Stellar Population Gradients in ULIRGs: Implications for Gas Inflow Timescales
aa r X i v : . [ a s t r o - ph . C O ] M a y Draft version November 23, 2018
Preprint typeset using L A TEX style emulateapj v. 11/10/09
STELLAR POPULATION GRADIENTS IN ULIRGS: IMPLICATIONS FOR GAS INFLOW TIMESCALES
Kurt T. Soto, Crystal L. Martin
Physics Department, University of California, Santa Barbara, CA 93106-9530
Draft version November 23, 2018
ABSTRACTUsing longslit, optical spectra of Ultraluminous Infrared Galaxies (ULIRGs), we measure the evo-lution in the star-formation intensity during galactic mergers. In individual galaxies, we resolve kpcscales allowing comparison of the nucleus, inner disk, and outer disk. We find that the strength ofthe H β absorption line increases with the projected distance from the center of the merger, typicallyreaching about 9 ˚Aaround 10 kpc. At these radii, the star formation intensity must have rapidly de-creased about 300-400 Myr ago; only stellar populations deficient in stars more massive than Type Aproduce such strong Balmer absorption. In contrast, we find the star formation history in the centralkpc consistent with continuous star formation. Our measurements indicate that gas depletion occursfrom the outer disk inwards during major mergers. This result is consistent with merger-induced gasinflow and empirically constrains the gas inflow timescale. Numerical simulations accurately calculatethe total amount of infalling gas but often assume the timescale for infall. These new measurementsare therefore central to modeling merger-induced star formation and AGN activity. Subject headings: galaxies: starburst — galaxies: evolution — galaxies: active — galaxies: formation INTRODUCTION
Ultraluminous Infrared Galaxies (ULIRGs) are someof the most luminous objects in the local universe withlog( L IR /L ⊙ ) >
12 – a result of starburst and nuclear ac-tivity triggered by major mergers. The number density ofULIRGs per unit comoving volume has evolved stronglycausing ULIRGs to become a rare phenomenon locally,but infrared-luminous sources appear to dominate thestar-forming activity beyond z ∼ . DATA
Observations
The Keck II echellete spectroscopy of 2 Jy ULIRGs, de-scribed previously in Martin (2005, 2006), is well suitedto our study of H Balmer emission/absorption. Thesedata resolve scales of 0.8 ′′ , about 1.5 kpc, spatially and ∼
70 km s − spectrally. The 20 ′′ slit is long enough tosufficiently sample the sky at both ends of the slit, en-suring accurate sky subtraction and continuum intensity.The slit position angle includes a second nucleus whenpresent (Table 1). We inspected 43 ULIRG spectra andselected 25 for analysis of any stellar H β absorption. Thissubsample presents H β emission such that the full widthof the emission line in their nuclear spectrum, defined bythe wavelengths where the trough changes from emissionto absorption, is no more than 1800 km s − .The resulting subsample includes a range of stages inthe merger progression. Veilleux et al. (2002) classifiedthe merger stage for 11 of the 25 objects. Using K-bandimages from Murphy et al. (1996), and their identifica-tion as multiple nuclei systems, we apply the Veilleuxscheme to the remainder finding 8 pre-mergers (i.e. two TABLE 1Object Morphology
IRAS name IR Opt Veil. log( L IR L ⊙ ) z PA(1) (2) (3) (4) (5) (6) (7)17028+5817 D D IIIa a a a a a a a a a a Note . — Col.(1): Name, Col.(2): Infrared Kband morphology. “S”, “D”, and “M” denote a sin-gle, double and multiple nucleus morphology respectively(Murphy et al. 1996). Col.(3): Optical R band morphol-ogy. Similar to R band morphology with an additional“t” to signify an extended tidal region, and “W” to de-note a widely separated pair (Murphy et al. 1996). Col.(4):Merger Classification based on Veilleux et al. (2002). Thisclassification separates wide binary pre-mergers (IIIa),close binary pre-mergers (IIIb), diffuse merger(IVa), com-pact merger (IVb), and old mergers (V). Col.(5): IR lu-minosity from (Murphy et al. 1996). Col.(6): redshift.Col.(7): Position angle of slit a Classifications directly from Veilleux et al. (2002) nuclei), a multiple merger (IRAS 10565+2448), 12 merg-ers, and 4 old mergers. The ULIRGS in the merger stagetend to present higher surface brightness, so our analy-sis of spatial gradients in Balmer absorption is heavilyweighted towards this merger phase.
Measurements
We examined the positional dependence of the H β lineprofile using a sliding aperture for spectral extraction.We matched the aperture width to a seeing element( ∼ ′′ .
8) and extracted a spectrum at each line of thespectrogram. At some spatial positions, as illustratedin Fig. 1, the spectra present a two-component H β lineprofile. A broad absorption trough underlies the narrowemission line. We fit the absorption and emission com-ponents simultaneously using non-linear least-squares fit-ting with MPFIT in IDL. These spectra exhibit continu-ous change from position to position due to atmosphericsmearing, so parameters found from fitting one apertureposition were supplied as the guess for the fit of the nextposition. The small, gray points in Figure 2 show the re-sult. Over scales of a few kpc, this analysis often reveals an increase in H β absorption equivalent width, W Hβ ,with distance from the peak continuum emission in theR band. Fig. 1.—
The aperture selected at 2 kpc north west from thecenter in IRAS17208-0014 is shown above with a fit to both H β components simultaneously. Emission component is a narrow linecontributed by photoionization of the ISM. The absorption compo-nent is provided by the presence of A stars in the stellar population.The emission line is narrow enough that the absorption line is notaffected in the fit. To eliminate effects due to correlated errors, we ex-tracted spatially separated spectra based on the shape ofthese radial H β absorption and continuum surface bright-ness trends. These spectra are shown in Figure 3, wherethe larger black points in Figure 2 show the H β absorp-tion equivalent width obtained by fitting these spectra.We measured H α surface brightness in these same aper-tures. The H δ lines could only be fitted in more than 3apertures for one object.To correct the H α emission for underlying stellarH α absorption, we measured the H β absorption equiv-alent width and assumed the H α absorption equiva-lent width was two-thirds as large. Inspection of syn-thetic spectra computed for stellar population mod-els (Bruzual & Charlot 2003) verified that this scal-ing works over a broad range of star formation mod-els, including those with truncated star formation his-tories. The maximum value of the correction, 3˚A, issmall compared to the emission equivalent width. Theflux in H α emission is also corrected for extinction.We determined the color excess, E(B – V), per aper-ture by measuring the Balmer decrement and usinga standard reddening law (R V = 4.05 = A V /E(B –V)) and the extinction law from Calzetti et al. (2000).When significant, we discuss the correction for at-tenuation by Galactic dust on a galaxy-by-galaxy ba-sis in Section 3.2. We estimate areal star forma-tion rate using the conversion SF R (M ⊙ yr − kpc − ) = Fig. 2.— key :
Upper:
The equivalent width of H β due to stellarabsorption is plotted versus position. The light grey points arethe measurements from the preliminary sliding aperture analysisof the spectrum. Black points are chosen apertures based on trendsin the preliminary analysis. The solid black line shows continuumintensity near H α as a function of position in units normalizedto the peak of the continuum, illustrating the alignment with theimages in the lower panel and an indication of the relative signalto noise in the extended regions. The solid grey line is the samefor the continuum near H β . Lower:
An R band image (courtesy ofT. Murphy) is shown with the position of the slit plotted to showwhere the 1” ×
20” slit was placed. We can see increasing W Hβ in the extended regions with a sharp drop in the nucleus. Figures2.a – 2.j are available in the online version of the Journal. [ L ( Hα ) /A (kpc )] / . × erg s − (Kennicutt 1998). RESULTS
The presence of H β emission and absorption reveal dif-ferent epochs of star formation. Recombination emissiontraces ionization by very massive stars, and Balmer ab-sorption is strongest in Type A stars. Strong Balmerabsorption dominates the stellar spectrum when hotterstars are absent. Figure 4 shows how the measured H β absorption equivalent width changes along the slit. Theprojected distance is measured from the position along Fig. 3.—
The extracted spectra used for measurement are plottedin grey with the fit to the absorption overplotted in black. Eachof these spectra correspond to one of the black points in Figure 2.Figures 3.a – 3.j are included in the online version. the slit with the highest R band surface brightness. This center typically coincides with the position of the nucleusmeasured in the K band. IRAS17208-0014 has two max-ima in R band continuum, however only the brightest ofthese maxima coincides with a K-band nucleus, whichwe regard as the center of the object. IRAS15245+1019has two nuclei resolved in K-band images (Murphy et al.1996) separated by ∼ ′′ , but we use the nucleus withthe highest R-band surface brightness to define the cen-ter. K-band imaging of IRAS20046-0623 does not resolvetwo nuclei, but has elongated structure over 4 ′′ from eastto west. The position of the maximum continuum sur-face brightness near H α differs from that of H β as well.For this object we chose to retain the definition of centerbased on the maximum R-band surface brightness sincethe K-band centroid is not directly measured in thesedata.The strength of the stellar Balmer absorption typicallyincreases with distance from the center. For example, inFigure 4 the H β absorption strength across IRAS 09111-1007 increases steadily out to 7 kpc. This behavior char-acterizes merger systems presenting relatively symmet-ric structure on either side of the nucleus. Objects suchas IRAS20046-0623 observed at an earlier stage in themerger progression often present more complicated mor-phology. The spatial gradient in H β equivalent width isnot smooth like that in IRAS09111-1007, but the absorp-tion strength is still largest away from the center. Thesurface brightness in the post mergers is generally too lowto look for the effect at large radii. The galaxies wheresuch gradients can be measured fall in or between theclose binary pre-merger stage (IIIa) to compact mergerstage (IVb). The 10 targets caught at these stages formour spatially resolved subsample, shown in Figure 4 anddiscussed individually in Section 3.4. Fig. 4.—
Equivalent width of H β absorption is plotted as a func-tion of distance from the maximum continuum surface brightness.The absorption strength generally increases away from the center. Emission effects the line profile and equivalent widthmeasurement less in H δ . However, IRAS 23365+3604 is the only galaxy where our data quality is sufficient tostudy their line profiles in more than 3 apertures. We fitthe H β and H δ emission/absorption profiles simultane-ously, requiring the absorption features to have the samewidth and Doppler shift but allowing different normaliza-tions. We apply the same requirement to the fit of theemission components. Figure 5 shows that the equiva-lent widths of these lines increases with distance from thecenter. Hence our results are not strongly influenced byour decomposition of the emission and absorption com-ponents of the H β profile. We can use the measuredstrength of Balmer absorption to constrain the recentstar formation history at different radii. The measure-ments in Figure 5 also show that the equivalent widthof H β and H δ are similar in the apertures that couldhave both balmer lines measured. The similarity is in-consistent with the population modeling which predictsslightly greater equivalent widths for H δ . Dominanceof spectral types A0 through A5 could possibly explainthis. A star formation history that allowed these starsto dominate might be required if measurements of moreULIRGs show this same inconsistency. Fig. 5.—
Equivalent widths of H β and H δ absorption are plot-ted as a function of distance from the central continuum peak inthe chosen apertures for IRAS23365+3604. The Balmer equivalentwidths appear to follow the same pattern. Truncation Timescales from Stellar PopulationSynthesis
Using GALAXEV (Bruzual & Charlot 2003), we syn-thesized model spectra of the stellar population for threedifferent star formation histories: (1) δ function, (2) con-tinuous star formation history that is extinguished after40 Myr, and (3) continuing star formation (CSF). Weassumed solar metallicities. We measured W Hβ in thesespectra using the same techniques applied to the obser-vations. We illustrate the evolution of the W Hβ as thestellar population ages in Figure 6. Around 850 Myr,continuous star formation reaches a maximum of 6 ˚Ain equivalent width as the population reaches an equilib-rium of births and deaths of A stars. In the other models,as O and B type stars die off, the stellar population dom-inating the stellar continuum emission gradually shifts tolater stellar types. A single A5V star has W Hβ near 20˚A, but the spread in stellar types does not allow the com-posite models to reach this limit. Once star formationis stopped, the W Hβ increases steadily towards a maxi-mum of 9˚A at 400 Myr. After 400 Myr, the W Hβ falls asthe number of A stars declines, reaching 6˚A after about850 Myr. An equivalent width greater than 6 ˚A clearlyrequires a very rapid decrease in the star formation ratein the past Gyr.The large values of W Hβ found at radii R > Hβ and time since truncation, i.e. Figure 6solid curve, we conclude that star formation ceased atleast 100 Myr ago at R > Hδ equivalentwidth varies much like H β in Figure 6 owing to its sim-ilar origin. The W Hβ gradient continues towards thecenter of IRAS 23365+3604, and we interpret this as ev-idence that the time since truncation decreases towardsthe center. In other words, star formation is apparentlybeing shutdown from the outside inwards.Another popular diagnostic of stellar age is the spec-tral break known as D4000. Our spectral bandpass doesnot completely cover this feature, but we can use thestellar population models along with our observed trendin Balmer equivalent width to predict the radial varia-tion in D4000. The D4000 index grows as the stellarpopulation ages. Measurements of D4000 break the agedegeneracy to either side of the maximum in W Hβ at400 Myr. Hence, we expect D4000 would show a steadyincrease with increasing radius in ULIRGs.Our simple stellar populations represent the most ex-treme limiting cases. A declining star formation rateproduces a rising W Hβ similar to our models with com-plete cessation. The stronger the rate of change in thestar formation rate, the closer the W Hβ comes to that ofthe truncated model. We can distinguish these two starformation histories by the absence/presence of Balmeremission in our ULIRG spectra. The presence of nebularemission indicates the level of ongoing star formation.Complete truncation of star formation activity reducesthe ionization rate, and Balmer emission ceases. We ex-amine the radial depedence of the implied star formationrate galaxy by galaxy. Spatially Resolved Star Formation Histories
In Section 3, we identified 10 galaxies having extractedspectra in a least four apertures with SNR ≥
5. Pre-merger and merger systems dominate this spatially re-solved subsample. We discuss the recent star formationhistory in each individually here. For reference, Table 2lists the measured star formation rate and stellar popula-tion ages for different apertures in each object. The starformation rates near the center should be treated cau-tiously due to two important systematic errors. First,an active galactic nucleus may contribute to the ioniz-ing photon luminosity, thereby lowering the inferred starformation rate. Second, the reddening measured by the Balmer decrement in the central aperature reflects onlythe least obsured regions; the extinction may be too highin some regions for Balmer photons to escape causing usto underestimate the central star formation rate. Thesefactors have little impact on our analysis of spectra ex-tracted outside of the central kiloparsec and our discus-sion of radial gradients.Our results suggest the following star formation histo-ries.
IRAS15245+1019:
This ULIRG has two K band nu-clei separated by ≈ . ± .
05 and 6 . ± . ⊙ yr − kpc − forthe brighter south-east nucleus and dimmer north westnucleus respectively. The relative dimness in the visibleband of the southeast nucleus is due to high levels of ex-tinction by dust. The brighter nucleus in the interactionhas W Hβ . ± . . ± .
02 Gyr. At 7 kpc,however, there is little to no star formation occurringcurrently. W Hβ indicates that 0 . ± . IRAS20046-0623:
From detailed analyses of galacticrelative velocities and tidal tail characteristics, it is deter-mined that IRAS20046-0623 has gone through pericen-ter in the last few times 10 years (Murphy et al. 2001).The separated nuclei are masked within the extendedcentral bulge that appears in both K band and R bandimaging. The morphology of the extended tails are rep-resentative of two nearly orthogonal disks in the processof merging. Peaks in star formation (3 . ± .
05 and8 . ± . ⊙ yr − kpc − ) occur in the nuclei indicatedin Murphy et al. (2001). As shown in figure 2.c, thesenuclear regions have W Hβ that indicate star formationhas continued for 0.18 ± ± ∼ Hβ is beyond that at-tainable by CSF. The truncation time scale indicated bythese measurements for these bins are from 80 to 100Myr. Galactic dust contributes ∼ α . IRAS20087-0308:
The morphology of the K-band im-age shows a single nucleus, implying that the two galaxieshave fully merged. Similarly we see one centrally locatedposition on the slit with continuing star formation. TheSNR for this object makes W Hβ difficult to measure,but the nuclear measurement in this object does appearslightly lower than the extended regions. This differenceis not significant enough to measure a gradient in thepopulation age, so we estimate an overall age of 100 Myrsince star formation stopped across the disk. Galacticdust contributes ∼ α . IRAS17208-0014:
This galaxy has a single K bandnucleus with many diffuse disturbed regions around itin the R band. The aperture with peak star formation(0 . ± .
03 M ⊙ yr − kpc − ) is adjacent to the regionwith the smallest W Hβ (4 . ± . ±
30 Myr of continuous star formation (Figure 2.e).Next to the K band centroid at the max of the R bandcontinuum W Hβ = 6 . ± . . ± .
003 M ⊙ yr − kpc − . This aper-ture shows an underlying population with a truncationtime scale of 57 ±
10 Myr. The furthest aperture witha center 4.25 kpc from the nucleus has a slightly largerW Hβ and very little star formation, making the timesince truncation of this population slightly larger at 100 ±
35 Myr. Galactic dust contributes ∼ α . IRAS18368+3549:
The K-band morphology is a sin-gle nucleus with the R band revealing a faint tidal regionencircling the nucleus. From this morphology we expectthat the nuclei have completely merged and tidal distor-tions have had time to circularize. The nucleus has anW Hβ of 6.8 ± ±
17 Myr in the truncated constant star formationhistory, and 95 ±
17 Myr in the δ -fn model. W Hβ in theextended region at 10 kpc from the nucleus is very largeputting it beyond the model describable with these starformation histories. IRAS23365+3604:
The morphology of this galaxy issimilar to that of IRAS18368+3549, in that it has abright nucleus with a tidal region that has had time tocircularize. The nuclear regions have a peak in star for-mation rate (2 . ± .
02 M ⊙ yr − kpc − ), and low W Hβ (3 . ± . ± ± ± δ -fn model). At a distance of 5.3 kpc from thenucleus, the slit intersects an extended tidal region, thecenter of which shows a dip in equivalent width (3 . ± . Hβ with radius. In thefurthest aperture no H α emission is measured, and themeasured W Hβ of absorption indicates that truncationoccurred 140 ±
25 Myr ago. Galactic dust contributes ∼ α . IRAS09111-1007:
This galaxy has undisturbed tidalregions characteristic of post-merger morphology (Figure2.a). The central kpc has W Hβ = 4 . ± . Hβ increases toward the edge of the object, upto 9 -10 ˚A at projected distances of 3 - 5 kpc. Theselarge equivalent widths imply that 300 Myr has passedsince the truncation of star formation. IRAS10378+1109:
This galaxy has a compact nucleus,but it is measurable in several apertures. The fit of theabsorption trough in the central aperture yields an un-physically narrow fit due to contamination by strongemission and is rejected from the rest of the analy-sis. This strong emission indicates strong star formation(6 . ± .
06 M ⊙ yr − kpc − ). The aperture 4.7 kpc hasW Hβ = 6 . ± . ∼
90 Myr.
IRAS11506+1331:
Star formation is maximum (2 . ± .
06 M ⊙ yr − kpc − ) in the central most aperture (1.3kpc). If a continuing star formation history were consid-ered, the absorption (W Hβ = 6 . ± .
25 ˚A) would indicatethat constant star formation has continued for 1 Gyr,however the presence of a strong absorption line indicatesthat there was a previous burst of star formation thatdominates the continuum over the stars being producedin the current epoch of star formation. In this case, theend of the previous burst of star formation would have ended 69 ± < ⊙ yr − kpc − )and truncation time scales up to 150 ±
35 Myr at 7 kpc.
IRAS10565+2448:
This object is classified as a mul-tiple merger by (Murphy et al. 1996), with a third com-ponent separated by 20 kpc from the brightest knot inour measurement. These data sample two of the nucleiand measure an increase in W Hβ with radius. In thestronger nucleus W Hβ drops to 1 . ± . . ± .
08 M ⊙ yr − kpc − )indicating a very young population. W Hβ increases to8 . ± . ∼
180 Myr. A dip in the trend occurs off center of theweaker nucleus. The truncation timescale in this regionis ∼ M yr .From analysis of the spatially-resolved sub-sample,we draw the following conclusions. Objects suchas IRAS20087-0308, IRAS18368+3459, IRAS23365-3604and IRAS10565+2448 have regions at large radii thatshow no H α , indicating that the star formation has beencompletely shut off. The others show examples of con-tinuing, but decreased, star formation activity at largerradii. Often the star formation rate per unit area dropsby at least a factor of 10 at the largest radii measured.Though the gradient varies among the galaxies, we seeregularity in that the time since truncation increases out-wards in radius. Other Sources of Spatial Gradients in W Hβ We considered other physical effects that could causelarge W Hβ values. Previous studies (Poggianti & Wu2000) attributed large Balmer equivalent width to in-termediate age stars having time to drift away from thedusty molecular clouds in which they were born. Thiseffect can have a major effect in integrated spectra of en-tire objects. For dust to be the cause of larger equivalentwidth at larger radii, the population of hidden youngstars must grow with distance from the nucleus. Thisconclusion contradicts measurements of the moleculargas distribution in ULIRGs. An active nucleus may con-tribute to the continuum emission in the central aper-ture. AGN typically contribute 15 −
30% of the totalbolometric luminosity from cool HII and LINER ULIRGsVeilleux et al. (2009). Boosting the continuum, however,only decreases the apparent stellar W Hβ , so an AGN can-not explain large W Hβ values. Moreover, the gradientsthat we find on spatial scales larger than the seeing can-not be caused by an AGN. Comparison of W Hβ among Nuclei Among the 25 spectra examined, the surface bright-ness of the merger remnant was not always high enoughto measure the spatial dependance of the line profile. Wecan, however, compare W Hβ among all the nuclei as il-lustrated in Figure 7. In the pre-merger subsample wemeasured W Hβ of both nuclei seen. Of these 18 nuclei, 4of them showed broad H β emission that did not allow themeasurement of an underlying absorption line. A typicalnucleus in the pre-merger sample presents W Hβ of 4 ˚A,although the full range in W Hβ is large. In the mergersample, the median W Hβ is higher. Our sample is toosmall to be certain that this difference is significant; but,if verified, would suggest the central star formation de- Fig. 6.—
Three stellar population age measures are compared us-ing solar metallicity burst models from Bruzual & Charlot (2003).W Hδ shows slightly larger equivalent widths than W Hβ . The 4000Angstrom break (D4000) is defined by Kauffmann et al. (2003) asthe ratio of fluxes in a bin redward of the break (4000 - 4100 ˚A) toa bin blueward of the break (3850 - 3950 ˚A). D4000 is monotonicand has an enhanced sensitivity over the range that the Balmerabsorption function turns over creating an ambiguity in age. Mea-surements of D4000 as a function of position for these objects willmore clearly define population age. clines once the nuclei merge. In the post merger phase,W Hβ would then continue to increase up to roughly 8 or9 ˚A before starting to decline. DISCUSSION
By resolving the recent star formation activity inULIRGs over scales of 1-10 kpc, we learned that star for-mation is being turned off in ULIRGs from the outsideinwards. Specifically, we measure an increase in W Hβ with increasing radius and argue that the only reason-able interpretation is an increasing fraction of Type Astars (relative to earlier types). We go one step furtherhere and interpret the reduced star formation rate as di-rect evidence for gas depletion. Above a threshold den-sity of approximately ∼
10 M ⊙ pc − , the star formationintensity empirically scales non-linearly with the gas sur- Fig. 7.—
Histograms of the nuclear W Hβ measurements are pre-sented for comparison between the different morphological classi-fications. The pre merger histogram does not include 4 of the 18nuclei that are present in the spectra due to the presence of emis-sion broad enough to make an absorption component can not bediscerned. The merger stage objects are predominantly between4 and 7 ˚A. The sample of old mergers is small, but the measurednuclei are greater than 5 ˚A. face density, Σ SF R ∝ µ . , where µ is the surface densityof gas (Martin & Kennicutt 2001; Kennicutt 1998). Ifstar formation were the primary mechanism removingthe gas, then the gas would be consumed more quicklyin the highest density regions, τ ∼ µ/SF R ∝ µ − . ,i.e. the nuclei. Star formation alone produces a gradi-ent oppositely sloped compared to the one we measured.Our result therefore provides direct evidence that an-other mechanism moves gas out of the outer disk. It isnatural to ask whether merger-induced inflows are thatmechanism.Simulations of gas-rich, galaxy – galaxy mergers pre-dict an increase in star formation during the first pas-sage and a second, stronger starburst at the time of ac-tual merger (Mihos & Hernquist 1994; Lotz et al. 2008;Hopkins et al. 2009). The torques imparted on the gasduring the merger cause the gas to flow inward, robbingthe outer regions of fuel to generate stars. The infall is awell known result of the separation of stars from gas inthe merger remnant; a consequence of the different colli-sional properties of stars and gas (Hopkins et al. 2009).This separation allows the stars to impart a torque onthe gas and cause infall. Models imply the correlationbetween the dynamical age of the merger and the age ofthe starburst population (Lotz et al. 2008) but have yetto predict how much of the star formation occurs in thecentral kpc. A key aspect of merger models should beto trace gas migration and the location of star formationactivity throughout the merger.Observations in Kewley et al. (2006) provide circum-stantial evidence relating gas inflow to the starburst andtherefore to the interaction. Galaxy mergers with greaterstarburst strength have lower nuclear metallicity. Thiseffect is expected to be due to the infall of pristine gasfrom the outer disk which dilutes the concentration ofmetals in the nucleus. We have measured the time of gasdepletion at radii from 1 to 10 kpc. The results constrainthe net infall velocity. Implication of Radial Gradients in Star FormationHistory for Gas Inflow Timescales
Table 2 lists the time elapsed since the star forma-tion rate or, equivalently, the gas surface density droppedabruptly. The time elapsed since peak star formation ac-tivity is typically 100 to 300 Myr at 5 to 10 kpc falling toless than 50 Myr within 5 kpc. These timescales indicatethe minimum time for gas inflow since the gas need onlyflow inward, not necessarily all the way to the center ofthe merger.In observations by Colina et al. (2005), thedynamical mass for objects included in ourstudy (IRAS17208+0014, IRAS20087-0308, IRAS23365+3604) are measured as similar to the Milky Waymass. The representation that we give for the orbitalvelocities may be reparameterized by substituting indistances and orbital velocities for other galaxies, aswell as changing the density values used in the freefallcalculation. Observations (Dasyra et al. 2006) haveindicated that ULIRGs are major mergers betweengalaxies with an average mass ratio of 1.5:1, indicatingthat the orbital velocities in each component are ofthe same order. This information allows us to choosethe Milky Way mass value as the fiducial mass scale;and we estimate maximum inflow speeds of order v inflow ∼
68 km s − ( R/ / ( t ∗ /
100 Myr).We compare this to two timescales in ULIRGs: (1)the free fall timescale at a given radius in an isother-mal sphere and (2) the orbital period. Both of thesetimescales increase linearly with radius. To determinethe freefall timescale we calculate the average densitywithin particular radii using a density profile of the form ρ ( r/r ) − with ρ = 1 . × M ⊙ kpc − the localhalo mass density (Gates et al. 1995). The average den-sity of the Milky Way interior to a radius of 8 kpc is ρ MW ≈ × M ⊙ kpc − giving the free fall timescale τ ff ( r ) ≈
40 Myr p ρ MW /ρ ( r ). The orbital period is τ orb ( R ) ≈
220 Myr ( R/ / ( v/
220 km s − ). In Figure8, we show the truncation timescale as a function of dis-tance from the center of the merger. The scatter in val-ues does not distinguish between freefall and orbital timescales. They do rule out the much longer time scales as-sociated with diffusion and lend empirical support to thetimescales assumed in simulations (Hopkins et al. 2008,2009). Picture Emerging from Comparison of Stellar andDynamical Ages
Gas migration toward the center slows down star for-mation in the outer radii. The paucity of massive stars al-lows W Hβ to grow. Regions closer to the nucleus are fedinfalling gas from exterior regions, prolonging star for-mation and allowing W Hβ to remain low longer. As themerger advances, the cessation of star formation trails in-ward. All 25 ULIRGs have star formation in their center,indicating the gas surface density there remains abovethe threshold density. As discussed in Section 3.4, the Fig. 8.—
The grey points are the measured values for age andposition in the truncated star formation history model. The blackpoints are the median values of age within bins that contain equalnumbers of points, where the bin size is represented by the errorbar. The median points show that the infall timescale is on thesame order as the orbital period and the free fall time scale. Thescatter at large radii may indicate differences in star formationhistory or inflow time scale. higher nuclear W Hβ in the merger and post-merger ob-jects may indicate the central star formation rate is de-clining by this stage. This result provides observationalevidence that the central gas density starts declining oncethe nuclei have merged. CONCLUSION
The physical processes regulating the rate of gas infallduring mergers is important because it influences the ageand metallicity gradients in the merger remnant as wellas the strength of supernova and AGN feedback. By ex-amining the positional dependence of stellar spectral in-dices in ESI long slit spectra , we determined the recentstar formation history across galaxy – galaxy mergers.Strong H β absorption indicates a diminished star forma-tion rate over the past few hundred Myr, which shouldbe accompanied by an increase in D4000, while hydrogenBalmer emission indicates the presence of massive starsfrom more recent star formation.We find the measured W Hβ increases with radius inthe sample of spatially resolved objects. We attribute thelarge W Hβ to a rapid decrease in star formation activityover 100 Myr ago at radii greater than 5 kpc. At radiiof a few kpc, the activity appears to have decreased just50 to 100 Myr ago based on the slightly lower, but stillprominent, W Hβ . Our measurements in the central kpcare consistent with star formation continuing unabated.The nature of the gradient implies that gas was removedfrom larger radii first allowing the stellar populations toage.We propose that these age gradients trace the inwardflow of gas from large radii, adding to the central gas sup-ply and prolonging the central starburst. Our data pri-marily address the pre-merger and early merger phases.At later stages, the surface brightness of a ULIRG istypically too low to make the equivalent measurement.We do, however, find preliminary evidence that thesespectral indices in the nuclei of late mergers indicate de-creasing star formation. We interpret these results asdirect evidence for strong gas inflow in the pre-mergerand early merger phases with a possible suppression bythe late merger phase.Assuming the time since star formation diminished re-flects the gas inflow timescale, our measurements indi-rectly constrain the gas infall rate. The time scales for in-fall to occur are within the range of the orbital timescaleand the freefall timescale of an isothermal sphere. The authors thank Omer Blaes, Dawn Erb, Philip Hop-kins, and Vivienne Wild for illuminating suggestions andcomments. This work was supported by the NationalScience Foundation under contract 080816. KS acknowl-edges additional support from the Department of Edu-cation through the Graduate Assistance in Areas of Na-tional Need program. The authors wish to recognize andacknowledge the very significant cultural role and rev-erence that the summit of Mauna Kea has always hadwithin the indigenous Hawaiian community. We are mostfortunate to have the opportunity to conduct observa-tions from this mountain. Facilities:
Keck
REFERENCESBruzual, G., & Charlot, S. 2003, MNRAS, 344, 1000Calzetti, D., Armus, L., Bohlin, R. C., Kinney, A. L., Koornneef,J., & Storchi-Bergmann, T. 2000, ApJ, 533, 682Colina, L., Arribas, S., & Monreal-Ibero, A. 2005, ApJ, 621, 725Dasyra, K. M., et al. 2006, ApJ, 638, 745Gates, E. I., Gyuk, G., & Turner, M. S. 1995, ApJ, 449, L123+Genzel, R., Tacconi, L. J., Rigopoulou, D., Lutz, D., & Tecza, M.2001, ApJ, 563, 527Hernquist, L., Spergel, D. N., & Heyl, J. S. 1993, ApJ, 416, 415Hopkins, P. F., Cox, T. J., Younger, J. D., & Hernquist, L. 2009,ApJ, 691, 1168Hopkins, P. F., Hernquist, L., Cox, T. J., Dutta, S. N., &Rothberg, B. 2008, ApJ, 679, 156Kauffmann, G., et al. 2003, MNRAS, 346, 1055Kennicutt, Jr., R. C. 1998, ApJ, 498, 541Kewley, L. J., Geller, M. J., & Barton, E. J. 2006, AJ, 131, 2004Le Floc’h, E., et al. 2005, ApJ, 632, 169Lotz, J. M., Jonsson, P., Cox, T. J., & Primack, J. R. 2008,MNRAS, 391, 1137 Martin, C. L. 2005, ApJ, 621, 227—. 2006, ApJ, 647, 222Martin, C. L., & Kennicutt, Jr., R. C. 2001, ApJ, 555, 301Mihos, J. C., & Hernquist, L. 1994, ApJ, 427, 112Murphy, Jr., T. W., Armus, L., Matthews, K., Soifer, B. T.,Mazzarella, J. M., Shupe, D. L., Strauss, M. A., & Neugebauer,G. 1996, AJ, 111, 1025Murphy, Jr., T. W., Soifer, B. T., Matthews, K., & Armus, L.2001, ApJ, 559, 201Poggianti, B. M., & Wu, H. 2000, ApJ, 529, 157Robertson, B., Cox, T. J., Hernquist, L., Franx, M., Hopkins,P. F., Martini, P., & Springel, V. 2006, ApJ, 641, 21Sanders, D. B., Scoville, N. Z., Young, J. S., Soifer, B. T.,Schloerb, F. P., Rice, W. L., & Danielson, G. E. 1986, ApJ,305, L45Veilleux, S., Kim, D.-C., & Sanders, D. B. 2002, ApJS, 143, 315Veilleux, S., et al. 2009, ApJS, 182, 628 TABLE 2Position Dependent Measurements
IRAS name r − r nuc W Hβ CSF TSFH DSFH SFR(1) (2) (3) (4) (5) (6)15245+1019 -7.0 ± ± +71 . − . +68 . − . ± a -5.0 ± ± +21 . − . +23 . − . ± a -1.9 ± ± +13 . − . +1 . − . +1 . − . ± ± ± +23 . − . +1 . − . +1 . − . ± ± ± +404 . − . +8 . − . +8 . − . ± ± ± +8 . − . +8 . − . ± ± ± +104 . − . +105 . − . ± a ± ± +34 . − . +37 . − . ± ± ± +35 . − . +3 . − . +1 . − . ± ± ± +31 . − . +2 . − . +1 . − . ± ± ± +33 . − . +3 . − . +2 . − . ± ± ± +63 . − . +3 . − . +5 . − . ± ± ± +16 . − . +16 . − . ± ± ± +19 . − . +19 . − . ± ± ± +65 . − . +60 . − . -2.4 ± ± +46 . − . +44 . − . ± a -0.4 ± ± +49 . − . +46 . − . ± ± +37 . − . +36 . − . ± ± ± +69 . − . +70 . − . ± ± ± +36 . − . +40 . − . ± a -3.1 ± ± +13 . − . +12 . − . ± ± ± +11 . − . +12 . − . ± ± ± +94 . − . +5 . − . +7 . − . ± ± ± +30 . − . +2 . − . +1 . − . ± ± ± +12 . − . +12 . − . ± ± ± ± ± ± b -2.5 ± ± ± a -0.4 ± ± +17 . − . +15 . − . ± ± +26 . − . +29 . − . ± ± ± +145 . − . +135 . − . ± a ± ± +29 . − . +26 . − . -6.8 ± ± +27 . − . +30 . − . -5.3 ± ± +42 . − . +4 . − . +3 . − . ± ± ± +92 . − . +4 . − . +7 . − . ± ± ± +23 . − . +2 . − . +1 . − . ± ± ± +9 . − . +3 . − . +0 . − . ± ± ± +8 . − . c c ± ± ± +64 . − . +2 . − . +5 . − . ± ± ± +9 . − . +8 . − . ± ± ± +57 . − . +3 . − . +4 . − . ± a ± ± +20 . − . +19 . − . ± ± +95 . − . +90 . − . ± ± ± +21 . − . +20 . − . ± ± ± +11 . − . +12 . − . ± ± ± +21 . − . +2 . − . +2 . − . ± ± ± +8 . − . +7 . − . ± ± ± +17 . − . +15 . − . ± ± ± +31 . − . +29 . − . ± ± ± ± a ± ± ± a ± ± +131 . − . +121 . − . ± TABLE 2 — Continued
IRAS name r − r nuc W Hβ CSF TSFH DSFH SFR(1) (2) (3) (4) (5) (6)-1.7 ± e N/A N/A N/A 0.60 ± ± ± +8 . − . +8 . − . ± ± ± +25 . − . +30 . − . ± ± ± +35 . − . +32 . − . ± ± ± +21 . − . +19 . − . ± ± ± +38 . − . +6 . − . +26 . − . ± ± e N/A N/A N/A 6.73 ± ± ± +16 . − . +2 . − . +1 . − . ± ± ± +27 . − . +30 . − . ± ± ± +14 . − . +19 . − . -6.3 ± ± +11 . − . +10 . − . -4.9 ± ± +529 . − . +24 . − . +25 . − . -3.4 ± ± +72 . − . +69 . − . ± a -2.3 ± ± +16 . − . +17 . − . ± ± ± +24 . − . +2 . − . +2 . − . ± ± ± c ± ± ± +123 . − . +5 . − . +6 . − . ± ± ± +28 . − . +26 . − . ± Note . — Col.(1): Projected distance from nucleus to center of measured region in kiloparsecs. The error corresponds to thespatial width of the aperture. Col.(2): Absorption equivalent width for the aperture measured in ˚A . Col.(3): Stellar populationage in continuing star formation history measured in Myr. In the apertures where the W Hβ is too large for the given stellarpopulation, the section is marked with ”N/A”. Col.(4): Stellar population age in Truncated star formation history measuredin Myr. Col.(5): Stellar population age in δ function star formation history measured in Myr. Col.(6): Star formation rateper aperture measured in M ⊙ yr − kpc − . a In these apertures, H β emission was not detected, indicating heavy extinction. These values represent a lower limit to theSFR based on assuming that the H β emission line has approximately the same peak as the noise, making it unmeasurable. b H α emission was not detected in these apertures. c Upper limit of age is presented for very low W Hβ . d These apertures have forbidden/Balmer line ratios that indicate a considerable contribution by AGN. Star formation rate isexcluded due to possible confusion. ee