bright side versus dark side of star formation: uv and ir views

Bright Side versus Dark Side of Star Formation: UV and IR Views

C. Kevin Xu, IPAC, Caltech

Veronique Buat, LAM, Marseille

Collaborators: J. Iglesias-Paramo, T. Tekeuchi, M. Rowan-Robinson, GALEX team, SWIRE team

Question: Do UV and IR surveys see the two sides of SF of the same population,

or SF of two different populations?

UV surveys

IR surveys

UV

sur

veys

IR surveys

Total SFR

Total SFR

1 population:

2 populations:

Talk plan• Local UV and IR galaxies: how much do they overlap?

• comparisons of IR/UV ratio, L_tot, Hubble type, mass, clustering• UV LF of IR galaxies and IR LF of UV galaxies • IR-quiet UV galaxies (low metallicity dwarfs)• UVLGs and ULIRGs

• LBGs and SCUBA galaxies: UV and IR galaxies at z ~ 3

• UV and IR galaxies at intermediate redshifts (0.5 < z < 0.7) --- early results from a GALEX/SWIRE comparison study

• evolution of attenuation in UV and IR selected galaxies• evolution of stellar mass in UV and IR selected galaxies

• Summary

FIR-UV bivariate luminosity function of local UV+FIR galaxies

A(FUV)=

1

Martin et al. 2005, ApJL, GALEX Edition

• Saturation of L_UV at ~ 2 x10^10 L_sun• bi-modality (of L_IR/L_UV ratio)

•Strong dependence of L_IR/L_UV (best A_FUV indicator) on L_tot =L_UV+L_IR.

L_tot LF of local UV+IR galaxies

Martin et al. 2005, ApJL, GALEX Edition

L_tot=L_FUV+L_60

Solid line -- log-normal fit

Blue -- UV selected (GALEX src)Red -- IR selected (IRAS src)

UV galaxies are

absent in high

L_tot (>10^11) end!

Local samples: IR selected versus UV selected

(details in J. Iglesias’ talk)

IR selected (126): f60 > 0.6 Jy

UV selected (61): NUV < 16 mag

UV: low L_60/L_FUVIR: high L_60/L_FUV

L_IR/L_FUV distributions of IR and UV selected galaxies

• Very different.• The overlap between the two samples is ~ 30%.•The mean ratio of IR galaxies is ~ 10 times higher than that of UV galaxies!

Mean attenuations from the Fdust/F(UV) ratio

NUV selected sample

<A(NUV)>=0.8+/-0.3 mag

<A(FUV)>=1.1+/-0.3 mag

FIR selected sample <A(NUV)>=2.1+/-1.0 mag

<A(FUV)>=2.9+/-1.0 mag

NUV

FUV

Confirm pre-GALEX results (Buat et al. 2005, ApJL, GALEX edition)

L_IR/L_FUV v.s. L_tot(`Adelberger plot’)

Two populations are separated:IR: high L_tot, high L_IR/L_UV ratio UV: low L_tot, low L_IR/L_UV ratio

Explanation:Consequence of selectioneffect on L_IR/L_UV ratio, and the strong correlation between the ratio and L_tot.

L_tot LFs of local UV and IR galaxies

Using L_IR/L_UV ratio to convert to L_tot

•The L_tot of UV galaxies has a sharp cutoff at ~ 10^11 L_sun

Comparison of Hubble type distributions of local UV and IR galaxies

Good overlap in the middle:both populations peak around~ Sbc

IR galaxies: • excess of interacting galaxies (~ 30%)• more early types (S0/Sa/Sb)

UV galaxies:• more late types (Sc/Sd/Ir/CB ~ 50%)

Comparison of correlation lengths

UV (FOCA sources):r_0=3.2 (+0.8, -2.3) h-1 Mpc

(Heinis et al. 2004)

IR (IRAS sources):r_0=3.9+-1.8 h-1 Mpc(Strauss et al. 1992)

UV galaxies seem to be less clustered than IR galaxies(confirmed by preliminary

GALEX results)

(FOCA result)

(Heinis et al. 2004 A&A 424, L9)

IRAS galaxies

Comparisons of stellar mass distributions

Mstars: estimated from L_K

(Cole et al. 2001 calibration, H_0= 70 km s-1 Mpc-1).

`Survival Tech.’ used.

Good overlap between twopopulations: medians differ

< 2, both are sub-M*.

Though:IR: slightly tilted for more massive end

UV: more galaxies with low mass (<10^10 L_sun).

L_tot v.s. mass L_IR/L_UV v.s. mass

• UV galaxies of lowest mass (< 10^10) have lowest L_tot and IR/UV• most massive IR galaxies (>10^12) are not galaxies with highest L_tot

• brightest IR galaxies (~ ULIRGs) have mass ~ M*• for given mass, UV galaxies have lower L_tot and IR/UV ratio than IR galaxies

60m LF of UV galaxies FUV LF of IR galaxies

• UV galaxies substantially under- represent galaxies of L_IR > 10^11 L_sun (LIGs).• ULIRGs (L_IR > 10^12) are completely absent in UV sample.

• IR galaxies can fully account for all UV galaxies of L_UV >

10^9 (L* ~ 4 10^9).

• some fainter UV galaxies (L_UV < 10^9) could be missing in IR sample.

L

IR-quiet UV galaxies

I Zw 18: prototype

metallicity = 1/50 solar(lowest known)

(from NED, Hubble Heritage Gallery image)

low mass (star+gas): ~ 2 10^8 M_sun

distance=15 Mpc

L(FUV)=2.5 10^8

not detected in FIR:L_dust/L_FUV < 0.25

SBS0335-052 another prototype IR-

quiet UV galaxy:

metallicity = 1/35 solar(2nd lowest)

(Houck et al. 2004, ApJS, Spitzer edition)

undetected by IRAS, but detected by both ISO and Spitzer: very different IR SED from normal galaxies

M82

mass (star+gas) ~ 2 10^9 M_sun

distance=58.3 Mpc

L(FUV) ~ 10^9

L_dust/L_FUV ~ 0.4

Characteristics of IR quiet UV galaxies

dwarf galaxies of low metallicity < ~1/10 solar

mass: a few 10^8 -- 10^9

UV lum: a few 10^8 -- 10^9

(~ 10 times < L* of FUV,

not z~0 LBG)

L_dust/L_FUV ~0.3(~ a few % of UV galaxies)

IR-quiet

IR-quiet

UV luminous galaxies (UVLG): z~0 LBGs(Heckman et al 05, ApJL GALEX edition)

•Nearby galaxies brighter than L_UV=2 1010 L(sun) with z<0.3

•10-5 galaxy/Mpc3 (~100 times less dense than LBGs)

(kpc

)

FUV Luminosity vs. Half-light Radius

(L๏/

kpc2

)

(L๏)(L๏

kpc-2

)

FUV Surface Brightness vs. Stellar Mass

(M๏)

Compact

Large

Population ComparisonLarge UVLGs, Compact UVLGs, LBGs

(Slide courtesy of Chris Martin)

12

11

10

9

Log LUV

1.5

1

0.5

9

Log rUV

12

11

10

9

M* 3

2

1

0

AUV

2

1

0

-1

Log b

9

8.5

8

7.5

[O/H]

A_FUV vs. SFR plot of UVLGs: comparison with ULIRGs and others

• UVLGs occupy the bright end of UV population, but still they have L_tot cutoff at ~ 2 10^11.

• Their attenuation (IR/UV ratio) spans the same range as that of major UV population.

• None of UVLGs is as bright as ULIRGs (>10^12).

/LIRGs

LBGs and SCUBA galaxies: UVLGs and ULIRGs at z~3

(Adelberger & Steidel 2001)•Blue dots: LBG galaxies. L_dust/L_1600 estimated using UV slope (very uncertain).

•Red squares: SCUBA galaxies (radio pre-selected) studied in Chapman et al. 2004.

• The overlap between the two populations is small:only 1 LBG detected by SCUBA (Chapman et al. 2000). Only 1 red square (SCUBA) has IR/UV < 100.

SCUBA galaxies: HST ACS images overlaid by radio contours

(Chapman et al. 2004, ApJ 611, 732)

Extended UV emission outside the radio/FIRemission region:unobscured UV light.

3”

Rest frame V-band luminosity and mass

(Smail et al. 2004, ApJ 616, 71)

LBG(Shapley et al 2001)SC

UB

A

SCUBA galaxies:stellar mass(estimated fromrest V-band lum.) plus gas mass ~ 5 10^10 M_sun,(Spitzer measurementsof rest frame K may be ~2 times higher).

A few times (~1.5 mag) more massive than LBGs (green curve).

Corrlations lengthsof SCUBA galaxies

and LBGs

(Blain et al. 2004, ApJ 611, 725)

SCUBA

LBGs

SCUBA galaxies:r_0=6.9+-2.1 h-1 Mpc

Significantly largerthan that of LBGs(~ 3 -- 4 h-1 Mpc)

UV and IR galaxies at intermediate redshifts (z ~ 0.6) --- early results of a

GALEX/SWIRE comparison study

• Why z=0.6?• close to the peak of cosmic SF suggested by some ISO and SDSS

fossil studies

• for z ~ 1 or larger , NUV is affected by rest frame Ly

emission/absorption (K-correction for L_UV very uncertain)

• at z~0.6:

• NUV ( 2300A) ---> rest frame FUV (1500A)

• MIPS 24m ---> rest frame 15m (L_IR indicator)

• IRAC 3.6m ---> rest frame K band (stellar mass indicator)

Field:

NUV sources: 8995F3.6 sources: 19100F24 sources: 2080Matches f24/NUV: 1086(52% of f24 srcs, 12% of NUV srcs)

GALEX ELAISE-N1_00 (~ 1 deg2) (inside SWIRE ELAISE-N1 ~ 9 deg2)

restrictions:- within 1 deg circle of GALEX field- exclude the SWIRE gap

Final area: 0.6 deg2

ELAIS-N1_00

NUV

ELAIS-N1, 24μ m

NUV

Sample selection of 0.5<z<0.7 galaxies • redshifts: photo-z catalog of ELAIS-N1 (ugriz + IRAC, by Rowan-Robinson)

• GALEX sources: 1124 (NUV < 24)

• MIPS sources: 396 (F24 > 0.2mJy)

• NUV/F24 matches: 159 ( 40% of F24 src, but only 14% of NUV src!!).

F24 ~ 0.2mJy, z~0.6 --> L_dust ~ 10^11 L_sunNUV ~ 24, z~0.6 --> L_FUV~ 10^9.5 L_sun

~

~

Mean f24 flux of z=0.6 UV sources from stacking

Stacked f24 image of UV sources in bin 9.4 < log(L_FUV) < 9.89.4 < log(L_FUV) < 9.8:

212 sources, <f24>=39 Jy

9.8 < log(L_FUV) < 10.2: 422 sources, <f24>=70 Jy

10.2 < log(L_FUV) < 10.5: 95 sources, <f24>=107 Jy

10.5 < log(L_FUV) < 10.8: 17 sources, <f24>=219 Jy (212 sources)

L_dust of z=0.6 UV galaxies:comparison with z=0 couterparts

<f24> --> < L(15m)> in rest frame•L_dust = 11.1 x L(15m) (Chary & Elbaz 2001, Elbaz et al. 2002)•Error bar estimated from fraction of F24 > 0.2mJy

In the 2 fainter L_UV bins,the means of z=0.6 andz=0 galaxies are close to each other, both are a factor of few below theSWIRE detection limit.

Comparison of mean L_dust/L_FUV ratios of z=0.6 and z=0 UV galaxies

• for galaxies of L_FUV < 10^10.2 L_sun, the IR/UV ratio does not show any evolution from z=0 to z=0.6.• for brighter galaxies of L_FUV > 10^10.2 L_sun, there seems to be a negative evolution in the sense that z=0.6 galaxies have lower ratios.

SEDs of the 160um source at z=0.6However, is the extrapolationfrom L(15m) to L_dustreliable???

Need to check the SEDs ofz=0.6 sources which are alsodetected in MIPS 70m bandand 160m band.

Only 1 z=0.6 source in24m sample (395 sources)is also detected in both70m and 160m band.It is a ULIRG with an SEDclose to that Arp220! (M82 SED is closer to Elbaz calib.)

F24 image of the f160 source at z=0.6

The green circle:160m beam (40”)

An isolated, clean source (no confusion).

SEDs of m sources at z=0.6Other 4 z=0.6sources are detectedin 70m, but notin 160m:2 have log(L_dust)< 12 and M82 likeSEDs.

2 have log(L_dust)>12 and SEDscloser to Arp220.

~

~

Conclusion:SEDs span a widerange.

effect of different calibrations

When Arp220 SEDis used in convertingL(15m) to L_dust,the mean IR/UV ratioof z=0.6 UV galaxiesin 2 bright bins (log(L_FUV)>10.2)is in good agreementwith that of z=0 galaxies.

Consistent with noevolution in the ratio!

L_dust/L_FUV ratio of z=0.6 IR galaxies

• Elbaz calibration• mean ratios derived from both stacking and ‘survival tech.’ (consistent with each other).• mean ratios of z=0.6 galaxies in all lum. bins are consistent with those of z=0 galaxies in the same bins.

Effect of Arp220 calibration• The arp220 calib. shifts the points along the IR/UV vs. L_dust correlation line, so does not change the result that the IR/UV ratio for given L_dust does not have any significant evolution.

Stellar mass of given L_FUV: comparison of z=0.6 and z=0 UV galaxies

pink points: z=0.6

blue squares: z=0

• Stellar mass: estimated from f3.6 (rest frame K).• SWIRE sensitivity limit of 3.6 m band (the green line).• The stellar mass of z=0.6 galaxies of given L_FUV is about ~2 times less than that of their z=0 counter- parts.

f3.6=3.7Jy

Stellar mass v.s. L_dust: comparison of z=0.6 and z=0 IR galaxies

red points: z=0.6

No evidencefor evolution in stellar massof IR selected galaxies.

z (dust)

(L0 Mpc-3)

(FUV)

(L0 Mpc-3)

(dust)/

(FUV)

A(FUV)

(mag)

0.06 6 107 1.8 107 3.3 1.2

0.5 30 107 4.1 107 7.4 1.9

0.7 50 107 7.7 107 6.5 1.8

1 96 107 6.9 107 13.9 2.1

• (dust): ~ (1+z)4, Spitzer results of Le Floch et al. (2005).

• (FUV): Schiminovich et al. 2005 (ApJL, GALEX Edition).

• Both IR and UV have luminosity evolution --> at high z galaxies on

average are more luminous, therefore with higher attenuation.

Evolution seen in IR and in UV:from z=0 to z=1

Summary

1. By selection, UV galaxies and IR galaxies have very different characteristic IR/UV ratios (the means differ by a factor of 10). 2. The morphological and stellar mass distributions of the two populations have good overlaps (> 70%). IR galaxies tend to be more massive and earlier types, with an excess of interacting galaxies, and UV galaxies to be less massive and later types.3. UV galaxies are less clustered than IR galaxies.

4. Galaxies with the highest SFR (>100 M /yr, Ltot > 1012 L ),

are missed in the UV samples.5. A population of low metallicity (< 1/10 solar), low mass (<10^9 M ) dwarf UV galaxies (prototype I Zw 18) are `IR quiet’, with the IR/UV ratio ~ 0.3 or less. They occupy only a few percent of a UV selected sample.

๏

๏ ๏

Summary (continue)5. The z~0 counterparts of LBGs are a population of compact luminous UV galaxies (UVLG). In terms of Ltot (SFR), UVLGs are more than

10 times fainter than ULIRGs. 6. LBGs and SCUBA galaxies (UV and IR selected galaxies at z~3) do not overlap with each other very much. SCUBA galaxies have significantly higher SFR, higher attenuation, higher stellar mass, and higher correlation length than LBGs.7. At intermediate redshifts of z~0.6, UV selected galaxies show moderate evolution in stellar mass in the sense that for a given luminosity, galaxies at z=0.6 have stellar mass ~2 times less than their z=0 counterparts. No evidence for any evolution in the IR/UV ratio (attenuation) for UV galaxies. For IR (24m) selected galaxies at z~0.6, no evidence is found for evolution of either the stellar mass or the IR/UV ratio for given LIR.

8. Both IR and UV evolve significantly from z=0 to z=1, and the ratio IR/UV increases by ~ 4. This is consistent with the scenario that high z galaxies are more luminous therefore with higher attenuation.

Star formation history measured in diff. wavebands

1) They have the same trend (rising from z=0 to z ~1, then becoming relatively flat).2) IR (ISO, IRAS, SCUBA) and rest frame UV (blue symbols and yellow shade) measurements agree with each other within a factor of ~2!!

Schiminovich et al. 2005

I Zw 18 does have dust:

Balmer decrement () study

(Cannon et la. 2001) found dust in

regions delineated by the boxes in the image, covering

only parts of the bubble-like star

formation regions: blow-away of dust? (Cannon et al. 2002, ApJ. 595, 931)

HST image