Whittle : EXTRAGALACTIC ASTRONOMY

 
     
 

1 : Preliminaries	6 :   Dynamics I	11 : Star Formation	16 : Cosmology
2 : Morphology	7 :   Ellipticals	12 : Interactions	17 : Structure Growth
3 : Surveys	8 :   Dynamics II	13 : Groups & Clusters	18 : Galaxy Formation
4 : Lum. Functions	9 :   Gas & Dust	14 : Nuclei & BHs	19 : Reionization & IGM
5 : Spirals	10 : Populations	15 : AGNs & Quasars	20 : Dark Matter

(1) Introduction

• Star Formation is an important and widespread phenomenon in the universe
  in Bill Keel's words, it is "galaxy evolution caught in the act"
  With this evolutionary perspective, one should distinguish between
  current star formation -- HII regions, OB associations, starbursts, etc
  past star formation -- which gave rise to the present day stellar population
  the current stellar mix is defined by the history of star formation

• There is an enormous (10⁷) range in galaxy star formation rates : 10^-4 - 10³ M yr^-1
  Loosely, we divide this range into two regimes :
  (i) normal galaxies (75% of local SF) have SFRs : 0 - few M yr^-1 [fig 1 from K98]
  note: integrated galaxy spectra varying mix of A-F V (<1 Gyr) and G-K III (3 - 15 Gyr)
  (ii) starburst galaxies (25% of local SF) range from :
  few M yr^-1 (SB) 50 M yr^-1(LIGs) 10^2-3 M yr^-1 (ULIGs)

• We must also distinguish between two rather different locations for SF :
  (i) galaxy disks -- predominantly normal SF
  (ii) near galaxy nuclei (circumnuclear, C-Nuc) -- both normal and starburst SF
  Note : high luminosity C-Nuc SF is qualitatively different from Disk SF.

(2) Definitions and Abbreviations

SF	Star Formation
SFR	Star Formation Rate, in M yr-1
SFR	surface SFR rate, in M yr-1kpc-2
gas	surface density of gas, in M pc-2
SF-History	time dependence of SFR (eg declining exponential; burst; constant; etc)
C-Nuc	Circumnuclear 100 - 1000 pc
IRAS	Infrared Astronomical Satellite (1983): S12 etc = fluxes at 12; 25; 60; 100 µ (in Jy)
PSC & FSC	Point (& Faint) Source Catalogs from IRAS all sky survey
FIR	Far-Infrared :   40 - 500 microns, depends on usage
NIR & Mid-IR	Near-IR (1-5µ) & Mid-IR (5-20µ)
FUV & NUV	Far (ionizing) UV & Near (1500-2800) UV
FFIR	FIR flux (40 - 500µ) = 1.26×10-14(2.58S60 + S100) W m-2
FIR	IR flux (8 - 1000µ) = 1.8×10-14(13.5S12 + 5.2S25 + 2.58S60 + S100) W m-2
LFIR & LIR	Luminosities corresponding to FFIR & FIR
Lcm	Radio luminosity at cm wavelengths (eg 5 GHz), mostly synchrotron
CR	Cosmic Rays associated with synchrotron radio emission
SN & SNR	Supernova & Supernova Remnant
SB	Starburst
LIG	Luminous Infrared Galaxy (LFIR > 1011L)
ULIG	Ultra-Luminous Infrared Galaxy (LFIR > 1012L)
LINER	Low Ionization Nuclear Emission Line Region (low luminosity AGN)
EW(Ha)	Equivalent width of Ha = f(Ha) / f(cont)   Angstroms
IMF	Initial Mass Function, usually PL :   N(M) M-x (eg x = 2.35 = Salpeter IMF)
Mlow & Mup	lower and upper mass cut-off for the IMF

(3) Emission From Star Formation Regions

Star formation yields an IMF with high mass stars dominating the luminosity
These yield, directly or indirectly, to a wide range of emission [image]
See Kennicutt and Evans ARAA 2012 for excellent review: [o-link]

• UV flux : high mass stars dominate UV luminosity visible if non-dusty

• H flux : B0 and hotter create ionizing flux <912 A 1 ionizing photon = 1 ionized H atom
  Photoionization rate (Q_H = dN_ion / dt   s^-1) balances recombination rate (3×10^-13n_e²V   s^-1)
  (n_e = electron density; V = total volume; 3×10^-13 = recombination coeff at T = 10⁴K)
  1 in 4 recombinations yield an H photon L_H = 1.3×10^-12Q_H erg/s

• Radio free-free flux : ionized gas also radiates free-free (Bremmstrahlung) at 5 GHz
  L (erg/s/Hz) = 7.3×10^-39n_e²V = 2.4×10^-26Q_H (at 10⁴K)

• FIR flux : dust absorbs UV very efficiently, reradiates in FIR (20-200µ; T_D = 15-30K)
  hence, IRAS selected galaxies often have high star formation rates
  L_FIR is acting like a bolometer

  Only the most massive stars are relevant and, for single stars, we have :

  Star	Mass	Log QH	Log Lff	Log LH	Log Lbol
  O5	40	50.0	24.4	38.1	39.0
  B0	16	48.7	23.1	36.8	38.0
  A0	4	42.7	17.1	30.8	35.5

• Radio Synchrotron flux : most radio emission is from diffuse cosmic rays
  (MW CRs are detected striking our upper atmosphere)
  the CRs are generated in SNR shocks; they diffuse out into the galaxy (which holds 99%)
  Despite rather indirect link to SFR -- L_cm is well correlated with other SFR diagnostics.

• CO and HI flux : molecular and atomic emission (CO acting for H₂; 21cm from HI) traces the gas component immediately prior to SF

Many correlations of various strengths support the above picture:
See section 3 in Kennicutt and Evans ARAA 2012: [o-link]

Good correlations include :   L_H vs L_FIR vs L_cm
Less good correlations include :   L_H vs colors vs L_CO

(4) Measurements of Current SFR

• Almost all the above observables depend on high mass stars
  these are a transient population, and so track current SFR

• Lower mass stars contribute essentially nothing to these observables
  however, these low mass stars dominate the mass and should not be forgotten!

1. evaluate evolutionary tracks L_bol & T_e & R_* as functions of Mass and Age
   
2. add stellar atmospheres (T_e & g) spectra (or UBV etc)
   
3. Sum over an IMF isochrone spectrum (or UBV)
   
4. Sum over a chosen star formation history current spectrum
   

Free parameters : SF-History; IMF; Metallicity.
in practice, main parameters are : burst age and/or e-folding decay; plus fraction of old pop

(b) Conversion Relations to find SFR

(i) Near-UV (1500 - 2800) Luminosity

hot high mass young stars dominate the NUV emission, yielding :

SFR   (M yr^-1)   =   1.4×10^-28 L_NUV   (erg s^-1Hz^-1)

strengths:   for moderate-strong SFR, very little contamination from non-SB stars;

useful for high-z galaxies (where UV is redshifted into optical)
weaknesses:   sensitive to IMF and to dust.

(ii) H Luminosity

In principle this applies to other recombination lines : eg Br & Pa & H109 etc
Significant ionizing radiation only comes from stars with M > 10M
lifetime of these stars is < 20 Myr H measures current SFR

SFR   (M yr-1)	=   7.9×10-42 LH   (erg s-1)
	=   8.2×10-40 LBr   (erg s-1)
	=   1.1×10-53 QH   (s-1)

strengths :   sensitive; direct; high spatial resolution; useful out to z 2
weaknesses :   sensitive to reddening (typical A_H0.5 - 1.5 mags), IMF slope and M_up
5 - 50% of the ionizing radiation escapes the HII regions
must include H from the diffuse ionized medium (DIM) emission
(only 3% ionizing flux escapes the galaxy)

at higher z (when H too redshifted), a less precise relation is :

SFR   (M yr^-1)   =   1.4±0.4×10^-41L_[OII]3727   (erg s^-1)

(iii) Equivalent Width : EW(H)

Recall EW(H) measures the relative strength of H to the continuum under the line
It therefore acts like a long baseline color index UV(H) 6550 A
Although it cannot be converted to a current SFR, it has another important use :
It measures the ratio of the current SFR (from H) to the integrated past SF (from the continuum)
Using synthesis models, this relation can be quantified, to give :
EW(H) (current SFR) / (mean past SFR)   ; written SFR/<SFR> or "b"


(iv) FIR Luminosity

For Starbursts, where SF dominates the FIR emission, we have :

SFR   (M yr^-1)   =   4.5×10^-44 L_IR (8 - 1000µ)   (erg s^-1)

Unfortunately, FIR can contain two other components :

1. cirrus : diffuse emission @ 100µ from dust warmed by normal optical starlight
   this may dominate in E, S0, Sa, Sab     so FIR is not good SFR measure for these early types
   However, for Sb and later, we have a rough relation :

   SFR   (M yr^-1)   =   8(+8/-3)×10^-44 L_IR (8 - 1000µ)   (erg s^-1)

2. AGN : important in Seyferts & many ULIGS
   AGN generates hotter dust, so spectrum is "warmer" (eg fig 2 SM 96)
   eg S₂₅/S₆₀ > 3   &/or   S₆₀/S₁₀₀ > ??

(v) Radio Free-Free Luminosity

SFR   (M yr^-1)   =   4.3×10^-28 L_ff   (erg s^-1Hz^-1 @ 5 GHz)

strengths :   direct link to HII regions (like H); zero reddening
weaknesses :   usually weak w.r.t. synchrotron; requires separation using spectral indices.

(vi) Radio Synchrotron Luminosity

This cannot be calibrated directly because of the uncertainties of SNR & CR production
    not to mention the synchrotron efficiencies
One could use the L_cm vs L_H   or   L_cm vs L_FIR   correlations to derive an SFR vs L_cm relation
    but it would not be an independent relation.

(5) Factors Affecting the SFR

• Recall, there are two significantly different regions to consider
  Both/either/neither can be significant
  • Galaxy Disks
  • Galaxy Circum-Nuclear Regions (C-Nuc)
    (the latter includes both "normal" and "starburst" phenomena)
  To a large extent, star formation is decoupled between these two regions

• Depending on the type of observations, measurements may refer to disk; C-Nuc; or both
  eg :
  • IRAS fluxes are usually integrated, except for the nearest galaxies
    (eg BGS; FPSC...)
  • small aperture spectoscopy yields C-Nuc fluxes
    (eg Stauffer '82; Keel '83; Ho et al '87)
  • wide aperture spectrophotometry yields integrated fluxes
    (eg H measurements of Kennicutt & Kent '83; Romanishin '94)

• Several measures of SFR are useful :
  • direct SFR, in M yr^-1
  • surface density of SFR :   _SFR, measured in M yr^-1 kpc^-2
  • relative SFR = (current SFR)/(mean SFR in past); where a continuum flux measures <SFR>
    eg EW(H) = f(H) / f(cont) Angstroms; or L_FIR / L_1m
    Synthesis models convert these to a quantitative measure : b = SFR/<SFR>

• Results can depend on the sample selection method, eg :
  
  • Optical selection usually includes low to moderate SFRs (eg NGC galaxies)
    
  • UV selection usually includes moderate SFRs (eg Markarian galaxies)
  • FIR selection usually includes moderate to high SFRs (eg IRAS galaxies)

(i) Integrated SF

• Loosely speaking, there is a strong dependency of SFR on Hubble type, with significant scatter :
  eg 10^-2 M yr^-1 for S0; up to 20 yr^-1 for gas rich spirals
  (on up to 1000 yr^-1 for merging starbursts with Hubble type "pec")
  Eg Fig 4 from K98 shows the FIR / H band luminosity for nearby galaxies
  (note: cirrus contributes to the early types, reducing the apparent trend with SFR)

  (ii) Disk SF

• A purer measure of relative SFR is EW(Ha), shown in Fig 3 from K98 for nearby galaxies
  (while EW(Ha) is integrated, in this sample, disk emission dominates)
• There is a strong trend with considerable scatter : SFR/<SFR> increases by ×10 from Sa to Sc
  (<EW> = 3 - 30A or SFR = 0.2 - 2 M yr^-1 for an L^* galaxy)
• This increase is due to two factors :
  • more HII region complexes
  • more luminous HII region complexes
    for Sa : HII cluster has a few OB stars
    for Sc : HII cluster has few hundred OB stars (see figure of NGC 604 in M33)
• Now consider the quantitative values of b = SFR / <SFR> (RHS of figure) :
  for Sa : <b> 0.1 while for Sc : <b> 1
  SFRs in Sa disks were significantly higher in the past
  SFRs in Sc disks have been roughly constant over cosmic history
• Although the actual SFR-Histories may be complex, figure 8 from K98 shows simple exponentials consistent with these numbers
  (in the heirarchical merger picture, these smooth trends would be punctuated by merger induced spikes)
  As a function of redshift, we expect :
  z = 0 (locally) : most SF is in late type disks
  z = 1 : SF is equally spread along the Hubble sequence Sa - Sb - Sc
  z > 1 : present day Sa galaxies increasingly dominate the SFRs
  
  Note, however, the Madau peak at z 1-2 is not represented here (mergers probably important)

  (iii) C-Nuc SF

• As for disk emission, there is a wide range 10^-4- 10² M yr^-1 (mean 0.1; median 0.02)
• Detection/classification of C-Nuc HII emission increases along the Hubble sequence :
  0% E;  8% S0;   22% Sa;   51% Sb;   80% Sc-Im (viewgraph from Ho et al 97)
  (but overestimates trend since LINER/AGN emission may mask HII emission in eary types)
• However, C-Nuc SFRs decrease down the Hubble sequence :
  most nuclear SF comes from early types (despite lower frequency)
  in early types, C-Nuc SFR often similar/surpasses disk emission
• Overall, not a clear Type dependence of C-Nuc SFR
• C-Nuc <EW(H)> 3 - 30 A <EW(H)> for Sc disks
    SFR/<SFR> 1     SF-History constant interspersed by bursts (see below)

• No difference, statistically, in SFRs, for either Disk or C-Nuc
    density waves are not themselves responsible for variation in SFR
• However, Grand Design Arms have higher SFR contrast than Arms in flocculants

(i) Disk SF

There is little/no dependence of disk SFR on presence/absense of bar (Fig 3 in K98)

(ii) C-Nuc SF

• HII detection/classification is independent of Bar
• However, mean SFR is significantly higher in barred galaxies
  tail in SFR distribution out to 0.2 - 8 M yr^-1 (absent in unbarred galaxies)
  especially true in early types, eg 30% of SB0/a - SBb galaxies are in this tail.
• Bars are effective at transporting gas to the nuclear regions
  especially in large bars, as found typically in early types

(i) Disk SF

• On average, interacting galaxies have disk SFRs ×2-3 higher than isolated galaxies
• However, the effect is very variable :
  gas poor galaxies show no enhancement
  in extreme cases the SFR is ×10-100 higher

  (ii) C-Nuc SF

• even stronger effect than disks : mean enhancement ×3-4 in interacting galaxies
• at higher luminosities (LIGs & ULIGs) interactions are clearly important :
  L_IR <10¹⁰ L   (<1 M yr^-1)   20 - 30% are interacting (75% of remainder have strong bars)
  L_IR >10¹² L   (>100 M yr^-1)   70 - 95% are interacting/merging
  

  (iii) Physical Effects of Tidal Interaction

• At low level of tidal interaction, we may have :
  -- induced density wave
  -- induced bar which removes AM from gas
  -- orbit crossing     cloud-cloud collisions
  -- modified rotation curve     drops Q     disk unstable
• For stronger interactions and mergers :
  -- star distributions drain AM from gas (Topic 12)
  -- cloud collisions due to interpenetrating galaxies
  -- gas falls into E-S0 galaxies     nuclear gas disk     SF

(i) Disk SF

• Fig 5 from K98 shows _SFR vs _gas for normal galaxy disks
  There is a clear trend for SFR to increase with surface gas density
• The trend holds within a given Hubble type (symbols divide sample by type)
  this accounts for some of the scatter in the EW(H) vs Type plot
• Expressed as an efficiency, in % per 10⁸yr, there is a large range
    1-30% per 10⁸yr
• Roughly, for disks with gas mass fraction 20%, <> 5% per 10⁸yr
    1% stellar disk added per 10⁸     stellar disk constructed in a Hubble time

  (ii) C-Nuc SF

  C-Nuc SF often occurs within dense gas disks, 100-1000pc in size
  In these disks, _gas   10²-10⁴ Mpc^-2, comparable to the cores of disk DMCs (eg 30 Doradus <10pc) but extended over 1 kpc
  These gas densities are much higher than in normal disks (by factors 10-10³)

• Fig 7 from K98 shows _SFR vs _gas for C-Nuc starbursts
  the correlations persists even at these higher rates
• In fact, not only is there more SF because there is more gas,
  but the efficiency is higher than in disks by factors 2-30

  (iii) Schmidt Law

  A relationship between surface density and star formation rate was postulated by Schmidt (1959)
• Fig 9 from K98 (LHS) shows the combined sample of galaxy disks and C-Nuc disks
  There is a single Schmidt law spanning 6 decades :
  _SFR   (M yr^-1kpc^-2)   =   2.5±0.7 ×10^-7 _gas^{1.4 ± 0.15}   (M pc^-2)
• We naively expect a PL index of 1.5 :
  SFR     (gas density)/(free fall time)     /^{-½     1.5}
• We can consider the SFR per orbit (_SFR × P_rot) by plotting _SFR vs _gas / P_rot
  where P_rot is the orbital period at ½ R_outer (see RHS of Fig 9 from K98)
• Remarkably, we obtain a graph of gradient unity
    for all systems, 10% gas is converted to stars per orbit
  _SFR   =   0.017 _{gas gas}
• This allows two alternative view of why LIGs and ULIGs are so efficient :
  1. due to higher densities :
     efficiency     _SFR / _{gas     gas}^0.4
     since _gas is 10²-10³ higher than disks     efficiency is 5-50 times higher
  2. due to shorter timescales :
     efficiency = _SFR / _gas = 0.017 _gas which is independent of _gas
     since 10% gas goes into SF per orbit, the higher efficiencies simply reflect shorter orbital times for the ULIG gas
• The Schmidt law fails at low SFRs, below a critical threshold (see 11.6 below)

  (iv) Consumption Timescales

• Returning to Fig 5 and Fig 7 from K98, we can reinterpret the three efficiency lines in terms of gas depletion times
  100%   10%   1%   SFR efficiencies per 10⁸yr correspond to
  10⁸yr   10⁹yr   10¹⁰yr   depletion timescales
  for the mean of 5%, the mean depletion timescale is 2Gyr
• For the longer times, these are underestimates because gas can be replenished from stellar has loss and from infall
• The starburst LIG and ULIGs have much shorter depletion timescales : 1-10×10⁸yr
  The maximum SFR arises from 100% conversion in a dynamical time (P_rot = P_{free fall}) :
  SFR_max     100 M yr^-1 M_gas,10 P_rot,8^-1   with corresponding :
  L_max     7×10¹¹ L M_gas,10 P_rot,8^-1
• Fig 6 from K98 shows L_FIR vs H₂ mass for LIGs (open circles) and ULIGs (filled circles)
  the points lie between the solid line for normal MW galaxies, and the dashed line for SFR_max for P_{free fall} = 10⁸yr.
    the most powerful ULIGs are converting 10¹⁰ M gas into stars on a dynamical timescale
  This can only occur during :
  1. violent interaction/mergers
     eg entire ISM of galaxy driven into nucleus & converted into stars over 10⁸yrs
  2. initial collapse of protogalaxy
     hence ULIGs may be thought of as local analogs to high z young forming galaxies

Property	Spiral disks	Circumnuclear regions (including starbursts)
Radius	1–30 kpc	0.2–2 kpc
Star formation rate (SFR)	0–20 Myr-1	0–1000 M yr-1
Bolometric luminosity	106–1011 M	106–1013 M
Gas mass	108–1011 M	106–1011 M
Star formation time scale	1–50 Gyr	0.1–1 Gyr
Gas density	1–100 M pc-2	102–105 M pc-2
Optical depth (0.5 µm)	0–2	1–1000
SFR density	0–0.1 Myr-1 kpc-2	1–1000 M yr-1 kpc-2
Dominant mode	steady state	steady state + burst
Type dependence	strong	weak/none
Bar dependence	weak/none	strong
Spiral structure dependence	weak/none	weak/none
Interactions dependence	moderate	strong
Cluster dependence	moderate/weak	moderate
Redshift dependence	strong	?

(6) SF Threshold & Toomre's Q Parameter

(7) Starburst Galaxies

• Three examples with decreasing luminosity:
  Arp 220 is the nearest ULIRG.
  The Antennae an interacting luminous starburst.
  M 82 is a nearby lower luminosity Starburst.
  Henize 2-10 is a nearby dwarf starburst galaxy.

• Although there is no formal definition of Starbursts, two aspects are key :
  -- intense SF which dominates the integrated luminosity
  -- short burst     gas depletion 10⁸yrs << age of galaxy

• other characteristics include :
  -- often nuclear location 100-1000pc
  -- fueled by central accumulation of dense molecular gas
  -- SFR significantly higher than in galaxy disks (by 10³)
  MW : SFR 1 M yr^-1 over entire disk     _SFR 0.01 M yr^-1kpc^-2
  SB : SFR 10 M yr^-1 within 500pc     _SFR 10 M yr^-1kpc^-2
  
  -- Locally, 25% SF occurs in SBs (75% occurs in spiral disks)
  -- at higher z this fraction increases

• Several methods have generated samples of SB galaxies
  1. UV continuum (objective prism) surveys, eg Markarian
     these SBs tend to be low-moderate luminosity, less dusty galaxies
  2. Emission line surveys (objective prism, or aperture), eg SBS, many redshift surveys
     again, these tend to find low-moderate luminosity, less dusty galaxies
     note : spectroscopy of magnitude limited sample gives a low yield of SBs
  3. FIR (IRAS) survey yield many samples
     eg BGS (Bright Galaxy Sample, Soifer et al 89),
     1Jy ULIGs (Ultra Luminous Infrared Galaxy, Kim 95)
     These tend to have high luminosity and be quite dusty.

• Dust relates to metallicity :
  larger galaxies tend to be more metal rich, more dusty, more prone to FIR detection
  dwarf galaxies tend to be metal poor, less dusty, often detected by UV or emission lines
  Fig 4 from SM96 shows 10 ULIRGs from the BGS
  Fig 8 from SM96 shows 4 (U)LIRGs mergers with HI and CO contours superposed
  

Fig 1 from SM96 shows the LF for starburst galaxies

• the overall form is not like the Schechter function
  it can be characterised by a double power law, slope -1 below 10^10.3L and -2.35 above
• at low luminosities, SBs are much less common than normal galaxies
  (eg only a small fraction of RSA or RC3 galaxies are SBs)
• at higher luminosities (eg L > 10¹¹L) LIGs dominate over all other galaxy types
• at the highest luminosities (eg L > 10¹²L) ULIGs are 2× more numerous than QSOs
  these are, of course, still quite rare : expect 1 within cz10⁴km/s (find 1 : Arp 220 at cz=6000)
• Emission from LIGs and ULIGs makes up 6% of the FIR in the local universe.

• while IR fluxes go up by ×10³, the optical fluxes only increase by ×3-4
    ULIGs are not particularly luminous optically
• IR colors change as L_IR increases : S₆₀/S₁₀₀ increases, S₁₂/S₂₅ decreases
  overall, the emission is becoming dominated by a 60µ component at high L_IR
• Different components dominate at different luminosities :
  • Normal Galaxies
    100-200µ (T_D 20K) emission from cirrus heated by old population starlight
    10µ (T_D 200K) peak from small hot dust grains near hot stars
    (note these high temps are non-eqlm since grain thermal capacity < UV photon energy)
  • Seyferts
    include a "warm" component @ 25µ (T_D 150-200K) heated by the AGN
  • LIGs
    the starburst component at 60µ (T_D 30-60K) becomes increasingly strong
    some ULIGs also have an AGN which adds a 25µ component making the spectrum "warmer"
    (see inset in fig 2)

• Optical spectra allow classification by emission lines (see Topic 14)
  Fig 5 from SM96 shows the changing classifications at higher IR luminosities :
  the fraction of HII nuclei gradually drops as the fraction of Seyferts spectra increases
  the high nuclear gas content is either creating or feeding a nuclear black hole
  the fraction of LINERs is approximately constant

• Without a doubt, interactions & mergers play a crucial role in triggering luminous SBs
  A nice example : M81/M82/NGC3077, (cf interaction more evident in HI)
  (the M82 starburst was triggered 600 Myr ago near closest approach)
• furthermore, the most strongly interacting tend also to be the most luminous
  Table 3 from Sanders & Mirabel 1996 ARAA 34 749 shows these trends nicely :

  		Luminosity Ranges : Log LIR/L
  		10.5–10.99	11.0–11.49	11.5–11.99	12.0–12.50
  No. of objectsa		50	50	30	40
  Morphology	merger	12%	32%	66%	95%
  	close pair	21%	36%	14%	0%
  	single (?)	67%	32%	20%	5%
  Separationb	[kpc]	36.	27.	6.4	1.2
  Opt Spectra	Seyfert 1 or 2	7%	10%	17%	34%
  	LINER	28%	32%	34%	38%
  	H II	65%	58%	49%	28%
  LIR/LBc		1	5	13	25
  LIR/LCOc	[L (K km s-1pc2)-1]	37	78	122	230

  ^aObjects in the IRAS BGS plus additional ULIGs from Kim & Sanders (1996)
  ^bMean projected separation of nuclei for mergers and close pairs only.
  ^cMean values.
  

  • Less Luminous SBs :
    often milder interactions; pairs at larger separation;
    lower fraction interacting, although remainder have strong bars
  • ULIGs :
    high fraction of strong mergers; close double nuclei; progenitors probably gas rich spirals
    (See fig 8 from SM96 for examples)
    10¹⁰M gas (eg MW ISM) goes to center 10^2-3pc   SF (+AGN) yields 10¹²L
    
• In all these cases, the interactions (and bars) result in the loss of AM from the gas
  (see Topic 12 for a more complete discussion of this process)
• Dwarf starbursts (eg BCDs) do not seem to involve interactions
  it is currently unclear what the trigger mechanisms are in these cases :
  possibilities include :
  
  self-propagating star formation
  widespread instability (many BCDs hardly rotate, hence Q is very low)
  

• HST sees 100 clusters (maybe ×20-40 more hidden)
  They reside in the central few 100 pc, each of size 3pc & luminosity 10⁶L
  They have a power law luminosity distribution, and ages 600 Myr (matching time since last encounter)
  This HST image shows some of the clusters (in V and NIR)

• Similar young blue super star clusters are ofen seen in merging galaxies
  Examples : Antennae fig A,   fig B,   NGC 5253,   NGC 1569

• They are larger than any MW star formation region (eg 30 Doradus) with M_V = -8 to -14

• These clusters are therefore similar to forming globular clusters
    maybe MW globulars were formed this way during early galaxy assembly
    lower mass clusters are destroyed by evaporation/disruption to leave the present (Log Gauss) distribution

(i) Sketch of Physical Mechanisms

• The principle energy source entering the ISM is from winds and supernovae
  Their relative contributions are 1:3, so SNe dominate
  Typical rates in SB galaxies are 1 per 10-20 yrs (few million per burst)
  The average KE input rate is 1% L_bol

• This energy dumped goes through a number of transformations :
  KE from SN explosion     thermalised in shocks     hot gas expanding
  SNRs overlap     superbubble which expands at 100 km/s

• This expanding shock/shell initially decelerates and is Rayleigh-Taylor (RT) stable
  The superbubble tends to expand perpendicular to the disk (lowest pressure gradient)
  When the bubble reaches a few scale heights, it accelerates and becomes RT unstable
  

• The shell breaks up into a poorly collimated bipolar flow, or superwind
  The wind has terminal velocity 1-few 1000 km/s
  It also incorporates colder ISM gas by turbulent entrainment &/or conductive evaporation

• Thus, hot, warm, cold (and relativistic) components are advected up into the galaxy halo
  in a loosely biconical outflowing wind.

  See Images from Simulations.

  (ii) Observational Signatures

  The most well studies examples include: M82, NGC 253; NGC 3079; NGC 1482

• H images show bi-conical filaments
  
• Long slit spectroscopy shows line-splitting indicating an expanding shell
  Velocities for this warm (10⁴K) component are 10²-10³ km/s

• Significant quantities of neutral gas (eg NaI D, OI, CII lines) can be seen in absorption
  Blueshifts unambiguously indicate outflow

• Hot (3-10×10⁶K) X-ray emission is seen 10-30 kpc along the minor axis

• Low surface brightness radio synchrotron is seen above and below the disk

• FUSE has detected 10^5-6K gas via absorption of the OVI1035 doublet
  This (and the X-ray luminosities) show the superwind does not suffer significant radiative losses

  (iii) Global Characteristics

• The above observations have yielded estimates of the total rates of mass and energy loss carried out by the superwind
  They match the SFR conversion and energy deposition rates of the nuclear starburst
  The wind is clearly driven by the nuclear starburst

• The threshold SFR which can drive a superwind is not very high
  approximately, _SFR 0.1 M yr^-1kpc^-2 seems to be the threshold (less for dwarf galaxies)
  This is only about 10× that of the MW disk
  (cf galactic fountains are, of course, related lower power phenomena --- see Topic 9)

• Similar phenomena have been seen in high-z galaxies
  UV absorption line blueshifts indicate outflowing superwinds

• Although starbursts only account for 25% of current star formation, in the past this fraction was probably much higher
    most current epoch stars may have formed in starbursts
  for this (and other reasons, see below) starbursts are cosmologically very important

  (i) Possible Importance of Superwinds

• Superwinds carry enriched gas into the galaxy halo and possibly beyond
  This has several consequences :
  • Narrow metal line absorption systems in QSO spectra originate in the halos of young(ish) galaxies [image]
    This halo gas is clearly quite metal rich --- how come?
    Starburst driven winds are likely the main pollutant

  • Galaxy clusters contain a massive metal rich ICM [image]
    The ICM contains as many metals as all the stars in all the cluster galaxies
    Abundance analysis suggests the metals were generated by Type II (core collapse) SN
      the ICM and its metals probably originated from starburst driven superwinds.

  • The metallicity vs mass relation in spheroidal galaxies seems ubiquitous and fundamental
    again, enriched SB driven winds can explain this relation
    gas is lost more easily from lower mass galaxies
    recall (Topic 7) that the correlation can be recast as metallicity vs escape velocity

• At temperatures of 1-10×10⁶K, the ICM contains a huge quantity of thermal energy
  there has been some discussion as to where this energy originated
  clearly, superwinds may be the answer --- the energy budget works out fine

• Superwinds may also clear out a path for UV ionizing radiation to escape from the center
  this UV flux may contribute to the UV background, especially at high-z
  obviously, QSOs also contribute, but SBs may also be important
  the UV background is important since it is responsible for ionizing the IGM

  (ii) Starbursts at low and high z

• Starbursts provide 25% of the local star formation (75% in spiral disks)
  they also provide 10% L_bol in the local universe

• These fractions are probably simliar out to z1

• At higher z, it seems many galaxies resemble SBs (see MDS galaxies)
  At z>2,   UV, V, FIR are redshifted to V, K, sub-mm   and are now observationally accessible
  the "UV" selected galaxies at high z resemble local UV SBs :
  
  • same _SFR;   same colors;   same spectra
    
  • however, SF regions are larger   few ×10² M yr^-1 over a few kpc

• clearly, local SBs seem to provide close analogs to many high-z galaxies

• Since starbursts are, by definition, short lived, we expect to find evidence of "past starburst" relics
  What might they look like ?

  • E+A galaxies are in the post-starburst phase, with A stars still present but no current SF.
    This SDSS example has red old spectrum, no emission lines, but strong Balmer absorption lines: [o-link].

  • As the A-type spectra fades, the last Balmer line to disappear is H 4101
      strong H may signify a slightly older SB relic

  • Kinematically decoupled cores (KDCs -- Topic 7.6d) are often found in E and S0 galaxies
    Rotation suggests they formed from a nuclear gas disk   a starburst
    Since KDCs usually have old population colors, the SB event was probably 5 Gyr ago