VELOCITY DISPERSIONS AND STELLAR POPULATIONS OF THE MOST COMPACT AND MASSIVE EARLY-TYPE GALAXIES AT REDSHIFT ∼1

The evolution of the structural properties of massive early-type galaxies (ETGs) is currently a topic of debate with many open questions. In recent years, studies have shown that galaxies with  M_⋆ ⩾ 10¹¹ M_☉ are smaller by a factor of four at z ⩾ 1.5 (Daddi et al. 2005; Trujillo et al. 2006; Longhetti et al. 2007; Buitrago et al. 2008; Damjanov et al. 2009) and a factor of two at z ∼ 1 (Trujillo et al. 2007, 2011) than the nearby population of the same mass. For the most compact and massive galaxies in particular, Trujillo et al. (2009) found that only ∼0.03% of SDSS galaxies with  M_⋆ ⩾ 8 × 10¹⁰ M_☉ at z ⩽ 0.2 are ETGs with  R_e ⩽ 1.5 kpc. This contrasts with the much higher fraction ∼15% of such galaxies found at z∼1 by Trujillo et al. (2007, hereafter T07). Most interestingly, the luminosity-weighted ages of the local compact ETGs from Trujillo et al. (2009) are very low (∼2 Gyr), which argues against this sample being the counterpart of a passively evolved population of high-redshift galaxies. Thus, this suggests that most high-redshift compact and massive galaxies must have undergone a systematic growth in size.

Among the different mechanisms proposed to explain this expansion, the one that reproduces best the observed evolution of the average-mass–size relation (Trujillo et al. 2011) is the minor merger scenario (Naab et al. 2009; Hopkins et al. 2010). With this mechanism, galaxies grow inside-out by accretion of smaller satellites that build up an extended envelope (van Dokkum et al. 2010; Hopkins et al. 2009). This model predicts a substantial growth in mass and effective radius, with a mild decrease in stellar velocity dispersion σ. Thus, measurements of σ at different redshifts are a direct way to constrain this evolutionary scenario. In addition, they allow for an independent mass estimate through the virial theorem.

The minor merger scenario is consistent with the common findings of σ ≲ 250  km s⁻¹ for general ETG samples at z = 1–2 (Gebhardt et al. 2003; Treu et al. 2005; van der Wel et al. 2005; di Serego Alighieri et al. 2005; Cenarro & Trujillo 2009; Cappellari et al. 2009; Fernández Lorenzo et al. 2011). Conversely, it is greatly challenged by the dramatic evolution required for the super-dense galaxies ( R_e < 1 kpc,  M_⋆ ⩾ 10¹¹ M_☉), which are expected to have much larger σ. van Dokkum et al. (2009) measured σ ≃ 500 ± 100  km s⁻¹ from a super-dense ETG at z = 2.2, yielding a dynamical mass that is in agreement with the photometric mass estimate. However, more σ measurements of these extreme galaxies are needed to support or weaken the need of a stronger evolutionary scenario. In this Letter we present the σ of four massive and compact ETGs at z ∼ 1, two of which are super-dense. We compare dynamical/stellar masses and estimate the evolutionary paths of these objects. Throughout this work we adopt a standard cosmology of Ω_m = 0.3, $\Omega _\Lambda =0.7$, and H₀ = 70  km s⁻¹ Mpc⁻¹.

2.1. Sample Selection

2.2. Observations and Reduction

We observed two galaxies at a time using optical long-slit spectroscopy with the OSIRIS camera spectrograph (Cepa 1998) at the 10.4 m Gran Telescopio Canarias. We used a 08 slit and R1000 grism in the spectral range of 5000–10000 Å under a seeing of ∼07. The total integration time per target was 12,800 s. The instrumental resolution was determined by Gaussian fitting of the strong sky emission lines yielding σ_inst = 120 km s⁻¹. However, on one of the observing nights the spectrograph suffered a defocus that decreased the resolution, affecting half of the exposures of 12019899 and 12024453. For these frames we found σ_inst = 210  km s⁻¹and reduced this data subset independently. The standard star L970-30 was also observed to flux calibrate the spectra. The data reduction was performed with the IRAF long-slit package and involved the standard steps of bias level correction, spectral flat-fielding, cosmic ray removal, sky subtraction, frame co-adding, and aperture extraction. The telluric absorption band around 7600 Å was present in all of our spectra, affecting the Ca H+K lines in targets 12019899 and 12024321. To correct for telluric absorption in these two galaxies, we used a F-type star observed in the same slit as 12019899. This star's spectrum has a signal-to-noise ratio (S/N) = 280 Å⁻¹ and a smooth continuum with no spectral features within the telluric band. After its continuum removal and normalization we obtained the atmospheric transmission profile in the spectral window and divided the galaxy spectra with affected Ca H+K lines by this transmission profile. We did not apply the telluric band correction on the other targets because the main spectral lines used to measure σ were not affected in them. Note that this correction increases the noise in the spectral window where it is applied. Thus, in these cases the correction did not contribute toward more precise σ measurements and for this reason we simply excluded the affected range in wavelength from the analysis. The final rest-frame spectra have S/N = 22–29 Å⁻¹ and are shown in Figure 1. Note the strong absorption in the Ca H+K lines and the Balmer series from H_γ to H₁₂. [O ii]λ3727 doublet emission was found in two objects. In 12024790, this emission shows a double peak with a separation of 4.1 Å as compared to 2.7 Å of the doublet.

Zoom In Zoom Out Reset image size

2.3. Velocity Dispersions

2.4. Stellar Populations

Dynamical masses are calculated under the assumption of homology as M_dyn = β R_eσ²/G with β = 5 (see Table 1), following studies of local ETGs (Cappellari et al. 2006). As noted in Section 2.2, the galaxy 12024790 shows a double peak in the [O ii] emission. If such feature is due to rotation of a gaseous disk, it yields a projected velocity of V_rsin(i) ≃ 165  km s⁻¹, which is comparable to its measured σ = 156  km s⁻¹. However, it is unlikely that this caused an important bias in the dynamical mass estimates. At z = 0 it has been found that β = 5 is robust against rotation up to V_r/σ ≈ 1 (Cappellari et al. 2006, 2007), and there is no clear evidence that this would be different at z = 1 (van der Wel & van der Marel 2008). Therefore, we assume that any deviations from homology due to possible rotational support are within the errors of our dynamical masses.

Our galaxies have values that are systematically smaller than the originally published stellar masses by an average factor of ∼6. This is much larger than the combined error of photometric stellar mass and radius (typically a factor of ∼3). Note that for the galaxy 12024321 these masses are consistent within uncertainties. The stellar population models we used with the RAINBOW data yield stellar masses in the range of log M_⋆ = 10.65–10.98, on average a factor of ∼1.8 larger than the dynamical masses. The masses from RAINBOW are derived with the same IMF and stellar population models as in T07 but with a broader spectral range (UV–FIR) as described in Section 2.4. In order to investigate the large differences between both sets of masses, we also derived masses using only BRIJK bands as in T07. We found no significant difference between our UV–FIR and BRIJK masses, implying that the discrepancy with T07 cannot be simply explained by the use of a different spectral range. We also checked that our mass determinations are weakly sensitive to the choice of extinction model and small variations in metallicity. Barro et al. (2011b) derive masses for hundreds of galaxies in T07's sample using the full RAINBOW photometry. They find that, on average, their masses are consistent with those from T07. Therefore, the cause of the inconsistency with our photometric masses is unclear. Given the large uncertainties in the systematics of photometric masses, we adopt dynamical masses as the correct total mass estimates. Following local lensing studies (Gavazzi et al. 2007), we apply a fiducial dark matter fraction f_DM = 0.3 and recalculate the stellar masses as 0.7 × M_dyn.

Comparison studies between stellar and dynamical mass of less dense ETGs than those in our sample have been conducted at low (Cappellari et al. 2006) and high redshifts (van der Wel et al. 2006) with the result of dynamical masses being on average equal or larger than stellar masses, as expected when considering a dark matter contribution. When van der Wel et al. (2006) calculate stellar masses using similar multi-band photometry and stellar population parameters as in T07, they find M_⋆/M_dyn = 0.66 (Salpeter IMF). Therefore, our results point toward a systematic overestimation of the stellar masses used to select the galaxies of our sample. We cannot rule out a deviation from homology that requires β ≠ 5. However, the theoretical expectation for objects with such high Sérsic index is β ≲ 5 (Cappellari et al. 2006; Bertin et al. 2002), which would yield even smaller dynamical masses.

Figure 2(a) shows the position of our objects in the stellar-mass–size relation with the old/new stellar mass estimates. Figure 2(b) shows our objects in the fundamental plane (FP). At a given  R_e, they have systematically larger masses and smaller combination of surface brightness (SB_e) and σ than the respective local relations. Moreover, our objects follow a trend in which they lie further from both local relations at smaller radii. This result has also been found in z ∼ 0.9 clusters by Jørgensen et al. (2006, 2007). Compared to the FP derived by these authors, at a given  R_e our field sample shows a similar slope and a relative offset of ∼0.3 mag toward brighter SB_e, which is in agreement with field galaxies having generally younger stellar populations than clusters of the same redshift.

Zoom In Zoom Out Reset image size

In order to understand how our galaxies relate to low-redshift ETGs, we investigate the simple evolutionary scenario based on passive luminosity fading of the stellar populations and growth via minor mergers that has been proposed for the general population of ETGs at z ∼ 1 (Gebhardt et al. 2003; Naab et al. 2009). To derive the amount of fading for our sample, we use Bruzual & Charlot (2003) models with Chabrier IMF and the redshift, ages, and metallicities from Table 1. Passive evolution from z ∼ 1 to z = 0 results in an average dimming of ∼2 mag in SB_e. We assume that the passive luminosity evolution does not affect  R_e. Minor mergers are very efficient growing the size at a given mass increment, slightly decreasing σ. According to the simulations by Naab et al. (2009), an ETG with M_⋆ = 10¹¹M_☉ and  R_e = 1.5 kpc would grow a factor of 1.5 in mass via minor mergers from z ∼ 1 to z = 0. We adopt this factor of mass growth for our galaxies. Following these authors, we model each merger as σ²_f/σ²_i = (1 + η)/(1 + η) and  R_ef/ R_ei = (1 + η)²/(1 + η), where the subindices (i)/(f) denote initial/final values of the host galaxy and = σ²_a/σ²_i and the mass ratio η = M_a/M_i take into account the accreted system (M_a, σ_a). We assume a history of four mergers with constant η = 0.1 and = 0.2 for all galaxies (note that η and the total mass growth set the total number of mergers). This choice of values is consistent with the simulations by Naab et al. (2009). The total change for  R_e and σ becomes a factor of 2 and 0.8, respectively. These results are robust against variations in the values of η and at the given total mass growth. We assume that the mass-to-light ratio does not change significantly after each merger. In addition, we assume that as the size increases, SB_e decreases by ΔSB_e∝2.94 Δlog R_e (Hamabe & Kormendy 1987).

Figure 2 shows the combined evolution of minor mergers and luminosity fading for our galaxies. Most of them reach positions that are consistent with the local FP and mass–size relations, considering the estimated uncertainties in our measurements. Therefore, this simple evolutionary scenario can plausibly describe the evolution in our sample galaxies to become normal ETGs at z = 0.

This work is partially funded by the Spanish MICINN under the Consolider-Ingenio 2010 Program grant CSD2006-00070: First Science with the GTC. J.M.M. acknowledges funding from a Graduate Alumni Fellowship at the University of Florida. The Spanish MICINN provides support to M.B., M.P., and A.H. through grant AYA2009-11137 and to J.G. through grant AYA2009-10368. Our observations were made with the GTC Telescope at the Spanish Observatorio del Roque de los Muchachos, which is managed by the Instituto de Astrofísica de Canarias. We are grateful to the personnel who took the observations and to Ana Matković for her help with the use of the $\mbox{R\raisebox {-0.35ex}{E}D \hspace{-0.50003pt}\raisebox {0.85ex}{uc}\hspace{-8.99994pt} \raisebox {-.35ex}{{m}}\hspace{0.50003pt}E}$ software. We also thank the anonymous referee for useful comments that improved the presentation of this Letter.