The Properties of the Massive Star-forming Galaxies with an Outside-in Assembly Mode

Our main sample is the MaNGA released data, including 1390 galaxies, which enables us to select a sample of massive SF galaxies with the outside-in assembly mode. Here we briefly introduce the MaNGA data. A full description of both observations and data reduction are presented in Bundy et al.(2015).

The MaNGA project is an ongoing large IFS survey aimed at mapping 10,000 nearby galaxies with redshifts from 0.01 to 0.14. The MaNGA instrument covers a wavelength range from 3600 to 10300 Å at a typical resolution R ∼ 2000. The absolute flux calibration is better than 5% for more than 89% of the MaNGA wavelength range (Yan et al. 2016). The raw data were reduced using the MaNGA Data Reduction Pipeline, which was developed by Law et al. (2015).

The MaNGA released sample is mainly composed of the "primary" (∼2/3) and "secondary" (∼1/3) samples (SDSS Collaboration et al. 2016), which are defined by two radial coverage goals: the former of 1.5${R}_{{\rm{e}}}$, and the latter of 2.5${R}_{{\rm{e}}}$. The released data cubes have a spaxel size of 05, and the typical effective spatial resolution can be described by a Gaussian with an FWHM ∼ 25.

The spectral fitting methods and parametric measurements are described in Li et al. (2015). Here we briefly summarize the fitting procedure and the measurements of the referred parameters. To fit the continuum, we adopt the χ² minimization fitting method developed by Li et al. (2005) that can carefully mask the emission-line regions by iteration. The templates we use are galactic eigenspectra generated by the principal component analysis method. This fitting code is efficient and stable, even for spectra with low signal-to-noise ratios (S/Ns). However, this code cannot fit the line-of-sight velocity automatically. Thus, we use the line-of-sight velocity measured by the public code, STARLIGHT (Cid Fernandes et al. 2005), as input for each spectrum. The templates used in STARLIGHT fitting are 45 single stellar populations from Bruzual & Charlot (2003) with a Chabrier (2003) initial mass function. The intrinsic stellar extinction law is from Cardelli et al. (1989) in both fittings. For each galaxy, we perform the spectral fitting to the spaxels separately without a binning scheme. Spectra with the continuum S/N < 5 at 5500 Å are not included in the analysis in this work.

We use the 4000 Å break (D_n(4000)) and the Hδ absorption (EW(Hδ_A)) and Hα emission (EW(Hα)) equivalent widths to trace the recent star formation histories (Bruzual & Charlot 2003; Kauffmann et al. 2003a; Li et al. 2015), since these parameters are more model-independent than the mean stellar age and star formation rate (SFR). Here D_n(4000) and the Hδ_A index are measured based on the best-fitting spectra, while EW(Hα) is measured based on the stellar component-subtracted spectrum by fitting a Gaussian profile to the Hα emission lines. A full description of the spectral fitting and parametric measurements are presented in Li et al. (2015).

In previous works, galaxies were usually separated according to their global properties, such as SFR, stellar mass, and color. The IFS surveys show that different regions of galaxies can have very different star formation properties, suggesting that only using global parameters is not enough. Thanks to the IFS survey MaNGA, we are able for the first time to select the galaxies with an outside-in mass assembly mode and study their global and resolved properties.

We use the 4000 Å break to select the targeted galaxies. The 4000 Å break is defined as the ratio of the flux in the red continuum to that in the blue continuum at the 4000 Å wavelength (Balogh et al. 1999; Brinchmann et al. 2004). The D_n(4000) has been widely used to trace the mean stellar ages of stellar populations younger than 1–2 Gyr. We note that mass assembly histories can be diverse in different periods (Ibarra-Medel et al. 2016), and the mass assembly mode referred to in this work is the recent assembly mode (<2 Gyr).

The targeted sample was selected from the SDSS-IV MaNGA released sample (SDSS Collaboration et al. 2016), which contains 1390 galaxies with 0.01 < z < 0.14. Since we are just interested in massive galaxies with an outside-in assembly mode, the sample was required (1) to have ${\mathrm{log}}_{10}({M}_{* }/{M}_{\odot })$ > 9.8, (2) to have a larger 4000 Å break in the 1.5${R}_{{\rm{e}}}$ radial bin (${D}_{n}{(4000)}_{1.5{R}_{{\rm{e}}}}$) than in the center (${D}_{n}{(4000)}_{\mathrm{cen}}$), and (3) to consist of SF galaxies or partially SF galaxies ($\mathrm{NUV}-r$ < 5). The reason that we used ${M}_{* }\ \gt {10}^{9.8}\,{M}_{\odot }$ rather than ${M}_{* }\ \gt {10}^{10}\,{M}_{\odot }$ and $\mathrm{NUV}-r$ < 5 rather than $\mathrm{NUV}-r$ < 4 is so that we could select more galaxies. This is useful for measuring the clustering properties of our sample in Section 3.4. In addition, we excluded mergers and heavily disturbed galaxies by visual inspection. In total, 77 galaxies were selected according to the above criteria, accounting for 14% of all galaxies with ${\mathrm{log}}_{10}({M}_{* }/{M}_{\odot })$ > 9.8 and $\mathrm{NUV}-r$ < 5.

For comparison, we built two control samples closely matched in stellar mass with the targeted sample but not in outside-in assembly mode. The control samples were randomly selected from the MaNGA released sample with several criteria. We call them the blue control (BC) sample and red control (RC) sample. The BC sample was selected to have ${D}_{n}{(4000)}_{\mathrm{cen}}$ > ${D}_{n}{(4000)}_{1.5{R}_{{\rm{e}}}}$ and $\mathrm{NUV}-r$ < 5, while the RC sample was selected to have $\mathrm{NUV}-r$ > 5. In total, 154 BC galaxies and 62 RC galaxies were selected.

Since we do not limit the galaxy inclination in the sample selection, 12 out of 77 targeted galaxies have an inclination angle greater than 70°, based on the minor-to-major axis ratio from the NASA-Sloan Atlas (NSA). One may worry that the measurements of D_n(4000) are heavily affected by the dust extinction for these highly inclined galaxies, which may pollute the identification of targeted galaxies. We have checked the D_n(4000) and ${\mathrm{log}}_{10}\mathrm{EW}({\rm{H}}\alpha )$ profiles of these highly inclined galaxies and found that they exhibit positive D_n(4000) gradients and prominently negative ${\mathrm{log}}_{10}\mathrm{EW}({\rm{H}}\alpha )$ gradients in general. The measurements of EW(Hα) are not sensitive to the dust extinction, which confirms that these highly inclined galaxies are in the outside-in assembly mode. In addition, we have reproduced our main results while excluding these highly inclined galaxies and found that the basic results are not changed. Thus, we keep these highly inclined galaxies in this paper.

The left and middle panels of Figure 1 show the targeted sample (green circles), BC sample (blue crosses), and RC sample (red triangles) on the diagrams of ${D}_{n}{(4000)}_{\mathrm{cen}}$ versus ${D}_{n}{(4000)}_{1.5{R}_{{\rm{e}}}}$ and ${\mathrm{log}}_{10}\mathrm{EW}{({\rm{H}}\alpha )}_{\mathrm{cen}}$ versus ${\mathrm{log}}_{10}\mathrm{EW}{({\rm{H}}\alpha )}_{1.5{R}_{{\rm{e}}}}$. For comparison, we plot all MaNGA galaxies with ${M}_{* }\ \gt {10}^{9.8}\,{M}_{\odot }$ as gray circles in these two panels. The right panel shows the fraction of our targeted galaxies with respect to massive SF galaxies ($\mathrm{NUV}-r$ < 5 and ${M}_{* }\ \gt {10}^{9.8}\,{M}_{\odot }$) as a function of stellar mass. The errors are estimated by the bootstrap method. The fraction of targeted galaxies has a peak at ${M}_{* }\ \sim \ {10}^{10.2}\,{M}_{\odot }$ and drops heavily at ${M}_{* }\ \gt {10}^{11.1}\,{M}_{\odot }$.

Zoom In Zoom Out Reset image size

As shown in Figure 1, these three samples are well separated in the diagrams of ${D}_{n}{(4000)}_{\mathrm{cen}}$–${D}_{n}{(4000)}_{1.5{R}_{{\rm{e}}}}$ and ${\mathrm{log}}_{10}\mathrm{EW}{({\rm{H}}\alpha )}_{\mathrm{cen}}$–${\mathrm{log}}_{10}\mathrm{EW}{({\rm{H}}\alpha )}_{1.5{R}_{{\rm{e}}}}$. Most galaxies have a higher 4000 Å break in their centers than in their 1.5${R}_{{\rm{e}}}$ radial bins, which is consistent with the previous findings of inside-out assembly mode for massive galaxies (Pérez et al. 2013; Pan et al. 2015; Goddard et al. 2017). However, we do see a small fraction of massive SF galaxies with ${D}_{n}{(4000)}_{\mathrm{cen}}$ < ${D}_{n}{(4000)}_{1.5{R}_{{\rm{e}}}}$, shown by green circles in the figure. As expected, they have larger ${\mathrm{log}}_{10}\mathrm{EW}({\rm{H}}\alpha )$ in their centers than in their 1.5${R}_{{\rm{e}}}$ radial bins, indicating the dramatic ongoing star formation activity in their centers. In contrast, the BC galaxies have a larger 4000 Å break and smaller EW(Hα) in their centers than in their 1.5${R}_{{\rm{e}}}$ radial bins, representing a normal SF population. The RC galaxies show no or very weak star formation activity, representing a quiescent population. They have a larger 4000 Å break and smaller EW(Hα) in both their centers and their 1.5${R}_{{\rm{e}}}$ radial bins than the targeted and BC galaxies.

Examples of a BC galaxy (top row), targeted galaxy (middle row), and RC galaxy (bottom row) are shown in Figure 2. In each row, the columns represent the SDSS color image, 4000 Å break, Hδ_A index, and ${\mathrm{log}}_{10}\mathrm{EW}({\rm{H}}\alpha )$ maps of the shown galaxy. The hexagon in each panel displays the area covered by IFS bundles. Here Hδ_A is the Lick/IDS index of the Hδ absorption line (Worthey & Ottaviani 1997), and ${\mathrm{log}}_{10}\mathrm{EW}({\rm{H}}\alpha )$ is the logarithm of the Hα emission line equivalent width. These three parameters are good indicators of recent star formation histories (Bruzual & Charlot 2003; Kauffmann et al. 2003a; Li et al. 2015). The Hδ_A index traces the star formation that occurred 0.1–1 Gyr ago, and EW(Hα) traces the strength of very recent (within 50 Myr) or ongoing star formation activity.

Zoom In Zoom Out Reset image size

As shown in Figure 2, the two-dimensional star formation properties of the targeted galaxy (3704@8618, the galaxy located on plate 8618 with a bundle ID of 3704) and the BC galaxy (6101@8256) are significantly different. The BC galaxy has a higher D_n(4000), lower Hδ_A index, and ${\mathrm{log}}_{10}\mathrm{EW}({\rm{H}}\alpha )$ in its center rather than its outer regions. The star formation activity has already been shut down in its center, where a bulge and bar can be seen. According to its ${\mathrm{log}}_{10}\mathrm{EW}({\rm{H}}\alpha )$ map, ongoing star formation activity can be seen at its outer regions. In contrast, the targeted galaxy (3704@8618) has a lower D_n(4000), higher Hδ_A index, and ${\mathrm{log}}_{10}\mathrm{EW}({\rm{H}}\alpha )$ in its center rather than its outer regions. It has a smooth morphology without spiral arms seen from the SDSS color image. It has strong star formation activity in its center, while very weak star formation activity is shown in its outer regions. The RC galaxy (3701@8440) is a quiescent galaxy with no prominent Hα emission lines.

In the previous section, we selected the galaxies with a recent outside-in mass assembly mode. Now questions arise as to whether their global properties, such as size, structure, SFR, and metal abundance, are different from those of control galaxies.

Figure 3 shows the comparison of the global properties of the targeted sample and two control samples. These four panels show the stellar mass versus size, concentration (${R}_{90}/{R}_{50}$), SFR, and gas-phase metallicity for the targeted sample (green circles), BC sample (blue crosses), and RC sample (red triangles). The measurements of these four parameters for the targeted galaxies are listed in Table 1. In the bottom right panel of the figure, only the targeted and BC samples are shown, because the gas-phase metallicity is computed using emission lines and RC galaxies usually do not have prominent emission lines. In the top left panel, the blue and red curves represent the mass–size relation for late-type galaxies (Sérsic n < 2.6) and early-type galaxies (Sérsic n > 2.6) taken from Shen et al. (2003). The gray dashed lines represent a constant stellar mass surface density of ${{\rm{\Sigma }}}_{e}={10}^{7}$, 10⁸, 10⁹, and 10¹⁰ ${M}_{\odot }$ kpc⁻² (top left to bottom right). In the bottom left panel, the gray solid line shows the star formation main sequence taken from Chang et al. (2015), and the gray dashed lines show the 14%−84% region. In the bottom right panel, the black solid line shows the mass–metallicity relation taken from Tremonti et al. (2004), and the black dashed lines show the 1σ region.

Zoom In Zoom Out Reset image size

Table 1.  List of Targeted Galaxies

IDa	R.A.	Decl.	log10M*	Re	log10SFR	Zgas	B/Tb	Bar/Unbarc	Morphologyc	Dpd
	(J2000.0)	(J2000.0)	(${M}_{\odot }$)	(kpc)	(${M}_{\odot }\,{{\rm{yr}}}^{-1}$)					(kpc)
7443–12701	15:22: 1.79	43:31:56.4	10.0	1.58	−1.4	8.96	0.33	unbar	smooth	⋯
8257–12701	11:01:59.00	45:13:40.9	10.7	3.55	0.8	9.11	⋯	unbar	disk	125
8465–12701	13:03:53.56	47:23:11.2	10.3	4.16	0.2	9.04	0.17	bar	disk	139
8318–12702e	13:05:46.57	46:27:42.7	10.5	6.13	0.1	9.08	0.01	⋯	disk	⋯
8138–12703	07:45:33.53	45:46:20.4	11.1	6.65	−0.9	9.10	⋯	bar	disk	159
8312–12703	16:28:50.17	39:50: 6.3	10.7	5.51	0.5	9.17	0.04	unbar	disk	⋯
8550–12703	16:30:41.86	40:31:45.8	10.6	4.91	0.2	9.16	0.25	bar	disk	41
7962–12704	17:23:31.94	27:27:31.4	10.2	3.48	−0.1	9.06	0.16	unbar	disk	⋯
8253–12704e	10:36:36.79	43:30:24.4	10.2	3.08	−1.0	9.17	0.05	⋯	smooth	⋯
8312–12704	16:29:12.99	41:09: 3.1	10.5	4.30	−0.3	9.21	0.26	bar	disk	110
8448–12704e	11:06:57.52	22:50: 8.9	10.0	2.89	−0.2	9.03	0.00	⋯	disk	107
8603–12704	16:31:34.53	40:33:56.2	10.9	4.52	0.5	9.19	⋯	unbar	disk	79
8315–12705	15:43:40.92	39:32:25.3	10.8	7.88	0.9	9.06	0.20	unbar	disk	⋯
8466–12705f	11:25:12.46	45:04: 0.4	10.2	4.72	−0.3	8.59	0.06	unbar	disk	⋯
8551–12705	15:35:45.82	44:50: 5.3	10.6	5.20	0.7	9.00	0.22	unbar	disk	49
8082−1901	03:13:15.16	−00:05:38.1	10.2	1.27	−1.7	9.03	0.94	unbar	smooth	⋯
8143−1901	07:54:38.45	42:32:38.8	10.1	1.27	−0.3	9.15	0.36	unbar	smooth	⋯
8249−1901g	09:08:52.64	44:55:56.2	9.9	1.27	0.0	8.99	0.72	unbar	smooth	⋯
8253−1901	10:32: 4.09	43:51:33.4	10.5	2.04	−0.6	9.09	0.80	unbar	smooth	130
8325−1901	14:11:11.16	46:23:26.0	10.1	1.14	−3.3	9.08	0.54	unbar	smooth	⋯
8452−1901h	10:23:32.53	46:03:27.9	10.0	1.95	−1.2	8.93	0.61	unbar	smooth	⋯
8453−1901	10:13:27.73	47:30:58.4	10.0	1.38	−1.4	9.03	0.32	unbar	smooth	⋯
8485−1901	15:30:10.73	48:41:24.7	9.9	1.07	−0.2	9.16	0.87	unbar	smooth	⋯
7990−1902	17:38: 5.40	57:07: 5.4	10.7	1.84	0.5	9.22	0.88	unbar	smooth	80
8254−1902g	10:52:52.59	43:25:42.3	10.1	1.13	−0.1	9.26	0.75	unbar	smooth	42
8259−1902	11:51:11.07	44:00:31.6	9.8	1.11	−0.5	9.19	0.10	unbar	smooth	⋯
8464−1902	12:25:36.21	45:05: 1.9	10.9	2.31	−0.2	9.07	0.96	unbari	smoothi	⋯
8482−1902	16:12:24.62	48:31:47.6	10.8	3.92	−0.2	9.21	0.36	bar	disk	110
8552−1902	15:07:50.95	43:31:21.7	10.1	1.53	−1.2	9.08	0.49	unbar	smooth	85
8263−3701	12:19: 3.44	46:06:24.5	10.0	1.82	−0.3	9.18	0.13	unbar	smooth	⋯
8332−3701	13:55: 5.68	42:56:50.5	11.0	3.38	⋯	9.15	0.94	unbari	smoothi	48
8550−3701	16:26:36.65	39:51:31.2	10.3	2.00	0.2	9.23	0.34	unbar	smooth	41
8554−3701g	12:12: 1.90	35:24:15.8	10.0	1.98	−0.4	9.11	0.27	unbar	smooth	⋯
7815−3702	21:11:36.77	11:29:49.0	10.3	2.11	0.3	9.16	0.32	unbari	diski	⋯
8244−3702	08:47:15.60	51:14:44.9	10.1	2.61	0.2	9.12	0.12	unbar	disk	⋯
8249−3702	09:08: 7.84	45:55:15.5	9.9	1.71	−0.1	8.96	0.44	unbar	smooth	⋯
8312−3702	16:28:58.87	41:15:18.9	10.2	2.80	−1.1	9.15	0.31	unbar	smooth	70
8143−3703	08:02:33.56	42:23:33.7	10.2	2.25	−0.9	8.92	0.43	unbar	smooth	⋯
8249−3703	09:18:52.91	45:43:40.0	10.0	2.45	−0.3	9.10	0.29	unbar	smooth	160
8253−3703	10:29:22.46	43:10:14.1	10.1	1.99	0.1	9.12	0.01	unbar	smooth	⋯
8257−3703	11:06:37.37	46:02:19.6	10.4	2.17	0.9	9.26	0.36	bar	disk	26
8606−3703	17:03:23.63	37:37:56.9	10.5	2.44	−4.3	9.16	0.42	unbar	smooth	60
8083−3704	03:25: 8.34	−00:32:41.8	9.9	1.65	−2.8	9.20	0.88	unbar	smooth	⋯
8252−3704g	09:41:13.95	47:41:19.0	10.5	2.71	0.7	9.23	0.43	unbar	smooth	⋯
8259−3704	11:58:22.39	43:48:55.0	10.9	3.54	0.7	9.21	0.98	unbar	smooth	⋯
8263−3704	12:26:13.49	45:32:19.4	9.9	2.29	−2.1	9.09	0.97	unbar	smooth	⋯
8332−3704e,g	13:58:33.55	41:43:27.1	10.1	1.80	−0.5	9.14	0.18	⋯	smooth	178
8453−3704	10:17:55.40	46:36:11.8	10.1	2.25	0.4	9.12	0.49	unbar	smooth	⋯
8554−3704	12:18:29.57	35:37:20.1	10.6	3.32	0.3	9.21	0.78	unbar	smooth	81
8618−3704	21:15:26.95	09:45:28.1	10.9	3.17	0.7	9.24	0.94	unbari	smoothi	⋯
7990−6101	17:24: 9.48	56:52:28.1	9.9	2.52	−1.4	9.13	0.19	unbar	smooth	47
8138−6101	07:43:51.36	44:43:27.6	10.3	2.53	−0.0	9.17	0.10	unbar	smooth	⋯
8249−6101	09:10:14.99	46:17:35.8	10.3	2.76	0.1	9.15	0.27	bar	disk	⋯
8250−6101	09:15: 0.76	42:01:27.8	10.4	2.75	1.0	9.14	0.58	bar	disk	60
8320−6101	13:46:30.61	22:42:21.7	10.4	2.88	−0.2	9.22	0.28	unbar	disk	48
7443−6102	15:26:25.61	42:21:14.5	10.5	2.92	⋯	9.14	⋯	unbari	diski	⋯
8312−6102	16:25:38.67	40:05:55.8	10.7	2.32	−0.3	9.11	0.50	unbar	disk	105
8318−6102	13:08:57.44	45:54:19.6	11.5	9.76	⋯	9.04	0.48	unbar	smooth	⋯
8326−6102h	14:20: 4.30	47:07:16.8	11.1	4.27	⋯	9.25	0.55	bar	disk	⋯
8439−6102	09:31: 6.76	49:04:47.1	10.8	3.29	1.3	9.22	0.42	bar	disk	⋯
8618−6102h	21:16:24.77	10:16:24.2	10.3	1.72	−1.6	8.99	0.33	unbari	smoothi	⋯
8082−6103	03:19: 7.72	00:57:21.5	10.0	2.55	0.0	9.01	0.43	unbar	disk	175
8261−6103	12:06:11.03	45:08:57.1	11.1	3.90	0.6	9.24	0.61	unbar	disk	⋯
8448−6103	11:06:31.53	22:46:39.5	10.2	1.81	−0.7	9.01	0.46	unbar	disk	146
7815−6104h	21:16:46.34	11:02:37.5	10.8	7.55	⋯	9.01	0.40	unbari	diski	⋯
7990−6104	17:26:25.83	58:35:19.9	10.3	3.37	0.4	9.01	0.36	unbar	disk	113
7992−6104	17:01: 7.08	64:40:36.8	10.3	3.37	0.4	9.20	0.13	bar	disk	⋯
8250−6104	09:21:38.74	43:43:34.1	10.7	4.25	⋯	9.21	0.00	unbari	diski	⋯
8548−6104	16:22:59.37	46:40:31.2	10.4	2.75	0.5	9.18	0.21	bar	smooth	⋯
7990–9101	17:19: 1.32	57:10:24.6	10.1	3.03	−0.6	9.18	0.03	bar	disk	150
8481–9101	15:44:36.61	54:12:33.6	11.0	5.65	−1.2	9.27	0.81	unbar	smooth	⋯
8549–9101	16:09: 6.35	46:40:16.3	11.3	7.69	−0.9	9.09	0.76	unbar	smooth	155
8243–9102	08:43:17.22	52:45:28.5	11.5	7.85	−3.4	9.22	0.60	unbar	smooth	95
8465–9102	13:12:45.40	46:56: 6.0	10.5	4.15	0.1	9.26	0.32	unbar	disk	73
8550–9102	16:28:50.17	39:50: 6.3	10.7	5.51	0.5	9.16	0.04	unbar	disk	⋯
8604–9102f	16:25:49.28	40:20:42.8	10.2	3.40	0.9	8.44	0.45	unbar	disk	59
8606–9102f	17:02:50.17	36:42:24.3	10.1	3.64	0.6	8.46	0.20	bar	disk	91

Notes. Properties of targeted galaxies. From left to right, the columns represent galaxy ID, right ascension, declination, base-10 logarithm of stellar mass, half-light radius, base-10 logarithm of SFR, gas-phase metallicity, bulge-to-total ratio, hosting a bar or not, disklike or smooth, and projected distance to the nearest neighbor.

^a ID of the galaxies, composed of the plate ID and IFS bundle ID. ^bBulge-to-total fraction based on the SDSS g band (Simard et al. 2011). ^cBar or unbar and disklike or smooth from Galaxy Zoo 2 (Willett et al. 2013). ^dProjected distance from the nearest neighbor. Values of D_p greater than 200 kpc are not shown. ^eHighly inclined galaxies with an inclination angle greater than 75° (Blanton et al. 2005). We note that two of them are classified as smooth, suggesting that the morphological classification in the galaxy zoo may not be reliable for individual sources. However, it is fine to use the catalog to do statistics. ^fOutliers with very low metallicity in the bottom right panel of Figure 3. ^g Counter-rotators with kinematically misaligned gas and stars in Chen et al. (2016). ^h Seyfert galaxies classified by Baldwin, Phillips, and Terlevich (BPT) diagram (Baldwin et al. 1981). ⁱ Visual classifications for galaxies not matched in the galaxy zoo catalog.

Download table as:  ASCIITypeset images: 1 2

For comparison, a volume-limited sample selected from the NSA (Blanton et al. 2011) is shown by the background contours in all panels of Figure 3 except the bottom right one. The volume-limited sample was selected to have an r-band magnitude brighter than −19.5 and a redshift between 0.01 and 0.055. We also present the distributions of the four parameters for the three samples (except for the gas-phase metallicity distribution of the RC sample). According to the distributions shown in Figure 3, we find significant differences in the size, concentration, SFR, and gas-phase metallicity distributions of the targeted sample and the two control samples. We perform Kolmogorov–Smirnov (K-S) tests to quantify the significance, and the K-S test probabilities are denoted in the top right corner of each panel. The blue ones are for the K-S probability between the targeted and BC samples, and the red ones are for the K-S probability between the targeted and RC samples.

The stellar masses and sizes in Figure 3 are from the NSA. The size is the half-light radius of the SDSS r-band image. The concentration is defined as ${R}_{90}/{R}_{50}$, where R₉₀ and R₅₀ correspond to the radii containing 90% and 50% of the light in the SDSS r band. The SFR is the global SFR from Chang et al. (2015), which is computed by the spectral energy distribution (SED) fittings with SDSS and WISE broadband fluxes.

The mass–size relation of galaxies has been well investigated in recent literature (e.g., Shen et al. 2003; McIntosh et al. 2005; Cappellari et al. 2013). Quiescent and SF galaxies have different mass–size relations, suggesting the galaxy structure evolves along with star formation cessation. In the top panels of Figure 3, the targeted galaxies broadly lie between the two control samples. This indicates that galaxies with the outside-in assembly mode have structural properties between those of normal SF galaxies and those of quiescent galaxies. We have checked the fraction of smooth and disklike galaxies in the targeted sample and two control samples by using the data from Galaxy Zoo 2 (Willett et al. 2013). We find an increasing fraction of smooth galaxies from the BC sample (36%), to the targeted sample (54%), to the RC sample (86%). In addition, the targeted galaxies have a median bulge fraction (Simard et al. 2011) of 0.36, which is higher than that of the BC galaxies (0.26) and lower than that of the RC galaxies (0.57). These agree well with the results in the top panels of Figure 3.

The bottom left panel of Figure 3 shows that the targeted galaxies tend to have a higher global SFR than the BC galaxies for about 0.30 dex. The majority of the targeted galaxies with $\mathrm{NUV}-r$ < 4 are above the median star formation main sequence from Chang et al. (2015). This is in contrast to the general knowledge that more-concentrated galaxies tend to be less active in forming stars (Bell et al. 2012; Fang et al. 2013; Bluck et al. 2014; Barro et al. 2017), suggesting that the targeted galaxies are indeed an unusual population with respect to normal SF galaxies.

The mass–metallicity relation is one of the basic relations in the galaxy formation and evolution field, which has been studied in many works (e.g., Lequeux et al. 1979; Garnett & Shields 1987; Tremonti et al. 2004; Gallazzi et al. 2005; Panter et al. 2008; Sánchez et al. 2013; Lian et al. 2015). The bottom right panel of Figure 3 shows that the targeted galaxies are more metal-rich than the BC galaxies. The oxygen abundance used in this panel is the Hα flux-weighted metallicity for all spaxels based on MaNGA data. For each spaxel, the oxygen abundance is calculated using the N2O2 indicator, adopting log₁₀(O/H) + 12 = log₁₀[1.54020 + 1.26602 × R + 0.167977 × DR²] + 8.93 where $R\,={\mathrm{log}}_{10}[{\rm{N}}\,{\rm{II}}]/[{\rm{O}}\,{\rm{II}}]$ (Dopita et al. 2013). Zhang et al. (2017) proposed that the N2O2 indicator is optimal in measuring metallicity for diffused ionized regions because it is not sensitive to the ionization parameter (Dopita et al. 2000, 2013; Kewley & Dopita 2002). As shown in this panel, the median gas-phase metallicity of the targeted sample is about 0.12 dex higher than that of the BC galaxies.

There are three targeted galaxies with dramatically low gas-phase metallicities of $12+{\mathrm{log}}_{10}({\rm{O}}/{\rm{H}})\lt 8.6$, which is about 0.4 dex lower than the median relation from Tremonti et al. (2004). One of them is an edge-on galaxy, and the other two show blue clumpy structures in the SDSS color image. Their low gas-phase metallicities suggest external gas acquisition. In contrast, the cold gas in other targeted galaxies may be pre-existing gas, rather than newly accreted gas from cosmic web, and the metal abundance is enriched by the recent/ongoing star formation activity (Chen et al. 2016).

In addition to the global properties investigated in the previous section, the IFS survey enables us to study the resolved properties of our samples. Figure 4 shows the radial profiles of the 4000 Å break (left panel), Hδ_A index (middle panel), and ${\mathrm{log}}_{10}\mathrm{EW}({\rm{H}}\alpha )$ (right panel) for these three samples. In each panel, the blue, green, and red lines show the median profile for the BC sample, targeted sample, and RC sample, respectively. The shaded regions correspond to the 20%–80% range. In presenting the radial profile, we scale the radii with the effective radius of the SDSS r band.

Zoom In Zoom Out Reset image size

For an individual galaxy, the radial profile of the 4000 Å break is computed in the following way. We separate all spaxels with the continuum S/N > 5 at 5500 Å into a set of radial intervals with a constant width of ${\rm{\Delta }}{\mathrm{log}}_{10}(R/{{R}}_{{\rm{e}}})=0.2$. At a given radial bin for an individual galaxy, the value of the 4000 Å break profile is estimated by the median D_n(4000) of the spaxels falling in that radial bin. In this process, we have corrected the inclination effect for each spaxel based on the minor-to-major axis ratio from the NSA (Blanton et al. 2011). Throughout the paper, we calculate the radial profiles in the way described above for all referred parameters.

As expected, the targeted and BC samples have lower D_n(4000), higher EW(Hδ_A), and higher ${\mathrm{log}}_{10}\mathrm{EW}({\rm{H}}\alpha )$ than the RC sample as a whole. The BC sample has negative D_n(4000) gradients and positive EW(Hδ_A) and ${\mathrm{log}}_{10}\mathrm{EW}({\rm{H}}\alpha )$ gradients within the effective radius. The radial profiles of the BC sample become flat when the radii are greater than ${R}_{{\rm{e}}}$. In contrast, the radial profiles of the targeted sample show slightly positive gradients of D_n(4000), slightly negative gradients of EW(Hδ_A), and pronounced negative gradients of ${\mathrm{log}}_{10}\mathrm{EW}({\rm{H}}\alpha )$. The gradients of these three parameters still exist when the radius is beyond ${R}_{{\rm{e}}}$. Figure 4 confirms that the targeted galaxies have stronger star formation activity in their centers than in their outer regions. This is in contrast to the previous findings of an inside-out assembly mode in massive SF galaxies (e.g., Muñoz-Mateos et al. 2007; Bezanson et al. 2009; Pérez et al. 2013; Pan et al. 2015; Tacchella et al. 2015). The targeted galaxies are very interesting, since they are associated with rapid central stellar mass assembly (or bulge growth). Investigating their properties would probably provide clues about the processes of bulge growth and star formation cessation (see further discussion in Section 4.3).

Figure 5 shows the median radial profiles of stellar surface mass density (${{\rm{\Sigma }}}_{* }$, left panel) and gas-phase metallicity (right panel) for our samples. The shaded regions correspond to the 20%–80% range. The stellar mass maps are the outputs from the STARLIGHT fittings (see Section 2.2). As shown in the left panel of Figure 5, the RC sample has the highest stellar surface density, and the BC sample has the lowest stellar surface density as a whole. The median ${{\rm{\Sigma }}}_{* }$ profiles of the three samples are nearly parallel in logarithmic space. The targeted sample has a higher median ${{\rm{\Sigma }}}_{* }$ profile than the BC sample for about 0.22 dex and a lower one than the RC sample for about 0.39 dex. This result is consistent with the results in Figure 3, indicating that the structural properties of the targeted sample are between the two control samples.

Zoom In Zoom Out Reset image size

The right panel of Figure 5 shows the gas-phase metallicity profile for the BC sample (blue line) and targeted sample (green line). As shown, the targeted sample is systematically more metal-rich than the BC sample for ∼0.07 dex within 1.5${R}_{{\rm{e}}}$. The median metallicity profile of the BC sample is slightly steeper than that of the targeted sample. The BC sample has a median metallicity gradient of −0.11 dex/${R}_{{\rm{e}}}$, which is slightly steeper than that of the targeted sample (−0.08 dex/${R}_{{\rm{e}}}$). The gradients measured in this work are comparable with the previous measurements of −0.1 dex/${R}_{{\rm{e}}}$ by Sánchez et al. (2014).

Recently, the ${{\rm{\Sigma }}}_{* }$ and local gas-phase metallicity have been found to be closely related, with a higher ${{\rm{\Sigma }}}_{* }$ indicating a higher metallicity (Barrera-Ballesteros et al. 2016; Zhu et al. 2017). Thus, one may ask whether the higher metallicity of the targeted sample is only due to the higher stellar surface density. To answer this question, we plot ${\mathrm{log}}_{10}{{\rm{\Sigma }}}_{* }$ versus metallicity for all spaxels in the BC sample (blue contours) and targeted sample (green contours) in Figure 6, excluding the spaxels with an S/N of [O ii] λ3727, Hβ, Hα, or [N ii] λ6584 flux less than 3. In Figure 6, the blue and green data points represent the median relations for the BC sample and the targeted sample, respectively. The targeted galaxies are more concentrated than the BC galaxies and have a larger stellar surface density as a whole. The targeted sample is more metal-rich than the BC sample at the high-${{\rm{\Sigma }}}_{* }$ end (${{\rm{\Sigma }}}_{* }\ \gt \ {10}^{8.5}$ ${M}_{\odot }$ kpc⁻²). The metallicities of these two samples become comparable when ${{\rm{\Sigma }}}_{* }\ \lt$ 10⁸ ${M}_{\odot }$ kpc⁻². We find that the inner regions of the targeted galaxies are more metal-rich than the inner regions of the BC galaxies, even at a fixed stellar surface mass density. With respect to the BC galaxies, the higher central oxygen abundance in the targeted galaxies may be due to enhancement from ongoing intense star formation activity, as proposed by Chen et al. (2016). By using the highly resolved IFS data of HGC 91c, Vogt et al. (2017) found that the metal enhancement has proceeded preferentially along the spiral arms and less efficiently across the arms, suggesting that the metal enhancement and intense star formation activity are associated.

Zoom In Zoom Out Reset image size

We have found that the structural and radial star formation properties of the targeted sample are significantly distinct from those of the BC sample. In this section, we investigate the environment of our samples and try to find out whether they reside in the same environment.

We use the projected two-point cross-correlation function (2PCCF; ${w}_{p}({r}_{p})$) to quantify the clustering properties of our samples. Here we briefly describe the method of calculating 2PCCFs; a full description of the methodology is presented in Li et al. (2006). The spectroscopic reference sample is from version dr72 of the NYU Value-Added Galaxy Catalog (NYU-VAGC). It is a magnitude-limited sample of about half a million galaxies with an r-band Petrosian apparent magnitude corrected for galactic extinction $r\lt 17.6$, an absolute magnitude range of $-23\lt {M}_{0.1r}\lt -17$, and redshifts in the range $0.01\lt z\lt 0.2$. Here ${M}_{0.1r}$ is the r-band Petrosian absolute magnitude, corrected for evolution and K-corrected to its value at z = 0.1. In calculating the 2PCCF, we have generated a random sample that has the same selection effects as the reference sample (Li et al. 2006; Lin et al. 2014). We cross-correlate the targeted sample and two control samples with respect to the reference and random samples, then define ${w}_{p}({r}_{p})$ as the ratio of the two pair counts minus one in the projected separation r_p. The effect of fiber collisions is carefully corrected following the method in Li et al. (2006).

The top panel of Figure 7 shows the 2PCCFs measured for the BC sample (${w}_{p}^{\mathrm{BC}}$; blue line), targeted sample (${w}_{p}^{{\rm{T}}}$; green line), and RC sample (${w}_{p}^{\mathrm{RC}}$; red line). For clarity, we also include the amplitude ratio, ${w}_{p}^{\mathrm{BC}}/{w}_{p}^{{\rm{T}}}$ and ${w}_{p}^{\mathrm{RC}}/{w}_{p}^{{\rm{T}}}$, in logarithmic space in the bottom panel. The errors in the ${w}_{p}({r}_{p})$ measurements are computed by using the bootstrap resampling technique (Barrow et al. 1984). The data points of ${w}_{p}({r}_{p})$ are not shown on scales smaller than 30 kpc, because they have very large uncertainties due to the limitation of the sample sizes.

Zoom In Zoom Out Reset image size

As expected, the RC galaxies are more strongly clustered than the BC galaxies at a large range of scales from 30 kpc to 3 Mpc. As a whole, the clustering amplitude of the targeted sample is between that of the BC sample and that of the RC sample. The targeted galaxies are more strongly clustered than the BC galaxies at almost the whole scale we considered, especially at 0.8–2 Mpc (0.3–0.5 dex higher). However, the targeted sample appears to be more weakly clustered than the RC sample, especially at the scale of 50–300 kpc (0.3–0.5 dex lower). Assuming that the environment effect is the main driver for the outside-in mode in the targeted galaxies, it is difficult to understand why the environment at a large scale (∼1 Mpc) could have an effect on the SF properties of the galaxies. At a scale of 30 kpc, the targeted sample appears to be more strongly clustered than the BC sample. However, we do not overinterpret the results in Figure 7 because of the limitation of the sample size.

In addition to 2PCCFs, we also use the three-dimensional reconstructed local mass density ($\delta =\rho /\bar{\rho }$, where ρ is the local matter density and $\bar{\rho }$ is the mean matter density) to quantify the environment for each galaxy in our sample. The mass density field is taken from the Exploring the Local Universe with the reConstructed Initial Density Field project (ELUCID; Wang et al. 2016) and matches well with the galaxy and group distribution (Yang et al. 2007) of the SDSS Data Release 7 (Abazajian et al. 2009). The local mass density we use is the smoothed density by a Gaussian kernel with a scale of 2 Mpc h⁻¹ at each galaxy position.

Figure 8 shows the density distribution of the BC sample (blue histogram), targeted sample (green filled histogram), and RC sample (red histogram). All three samples have broad density distributions. To quantify the significance of the distribution difference between the targeted sample and the two control samples, we present the K-S test probabilities in the top right corner. The K-S test probabilities suggest that significant differences between the density distribution of the targeted sample and the two control samples are not found. We note that about 17% of the RC galaxies, 11% of the targeted galaxies, and 2% of the BC galaxies reside in the dense regions with $\rho /\bar{\rho }\sim 100$.

Zoom In Zoom Out Reset image size

The results of Figures 7 and 8 suggest a question as to whether the large distinction of clustering amplitude between the targeted and BC samples at a scale of 0.8–2 Mpc is mainly contributed by the 11% of targeted galaxies residing in big clusters. To answer this question, we have checked the clustering properties of the targeted and BC samples while excluding the galaxies residing in the most massive halos (${M}_{h}\gt {10}^{14}\,{M}_{\odot }$). The result shows that the clustering amplitude of the targeted galaxies has been reduced to slightly higher than (or comparable with, if considering the errors) that of the BC sample at a scale of 0.8–2 Mpc. However, the targeted galaxies still appear to be more strongly clustered than the BC sample at a scale of 30 kpc.

Furthermore, we have calculated the projected separation of the nearest neighbor for the targeted and BC galaxies with velocity differences less than 500 km s⁻¹. In this process, we ignore the spectroscopic incompleteness effect, since we just care about the comparison of the companion fraction between the targeted and BC galaxies. We find that 19 (∼25%) targeted galaxies and 14 (∼10%) BC galaxies have one or more companions within 100 kpc. This suggests that interaction with close companions is likely to be one of the drivers of the outside-in assembly mode. Since the majority of the targeted galaxies do not have companions within 100 kpc, it appears that this mechanism is not the main driver. We will discuss the possible mechanisms for the formation of the targeted galaxies in Section 4.2.

The IFS survey enables us to select a sample of massive SF galaxies with a recent outside-in stellar mass assembly mode. They are selected to have ${D}_{n}{(4000)}_{1.5{R}_{{\rm{e}}}}$ > ${D}_{n}{(4000)}_{\mathrm{cen}}$ and account for 14% of the total massive SF galaxies ($\mathrm{NUV}-r$ < 5 and ${M}_{* }\ \gt {10}^{9.8}\,{M}_{\odot }$). For comparison, we have built two control samples matched in stellar mass. On average, the targeted galaxies appear to have smaller sizes and higher concentration, SFR, and oxygen abundance than normal SF galaxies.

We have generated D_n(4000), EW(Hδ_A), and ${\mathrm{log}}_{10}\mathrm{EW}({\rm{H}}\alpha )$ maps and profiles for our samples. The inside-out mass assembly scenario has been found in observations of massive SF galaxies (e.g., Sánchez-Blázquez et al. 2007; Pérez et al. 2013; Ibarra-Medel et al. 2016). However, the targeted galaxies appear to have significantly different two-dimensional mass assembly patterns from those of normal SF galaxies. They have lower D_n(4000) and higher ${\mathrm{log}}_{10}\mathrm{EW}({\rm{H}}\alpha )$ in their inner regions than their outer regions, suggesting that they host younger stellar populations and more intense star formation activity in the inner regions than the outer regions. The higher metallicities in the inner regions of the targeted galaxies are likely due to the enhancement of recent intense star formation activity.

Lin et al. (2017) identified a sample of 17 "turnover" galaxies from 57 nearly face-on spiral CALIFA galaxies whose central regions exhibit a significant drop in D_n(4000). The central regions of these galaxies have experienced star formation in the past 1–2 Gyr. Nearly all of the turnover galaxies are barred, suggesting a bar-induced star formation scenario in the galaxy centers (e.g., Hawarden et al. 1986; Ho et al. 1997; Wang et al. 2012; Cole et al. 2014). The targeted galaxies in this work are different from their population, because the D_n(4000) profiles of the targeted galaxies do not show turnover features at galactic center. We have checked the bar fraction of the targeted galaxies and found that about 19% of them appear to have a bar-like structure (Willett et al. 2013). This indicates that the existence of a bar is not mainly responsible for the outside-in mass assembly mode in the targeted galaxies.

From 489 SF galaxies of the MaNGA sample, Chen et al. (2016) identified nine counter-rotators, whose gas and stars are kinematically misaligned. They also have a positive slope of their D_n(4000) profiles and more intense, ongoing star formation in their central regions than their outer regions (Chen et al. 2016; Jin et al. 2016). Five of them are selected in our targeted sample, and the other four have a stellar mass that is lower than our selection criteria. Chen et al. (2016) proposed that the formation of SF counter-rotators is due to gas accretion from gas-rich dwarfs or the cosmic web. The counter-rotators tend to reside in more isolated environments (Jin et al. 2016), while our targeted galaxies live in a variety of environments (see Figure 8). This suggests that gas accretion may be one of the mechanisms for the outside-in assembly mode, though it is not known whether it is the main driver.

The targeted galaxies can be formed in two ways according to their abnormal resolved star formation properties. One is that normal SF galaxies suffer from some kind of process, such as interaction or gravitational instability of the disk, leading to gas inflow and triggering central star formation, and subsequently the galaxies with the outside-in assembly mode are formed. The other is that quenched galaxies accrete cold gas and re-form stars in their centers, which can also lead to the observed outside-in assembly picture. Figure 5 shows that the central ${{\rm{\Sigma }}}_{* }$ of the targeted galaxies is between that of the BC galaxies and the RC galaxies. If the targeted galaxies are mainly formed via gas accretion of quenched galaxies, their central ${{\rm{\Sigma }}}_{* }$ should be, on average, comparable with (or higher than) that of the RC galaxies. Thus, we deduce that the progenitors of the targeted galaxies are probably normal SF galaxies, which might later undergo cold gas inflow and morphological transformation.

In Table 1, we list the properties of targeted galaxies, as well as their morphologies and the projected distance to the nearest neighbor. We find that 14 out of 73 targeted galaxies have a bar-like structure (four galaxies with an inclination angle greater than 75° are not included in counting the bar fraction), and 19 out of 77 targeted galaxies have one or more companions within 100 kpc. It is well established that both the existence of the bar and interaction with companions can effectively cause gas inflow and enhance star formation in the galaxy center (Li et al. 2008; Wang et al. 2012), although they do not always work (Lin et al. 2017). Among galaxies with companions, some are located in very massive clusters that may suffer from a variety of environmental processes, such as tidal stripping, ram-pressure stripping, and galaxy harassment (e.g., Moore et al. 1996; Cox et al. 2006; Cheung et al. 2012). However, in our case, both the environmental effect and the existence of the bar are not enough to explain most of the individual galaxies, suggesting that other mechanisms are needed. Specifically, there are 44 (∼60%) out of 73 targeted galaxies that do not have a bar-like structure or companions within 100 kpc. Four of them are the counter-rotators in Chen et al. (2016), which are proposed to undergo gas accretion from gas-rich dwarfs or the cosmic web. It is possible that more targeted galaxies are undergoing the same process, since gas accretion does not always cause the kinematical misalignment of gas and stars.

Recently, in addition to mergers, another channel for bulge growth via disk instabilities was proposed in a semianalytical model (Porter et al. 2014), which brought the model into better agreement with the observed galaxy stellar mass function of spheroid-dominated galaxies. Furthermore, Brennan et al. (2015) confirmed that a model adding the disk instabilities for bulge growth agrees better with the observed galaxy distribution on the speciific SFR (sSFR)–Sérsic index plane than a model in which bulges grow only through mergers. By implementing the bulge growth model with the mergers and disk instabilities into the semianalytical galaxy formation code, Tonini et al. (2016) produced two distinct populations of bulges: merger-driven bulges, which are similar to classical bulges, and instability-driven bulges, which are similar to pseudo-bulges. The instability-driven bulges dominate the population of galaxies with intermediate stellar masses ($10\lt {\mathrm{log}}_{10}({M}_{* }/{M}_{\odot })\lt 11$), which are in the mass range similar to that of our targeted galaxies. Considering the very low major merger rate in the local universe, disk instabilities and minor mergers may be responsible for the bulge growth of the targeted galaxies.

The targeted galaxies are found to assemble their central stellar mass rapidly (see Figure 4), indicating a rapid (pseudo) bulge growth. A massive bulge is believed to be the key factor for quenching star formation for local galaxies (Bell et al. 2012; Fang et al. 2013). Bluck et al. (2014) found that the bulge mass is more critical in quenching star formation than the bulge-to-total stellar mass ratio for local galaxies. More recently, Barro et al. (2017) investigated the evolution of the stellar surface mass density within the central 1 kpc (${{\rm{\Sigma }}}_{1\mathrm{kpc}}$) of SF and quiescent galaxies since a redshift of 3. They found that the growth of a dense stellar core is the necessary condition for star formation quenching.

Figure 9 shows ${{\rm{\Sigma }}}_{1\mathrm{kpc}}$ versus stellar mass for the BC sample (blue crosses), targeted sample (green circles), and RC sample (red triangles). The red line is taken from Fang et al. (2013) for quenched galaxies. For comparison, the ${{\rm{\Sigma }}}_{1\mathrm{kpc}}$ used here is computed following the same procedure as in Fang et al. (2013). The green and blue lines are the best fits of the targeted and BC galaxies. For comparison, a volume-limited sample selected from SDSS is shown by the background contours (the same data set as in Figure 3). We find that the targeted galaxies are broadly located between two control samples on the diagram of ${{\rm{\Sigma }}}_{1\mathrm{kpc}}$–${M}_{* }$. Considering the intense star formation activity at the center, they are undergoing rapid (pseudo) bulge growth and will meet the quenching requirements on a more rapid timescale than normal SF galaxies.

Zoom In Zoom Out Reset image size

It is well known that the bulge mass has a tight relation with the central massive black hole mass (e.g., Daddi et al. 2007; Hu 2008; McConnell & Ma 2013), suggesting the coevolution of bulge growth and black hole accretion. The active galactic nucleus (AGN) feedback is supposed to be one of the mechanisms for quenching star formation in galaxies (e.g., Kauffmann et al. 2004; Schawinski et al. 2007; Dekel & Burkert 2014). We find four (∼5%) targeted galaxies that are classified as Seyfert galaxies according to the BPT diagram (Baldwin et al. 1981; Kauffmann et al. 2003b), while only three (∼2%) BC galaxies are classified as Seyfert galaxies. Because of the limitation of sample size, we do not overinterpret this result.