STELLAR POPULATION GRADIENTS IN ULTRALUMINOUS INFRARED GALAXIES: IMPLICATIONS FOR GAS INFLOW TIMESCALES

The Keck II echellete spectroscopy of 2 Jy ULIRGs, described previously in Martin (2005, 2006), is well suited to our study of H Balmer emission/absorption. These data resolve scales of 08, about 1.5 kpc, spatially and ∼70 km s⁻¹ spectrally. The 20'' slit is long enough to sufficiently sample the sky at both ends of the slit, ensuring accurate sky subtraction and continuum intensity. The slit position angle includes a second nucleus when present (Table 1). We inspected 43 ULIRG spectra and selected 25 for analysis of any stellar Hβ absorption. This subsample presents Hβ emission such that the full width of the emission line in their nuclear spectrum, defined by the wavelengths where the trough changes from emission to absorption, is no more than 1800 km s⁻¹.

Table 1. Object Morphology

IRAS Name	IR	Opt	Veil.	log($\frac{L_{\rm IR}}{L_\odot }$)	z	P.A.
(1)	(2)	(3)	(4)	(5)	(6)	(7)
17028+5817	D	D	IIIaa	12.38	0.1061	94.5
23327+2913	D	D	IIIaa	12.03	0.1075	−4.3
05246+0103	D	D	IIIb	12.05	0.0971	109.5
00153+5454	D	D	IIIb	12.10	0.1116	−22.0
16487+5447	D	D	IIIba	12.12	0.1038	66.2
16474+3430	D	D	IIIba	12.12	0.1115	161.8
15245+1019	D	St	IIIb	11.96	0.0755	127.8
20046−0623	S	St	IIIb	12.02	0.0843	72.0
20087−0308	S	St	IVb	12.39	0.1057	85.5
17574+0629	S	S	IVb	12.10	0.1096	51.2
17208−0014	S	St	IVb	12.38	0.0428	166.7
18368+3549	S	S	IVb	12.19	0.1162	−31.0
23365+3604	S	St	IVb	12.13	0.0645	−19.5
09111−1007	S	W	IVb	11.98	0.0542	73.5
01298−0744	S	S	IVba	12.29	0.1362	−89.3
11506+1331	S	S	IVba	12.27	0.1273	80.3
10378+1109	S	S	IVba	12.23	0.1362	11.3
11095−0238	S	St	IVba	12.20	0.1065	10.0
18443+7433	S	St	IVb	12.23	0.1347	29.6
16090−0139	S	S	IVaa	12.48	0.1336	107.6
00262+4251	S	S	V	12.08	0.0972	14.1
08030+5243	S	S	V	11.97	0.0835	0.0
00188−0856	D	D	Va	12.33	0.1285	−3.3
10494+4424	S	S	V	12.15	0.0923	25.0
10565+2448	M	M	M	11.98	0.0431	109.0

Notes. Column 1: name. Column 2: infrared K-band morphology. "S," "D," and "M" denote a single, double, and multiple nucleus morphology, respectively (Murphy et al. 1996). Column 3: optical R-band morphology. Similar to R-band morphology with an additional "t" to signify an extended tidal region, and "W" to denote a widely separated pair (Murphy et al. 1996). Column 4: merger classification based on Veilleux et al. (2002). This classification separates wide binary pre-mergers (IIIa), close binary pre-mergers (IIIb), diffuse merger (IVa), compact merger (IVb), and old mergers (V). Column 5: IR luminosity from Murphy et al. (1996). Column 6: redshift. Column 7: position angle of slit. ^aClassifications directly from Veilleux et al. (2002).

Download table as:  ASCIITypeset image

The resulting subsample includes a range of stages in the merger progression. Veilleux et al. (2002) classified the merger stage for 11 of the 25 objects. Using K-band images from Murphy et al. (1996), and their identification as multiple nuclei systems, we apply the Veilleux scheme to the remainder finding eight pre-mergers (i.e., two nuclei), a multiple merger (IRAS 10565+2448), twelve mergers, and four old mergers. The ULIRGs in the merger stage tend to present higher surface brightness, so our analysis of spatial gradients in Balmer absorption is heavily weighted toward this merger phase.

We examined the positional dependence of the Hβ line profile using a sliding aperture for spectral extraction. We matched the aperture width to a seeing element (∼08) and extracted a spectrum at each line of the spectrogram. At some spatial positions, as illustrated in Figure 1, the spectra present a two-component Hβ line profile. A broad absorption trough underlies the narrow emission line. We fit the absorption and emission components simultaneously using nonlinear least-squares fitting with MPFIT in IDL. These spectra exhibit continuous change from position to position due to atmospheric smearing, so parameters found from fitting one aperture position were supplied as the guess for the fit of the next position. The small, gray points in Figure 2 show the result. Over scales of a few kiloparsecs, this analysis often reveals an increase in Hβ absorption equivalent width, W_Hβ, with distance from the peak continuum emission in the R band.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

View extended figure (10 panels)

To eliminate effects due to correlated errors, we extracted spatially separated spectra based on the shape of these radial Hβ absorption and continuum surface brightness trends. These spectra are shown in Figure 3, where the larger black points in Figure 2 show the Hβ absorption equivalent width obtained by fitting these spectra. We measured Hα surface brightness in these same apertures. The Hδ lines could only be fitted in more than three apertures for one object.

Zoom In Zoom Out Reset image size

View extended figure (10 panels)

To correct the Hα emission for underlying stellar Hα absorption, we measured the Hβ absorption equivalent width and assumed the Hα absorption equivalent width was two-thirds as large. Inspection of synthetic spectra computed for stellar population models (Bruzual & Charlot 2003) verified that this scaling works over a broad range of star formation models, including those with truncated star formation histories. The maximum value of the correction, 3 Å, is small compared to the emission equivalent width. The flux in Hα emission is also corrected for extinction. We determined the color excess, E(B − V), per aperture by measuring the Balmer decrement and using a 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 attenuation by Galactic dust on a galaxy-by-galaxy basis in Section 3.2. We estimate areal star formation rate using the conversion SFR(M_☉ yr^-1 kpc⁻²) = [L(Hα)/A(kpc²)]/1.26 × 10⁴¹ ergs^-1 (Kennicutt 1998).

Using GALAXEV (Bruzual & Charlot 2003), we synthesized model spectra of the stellar population for three different star formation histories: (1) δ function, (2) continuous star formation history that is extinguished after 40 Myr, and (3) continuing star formation (CSF). We assumed solar metallicities. We measured W_Hβ in these spectra using the same techniques applied to the observations. We illustrate the evolution of the W_Hβ as the stellar population ages in Figure 6. Around 850 Myr, continuous star formation reaches a maximum of 6 Å in equivalent width as the population reaches an equilibrium of births and deaths of A stars. In the other models, as O and B type stars die off, the stellar population dominating the stellar continuum emission gradually shifts to later stellar types. A single A5V star has W_Hβ near 20 Å, but the spread in stellar types does not allow the composite models to reach this limit. Once star formation is stopped, the W_Hβ increases steadily toward a maximum of 9 Å at 400 Myr. After 400 Myr, the W_Hβ falls as the number of A stars declines, reaching 6 Å after about 850 Myr. An equivalent width greater than 6 Å clearly requires a very rapid decrease in the star formation rate in the past Gyr.

Zoom In Zoom Out Reset image size

The large values of W_Hβ found at radii R>5 kpc in IRAS 23365+3604 are inconsistent with the continuous star formation rate model. Applying the relation between W_Hβ and time since truncation, i.e., Figure 6 solid curve, we conclude that star formation ceased at least 100 Myr ago at R>5 kpc. The Hδ equivalent width varies much like Hβ in Figure 6 owing to its similar origin. The W_Hβ gradient continues toward the center of IRAS 23365+3604, and we interpret this as evidence that the time since truncation decreases toward the center. In other words, star formation is apparently being shutdown from the outside inward.

Another popular diagnostic of stellar age is the spectral break known as D4000. Our spectral bandpass does not completely cover this feature, but we can use the stellar population models along with our observed trend in Balmer equivalent width to predict the radial variation in D4000. The D4000 index grows as the stellar population ages. Measurements of D4000 break the age degeneracy to either side of the maximum in W_Hβ at 400 Myr. Hence, we expect D4000 would show a steady increase with increasing radius in ULIRGs.

Our simple stellar populations represent the most extreme limiting cases. A declining star formation rate produces a rising W_Hβ similar to our models with complete cessation. The stronger the rate of change in the star formation rate, the closer the W_Hβ comes to that of the truncated model. We can distinguish these two star formation histories by the absence/presence of Balmer emission in our ULIRG spectra. The presence of nebular emission indicates the level of ongoing star formation. Complete truncation of star formation activity reduces the ionization rate, and Balmer emission ceases. We examine the radial dependence of the implied star formation rate galaxy by galaxy.

In Section 3, we identified 10 galaxies having extracted spectra in a least four apertures with signal to noise ratio (S/N) ⩾ 5. Pre-merger and merger systems dominate this spatially resolved subsample. We discuss the recent star formation history in each individually here. For reference, Table 2 lists the measured star formation rate and stellar population ages for different apertures in each object. The star formation rates near the center should be treated cautiously due to two important systematic errors. First, an active galactic nucleus (AGN) may contribute to the ionizing photon luminosity, thereby lowering the inferred star formation rate. Second, the reddening measured by the Balmer decrement in the central aperture reflects only the least obscured regions; the extinction may be too high in some regions for Balmer photons to escape causing us to underestimate the central star formation rate. These factors have little impact on our analysis of spectra extracted outside of the central kiloparsec and our discussion of radial gradients.

Our results suggest the following star formation histories.

IRAS 15245+1019. This ULIRG has two K-band nuclei separated by ≈4 kpc. These two nuclei host star formation of 4.83 ± 0.05 and 6.8 ± 2.1 M_☉ yr^-1 kpc⁻² for the brighter southeast nucleus and dimmer northwest nucleus, respectively. The relative dimness in the visible band of the southeast nucleus is due to high levels of extinction by dust. The brighter nucleus in the interaction has W_Hβ 5.0 ± 0.1 Å which indicates that star formation has been going on for 0.22 ± 0.02 Gyr. At 7 kpc, however, there is little to no star formation occurring currently. W_Hβ indicates that 0.2 ± 0.1 Gyr has passed since truncation.

IRAS 20046−0623. From detailed analyses of galactic relative velocities and tidal tail characteristics, it is determined that IRAS 20046−0623 has gone through pericenter in the last few times 10⁷ years (Murphy et al. 2001). The separated nuclei are masked within the extended central bulge that appears in both K-band and R-band imaging. The morphology of the extended tails is representative of two nearly orthogonal disks in the process of merging. Peaks in star formation (3.95 ± 0.05 and 8.3 ± 0.7 M_☉ yr⁻¹ kpc⁻²) occur in the nuclei indicated in Murphy et al. (2001). As shown in Figure 2(c), these nuclear regions have W_Hβ that indicate star formation has continued for 0.18 ± 0.03 Gyr and 0.36 ± 0.07 Gyr. In the outer regions ∼2 kpc beyond both nuclei, star formation is weaker, and their W_Hβ is beyond that attainable by CSF. The truncation timescale indicated by these measurements for these bins is from 80 to 100 Myr. Galactic dust contributes ∼0.16 mag of extinction at the redshifted wavelength of Hα.

IRAS 20087−0308. The morphology of the K-band image shows a single nucleus, implying that the two galaxies have fully merged. Similarly, we see one centrally located position on the slit with CSF. The S/N for this object makes W_Hβ difficult to measure, but the nuclear measurement in this object does appear slightly lower than the extended regions. This difference is not significant enough to measure a gradient in the population age, so we estimate an overall age of 100 Myr since star formation stopped across the disk. Galactic dust contributes ∼0.31 mag of extinction at the redshifted wavelength of Hα.

IRAS 17208−0014. This galaxy has a single K-band nucleus with many diffuse disturbed regions around it in the R band. The aperture with peak star formation (0.64 ± 0.03 M_☉ yr^-1 kpc⁻²) is adjacent to the region with the smallest W_Hβ (4.7 ± 0.2 Å) which corresponds to 170 ± 30 Myr of continuous star formation (Figure 2(e)). Next to the K-band centroid at the max of the R-band continuum W_Hβ = 6.6 ± 0.3 Å and the star formation is diminished to 0.016 ± 0.003 M_☉ yr^-1 kpc⁻². This aperture shows an underlying population with a truncation timescale of 57 ± 10 Myr. The furthest aperture with a center 4.25 kpc from the nucleus has a slightly larger W_Hβ and very little star formation, making the time since truncation of this population slightly larger at 100 ± 35 Myr. Galactic dust contributes ∼0.67 mag of extinction at the redshifted wavelength of Hα.

IRAS 18368+3549. The K-band morphology is a single nucleus with the R band revealing a faint tidal region encircling the nucleus. From this morphology, we expect that the nuclei have completely merged and tidal distortions have had time to circularize. The nucleus has a W_Hβ of 6.8 ± 0.4 Å putting it beyond the maximum possible in CSF. This measurement represents a timescale of 77 ± 17 Myr in the truncated constant star formation history, and 95 ± 17 Myr in the δ-fn model. W_Hβ in the extended region at 10 kpc from the nucleus is very large putting it beyond the model describable with these star formation histories.

IRAS 23365+3604. The morphology of this galaxy is similar to that of IRAS 18368+3549, in that it has a bright nucleus with a tidal region that has had time to circularize. The nuclear regions have a peak in star formation rate (2.60 ± 0.02 M_☉ yr^-1 kpc⁻²), and low W_Hβ (3.3 ± 0.1 Å). We measure 45 ± 5 Myr since continuous star formation began. Apertures exterior to this have similar times since a truncation of star formation (37 ± 3 Myr in a truncated constant star formation history, 52 ± 4 Myr in a δ-fn model). At a distance of 5.3 kpc from the nucleus, the slit intersects an extended tidal region, the center of which shows a dip in equivalent width (3.9 ± 0.4 Å) in an otherwise increasing W_Hβ with radius. In the furthest aperture no Hα emission is measured, and the measured W_Hβ of absorption indicates that truncation occurred 140 ±  25 Myr ago. Galactic dust contributes ∼0.19 mag of extinction at the redshifted wavelength of Hα.

IRAS 09111−1007. This galaxy has undisturbed tidal regions characteristic of post-merger morphology (Figure 2(a)). The central kiloparsec has W_Hβ = 4.5 ± 0.2 Å, which is indicative of younger stellar populations. From this point, W_Hβ increases toward the edge of the object, up to 9–10 Å at projected distances of 3–5 kpc. These large equivalent widths imply that 300 Myr has passed since the truncation of star formation.

IRAS 10378+1109. This galaxy has a compact nucleus, but it is measurable in several apertures. The fit of the absorption trough in the central aperture yields an unphysically narrow fit due to contamination by strong emission and is rejected from the rest of the analysis. This strong emission indicates strong star formation (6.73 ± 0.06 M_☉ yr^-1 kpc⁻²). The aperture 4.7 kpc has W_Hβ = 6.9 ± 0.4 Å indicating a truncation timescale of ∼90 Myr.

IRAS 11506+1331. Star formation is maximum (2.23 ± 0.06 M_☉ yr^-1 kpc⁻²) in the central most aperture (1.3 kpc). If a CSF history were considered, the absorption (W_Hβ = 6.2 ± 0.25 Å) would indicate that constant star formation has continued for 1 Gyr; however, the presence of a strong absorption line indicates that there was a previous burst of star formation that dominates the continuum over the stars being produced in the current epoch of star formation. In this case, the end of the previous burst of star formation would have ended 69 ± 9 Myr ago. Exterior to this, regions have diminished star formation rates (< 0.1 M_☉ yr^-1 kpc⁻²) and truncation timescales up to 150 ± 35 Myr at 7 kpc.

IRAS 10565+2448. This object is classified as a multiple merger by Murphy et al. (1996), with a third component separated by 20 kpc from the brightest knot in our measurement. These data sample two of the nuclei and measure an increase in W_Hβ with radius. In the stronger nucleus, W_Hβ drops to 1.8 ± 0.1 and has a large amount of star formation (7.28 ± 0.08 M_☉ yr^-1 kpc⁻²) indicating a very young population. W_Hβ increases to 8.3 ± 0.5 Å, corresponding to a truncation timescale of ∼180 Myr. A dip in the trend occurs off center of the weaker nucleus. The truncation timescale in this region is ∼50 Myr.

From analysis of the spatially resolved subsample, we draw the following conclusions. Objects such as IRAS 20087−0308, IRAS 18368+3459, IRAS 23365−3604, and IRAS 10565+2448 have regions at large radii that show no Hα, indicating that the star formation has been completely shut off. The others show examples of continuing, but decreased, star formation activity at larger radii. Often the star formation rate per unit area drops by at least a factor of 10 at the largest radii measured. Though the gradient varies among the galaxies, we see regularity in that the time since truncation increases outward in radius.

We considered other physical effects that could cause large W_Hβ values. Previous studies (Poggianti & Wu 2000) attributed large Balmer equivalent width to intermediate age stars having time to drift away from the dusty molecular clouds in which they were born. This effect can have a major effect in integrated spectra of entire objects. For dust to be the cause of larger equivalent width at larger radii, the population of hidden young stars must grow with distance from the nucleus. This conclusion contradicts measurements of the molecular gas distribution in ULIRGs. An active nucleus may contribute to the continuum emission in the central aperture. AGN typically contributes 15%–30% of the total bolometric luminosity from cool H ii and LINER ULIRGs Veilleux et al. (2009). Boosting the continuum, however, only decreases the apparent stellar W_Hβ, so an AGN cannot explain large W_Hβ values. Moreover, the gradients that we find on spatial scales larger than the seeing cannot be caused by an AGN.

Among the 25 spectra examined, the surface brightness of the merger remnant was not always high enough to measure the spatial dependance of the line profile. We can, however, compare W_Hβ among all the nuclei as illustrated in Figure 7. In the pre-merger subsample, we measured W_Hβ of both nuclei seen. Of these 18 nuclei, 4 of them showed broad Hβ emission that did not allow the measurement of an underlying absorption line. A typical nucleus in the pre-merger sample presents W_Hβ of 4 Å, although the full range in W_Hβ is large. In the merger sample, the median W_Hβ is higher. Our sample is too small to be certain that this difference is significant; but, if verified, would suggest the central star formation declines once the nuclei merge. In the post-merger phase, W_Hβ would then continue to increase up to roughly 8 or 9 Å before starting to decline.

Zoom In Zoom Out Reset image size

Table 2 lists the time elapsed since the star formation rate or, equivalently, the gas surface density dropped abruptly. The time elapsed since peak star formation activity is typically 100–300 Myr at 5–10 kpc falling to less than 50 Myr within 5 kpc. These timescales indicate the minimum time for gas inflow since the gas needs only flow inward, not necessarily all the way to the center of the merger.

In observations by Colina et al. (2005), the dynamical mass for objects included in our study (IRAS 17208+0014, IRAS 20087−0308, and IRAS 23365+3604) is measured as similar to the Milky Way mass. The representation that we give for the orbital velocities may be reparameterized by substituting in distances and orbital velocities for other galaxies, as well as changing the density values used in the free-fall calculation. Observations (Dasyra et al. 2006) have indicated that ULIRGs are major mergers between galaxies with an average mass ratio of 1.5:1, indicating that the orbital velocities in each component are of the same order. This information allows us to choose the Milky Way mass value as the fiducial mass scale; and we estimate maximum inflow speeds of order v_inflow ∼ 68 km s^-1(R/7 kpc)/(t_*/100 Myr).

We compare this to two timescales in ULIRGs: (1) the free-fall timescale at a given radius in an isothermal sphere and (2) the orbital period. Both of these timescales increase linearly with radius. To determine the free-fall timescale, we calculate the average density within particular radii using a density profile of the form ρ₀(r/r₀)⁻² with ρ₀ = 1.3 × 10⁷ M_☉ kpc⁻³ the local halo mass density (Gates et al. 1995). The average density of the Milky Way interior to a radius of 8 kpc is ρ_MW ≈ 4 × 10⁷ M_☉ kpc⁻³ giving the free-fall timescale $\tau _{\rm ff}(r) \approx 40\,{\rm Myr} \sqrt{\rho _{\rm MW}/\rho (r)}$. The orbital period is τ_orb(R) ≈ 220 Myr(R/8 kpc)/(v/220 km s^-1). In Figure 8, we show the truncation timescale as a function of distance from the center of the merger. The scatter in values does not distinguish between free-fall and orbital timescales. They do rule out the much longer timescales associated with diffusion and lend empirical support to the timescales assumed in simulations (Hopkins et al. 2008, 2009).

Zoom In Zoom Out Reset image size

Gas migration toward the center slows down star formation in the outer radii. The paucity of massive stars allows W_Hβ to grow. Regions closer to the nucleus are fed infalling gas from exterior regions, prolonging star formation and allowing W_Hβ to remain low longer. As the merger advances, the cessation of star formation trails inward. All 25 ULIRGs have star formation in their center, indicating the gas surface density there remains above the threshold density. As discussed in Section 3.4, the higher nuclear W_Hβ in the merger and post-merger objects may indicate the central star formation rate is declining by this stage. This result provides observational evidence that the central gas density starts declining once the nuclei have merged.