M. N. et al. 2004,Mon. Not. R. Astron. Soc., 348, L1Mac Low, M.-M., & Klessen, R. S. 2004, Reviews of Modern

Physics, 76, 125Masset, F. S. et al. & Ferreira, J. 2006,Astrophys. J., 642, 478Matsumoto T., & Tomisaka K. 2004,Astrophys. J., 616, 266Matsumura, S., & Pudritz, R. E. 2006,Mon. Not. R. Astron. Soc.,

365, 572Matsumura, S. et al. 2009,Astrophys. J., 691, 1764Maury, A. J. et al. 2010,Astron. Astrophys., 512, A40Mayer, L. et al. 2002, Science, 298, 1756McCaughrean, M. J., & O’Dell, C. R. 1996,Astron. J., 111, 1977McKee, C. F., & Ostriker, E. C. 2007,Annu. Rev. Astron. Astro-

phys., 45, 565Mellon, R. R., & Li, Z.-Y. 2008,Astrophys. J., 681, 1356Mellon, R. R., & Li, Z.-Y. 2009,Astrophys. J., 698, 922Men’shchikov, A. et al. 2010,Astron. Astrophys., 518, L103Mohanty, S. et al. 2007,Astrophys. J., 657, 1064Molinari, S. et al. 2010,Astron. Astrophys., 518, L100Motogi, K. et al. 2013,Mon. Not. R. Astron. Soc., 428, 349Mouschovias, T. C., & Ciolek, G. E. 1999, in: The Origin of Stars

and Planetary Systems (eds. C. J. Lada & N. D. Kylafis), 305Mouschovias, T. C., & Paleologou, E. V. 1979,Astrophys. J., 230,

204Mouschovias, T. C., & Paleologou, E. V. 1980,Astrophys. J., 237,

877Murillo, N. M. et al. 2013,Astron. Astrophys., 560, A103Myers, A. T. et al. 2013,Astrophys. J., 766, 97Nakamura, F., & Li, Z.-Y. 2007,Astrophys. J., 662, 395Nakano, T. 1984, Fund. Cosmic Phys., 9, 139Nakano, T., & Nakamura, T. 1978, PASJ, 30, 671Nakano, T. et al. 2002,Astrophys. J., 573, 199Natta, A. et al. 2007, in: Protostars and Planets V, (eds. B.

Reipurth, D. Jewitt, & K. Keil), 767Nishi, R. et al. 1991,Astrophys. J., 368, 181Norman, C., & Silk, J. 1980,Astrophys. J., 238, 158O’Dell, C. R., & Wen, Z. 1992,Astrophys. J., 387, 229Ouyed, R., & Pudritz, R. E. 1997,Astrophys. J., 482, 712Padovani, M. et al. 2013,Astron. Astrophys., in press

(arXiv:1310.2158)Palau, A. et al. 2013,Mon. Not. R. Astron. Soc., 428, 1537

22

Pelletier, G., & Pudritz, R. E. 1992,Astrophys. J., 394, 117Persson, M. V. et al. 2012,Astron. Astrophys., 541, A39Peters, T. et al. & Mac Low, M.-M. 2011,Astrophys. J., 729, 72Peters, T. et al. Mac Low, M.-M., Galvan-Madrid, R., & Keto, E.

R. 2010,Astrophys. J., 711, 1017Peters, T. et al. 2012,Astrophys. J., 760, 91Pilbratt, G. L. et al. 2010,Astron. Astrophys., 518, L1Pineda, J. E. et al. 2012,Astron. Astrophys., 544, L7Pollack, J. B. et al. 1996, Icarus, 124, 62Price, D. J., & Bate, M. R. 2007,Astrophys. Space Sci., 311, 75Price, D. J. et al. 2012,Mon. Not. R. Astron. Soc., 423, 45Pudritz, R. E. et al. 2009, in: Structure Formation in Astrophysics,

(ed. G. Chabrier), 84Pudritz, R. E., & Norman, C. A. 1983,Astrophys. J., 274, 677Pudritz, R. E., & Norman, C. A. 1986,Astrophys. J., 301, 571Pudritz, R. E. et al. 2007, in: Protostars and Planets V, (eds. B.

Reipurth, D. Jewitt, & K. Keil), 277Puga, E. et al. 2006,Astrophys. J., 641, 373Qiu, K. et al. 2011,Astrophys. J. Lett., 743, L25Qiu, K. et al. 2007,Astrophys. J., 654, 361Qiu, K. et al. 2008,Astrophys. J., 685, 1005Rafikov, R. R. 2009,Astrophys. J., 704, 281Ray, T. P. 2009, Revista Mexicana de Astronomıa y Astrofısica

Conference Series, 36, 179Ray, T. et al. 2007, in: Protostars and Planets V, (eds. B. Reipurth,

D. Jewitt, & K. Keil), 231Reipurth, B., & Bally, J. 2001,Annu. Rev. Astron. Astrophys., 39,

403Ricci, L. et al. 2010,Astron. Astrophys., 521, 66Ridge, N. A., & Moore, T. J. T. 2001,Astron. Astrophys., 378, 495Rodrıguez, T. et al. 2012,Astrophys. J., 755, 100Rogers, P. D., & Wadsley, J. 2012,Mon. Not. R. Astron. Soc., 423,

1896Santos-Lima, R. et al. 2012,Astrophys. J., 747, 21Santos-Lima, R. et al. 2013,Mon. Not. R. Astron. Soc., 429, 3371Seifried, D. et al. 2011,Mon. Not. R. Astron. Soc., 417, 1054Seifried, D. et al. 2012b,Mon. Not. R. Astron. Soc., 423, L40Seifried, D. et al. 2013,Mon. Not. R. Astron. Soc., 432, 3320Seifried, D. et al. 2012,Mon. Not. R. Astron. Soc., 422, 347Shepherd, D. S., & Churchwell, E. 1996a,Astrophys. J., 457, 267Shepherd, D. S., & Churchwell, E. 1996b,Astrophys. J., 472, 225Shibata, K., & Uchida, Y. 1985, PASJ, 37, 31Shu, F. H. et al. 1987,Annu. Rev. Astron. Astrophys., 25, 23Shu, F. H. et al. 2006,Astrophys. J., 647, 382Sicilia-Aguilar, A. et al. 2010,Astrophys. J., 710, 597Smith, R. J. et al. 2013,Astrophys. J., 771, 24Smith, R. J. et al. 2012,Astrophys. J., 750, 64Soker, N. 2005,Astron. Astrophys., 435, 125Sollins, P. K. et al. 2004,Astrophys. J. Lett., 616, L35Su, Y.-N. et al. 2012,Astrophys. J. Lett., 744, L26Surcis, G. et al. 2011,Astron. Astrophys., 527, A48Takakuwa, S. et al. 2012,Astrophys. J., 754, 52Tang, Y.-W. et al. 2009,Astrophys. J., 700, 251Tassis, K., & Mouschovias, T. C. 2005,Astrophys. J., 618, 783Tassis, K., & Mouschovias, T. C. 2007,Astrophys. J., 660, 388Terebey, S. et al. 1993,Astrophys. J., 414, 759Tobin, J. J. et al. 2012a,Astrophys. J., 748, 16Tobin, J. J. et al. 2012b, Nature, 492, 83Tobin, J. J. et al. 2010,Astrophys. J., 712, 1010Tomida, K. et al. 2013,Astrophys. J., 763, 6Tomisaka, K. 1998,Astrophys. J. Lett., 502, L163Tomisaka, K. 2000,Astrophys. J. Lett., 528, L41

Tomisaka, K. 2002,Astrophys. J., 575, 306Torrelles, J. M. et al. 2003,Astrophys. J. Lett., 598, L115Troland, T. H., & Crutcher, R. M. 2008,Astrophys. J., 680, 457Visser, R. et al. 2013,Astrophys. J., 769, 19Visser, R. et al. 2009,Astron. Astrophys., 495, 881Vorobyov, E. I., & Basu, S. 2005,Astrophys. J. Lett., 633, L137Vorobyov, E. I., & Basu, S. 2006,Astrophys. J., 650, 956Wang, P. et al. 2010,Astrophys. J., 709, 27Wardle, M., & Ng, C. 1999,Mon. Not. R. Astron. Soc., 303, 239Weidenschilling, S. J. 1977,Astrophys. Space Sci., 51, 153Whelan, E. T. et al. 2009,Astrophys. J. Lett., 691, L106Whelan, E. T. et al. 2006, New A Rev., 49, 582Whelan, E. T. et al. 2005, Nature, 435, 652Whelan, E. T. et al. 2012,Astrophys. J., 761, 120Whelan, E. T. et al. 2007,Astrophys. J. Lett., 659, L45Wu, Y. et al. 2004,Astron. Astrophys., 426, 503Zhang, Q. et al. 2007,Astron. Astrophys., 470, 269Zhao, B. et al. 2011,Astrophys. J., 742, 10Ziegler, U. 2005,Astron. Astrophys., 435, 385

This 2-column preprint was prepared with the AAS LATEX macros v5.2.

23