A 30 kpc CHAIN OF "BEADS ON A STRING" STAR FORMATION BETWEEN TWO MERGING EARLY TYPE GALAXIES IN THE CORE OF A STRONG-LENSING GALAXY CLUSTER

We present the new HST data as an RGB composite in Figure 1. The images used in this composite are shown individually in Figure 2, which uses contours and labels to highlight various features of interest. As is evident from these figures, the cluster center harbors two elliptical galaxies whose nuclei are separated ∼7 kpc (15) in projection from one another. The combined projected stellar envelope of both galaxies extends ∼100 kpc (∼22'') across the major axis of the lowest surface brightness isophote, though their light is very centrally concentrated with an r-band Petrosian radius of only ∼30 kpc (∼7''). Both new and archival optical spectroscopy confirms that this is no mere projection effect—the galaxies indeed share a common redshift, such that their stellar halos must be deeply embedded in one another over at least ∼20 kpc scales.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

This is evident in archival Sloan Digital Sky Survey (SDSS; York et al. 2000) optical spectroscopy, which we present in the leftmost panels of Figure 3. While the SDSS spectrum is from a fiber that covers the nuclei of both galaxies (see top left panel of Figure 3), it features a singular line system including Mg, Na D, and Ca ii h+k absorption tracing the old stellar component, all at a common redshift of z = 0.3350 ± 0.0002 (consistent with the SDSS photometric redshift). This serves as strong evidence that the galaxies are indeed merging, as they would have been bright enough to manifest as two distinct-redshift line systems if indeed they were two otherwise unrelated objects seen in projection.

Zoom In Zoom Out Reset image size

The merger hypothesis is confirmed by follow-up slit spectroscopy (see Table 1) obtained with the ALFOSC spectrograph on the NOT, the results of which are presented in the rightmost panels of Figure 3. As is evident from the [O ii] and Ca ii lines extracted from the southeastern (SE) and northwestern (NW) galaxy nuclei (see bottom right panel), the maximum spectroscopically permissible velocity offset between the galaxies is ∼280–300 km s⁻¹, far too small for these to be unrelated galaxies in the Hubble flow. Any three-dimensional (3D) configuration consistent with spectral constraints results in their stellar envelopes being deeply embedded in one another over ∼20 kpc scales. We therefore conclude that the two ellipticals (which are of roughly equal stellar mass, sharing similar surface brightnesses) are undergoing a major merger.

As noted above, the cluster mass derived from strong+weak-lensing analysis is $5.13^{+1.33}_{-1.19} \times 10^{14}\ h^{-1}$ M_☉ (Oguri et al. 2012). The implied projected one-dimensional velocity dispersion from such a mass would be of order ∼750 km s⁻¹, consistent with the measured velocity dispersion from 11 member galaxies of $998^{+120}_{-194}$ km s⁻¹ (Bayliss et al. 2011). The expectation value for pairwise velocities along the line of sight would therefore be ≳ 1000 km s⁻¹, rather more than the ∼280 km s⁻¹ permitted by our follow-up spectroscopy. The NOT data therefore suggest that the trajectories of the merger participants lie mostly in the plane of the sky. While the galaxy stellar isophotes are mostly smooth, unsharp masks and a F606W/F160W color map reveal residual surface brightness fluctuations between and around both nuclei, suggestive of some dynamical disturbance in the old stellar component as a result of the merger.

Far more dramatic than the merging galaxies is the F390W image of rest-frame NUV continuum emission mostly arising from young (≲ 300 Myr), massive (≳ 5M_☉) stars. This is seen in blue on the rightmost panel of Figure 1, and shown in the bottom left panel of Figure 2. Not counting the nuclei of the two elliptical galaxies, the emission consists of at least 19 stellar superclusters (or perhaps tidal dwarf galaxies) ∼0.5–1 kpc in diameter. The 19 stellar clusters are enhanced in NUV surface brightness by factors of 2-10 over the filaments in which they are embedded. Each of these superclusters are numerically labeled in the bottom right panel of Figure 2, which shows an unsharp mask revealing residual [O ii] and blue excess emission in the F606W image.

The young stellar superclusters are roughly equally spaced along 3-4 narrow (1–2 kpc wide), straight filaments roughly ∼10 kpc in projected length, resembling "beads on a string" (discussed in Section 4). As we show in Figure 4, a histogram of projected separations between neighboring stellar clusters reveals a strong peak around ∼1–2 kpc. All 19 clusters are at least marginally resolved, and while most are roughly circular, a few (especially cluster $\#$2) show evidence for ∼500 pc scale asymmetries reminiscent of tidal arms.

Zoom In Zoom Out Reset image size

Two filamentary strings of clusters (containing clusters numbered 6-9 and 15-19 on the bottom right panel of Figure 2) are aligned along a position angle (P.A.) that is roughly parallel to the axis joining the nuclei of the two merging galaxies (P.A. ∼ − 45°, measured north through east). The other two strings (containing clusters 1-4 and 11-14, excluding #13) lie along a P.A. ≈45° that is roughly perpendicular to this axis. The clusters are entirely contained within the combined projected old stellar envelope of both galaxies, within a circle of diameter ∼27 kpc (56). This can be seen in the top left panel of Figure 2, in which the NUV continuum emission is plotted in green contours over the F160W image, showing the old stellar population.

We consider three scenarios for the physical nature and origin of the clumpy NUV continuum emission.

• 1.  
  It is a gravitationally lensed image (or a set of images) from a higher redshift background galaxy (or multiple galaxies).
• 2.  
  It is a chance superposition along the line of sight, arising instead from a projection effect due to unrelated foreground sources.
• 3.  
  It arises from ongoing or very recent star formation taking place within the stellar envelope of the interacting elliptical galaxies.

Scenarios (1) and (2) are entirely ruled out by spectroscopic evidence. The same SDSS and NOT spectroscopy used to confirm that the two elliptical galaxies are indeed merging also confirms that Hα and [O ii] emission from star formation is mapped to the same redshift of the galaxies at z = 0.3350 ± 0.0002 (see Figure 3). The stellar superclusters therefore must inhabit the combined stellar envelope of the merger participants. The integrated flux density of the NUV continuum in the F390W bandpass is (5.161 ± 0.512) × 10⁻¹⁷ erg s⁻¹ cm⁻² Å⁻¹, suggesting that line emission from the young stellar superclusters is easily bright enough to be detected in the spectroscopy. If at appreciably different redshift than the two galaxies with which the emission is cospatial, the redshift of the Balmer lines should be offset from that of the Mg and Na D absorption as well as the calcium break. Moreover, strong-lensing analysis (Sharon et al. 2014) indicates that possible lensed images of background sources do not contribute significantly to the flux seen in the NUV emission. For the remainder of this Letter, we therefore adopt the most evidence-supported scenario (3), and conclude that (1) the two elliptical galaxies seen in close separation are indeed undergoing a major merger and (2) the clumpy NUV continuum emission arises from ongoing or recent star formation taking place within their combined stellar envelope.

The integrated SDSS Hα line flux from the 3'' diameter fiber, which covers only the innermost region of the star forming chain (see Figure 3, top right), is (6.335 ± 1.77) × 10⁻¹⁶ erg s⁻¹ cm⁻² with a FWHM of 152 ± 15 km s⁻¹. We estimate internal extinction via the Balmer decrement (Hα/Hβ flux ratio), following the procedure described in Tremblay et al. (2010) and adopting "case B" recombination and the R_V = 3.1 reddening-to-extinction law of Cardelli et al. (1989). After an additional correction for foreground Milky Way extinction using E(B − V) = 0.023, we estimate an extinction-corrected Hα luminosity of (3.65 ± 1.1) × 10⁴¹ erg s⁻¹ from the region covered by the SDSS fiber. As part of our NOT/ALFOSC observations, we used a 25 wide slit to observe the stellar superclusters in Hα, as detailed in Figure 3. This slit encompasses all of the NUV identified clumps, even after accounting for minimal slit losses given ground-based seeing. The derived Hα line flux from our ALFOSC data is 1.7 times greater than that reported by SDSS, approximately as expected given the more complete spatial coverage of the clumps by the wide slit observation. Including a 30% uncertainty to account for spectrograph CCD fringing and our [N ii] contamination correction, the implied total extinction-corrected Hα luminosity for the entire system is thus (6.21 ± 1.9) × 10⁴¹ erg s⁻¹.

By Equation (2) in Kennicutt (1998), this extinction-corrected Hα luminosity corresponds to a (very rough and assumption-heavy) star formation rate (SFR) of ∼5 ± 2 M_☉ yr⁻¹. Automated modeling of the SDSS spectrophotometric data by Maraston et al. (2006, 2009, 2013), which matches spectral energy distribution templates to extinction-corrected SDSS ugriz magnitudes, yields an SFR of 9.55 ± 3.4 M_☉ yr⁻¹ and a total stellar mass of (3.35 ± 0.04) × 10¹¹ M_☉ (for more details, see Maraston et al. 2006). Absent the follow-up data needed for a more robust analysis, the remainder of this Letter assumes that the SFR is roughly in the range of ∼5–10 M_☉ yr⁻¹. Each clump contributes approximately ∼2%–6% of the total NUV flux, which should roughly scale to its relative contribution to the total SFR. We therefore estimate that the clumps have an associated SFR in the range of ∼0.1–0.6 M_☉ yr⁻¹.

Inverting the Bigiel et al. (2008) calibration of the molecular Schmidt law (Kennicutt 1998) and assuming gas depletion times in the range of 1–2 Gyr, the total SFR translates to a rough associated molecular gas mass of $M_{\mathrm{H}_2} \approx 0.5\hbox{--}2 \times 10^{10}$M_☉. We obtained an IRAM 30 m telescope CO(1-0) observation in an attempt to measure the actual cold gas mass. No line was detected at a sensitivity of 0.9 mK T_mb per 80 km s⁻¹ channel, setting a 1σ upper limit of ∼1 × 10¹⁰ M_☉ of molecular gas, assuming a line width of 220 km s⁻¹ and a Galactic X_CO factor. A molecular gas mass in the range of ∼0.5–1 × 10¹⁰ M_☉would be ∼1%–3% that of the galaxies' stellar mass, consistent with that observed in large samples of early type galaxies (e.g., Young et al. 2011).

The observed star formation in SDSS 1531 is strongly reminiscent of the well-known "beads on a string" mode of star formation (e.g., Chandrasekhar & Fermi 1953; Toomre 1977; Elmegreen & Efremov 1996; Renaud et al. 2013) frequently observed in the arms of spiral galaxies (Elmegreen & Elmegreen 1983), resonance rings (Elmegreen 1994), and tidal arms that bridge interacting galaxies (Duc et al. 2000; Smith et al. 2010). Beads on a string star formation is a kpc-scale manifestation of the Jeans length, and the physics governing the system is analogous to the Plateau–Rayleigh instability causing (e.g.,) a continuous column of falling water to disrupt, explaining why rain falls in drops rather than in unbroken filaments from the sky (e.g., Quillen & Comparetta 2010). Despite being broadly categorized as "red and dead," it has been known for a number of years that star formation can be relatively abundant in both cluster and field ellipticals (Yi et al. 2005; O'Dea et al. 2008; Jeong et al. 2009; Davis et al. 2011; Tremblay et al. 2012). Nevertheless, the formation of "beads on a string" stellar superclusters in major, likely gas-rich mergers between giant ellipticals is rarely (if ever) observed, regardless of whether or not the star formation is driven by a merger or a cooling flow (e.g., G. R. Tremblay et al. 2014a, in preparation). The unusual system in SDSS 1531 is likely a serendipitous snapshot of what may be a very short-lived morphology, presenting a unique opportunity to study star formation and gas dynamical response in a tidal field governed by dynamical friction, shear, and gravitational torques associated with the two merging ellipticals. Forthcoming cluster mass reconstruction from strong lensing (Sharon et al. 2014) will provide a well-constrained canvas against which these kinematics may be studied.

These quantitative analyses, however, await the necessary multiwavelength follow-up data. Recently obtained Gemini GMOS-N integral field unit observations (G. R. Tremblay et al. 2014b, in preparation) will disentangle the kinematics, dynamical timescales, and 3D geometry of the system, as well as local internal extinction corrections required for a better SFR estimate. ALMA is the only facility with the sensitivity and resolution needed to independently constrain the clump-by-clump SFR and molecular gas masses.

Finally, Chandra X-ray observations would determine whether or not the observed star formation might originate from gas that has condensed from the ambient hot atmosphere via a cooling flow, or a shock in the X-ray gas driven by colliding hot halos from the galaxy merger. The location of the stellar superclusters leaves this origin question open, as the star formation is not occurring in a region between the merging galaxies, where one would most obviously expect it if the star formation is shock triggered. However, viscous drag effects preferentially influencing the gas may be responsible for the apparent dislocation of the stellar (and presumably dark matter mass) components from the gas, as has been observed in other galaxy cluster mergers (e.g., Clowe et al. 2006).