The$interstellar$medium$in$metal/poor$ blue...The$interstellar$medium$in$metal/poor$...
Transcript of The$interstellar$medium$in$metal/poor$ blue...The$interstellar$medium$in$metal/poor$...
The interstellar medium in metal-‐poor blue dwarf galaxies
Leslie Hunt INAF-‐Osservatorio Astrofisico di Arcetri,
Firenze
with S. Garcia-‐Burillo, V. Casasola, P. Caselli, F. Combes, C. Henkel, A. Lundgren, R. Maiolino, K. M. Menten, L. TesN, A. Weiss
Outline of talk
! Low-‐metallicity low-‐z starbursts as a local approach to the study of chemically unevolved high-‐redshiQ galaxies
! The ISM in metal-‐poor starbursts (IRAM 30m observaNons)
ü molecular depleNon Nmes ü conversion factor of CO luminosity to H2 mass, αCO ü HI, total gas content
! Comparison of observaNons with models for H2 formaNon
! InvesNgate discrepancies of models versus observaNons: the role of gas volume density
Difference of high-‐z populaNons due to global properNes (i.e., Mstar, morphology, SFR, Z)
or characterisNcs of the ISM? For given Mstar and SFR/Mstar (specific SFR, sSFR): Ø ionized gas densiNes were 7 Nmes higher at z~3 than z~0 (Shiraz+ 2014)
Ø G0 was 5 Nmes higher with consequently higher Tdust at z~3 than z~0 (Magdis+ 2012)
Low-‐z metal-‐poor starbursts have warmer dust spectral energy distribuNons, harder and more intense radiaNon fields, and frequently high gas densiNes, similar to high-‐z populaNons.
scaling relaNons of metallicity (O/H), stellar mass, and SFR
MS z~1 MS z~2
MS z~0
Le#: sSFR and O/H are clearly correlated for galaxies in the Local Universe. Slope corresponds to what would be expected if metallicity is governed by momentum-‐driven winds (e.g., Dave’ et al. 2011, Dayal et al. 2013). Right: star formaNon “main sequence” (e.g., Salim+ 2007, Noeske+ 2007, Karim+ 2011)
Adapted from
Hun
t+ 2012)
IZw18
SBS0335-052
scaling relaNons of metallicity (O/H), stellar mass, and SFR
MS z~1 MS z~2
MS z~0
CorrelaNons degrade when high-‐redshiQ samples are added, suggesNng that selecNon effects (conNnuum, line flux limits) are important for characterizing scaling relaNons. Metal-‐poor starbursts (mostly Blue Compact Dwarfs) in the Local Universe can have sSFRs that are typical of galaxies at z~2.
Adapted from
Hun
t+ 2012)
how can we characterize a metal-‐poor ISM?
AV (magnitudes) 1 10
H C+ O
H/H2
C+/C/CO
O/O2
T~100-‐1000K T~10-‐100K
UV flux
UV flux
UV flux
H+ H2 CO
H+ HI
AV<0.1
PhotodissociaNon region (PDR)
(adapted from Wolfire & Kaufman 2011)
PDR structure changes at low metallicity
CO emission comes from an increasingly smaller region as metallicity decreases (e.g., Maloney & Black 1988), and is expected to disappear at sufficiently low metalliciNes.
(taken from Bolaoo+ 2013, Bolaoo+ 1999)
decreasing metallicity
CO
C+ C
H2
CO-‐dark H2 gas prevalent at low metallicity
RaNo of atomic carbon [CI] (3P1-‐3P0 , 610μm) to CO(1-‐0) larger at low O/H
PDR model by Bolaoo+ 1999
Bolaoo+ 2000 fit to data
(APEX, Hunt+ 2015)
MOlecules and DUst and LOw metallicity (MODULO) 155 dwarf galaxies imaged with Spitzer/Herschel and Z<0.4 Z¤; median Z=0.19 Z¤ Metallicity decreases è ordered disks become clumpy knots of star formaNon.
Z=0.03 Z¤
12+logO/H = 7.5
7.79 7.99
8.15 8.30
CGCG007-‐025, 0.1 Z¤
UM462, 0.2 Z¤ UM448, 0.2 Z¤ NGC7077, 0.2 Z¤ Mrk996, 0.2 Z¤
12CO(1-‐0) IRAM 30m observaNons Detected 8 out of 8 galaxies observed, including a tentaNve 3σ detecNon at 0.1 Z¤. Largest collecNon of low-‐metallicity (≤ 0.2 Z¤) CO(1-‐0) detecNons outside the Local Group.
Haro3, 0.5 Z¤ NGC1140, 0.3 Z¤ NGC1156, 0.3 Z¤
Hunt+ 2015, submioed
CGCG007-‐025, 0.1 Z¤
UM462, 0.2 Z¤ UM448, 0.2 Z¤ NGC7077, 0.2 Z¤ Mrk996, 0.2 Z¤
IZw18, 0.03 Z¤ SBS0335-‐052, 0.03 Z¤
Haro3, 0.5 Z¤ IIZw40, 0.3 Z¤ NGC1140, 0.3 Z¤ NGC1156, 0.3 Z¤
The following analysis also includes 3
addiNonal galaxies: IIZw40 at 0.3 Z¤ and
SBS0335-‐052 and IZw18 at 0.03 Z¤.
(taken
from
Hun
t+ 2015, A&A, su
bmioed
)
The 12CO(1-‐0) spectra
molecular depleNon Nmes and the conversion factor of CO luminosity to H2 mass, αCO
CO traces SFR even at low metallicity
For our sample (with Z<0.5Z¤), there is a factor of 30 offset to lower L’(CO) for a given SFR relaNve to the relaNon found by Gao &
Solomon (2004).
Given the relaNvely narrow range of metalliciNes in our sample, we included addiNonal galaxies (Schruba+ 2012): significant dependence of SFR/L’(CO) on O/H.
examine L’(CO)/SFR raNos to determine molecular depleNon Nmes at low metallicity
Molecular τdep = αCO L’(CO)/SFR
Regression line by Saintonge+ (2011) for mass-‐selected COLDGASS sample with constant αCO. Is offset of L’(CO)/SFR vs sSFR due to αCO variaNon with Z?
(taken from Hunt+ 2015, submioed)
residuals from L’(CO)/SFR fit with sSFR show correlaNon with O/H
(αCO)-‐1 ≈ (Z/Z¤)1.9 Consistent with that found by measuring dust methods (e.g., Leroy+ 2011, Bolaoo+ 2011), but shallower than the metallicity dependence found by assuming constant depleNon Nmes (e.g., Schruba+ 2012, Genzel+ 2012: (αCO)-‐1 ≈ (Z/Z¤)2-‐3).
molecular depleNon Nmes are shorter for metal-‐poor dwarf starbursts
τdep varies by a factor of > 200 down to 100 Myr over a spread of 103 in sSFR and Mstar. Range of parameter space covered by our observaNons extends to higher sSFR and lower stellar mass than Saintonge+ (2011).
Short molecular gas depleNon Nmes τdep ≤ 100 Myr found at the highest sSFRs (≤ 100 Myr) are shorter than those found for “main sequence” galaxies at similar redshiQs (Genzel+ 2015), but comparable to those by Scoville+ (2014). TheoreNcal expectaNon is that low metallicity means low star-‐formaNon efficiency because of lack of cooling, but these results would suggest the opposite.
molecular depleNon Nmes are shorter for metal-‐poor dwarf starbursts
what about atomic gas? total gas content
gas -‐to-‐ baryonic mass fracNons higher for high sSFR and low metallicity
Including HI gives significant correlaNon of Mgas (=MHI+MH2)/Mbaryonic (=Mgas+Mstar) with sSFR and metallicity.
gas -‐to-‐ baryonic mass fracNons higher for high sSFR and low metallicity
Outliers (Mrk996, N7077), with ellipNcal morphology, may be in a quenching phase with liole HI but high H2/HI fracNons (~40%).
gas -‐to-‐ baryonic mass fracNons also higher for low stellar mass
Mgas/Mbaryonic fracNons and molecular depleNon Nme τdep behave in similar ways for the
galaxies in our combined sample:
metal-‐poor, low-‐mass, high sSFR galaxies tend to have
shorter molecular depleNon Nmes and higher gas-‐mass
fracNons.
Primary driver seems to be stellar mass since these
correlaNons tend to show less scaoer than those with O/H.
unlike molecular gas τdep HI depleNon Nmes constant
τdep(HI) ~ 3.4 Gyr (see Schiminovich+ 2010 for GASS galaxies)
ImplicaNon is that HI only indirectly involved with SF
comparison of observaNons with models for H2 formaNon
observaNons broadly consistent with models
Krumholz+ (2009), McKee & Krumholz (2010), Krumholz+ (2011) predict that the fracNon of molecular gas fH2 depends on metallicity through cloud opNcal depth τ assumed to be linearly proporNonal to Z/Z¤. FracNon of gas mass where carbon is in CO also depends on metallicity through AV, assumed to vary linearly with Z/Z¤ (see also Wolfire+ 2010).
Here global values (not surface densiNes) by including a resoluNon correcNon.
observaNons broadly consistent with models
Here global values (not surface densiNes) by including a resoluNon correcNon.
Despite general good agreement, at Z/Z¤≤0.2, some data exceed by a factor of 3 predicted variaNon of αCO.
invesNgaNng discrepancies of models versus observaNons: the role of gas volume density
(Adapted from Hunt+ 2012)
(Taken from Rem
y-‐Ruyer+ 2014)
dust-‐to-‐gas raNos not linear with metallicity
Gas/du
st m
ass raN
o 12+Log(O/H) Open blue circles correspond to gas mass total of ionized
gas + HI; filled circles ionized gas + HI + inferred H2. Dust mass inferred from SED fi}ng (Hunt+ 2014).
If dust-‐to-‐gas raNos are not uniquely linearly dependent on metallicity, then many of the model assumpNons for gas scaling relaNons are called into quesNon.
radio signature of dense gas in SBS 0335-‐052
3.6 cm 6 cm
2 cm 1.3 cm
High-‐resoluNon VLA images of SBS0335-‐052E show that only the two brightest clusters have high-‐ν free-‐free emission (Johnson, Hunt, Reines 2009).
SED fi}ng of the radio conNnuum gives SSC sizes of 3-‐6 pc at high-‐resl’n, and 8-‐15 pc at low resl’n (Hunt+ 2004). Inferred electron densiNes are high, ~3000 cm-‐3, about 6 Nmes higher than esNmated from the opNcal spectrum (500 cm-‐3).
low-‐frequency turnover
radio conNnuum in I Zw 18, no dense gas
Unlike for SBS 0335-‐052, SED fi}ng of the radio conNnuum in I Zw 18 shows no evidence for high emission measure (i.e., high densiNes along the line-‐of-‐sight). From fit of radio spectrum ionized gas densiNes ~ 10 cm-‐3 (Hunt+ 2005, see also Cannon+ 2005) about 10 Nmes lower than opNcal spectrum (~ 100 cm-‐3).
21 cm 6 cm
no low-‐frequency turnover
! Schneider+ (2015) show that dust mass in SBS0335-‐052 is grown mainly through grain accreNon in clouds, while IZw18 mostly by stellar sources.
! With few excepNons (e.g., Glover & MacLow 2011), most modeling efforts to determine H2 fracNon assume that the capacity of a cloud to avoid photo-‐dissociaNon of CO depends linearly on metallicity; such an assumpNon is implicit in the noNon that the dust opNcal depth and dust-‐to-‐gas raNo vary linearly with metallicity (e.g., Krumholz+ 2009, 2011, Wolfire+ 2010).
! If the dust opNcal depth does not vary linearly with metallicity, then these model predicNons may be incorrect.
! Together with metallicity, gas density may play an important role in shaping the properNes of the ISM
high gas density implies more efficient grain growth