Numerical Study of the IMF - Paolo Padoan Troels Haugbølle and Åke Nordlund

Page created by Randall Lee
 
CONTINUE READING
Numerical Study of the IMF - Paolo Padoan Troels Haugbølle and Åke Nordlund
Numerical Study of the IMF
                                                   Paolo Padoan
                             (ICREA - Institute of Cosmos Sciences - University of Barcelona)

                                  Troels Haugbølle and Åke Nordlund
                             (STARPLAN and Niels Bohr Institute - University of Copenhagen)

Friday, November 30, 12
Numerical Study of the IMF - Paolo Padoan Troels Haugbølle and Åke Nordlund
The Effect of Turbulence
                              If u >> cS and L >> λJ --> Turbulent Fragmentation:

            •     Velocity fields fragment the medium “independently” of gravity

            •     The strongest density peaks are then taken over by gravity

            •     We can use the universal statistical properties of the turbulence for a
                  statistical model of the SFR and the stellar IMF

Friday, November 30, 12
Numerical Study of the IMF - Paolo Padoan Troels Haugbølle and Åke Nordlund
density
                P    enhancements,
                     = MP  BE  =
                               +  1.182
                                  P        but
                                            =   Pfar from(1   +! clear
                                                                 M    2MBE,t
                                                                        )   that ⌘  turbulent
                                                                                       1.182       pressure can be used    (11)
         The     universality
                  0       th,0        of
                                    dyn,0
          as a source of internal support   G PIMF
                                            the
                                              3/2     1/2
                                                   th,0
                                                       to      peak
                                                             define     a   from
                                                                            critical  turbulent
                                                                                        mass    G 3/2
                                                                                                for    P  fragmentation
                                                                                                          1/2
                                                                                                     collapse).
                                                                                                          0
                                                  2 th,0
   P0 = Pth,0 + Pdyn,0 = Pth,0 (1 + M )
    The   characteristic
                Substituting
     If turbulence             mass
                                  intocan
                        is present,     we    be
                                         thehave    derived
                                               previous           as the
                                                               expression
                                                      to include                “turbulent”    , weBonnor-Ebert
                                                                                    of MBEpressure
                                                                          its dynamic                 get in a modified
                                                                                                                the totalmass:
                                                                                                                           turbu-
                                                                                                                            ex-
     ternallent
   Substituting   BE
              pressure,mass:
                    into    and
                          the     if the turbulence
                               previous     expression4        is M
                                                              of   highly
                                                                       BE   , wesupersonic,
                                                                                    get   a        its
                                                                                             modified   dynamic
                                                                                                        4 turbu-     pressure
                                                      th                                                th
     is dominant,
 nt BE mass:                M  BE  =  1.182               1/2
                                                                  !       M   BE,t   ⌘   1.182             1/2
                                                                                                                           (11)
                                                  3/2
                                               G MBE,0 Pth,0(1 + M )        2    1/2                3/2
                                                                                                 G MP0     1
                                    MBE,t =                                           ⇡ MBE,0                                 (12)
                                                       2     1/2                        1
          P   =This  M
          If0turbulence  + is
                  Pth,0modified
                        BE,t   = M
                               present,
                             Pdyn,0   =
                                     BE,0 P(1
                                            we+   M
                                                 have
                                                  (1   + )to
                                                           M   2
                                                                 )⇡
                                                              include
                                                                    ⇡ M  P   its
                                                                           BE,0  MM   2
                                                                                 dynamic    )   pressure
                                                                                                  M           (12)
                                                                                                             in
                                                                                                              ⇡  the
                                                                                                                 M   total
                                                                                                                         M    1
                                                                                                                            ex-
                                    turbulent
                                            th,0 BE mass is the                characteristic mass
                                                                            th,0                      BE,t of turbulent
                                                                                                                    BE,0     frag-
          ternal   pressure,
           mentation,           and   if
                           and predicts  the   turbulence         is  highly       supersonic,      its  dynamic      pressure
   This modified     turbulent    BE massquite  is theaccurately
                                                           characteristic    themassIMFofpeak.turbulent frag-
    So, even  if the functional
          is dominant,
          Substituting       into form
                                   the    of the IMFexpression
                                         previous          were universal     of  (for
                                                                                  M   BEexample
                                                                                          ,  we    universal
                                                                                                 get    a       Salpeter’
                                                                                                           modified      s slope),
                                                                                                                        turbu-
mentation, and predicts quite accurately the IMF peak.
    the
     lentIMF
           BEpeakmass:MUST be sensitive to the environment. But under average Larson relations, the IMF
    peak is roughly
               P0 = Pconstant,
                        th,0  + Pso  no
                                  dyn,0  obvious
                                          =   P th,0IMF
                                                      (1   +variations
                                                              M   2
                                                                    ) in our neighborhood:
                                                                    2       1/2 1/2                   1
              MBE,0 ⇠ ⇢ BE,t  M1/2      =   M
                                          1/2
                                    ⇠ L BE,0           (1   +
                                                and M ⇠ v ⇠ L  M       )         ⇡   M  )       M
                                                                                         BE,0MBE,t = const ⇥ T0(12)           (13)
                 1/2
               Substituting1/2 into the previous 1/2         expression         of M=       , we get     a modified     turbu-
  MMore   ⇠           ⇠
    BE,0 generally, let’s justand
             ⇢           L             M    ⇠        ⇠   L         )         M         BEconst    ⇥   T
                                 assume that stars are formed in clusters whose parent gas clouds have rms
                                                  v                             BE,t                    0     (13)
          This
          lent   modified
                Describe
                BE    mass:   turbulent
                             density         BE mass is the characteristic mass of turbulent frag-
                                         dependence
    velocity of order virial (so we do not hide the normalization of the Larson relations):
     mentation, and predicts quite 4accurately the IMF peak.
                                                                            1
                                )     M  BE,t  = th const     ⇥  T 0 N            (Padoan
                                                                                       2      et al. 2007) (14) 1
                                                                                            1/2
                      MBE,t ⇡ 1.182                  1/2
                                                            = MBE,0 (1 + M )
                                                                         col                     ⇡ MBE,0 M                 (12)
                                          G3/2 P0
    Perhaps SF in the main halo of an early-type galaxy                 2 can be viewed as a single turbulent
                         1/2        1/2             2                      1/2
        M
    systemBE,0 This
                  ⇠   modified
                      ⇢        ⇠ L turbulent
                                           and    M BE   ⇠  mass
             (huge turbulence outer scale), hence vlarge column   ⇠  isL   the characteristic
                                                                                    )      MBE,t
                                                                                     density,   hence=mass
                                                                                                         lowerof
                                                                                                         const    ⇥ turbulent
                                                                                                                  IMFT0peak?
                                                                                                                           (13)
          fragmentation, and predicts quite accurately the IMF peak.
   Friday, November 30, 12
Numerical Study of the IMF - Paolo Padoan Troels Haugbølle and Åke Nordlund
General conditions to study the IMF numerically

        1. Good AMR code (unless you can do MHD turbulence some other ways, e.g. with SPH)
        2. Sink particles with large creation density, ρsink > 14 max(ρturb), to avoid artificial sink
           creation. At Mach=40 (pre-gravity MHD) we can reach 3e5 , so ρsink~ 1e7.
        3. Well resolved turbulence, from large enough root grid and/or shear refinement criterion
           (to be quantified soon).
        4. Huge dynamic range:
           a. We must resolve both the IMF peak (possibly BDs) and measure the Salpeter slope,
              while keeping the SFE realistically low.
           b. We must create clusters with realistic initial and boundary conditions, so they must
              form out of a much larger region.

Friday, November 30, 12
Numerical Study of the IMF - Paolo Padoan Troels Haugbølle and Åke Nordlund
Most massive region simulated to date has M=1e4 M⊙: Full IMF with SFE=0.15
                                Bonnell et al. 2010 (mass resolution 0.017 M⊙):

   Problem: IMF peak 0.08 M⊙. Including winds and outflows should be 0.04 M⊙
Numerical Study of the IMF - Paolo Padoan Troels Haugbølle and Åke Nordlund
Next most massive regions simulated to date have M=1000 M⊙:
              They barely reach 10 M⊙ stars with SFE~20%, and barely resolve the IMF peak.

                              Bonnell et al. 2006 (mass resolution 0.15 M⊙):

                                      SFE=0.33     SFE=0.32      SFE=0.20

                             Krumholz et al. 2012 (mass resolution 0.05 M⊙):

Friday, November 30, 12
Numerical Study of the IMF - Paolo Padoan Troels Haugbølle and Åke Nordlund
Bate 2012, 500 M⊙, mass resolution 0.001 M⊙, SFE=0.18:

        •   Too small total mass to constrain Salpeter’s slope
        •   Too short evolution (0.1 Myr of star formation) to get the IMF peak settled (only 0.05 M⊙ for fully formed stars!)
        •   Unphysical initial conditions (random initial velocity not correlated with density, because there is no initial
            turbulence-developing phase prior to gravity)
        •   Unphysical boundary conditions (isolated system, no accretion).
        •   No magnetic fields

        Krumholz et al. 2012: they have a pre-gravity turbulence-developing phase, but SFR still huge (as it should be
        without B field).
        They actually run for only 0.02 Myr, so they must have an even larger ambiguity in the IMF peak, between
        accreting stars and fully formed ones.

Friday, November 30, 12
Numerical Study of the IMF - Paolo Padoan Troels Haugbølle and Åke Nordlund
Krumholz et al. 2012: much too large SFR in HD even if you
             • develop the initial turbulence
             • account for radiation
             • include outflows
             Padoan et al. 2011,2012: you need B field (and turbulence driving) for realistic SFR.

             Including radiation or B fields has the same effect? Not even close!
             The SFR is quite wrong without B fields, so why should the IMF be right?

                                                              OUR APPROACH:

             •   BIG volume, 20-60 pc, with 6.e4-2.e5 M⊙ (depending on Larson relations), for realistic boundary and initial
                 conditions for clusters
             •   Run for a long time, approaching 1 Myr, so no IMF ambiguity between accreting and fully formed stars
             •   Realistic B field, so realistic SFR
             •   Develop the turbulence for many dynamical times pre-gravity, for self-consistent ICs
             •   Continue to drive after including self-gravity (even GMCs are not isolated)
             •   BIG dynamic range 32-128 times larger than Krumholz et al. 2012 in linear size (or 5-7 extra AMR levels),
                 so comparable spatial resolution, 30-90 AU
             • We         form over 2,000 stars with realistic SFE~0.04, from BDs to 100 M⊙, so huge sample for the IMF.

Friday, November 30, 12
Numerical Study of the IMF - Paolo Padoan Troels Haugbølle and Åke Nordlund
AMR Simulations of Star Formation
                              (Vast convergence study, several Mhr on NASA/Ames Pleiades supercomputer)

               • Periodic boundary conditions
               • Random, large-scale force
               • Uniform initial magnetic and density fields
               • Random initial velocity field
               • Isothermal equation of state
               • Root grid: 2563-1,0243
               • AMR levels: 9-11
               • Sink particles, ρsink~ 1e7
               AMR code:
               Ramses, with a hybrid layout, using
               OpenMP within individual nodes,
               and MPI between nodes.

Friday, November 30, 12
Numerical Study of the IMF - Paolo Padoan Troels Haugbølle and Åke Nordlund
Ramses AMR simulation of star formation - 2563 + 9 AMR levels

                                                             0.2 pc

                    Maximum resolution
                    equivalent to 131,0723:

                    Range of length scales:
                      107 AU -- 80 AU
                                                                               4pc
                    Range of time scales:
                      106 yr -- 1 yr

                    We simulate a full GMC for
                    ~1 Myr, forming over 2,000
                    stars and brown dwarfs,
                    with SFE~0.04                                       20pc

Friday, November 30, 12
The birth of 2,000 stars, from 0.01 to 100 M⊙
                          (Full box)

Friday, November 30, 12
The birth of 2,000 stars, from 0.01 to 100 M⊙
        (Main cluster)

Friday, November 30, 12
Numerical convergence versus root-grid size:

                           Numerical convergence versus AMR levels:

Friday, November 30, 12
The Next Challenge

 Resolve the formation of individual stars within a galactic fountain volume, for realistic
 initial and boundary conditions for whole GMCs

Friday, November 30, 12
What I don’t know (yet)
                          Effect of realistic initial and boundary conditions of whole GMCs, not
                          just of clusters inside GMCs

                          Convergence with respect to various numerical parameters still
                          underway

                          Convergence with respect to magnetic field strength

                          Convergence with both B and radiation

              Some more work is needed, but we are close to constrain (and
              understand) the IMF numerically.
Friday, November 30, 12
Friday, November 30, 12
Friday, November 30, 12
The Cosmic History of the IMF and its M/L

                                     z~4

                                               z~2➞0

                                                       z~10

Friday, November 30, 12
You can also read