EVIDENCE FOR MORPHOLOGY AND LUMINOSITY TRANSFORMATION OF GALAXIES AT HIGH REDSHIFTS

We used a spectroscopic sample of galaxies in GOODS. GOODS is a deep multiwavelength survey covering two carefully selected regions including the Hubble Deep Field-North (HDF-N) and the Chandra Deep Field-South (CDF-S). Hereafter, two GOODS fields centered on HDF-N and CDF-S are called GOODS-North and GOODS-South, respectively. Total observing area is approximately 300 arcmin² and each region was observed by NASA's Great Observatories (Hubble Space Telescope (HST), Spitzer, and Chandra), ESA's XMM-Newton, and several ground-based facilities. HST observations with Advanced Camera for Surveys (ACS) were conducted in four bands: B (F435W, 7200 s), V (F606W, 5000 s), i (F775W, 5000 s), and z (F850LP, 10,660 s). The drizzled images have a pixel scale of 003 pixel⁻¹ and point-spread function FWHM of ∼01. Spectroscopic data for GOODS sources are enormous in the literature: GOODS-North (Cohen et al. 2000; Cowie et al. 2004; Wirth et al. 2004; Reddy et al. 2006) and GOODS-South (Szokoly et al. 2004; Le Fèvre et al. 2004; Mignoli et al. 2005; Vanzella et al. 2005, 2006, 2008; Ravikumar et al. 2007; Popesso et al. 2009). Among the sources in the ACS photometric catalog, we used 4443 galaxies whose reliable redshifts are available in the literature for further analysis.

Since we are going to investigate the effects of the nearest neighbor galaxies, it is important to identify genuine neighbor galaxies. In addition, since we use a spectroscopic catalog of galaxies combined from various redshift surveys with diverse selection criteria, our sample is heterogeneous. Therefore, it is necessary to know the completeness of our spectroscopic sample accurately. To compute the spectroscopic completeness, we used a sample from the ACS photometric catalog with the objects having SExtractor stellarity class greater than 0.79. The value was chosen from the stellarity distribution of genuine stars confirmed by the spectroscopic observation. In Figure 1, we plot the completeness of each survey as a function of the observed magnitude (SExtractor "BEST" magnitude) and color. It is seen that the completeness decreases significantly near i_775W,BEST ∼ 23.5. The mean completeness at i_775W,BEST ⩽ 23.5 is 69% and 63% for GOODS-North and -South, respectively. It is noted that the completeness below the magnitude limit changes significantly with color, which can cause a bias in the studies of galaxy morphology.

Zoom In Zoom Out Reset image size

The rest-frame B-band absolute magnitude M_B of galaxies is computed based on the ACS photometry with Galactic reddening corrections (Schlegel et al. 1998) and K-corrections (Blanton & Roweis 2007). The evolution correction (an increase of 1.3M_B per unit redshift) was applied to compute the final rest-frame M_B (Faber et al. 2007). In Figure 2, we show the evolution-corrected, rest-frame M_B as a function of redshift for the sample of galaxies with i_775W,BEST ⩽ 23.5. We finally define a volume-limited sample of 1332 galaxies with M_B ⩽ −18.0 and 0.4 ⩽ z ⩽ 1.0 for further analysis.

Zoom In Zoom Out Reset image size

For the volume-limited sample of galaxies shown in Figure 2, we visually inspected the images in individual (Bviz) bands and Bvi pseudocolor images. We divided the galaxies into two morphological types: early types (E/S0) and late types (S/Irr). Early-type galaxies are those with little fluctuation in the surface brightness and color and with good symmetry, while late-type galaxies show internal structures and/or color variations in the pseudocolor images.

However, the total color itself is not used as a classification criterion. We checked our results by comparing with the morphological classification of Bundy et al. (2005) who used the same ACS images as ours, and found that 98% of our classifications agree with those of Bundy et al. Some galaxies that were classified as early types in Bundy et al. (2005) are found to be late types in this study because of color variations in the pseudocolor images.

To see how often we classify galaxies differently because we classified galaxies in different wavelength bands across redshift, we made the following experiment. We classified galaxies at 0.4 < z < 0.6 brighter than M_B = −18.0 using their v- or i-band images separately, and checked if the morphological types of each galaxy agree with each other. At z = 0.9, the i band is centered at ∼4000 Å in the rest frame, and this wavelength roughly corresponds to the v band at z ≈ 0.5. We found that no galaxy was assigned different morphology. Therefore, our morphology classification is not expected to be affected by the redshift effects.

We consider two kinds of environmental factors: a surface galaxy number density estimated from five nearest neighbor galaxies (Σ₅) as a large-scale environmental parameter and the distance to the nearest neighbor galaxy (r_p) as a small-scale environmental parameter.

The background density, Σ₅, is defined by Σ₅ = 5(πD²_p,5)⁻¹. D_p,5 is the projected proper distance to the 5th-nearest neighbor. The 5th-nearest neighbor of each target galaxy was identified among the neighbor galaxies with M_B ⩽ −18.0 that have velocities relative to the target galaxy less than 1000 km s⁻¹ to exclude foreground and background galaxies.

To define the small-scale environmental parameter attributed to the nearest neighbor, we first find the nearest neighbor of a target galaxy that is closest to the target galaxy on the projected sky and satisfies the conditions of magnitude and relative velocity. We searched for the nearest neighbor galaxy among galaxies that have magnitudes brighter than M_B = M_B,target + 0.5 and have relative velocities less than Δv = |v_neighbors − v_target| = 600 km s⁻¹ for early-type target galaxy and less than Δv = 400 km s⁻¹ for late-type target galaxy. These values are the same as those used for selecting the nearest neighbor galaxy in the SDSS data (Park et al. 2008). Since we use the volume-limited sample of galaxies with M_B ⩽ −18.0, we study only the target galaxies brighter than M_B,target = −18.5 so that their neighbors are complete.

In Figure 3, we plot the velocity difference for all neighbor galaxies with velocity difference less than 900 km s⁻¹. To fit the velocity distribution, we used the Gaussian plus constant model for each type of target galaxies by combining the data in two surveys. We obtained, for the galaxies at projected proper distance r_p < 100h⁻¹ kpc, the best-fit values σ_Δv = 351 ± 48 and 216 ± 103 km s⁻¹ for early- and late-type target galaxies, respectively. It indicates that the velocity limit used for selecting the nearest neighbor is large enough not to miss the neighbor galaxies.

Zoom In Zoom Out Reset image size

The spectroscopic completeness can affect the identification of the genuine nearest neighbor. The completeness that depends on the apparent magnitude and color as shown in Figure 1, can also depend on the distance between galaxies due to the difficulty in observing galaxies close to each other using multiobject spectrograph (MOS). We checked the completeness as a function of the projected distance to the target galaxy, and found that it does not change with the projected distance. It might be because we combined spectroscopic data from numerous references, therefore, the difficulty in observing nearby galaxies using MOS is significantly reduced.

The virial radius of a galaxy is defined as the radius within which the mean mass density is 200 times the critical density of the universe (ρ_c), and is given by

$$\begin{equation} r_{\rm vir} = (3 \gamma L / 4\pi / 200{\rho }_c)^{1/3}, \end{equation} \tag{ 1 }$$

where L is the galaxy luminosity and γ is the mass-to-light ratio. We assume that the mass-to-light ratio of early-type galaxies is on average twice as large as that of late-type galaxies at the same absolute magnitude M_B, which means γ(early) =2γ(late) (see Section 2.5 of Park & Choi 2009 and Section 2 of Park et al. 2008). The critical density of the universe ρ_c is a function of redshift z and $\Omega _m(z)= \rho _c (z)/\rho _b(z)=\rho _c(z)/\overline{\rho }(1+z)^3$, where ρ_b and $\overline{\rho }$ are the mean matter densities in proper and comoving spaces, respectively. Then, the virial radius of a galaxy at redshift z in proper space can be rewritten by

$$\begin{equation} r_{\rm vir} (z) = [3 \gamma L \Omega _{m,0} / 800\pi \overline{\rho } / \lbrace \Omega _{m,0}(1+z)^3 + \Omega _{\Lambda,0} \rbrace ]^{1/3}. \end{equation} \tag{ 2 }$$

Ω_m,0 and Ω_Λ,0 are the density parameters at the present epoch. We compute the mean mass density $\overline{\rho }$ using the galaxies at z = 0.4–0.7 with various absolute magnitude limits varying from M_B = −17.5 to −20.0. We found that the mean mass density appears to converge when the magnitude cut is fainter than M_B = −18.0, which means that the contribution of faint galaxies is not significant because of their small masses. In this calculation, we weigh each galaxy by an inverse of completeness according to its apparent magnitude and color (see Figure 1). We obtain $\overline{\rho } = 0.0145 (\gamma L)_{-20}$ (h⁻¹ Mpc)⁻³ where (γL)₋₂₀ is the mass of a late-type galaxy with M_B = −20. This final value we adopted is computed using all galaxies with i_775W,BEST ⩽ 23.5 and 0.4 ⩽ z ⩽ 0.7 to account for the contribution of faint galaxies as much as possible.

If we adopt Ω_m,0 = 0.27 and Ω_Λ,0 = 0.73, the virial radii of galaxies with M_B = −18.5 and −20.0 at z = 0 are 220 and 350 h⁻¹ kpc for early types, and 180 and 280 h⁻¹ kpc for late types, respectively. The proper-space virial radii of galaxies with the same luminosities as above, but located at z = 1, are 160 and 250 h⁻¹ kpc for early types, and 120 and 200 h⁻¹ kpc for late types, respectively.

In Figure 4, we plot the fraction of early-type galaxies as a function of the background density, Σ₅. To account for the incompleteness shown in Figure 1, the early-type fraction is computed by weighing each galaxy by the inverse of completeness corresponding to its apparent magnitude and color. Each of the volume-limited samples is divided into brighter (−19.25 ⩾ M_B> − 20.00) and fainter (−18.50 ⩾ M_B> − 19.25) subsamples. The solid and dotted lines are early-type fractions (f_E) for brighter and fainter subsamples, respectively. The uncertainties of the fraction represent 68%(1σ) confidence intervals that are determined by the bootstrap resampling method. Figure 4 clearly shows that the early-type fraction increases along with Σ₅ (i.e., MDR), which is already known in similar high-redshift surveys (e.g., Capak et al. 2007). We note that the overall early-type fraction is higher for the brighter subsamples in all surveys. It implies that the morphology–luminosity relation is already well established at the redshifts under study. The background density dependence of morphology seems stronger for the brighter galaxies. In particular, the early-type fraction of the brighter sample rises sharply at high densities of Σ₅ ≳ 10h² Mpc⁻².

Zoom In Zoom Out Reset image size

To measure the effects of the nearest neighbor galaxy on the galaxy morphology, we plot, in Figure 5, the fraction of early-type galaxies as a function of the distance to the nearest neighbor. We can see the probability that a galaxy is found to be an early type, strongly depends on the projected distance to the nearest neighbor galaxy (r_p) as well as neighbor's morphology. When a galaxy is located farther than the virial radius from its nearest neighbor galaxy (r_p ≳ r_vir,nei), the early-type fraction slowly increases as the distance to the neighbor decreases, but its dependence on neighbor's morphology is weak.

Zoom In Zoom Out Reset image size

On the other hand, when r_p ≲ r_vir,nei, the early-type fraction increases as the target galaxy approaches an early-type neighbor, but decreases as it approaches a late-type neighbor. It is important to note that the bifurcation of the early-type fraction occurs at r_p ∼ r_vir,nei. In the case of the cosmology, we adopt that the radius r_p = r_vir,nei corresponds to the local mass density due to the neighbor of $\rho _n=740\overline{\rho }$. The bifurcation of the early-type fraction at r_p ∼ r_vir,nei is similarly found by Park et al. (2008) using nearby (z < 0.1) SDSS galaxies.

To investigate the effects of large-scale environment on the morphology in company with the effects of the nearest neighbor galaxy, we study the early-type fraction in the two-dimensional environmental parameter space as shown in Figure 6. In this case, the combined sample of GOODS-North plus South is used, and the luminosity of target galaxies is limited to −18.5>M_B> − 19.5 to reduce the luminosity dependence of morphology. Galaxies are distributed along the diagonal in this figure due to the statistical correlation between r_p and ∑₅. But there is a significant dispersion in r_p at a fixed ∑₅. It is noted that galaxies are located at a wide range of ∑₅ for both early- and late-type neighbor cases. In this figure, one can see the dependence of f_E on r_p at each fixed value of ∑₅ by moving along a vertical line. Figure 6 clearly shows that the dependence of morphology on the background density and the nearest neighbor distance is completely different when the neighbor's morphology is different. When the neighbor is a late type, f_E of target galaxies is about 0.2 and is almost independent of r_p or ∑₅. But when the neighbor is an early type, f_E can rise up to 0.7 when the target galaxy approaches the neighbor closer than 0.3r_vir,nei or enters a region with the background density exceeding ∑₅ ∼ 10 h² Mpc⁻². When r_p>r_vir,nei, galaxy morphology is independent of all environmental factors and f_E is about about 0.2. It should be noted that difference between the left and right panels of Figure 6 is manifest only when r_p is less than r_vir,nei. And the dependence on neighbor morphology is evident at the neighbor distance of r_p ≈ 0.3r_vir,nei. This confirms again the effects of the nearest neighbor on morphology for galaxies in the GOODS sample.

Zoom In Zoom Out Reset image size

In Figure 7, we plot the evolution-corrected, rest-frame M_B as a function of the distance to the nearest neighbor for the combined sample of GOODS-North plus South with M_B ⩽ −18.5. The lines show the median value in each projected separation bin. Top panels show that the early-type galaxies having r_p>r_vir,nei are significantly brighter than those at r_p < r_vir,nei, while the magnitude difference for late-type galaxies in the two regions is smaller. If we divide the galaxies into those in relatively higher density region (Σ₅>6h² Mpc⁻²) and those in lower density region (Σ₅ ⩽ 6h² Mpc⁻²), the increase of galaxy luminosity with increasing r_p is manifest in the high-density region.

Zoom In Zoom Out Reset image size

We found in Figure 5 that the morphological types of high-redshift galaxies depend critically on the small-scale environment, which is characterized by the morphology of the nearest neighbor galaxy and the distance to the nearest neighbor galaxy. If our results are affected by the trend that the early-type fraction (f_E) is higher when the local density is higher (i.e., MDR), f_E should be a function of r_p which is independent of the neighbor morphology, and the two curves in the top panel of Figure 5 have the same amplitude at each neighbor separation. The fact that f_E is independent of neighbor morphology at r_p>r_vir,nei but starts to show a significant dependence at r_p < r_vir,nei, demonstrates that the neighbor effects are the dominating factor of the change in f_E and the large-scale background density is not.

Increase of f_E with decreasing the neighbor distance at r_p ≳ r_vir,nei, can be explained by the tidal effects of neighbor galaxy. Park et al. (2008) showed that tidal energy deposit relative to the binding energy for the dark halos of equal mass galaxies, is not negligible at the separation of virial radius. The tidal effects can accelerate the consumption of cold gas, changing late types to early types. In fact, Park & Choi (2009) showed that the center of late-type galaxies becomes bluer by the existence of neighbor galaxies even when r_p>r_vir,nei independently of the morphological type of the neighbor. Their surface brightness and central velocity dispersion increase as the neighbor distance decreases, implying the growth of the bulge component.

The bifurcation of f_E at r_p ∼ r_vir,nei was interpreted as due to the hydrodynamic effects of the nearest neighbor (Park et al. 2008). If a target galaxy approaches a late-type neighbor within one virial radius of the neighbor, the cold gas of the neighbor can flow into the target galaxy and the target galaxy tends to become a late type. On the other hand, if the target galaxy approaches an early-type neighbor within one virial radius of the neighbor, the hot gas and the tidal force of the neighbor accelerate the consumption of cold gas so that the target galaxy tends to evolve to an early type.

Moreover, it is noted in Figure 6 that the galaxies tend to be early types when they have close early-type neighbors even though they are in low-density regions, and the galaxies are likely to be late types when they have close late-type neighbors even though they are in high-density regions. The galaxy morphology appears to depend on the large-scale background density. But it may be the result of the cumulative effects of neighbor interaction that is stronger in high-density regions (Park et al. 2008).

We also found in Figure 7 that isolated galaxies are brighter than less isolated ones. These results can provide important hints for the evolution of galaxies. Previously, Park et al. (2008) obtained results similar to ours using the SDSS data (see their Figure 6), and suggested a unified scenario as follows. Once the separation of two galaxies becomes small enough, namely, r_p ≲ 0.05r_vir,nei (Park & Choi 2009), they undergo a merger and the merger product will be more massive. The merger product will typically find itself isolated from its neighbors of comparable mass. As galaxies experience a series of merger events, the cold gas will be exhausted due to the merger-induced star formation. Thus, the massive galaxies are likely to be early types. As a supporting evidence for this scenario, they showed that, at fixed background density, postmerger features such as large displacement of the galaxy nucleus from the center, turmoil features, and/or very close double cores, are more frequently seen in the isolated galaxies compared to the less isolated ones.

We extended our analysis to another multiwavelength survey, AEGIS, of which survey area (∼710 arcmin²) is larger than GOODS (∼300 arcmin²). Using AEGIS data, we measured the early-type fraction of target galaxies as a function of the nearest neighbor distance (e.g., Figure 5), and found that the early-type fraction does not change significantly with the neighbor distance even though it showed dependence on the large-scale background density. Furthermore, it does not show the bifurcation in accordance with neighbor's morphology. These results are inconsistent with those from GOODS and SDSS samples (e.g., Park et al. 2008; Park & Choi 2009). The results from AEGIS may have been significantly affected by the low completeness of DEEP2 survey. Since the completeness of DEEP2 spectroscopic survey is only about 0.5, it is expected that the nearest neighbor is seriously misidentified compared to the case using the GOODS or SDSS data. Our Monte Carlo experiment shows that the fraction of the misidentified nearest neighbor reaches about 50% when the sample completeness is 50%. We therefore conclude that it is not appropriate to study galaxy interactions using the DEEP2/AEGIS data.

If galaxies transform their morphology through a series of interactions and mergers, it is expected that the high-redshift galaxies, on average, are less massive and richer in cold gas. Successive consumption of cold gas in galaxies through interaction and merger events is likely to transform late types into early types. And then the early-type fraction at high redshifts is expected to be lower compared to that at low redshifts, and the fraction of blue early types among early types to be higher at higher redshifts (Park et al. 2008; Capak et al. 2007; see also Conselice et al. 2005, 2008; Lotz et al. 2008; Menanteau et al. 2004; Lee et al. 2006; Puzia et al. 2007). Moreover, the early-type fraction is expected to evolve differently depending on the background density (Park et al. 2008). Since the mean separation between galaxies is relatively smaller and the merger/interaction rate is higher in high-density regions, galaxies are expected to show a stronger time evolution in morphology as well as luminosity, color, and SFR in high-density regions. In contrast, in low-density regions where the mean separation between galaxies is larger, the time evolution of galaxies is expected to be slower because the interactions and mergers among galaxies are less frequent.

There have been several studies of the evolution of morphology of high-redshift galaxies focusing on the role of environment. For example, Nuijten et al. (2005) used Sérsic index n and the u* − g' color of galaxies as proxies for galaxy morphology to study galaxy evolution. They found that the number fraction of "bulge-dominated" galaxies with n>2 in the field does not change with redshift, while that in the high-density region tends to increase toward the low redshift. When they use red galaxies with u* − g'>1, the fraction of red galaxies increases as redshift decreases in all environments. However, since they used photometric redshifts to estimate the local density, the environment was not accurately determined. In fact, Cooper et al. (2007) found no time evolution of the fraction of red galaxies in low-density regions at 0.4 < z < 1.35 using spectroscopic redshift data of DEEP2 galaxy redshift survey, though the number of red galaxies in their sample may be too small to draw a statistically significant conclusion.

Recently, Capak et al. (2007) used Gini parameter to select early-type galaxies in COSMOS data. They found that the early-type fraction in all environments grows as redshift decreases, and the growth rate is larger in high-density regions compared to that in low-density regions. However, in low-density regions with <100 galaxies Mpc⁻², the early-type fraction at z>0.4 changes little with redshift. van der Wel et al. (2007) adopted a combination of Sérsic index n and bumpiness of galaxies as a proxy of galaxy morphology in SDSS and GOODS-South data. They found no time evolution of the early-type fraction at z < 0.8 for both cluster and field regions. They suggested that the reason why they could not find the evolution of the early-type fraction even in high-density regions, is because they used mass-selected samples and different morphological classification from the other studies.

To confirm the environmental dependence of evolution of galaxy properties, we present the early-type fraction as a function of redshift in Figure 8. We divide our sample according to the redshift and the background density. It shows that the early-type fraction in high-density regions is higher than that in low-density regions in all redshift intervals explored. Moreover, the early-type fraction in high-density regions increases much more rapidly as redshift decreases. It indicates a strong evolution of galaxy morphology in high-density regions, which is consistent with the observational results of some of previous studies and also with the predictions of Park et al. (2008).

Zoom In Zoom Out Reset image size

Since our study indicates that the nearest neighbors derive the transformation of galaxy morphology, it is important to check galaxy morphology has evolved under strong influence of the nearest neighbors. Figure 9 is such an examination. We divided our sample into two subsets containing galaxies with early- or late-type neighbors. In this analysis, we included only those galaxies that are interacting with neighbors (r_p < r_vir,nei). For the early-type fraction at low redshifts, we used the SDSS galaxies with −20.0>M_r> − 21.5 and z < 0.1 and whose magnitudes roughly fall into our M_B magnitude range (Choi et al. 2007; Park et al. 2008). Figure 9 clearly shows that the early-type fraction monotonically increases as redshift decreases, and the early-type fraction for the case of early-type neighbor is always larger than that for the late-type neighbor case. Primordial origin is not likely to be the reason for this systematic difference because the neighbor effects are limited within relatively tiny volume of the universe associated with each galaxy (see Figure 5). As a comparison, the rms displacement of matter in the flat ΛCDM universe is 7.7 h⁻¹ Mpc by z = 0.5, more than an order magnitude larger than the virial radius of typical galaxies (Park & Kim 2009). This result demonstrates an important role of the nearest neighbor in the evolution of galaxy morphology, and is consistent with the prediction that the early-type fraction from high to low redshifts will increase because a series of interactions and mergers tend to transform late types into early types.

Zoom In Zoom Out Reset image size

In summary, the early-type fraction in high-density regions increases significantly as redshift decreases, while that in low-density regions increases much more gently. The evolution of galaxy morphology is also found to depend critically on the small-scale environment, which is characterized by the morphology of the nearest neighbor galaxy and the distance to the nearest neighbor galaxy, in addition to the large-scale background density. A large sample of high-redshift galaxies is needed to separate between the roles of small- and large-scale environments. We also found that isolated galaxies are brighter than less isolated ones. All these results are consistent with the predictions of Park et al. (2008), and confirm their unified scenario of transformation of galaxy morphology and luminosity class at high redshifts. It implies that galaxy–galaxy interactions play an important role in the evolution of morphology and luminosity classes of galaxies over a long period of time.