MIPS Observations of Galaxies, Near and Far

G. H. Rieke (University of Arizona), for

Almudena Alonso-Herrero (CSIC) Lei BaiEric Bell (MPIA)Karina Caputi (IAS)Hervé Dole (IAS)Jennifer DonleyEiichi EgamiChad EngelbrachtKarl GordonDean HinesJoannah HinzEmeric LeFloc’hGerry NeugebauerCasey PapovichPablo Pérez GonzálezMarcia RiekeJane RigbyYong ShiXianzhong Zheng (MPIA)

Galaxies Near and Far

• Even Simple Galaxies are Complex! Example of M31

• High-z Star Formation: Relevant Properties from Nearby Galaxies– How does metallicity affect galaxy spectral energy distributions?– What powers the far infrared emission in galaxies?– How can we calibrate the star formation rate from infrared measurements?

• High-z Star Formation: Relevant Properties from Far-Away Galaxies– How many hidden AGN might be confused with star forming galaxies?– Given confusion limits, how can we get far-infrared properties of high-z galaxies?

If we knew all the answers, we could determine star forming properties at high-z well!!

• High-z Star Formation with MIPS– How do the new “Madau” plots look?– Does the star formation vs. redshift match our models?– What happens at low luminosity?– What kinds of galaxies dominate the SFR?

M31Beck et al.

M31Thilker et al.

M31Guélin et al.

M31Devereux

M31MIPS GTO

M31MIPS GTO.

M312MASS

M31APOD/J. Ware

M31GALEX

M31X-Ray ROSAT

Stars, Dust, and Gas in M31Complexities show that abstracting to an unresolved point is error-prone!

Stars

blue = UVgreen = redred = J-band

Gas

blue = COgreen = HIred = H

Dust

blue = 24mgreen = 70mred = 160m

(Karl Gordon et al.)

High Redshift Galaxies: Relevant Properties from Nearby Galaxies

• Behavior of “Polycyclic Aromatic Hydrocarbon” features with metallicity– PAHs may dominate Spitzer signals at 24m out to z ~ 3– A change with metallicity/redshift would affect interpretation

of high-z results– Dramatic change seen at O/H ~ 25% solar

• What powers the far infrared emission in galaxies?– Is it powered by the youngest stars? Interstellar radiation field?– Close correspondence on all scales of H, 24 & 70m– 160m has different distribution, power source less clear (ISRF?)

• Calibration of star formation rate vs. infrared emission– Find close correlation between extinction-corrected Paschen

and 12m, 24m, and total far infrared

Behavior of “Polycyclic Aromatic Hydrocarbon” features with metallicity

Individually Detected ULIRGS have SEDs Like Local OnesEiichi Egami et al.

“PAH” featuresdominate restframe SED, ~ 7to ~ 9m. Hence, dominatesSpitzer 24m signals for1 < z < 3

PAH missingin AGN

Behavior of PAH Features with MetallicityChad Engelbracht et al.

• IRAC and MIPS colors can be combined to identify PAH emission– Subtract “stars” @ 3.6m off 4.5m, ratio to 8m and compare with 8/24m– SED models including PAH feature agree with locus of galaxies with

spectroscopically confirmed features

Behavior of PAH Features with Metallicity

• IRAC and MIPS colors can be combined to identify PAH emission– Subtract “stars” @ 3.6m off 4.5m, ratio to 8m and compare with 8/24m– SED models including PAH feature agree with locus of galaxies with

spectroscopically confirmed features

– Galaxies withoutPAH featuresoccupy adifferent region

Behavior of PAH Features with Metallicity

• IRAC and MIPS colors can be combined to identify PAH emission– Subtract “stars” @ 3.6m off 4.5m, ratio to 8m and compare with 8/24m– SED models including PAH feature agree with locus of galaxies with

spectroscopically confirmed features

– Galaxies withoutPAH featuresoccupy adifferent region

– Other galaxies divide into two areas, similar to those with spectral data

PAH Feature Disappears Abruptly at ~ 25% Solar Metallicity

• Use photometric technique to separate PAH SEDs from non-PAH ones• Plot metallicity vs. 8/24m• Separation in 8/24 predicted from previous plot• Find non-PAH SEDs lie below 25% solar O/H

(or 12 + log (O/H) < 8.2)• Probably not an

issue to z ~ 3 for the highly luminous galaxiesthat can bereached withSpitzer

(This figure differs from the similar one shown by Jessica Rosenberg because it goes to much lower metallicity)

SBS0335

Caution: individual“low metallicity” galaxies may have regions ofhigher metallicitydue to localizedstar formation!

What powers the far infrared?

M33before

JoannahHinzet al.

Whatpowersthe farinfrared?

Detailedimagecomparisonshows thatcompact24 & 70memission isentirelyfrom HIIregions.

Low passfilteringisolates theextended emission

afterM33

M33after

Extended 24m and70m arevirtuallyidentical toextendedH.

after

Extended 24m and70m arevirtuallyidentical toextendedH.

160m isdistinctlydifferent

M33

M33Effelsberg 17cm 160m

160m is a good match tofilled aperture radio

M33

J. Hinz

8-3m

Extended PAH and 160m are morphologically similar to extended non-thermal radio; powered by diffuse

interstellar radiation field from modest-age stars (recall Dopita talk)

Low Surface Brightness Galaxy UGC 10445Has Extended Cold Halo at 160m

(Joannah Hinz et al.)

a. 3.6mb. 4.5mc. 5.8md. 8me. 24mf. 70mg. 160m

Radial Profiles Clearly Show the Low Temperature Halo

• Lower two frames show the observed radialprofiles at 24 and 70m

• Upper frame shows 160mprofile, plus 3.6 and 24m ones convolved to the 160m resolution

• The 160m “halo” extendswell beyond the galaxyas defined at the twoother wavelengths

• It is not yet understoodwhat is warming thedust so far from thegalaxy

Calibration of star formation rate vs. infrared emission

Correcting from 24m to Far Infrared/Total Infrared Can Introduce Significant Errors in Star Formation Rates Based on FIR/TIR

Based on nearbygalaxy SEDs, thereare large uncertaintiesin extrapolating anobserved 24m fluxdensity to either a70m one, or to totalinfrared output.

From Danny Dale et al.

However, Accurate SFRs Can Be Determined from Mid-IR Alone

Extinction-correctedPaschen correlatesvery closely with 24mflux density in M51(Daniela Calzetti et al.)

Recall the behaviorin M33, where wefound that H and 24m flux densityhave identical distributions over the galaxy.

• A tight correlation is also seen between extinction-corrected Paschen and 12m (from IRAS) flux density for normal galaxiesand LIRGs (at least up to 5 X 1011 Lsun).

• The correlation is of similar quality to that with total IR and thedispersion is small. (from Almudena Alonso-Herrero et al.)

Accurate SFRs Can Be Determined from Mid-IR Alone

High Redshift Galaxies: Relevant Properties from Far-Away Galaxies

• How many hidden AGN might be confused with star-forming galaxies?– Sample of 27 galaxies in 2Msec region of CDFN selected as

AGN from radio/24m properties - independent of X-rays– 16 of the 27 are not cataloged as X-ray sources– 5 - 7 are not detected even at 2- level– Sample of ~ 100 power law galaxies in CDFS has ~ 40%

not detected in X-ray

• Given confusion limits, how can we get far-infrared properties of high-z galaxies?– How can one associate 70 & 160m sources with 24m ones?– 70 & 160m identifications of 24m sources account

for > 75% of total 70 & 160m infrared background– Validates using 24m sources to locate 70 & 160m ones

How many hidden AGN might be confused with star-forming galaxies?

Identification of X-Ray-Free AGNJennifer Donley et al.

• Galaxies too bright in radio for star-forming IR/Radio ratio have AGN• Allows identification of AGN sample without using X-ray selection

Star-forminggalaxies andradio-quietAGN

Identification of X-Ray-Free AGNJennifer Donley et al.

• Galaxies too bright in radio for star-forming IR/Radio ratio have AGN• Allows identification of AGN sample without using X-ray selection

Radio-loud andradio-intermediateAGN

• Of 27 AGN selected thisway with > 1 Ms Chandra exposure, 16 are not listed in CDFN catalog

• 9 more are detected at the 2 to 5- level

• 5 are not detected, even with stacking to achieveexposure of > 7Ms

• 2 more are undetected but may be confused

• SEDs of X-ray faint objects(left) are indistinguishablefrom X-ray bright ones(right)

• All are dominated by stars,• Half are detected at 24m

• In CDFS, ~ 100 sourcesdetected at 24m andwith power law SEDs inIRAC bands

• 40% are not detected inX-ray (even in center ofChandra field) (AlmudenaAlonso-Herrero et al.)

• Of 27 AGN selected thisway with > 1 Ms Chandra exposure, 16 are not listed in CDFN catalog

• 9 more are detected at the 2 to 5- level

• 5 are not detected, even with stacking to achieveexposure of > 7Ms

• 2 more are undetected but may be confused

• SEDs of X-ray faint objects(left) are indistinguishablefrom X-ray bright ones(right)

• All are dominated by stars,• Half are detected at 24m

Given confusion limits, how can we get far-infrared properties of high-z galaxies?

Stacking on 24m positionsyields strong detections at70 and 160m

Stacking on 24m positionsyields strong detections at70 and 160m

With stacking,most of the far infraredbackgroundis resolved

Using 24m Detections to Probe Far Infrared Galaxy Properties

S(24m) 24m 70m 160m> 80Jy 70% 75% 63%> 30Jy* >77% >84% >74%

* 24m detection is very incomplete at this flux density

• Stacking on the 24m positions yields a detection in an average sense accounting for ~75% of the cosmic background in both far infrared bands

Validates using 24m sources to locate 70 & 160m ones(although a few objects have very large 70/24 flux ratios)

• How do the new “Madau” plots look?– No surprises, error range is coming down

• Does the star formation vs. redshift match our models?– No!

• What happens at low luminosity?– Luminosity function is similar to local one

• What kinds of galaxies dominate the star formation?– To z ~ 1.5, dominated by LIRGs– For z > 1.5, ULIRGs important, SEDs similar to local ones

(with significant contamination from AGNs)– z ~ 1 LIRGs have similar morphology to local ones, – z ~ 1 LIRGs also have similar morphology to z ~ 1 IR-inactive ones– Suggests LIRG-level activity accompanies a certain level of asymmetry– This asymmetry is common at z ~ 1 (and rare locally), consistent with

LIRG-level star formation episodes also being common there

High-z Star Formation with MIPS

How do the new “Madau” plots look?

Evolution of Comoving IR Energy DensityEmeric leFloc’h et al.

Green is total, blueis galaxies below1011 Lsun, yellowis LIRGs (1011

< L < 1012) andred is ULIRGs(1012 < L).

Solid line evolvesas (1+z)3.9

Dashed line is UVwithout extinctioncorrection, dottedline is IR+UV

Evolution of Infrared Luminosity FunctionLe Floc’h et al.

Fits are local LF evolved in density and luminosity as (1+z)with D=1.0 & L=3.15: can trade off density and luminosity evolution, but in all cases L* increases rapidly with z

Evolution of Star Formation RatePablo Pérez-González et al.

The symbols show the results with various functional forms for the LF that arethought to bracket the uncertainties due to it.

Recent work byZheng et al.would favor thelower side of theindicated range.

However, thereare additionaluncertainties dueto the conversionfrom IR to SFR,and other causes.

Does the star formation vs. redshift match our models?

Fit to Models is Generally PoorPablo Pérez-González et al.: photo-z’s from empirical templates,

combines both GOODS - CDFS & HDFN

Result Confirmed with More Accurate Photo-z’s in GOODS-CDFSKarina Caputi et al.

• More accurate redshifts use ultra-deep JHK of Franx et al.• Show the influence of PAHs (in comparison with galaxies not detected at 24m)• Fraction of galaxies at z > 1.2 agrees well with Pérez-González et al.• Agreement shows that cosmic variance is not a factor• Model fit is nearly as poor for both

What happens at low luminosity?

Stacking Lets Us Probe IR Luminosities to Faint LevelsXianzhong Zheng et al.

• Images below at a range of redshifts show detections at effective noiselevel only ~ 10% of the conventional confusion limit

Low Luminosity Luminosity Function is Similar to Local One

• Slope of 1.2 + 0.3 ingood agreement withlocal value

• Low and moderate luminositygalaxies have similarnear UV/IR behavior aslocal galaxies*

• Can use well-determinedB-band LF for lowluminosity galaxies

• Resulting luminosity densityagrees “perfectly”with result assuming local LF by LeFloc’h et al.

* but compare with talkon GALEX by Kevin Xu

What kinds of galaxies dominate the high-z star formation?

Individual Detections Dominated by LIRGs to z > 1Emeric LeFloc’h et al.

At z > 1.5, individual detections dominated by ULIRGsPablo Pérez-González et al.

Individually Detected ULIRGS have SEDs Like Local OnesEiichi Egami et al.

• Individual high-zradio/Spitzer detections havebeen shifted tothe rest frame

• “Cold” objects matchthe SED of Arp220 closely*

• “Warm” objects have a compositepower-law-likeSED similar togalaxies withAGNs

* Remember the discussion inspiredby Mike Dopita onTuesday afternoon

At z ~ 0.7, mostof the IR-emittinggalaxies are visually classifiedas spirals. Thedistribution amongmorphological typescan be seen from these observed luminosity functions(Eric Bell et al.).

What a High-z LIRG Looks Like, I

What a High-z LIRG Looks Like, IIYong Shi et al.

Morphology Classification Using Concentration, C, and Asymmetry, A

20

80log5r

rC

Where r20 and r80 arethe radii enclosing 20and 80% of the galaxy flux.

0

1800

I

IIA

I0 and I180 are theintensity of the galaxyimage, and of the imagerotated by 180o,respectively. The valueis minimized by selectingthe optimum center forthe rotation and is also corrected for noise.

Morphology Classification Using Concentration, C, and Asymmetry, A

z ~ 1 LIRGs tend to be relatively asymmetric and modestly concentrated.Concentration is similar to local late-type spirals, but asymmetry is muchgreater than local “normal” galaxies (that have typical log A ~ -1.4). However, z ~ 1 IR-inactive galaxies are also highly asymmetric.

20

80log5r

rC

Where r20 and r80 arethe radii enclosing 20and 80% of the galaxy flux.

0

1800

I

IIA

I0 and I180 are theintensity of the galaxyimage, and of the imagerotated by 180o,respectively. The valueis minimized by selectingthe optimum center forthe rotation and is also corrected for noise.

• Local LIRGS have similar C & A to z ~ 1 LIRGs• Normal local galaxies have much lower asymmetry• It appears that a certain degree of asymmetry generally accompanies

LIRG-level star formation• Locally, it requires an interaction or some other extraordinary event

to set up this level of asymmetry• At z ~ 1, LIRGs and IR in-active galaxies have similar and high

levels of asymmetry• This behavior is consistent with LIRGs and IR in-active galaxies

coming from the same parent population• This population may be sufficiently asymmetric that it is subject

to spontaneous episodic LIRG-level star formation

A Speculative Explanation

LIRGs Inactive LIRGs

z ~ 1 z ~ 1 z ~ 0A 0.30 + 0.02 0.20 + 0.01 0.26 + 0.03C 3.0 + 0.1 2.9 + 0.04 2.7 + 0.08

Galaxies Near and Far

• Even Simple Galaxies are Complex! Example of M31• Abstraction of high-z galaxies to unresolved points is error-prone!

• High-z Star Formation: Relevant Properties from Nearby Galaxies– Metallicity < 25% solar has weak PAHs - detectable IR galaxies not affected– Very young stars power 24 & 70m - longer wavelengths powered by ISRF(?) – Young star formation (H recombination) closely correlated with 12 & 24m

• High-z Star Formation: Relevant Properties from Far-Away Galaxies– Up to half the high-z AGN are not detected in the deepest X-ray surveys– Most (> 75%) of the cosmic far infrared background is from galaxies detected

with Spitzer at 24m

• High-z Star Formation with MIPS– New “Madau” plots are similar to extinction-corrected UV ones -

uncertainties are coming down– Phenomenological models match star formation vs. z poorly, esp. for z > 1.2– Luminosity function at low luminosity similar to local LF, slope ~ 1.2 ---

LF shape at high luminosity also similar, but L* increases rapidly with z– High-z LIRGs are relatively asymmetric, like local ones -- high-z field galaxies

are similarly asymmetric unlike local ones, implying that the high-z onesare sufficiently “disturbed” that they can enter the LIRG state spontaneously