SHOCKED POSTSTARBURST GALAXY SURVEY. II. THE MOLECULAR GAS CONTENT AND PROPERTIES OF A SUBSET OF SPOGs

We observed the CO(1–0) line at the central position of our sample galaxies on 2014 September 9–14, October 15–18, and November 6–9 with the IRAM 30 m telescope on Pico Veleta. We used the dual-polarization Eight MIxer Receiver (emir; Carter et al. 2012) in combination with the autocorrelator Fourier Transform Spectrometers at a frequency resolution of 0.195 MHz at CO(1–0) (providing a velocity resolution of 0.57 km s⁻¹). The observations were done in wobbler switching mode with a wobbler throw of 240'' in the azimuthal direction.

The broad bandwidth of the receiver (16 GHz) and backend allowed us to group the observations of galaxies with similar redshifts. The central sky frequencies, taking into account the redshift of the objects, ranged between 96 and 111 GHz. Each object was observed until it was detected with an S/N of at least 5 or until an rms of 1.5 mK (${{\rm{T}}}_{{\rm{A}}}^{* }$) was achieved for a velocity resolution of 20 km s⁻¹. The integration times per object ranged between 0.5 and 2 hr, with a mean value of 80 minutes. Pointing was monitored on nearby quasars every 60–90 minutes. During the observation period, the weather conditions were generally good, with a pointing accuracy better than 4''. The mean system temperature for the observations was 116 K on the ${T}_{{\rm{A}}}^{* }$ scale. At 115 GHz, the IRAM forward efficiency, ${F}_{{\rm{eff}}}$, was 0.95; the beam efficiency, ${B}_{{\rm{eff}}}$, was 0.79; and the half-power beam size ranges between 223 (for 110 GHz) and 256 (for 97 GHz). All CO spectra and luminosities are presented on the Jansky scale, converted from the main-beam temperature scale (${T}_{{\rm{mb}}}$), which is defined as ${T}_{{\rm{mb}}}=({F}_{{\rm{eff}}}/{B}_{{\rm{eff}}})\times {T}_{{\rm{A}}}^{* }$, using a conversion factor of 5 Jy K⁻¹.

For the data reduction, we first discarded poor scans and then subtracted a constant or linear baseline. A large number of scans were affected by platforming, i.e., the baseline level changed abruptly at one or two positions along the band. This effect could be reliably corrected because the baselines in between these (clearly visible) jumps were flat and allowed to determine the (order 0 or 1) baseline that had to be subtracted from the different parts in order to move the baselines to a zero level along the entire band. After this correction, we summed the spectra of each source, smoothed them to resolutions between 20 and 40 km s⁻¹ in order to increase the S/N per channel, and visually determined the zero-level widths (the boundaries beyond which the spectra drop to zero). CO-SPOGs with S/N > 3 are considered detected, and CO-SPOGs with S/N > 5 are considered strongly detected. Figure 4 shows the integrated spectra of the 35 SPOGs observed with the IRAM 30 m (shaded turquoise). The systemic velocity of the line is set to zero, and the total line width is represented by a turquoise line and is shaded in the spectrum. The optically defined systemic velocity is shown as a black dotted line. The velocity-integrated spectra were calculated by summing the individual shaded channels and multiplying by the velocity width of each channel (21 or 42 km s⁻¹).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

As an alternative approach, we fit the spectra with a Gaussian profile and integrated it over velocity (details of the Gaussian fitting can be found in Appendix). The velocity-integrated intensities determined by the two methods are in good agreement, with a mean difference of 5%, confirming that our velocity integration is reliable. The line flux rms was then calculated by multiplying the rms per channel by the channel velocity width and the square root of the total number of channels determined to contain line emission (those that are shaded).

The CARMA SPOG observations were taken between 2014 June and December using CARMA, an interferometer of 15 radio dishes (6 × 10.4 and 9 × 6.1 m) located in the Eastern Sierras in California (Bock et al. 2006).¹¹  A total of 19 SPOGs were observed (with a 100% detection rate), in either D- or E-array, with baselines in the range 11–150 m and 8–66 m, respectively. Standard reduction and calibration techniques (as described in detail in Alatalo et al. 2013) were used on all targets.

Our SPOGs were unresolved in all cases except J1315+2437 (IC 860, which was also observed in CARMA C-array by McBride et al. 2014). The observing parameters associated with each of the CARMA SPOGs are listed in Table 2. A moment0 map was created for each CARMA SPOG (using the miriad task moment; Sault et al. 1995), in which a sigma clip was applied to velocity channels determined to have emission (see Alatalo et al. 2013 for details). We did not apply a standard sigma clip to the channel maps, instead iteratively determining the correct sigma value to maximize the detection of real emission to the exclusion of noise in each individual object. Figure 4 shows the integrated spectra of the 19 SPOGs observed with CARMA (shaded green). The systemic velocity of the line is set to zero, with the total line width represented by a green line, as well as in the shaded region of the spectrum. The integrated spectrum of each SPOG is determined by integrating the flux within an aperture determined by the moment0 map created for each CARMA SPOG.

Table 2.  CARMA Observational Parameters

	SPOG	Total	Gain	${\theta }_{{\rm{maj}}}\times {\theta }_{{\rm{\min }}}$	${\rm{\Delta }}v$	rms	Area
	Name	Hrs	Calibrator	(arcsec)	(km s−1)	(mJy bm−1)	($\square ^{\prime\prime}$)
	(1)	(2)	(3)	(4)	(5)	(6)	(7)
186	J0914+3753	2.98	0927+390	$10.1\times 6.3$	20	1.4	222
191	J0918+4200	5.44	0927+390	$5.5\times 4.4$	50	1.0	19
200	J0925+0623	3.16	0825+031	$8.4\times 6.9$	20	1.4	142
224	J0938+1819	2.11	0956+252	$7.7\times 7.0$	20	1.9	324
253	J0957−0012	3.35	1058+015	$8.7\times 7.4$	20	1.3	125
267	J1008+0936	2.44	1058+015	$7.1\times 6.6$	20	1.2	71
305	J1026+4340	2.17	0927+390	$9.2\times 6.5$	20	1.5	242
437	J1126+1913	1.85	1159+292	$9.9\times 6.4$	20	1.7	290
439	J1127+1256	3.37	1058+015	$8.4\times 7.3$	40	2.7	67
462	J1136+2453	0.82	1224+213	$4.1\times 3.5$	20	4.7	36
470	J1139+4631	3.63	1153+495	$8.4\times 6.9$	20	0.8	180
498	J1153+0930	6.69	3C273	$9.3\times 6.8$	20	0.6	133
578	J1229+3224	5.82	1159+292	$9.9\times 6.7$	20	0.9	245
662	J1314+2106	4.02	1310+323	$7.9\times 6.1$	20	1.4	232
663	J1315+2437	4.97	1310+323	$2.6\times 2.1$	20	15.1	134
689	J1326+1922	2.84	1310+323	$8.2\times 7.5$	40	0.9	201
704	J1336+3008	4.99	1310+323	$7.2\times 5.9$	30	1.3	172
711	J1339+4422	2.41	1419+543	$8.5\times 6.1$	20	2.2	343
745	J1356+2816	5.03	1310+323	$8.5\times 6.5$	20	0.8	48

Note. Column (1): SPOG name. Column (2): total time on-source. Column (3): gain calibrator used. Column (4): CARMA beam FWHM. Column (5): channel velocity width. Column (6): intensity rms per beam of CARMA images. Column (7): moment0 aperture area.

Download table as:  ASCIITypeset image

The rms per channel is calculated by (1) taking the standard deviation of all pixels within the cube that were outside the moment0 aperture, (2) applying an additional noise up-correction of 30% to account for the oversampling of the maps (see Alatalo et al. 2015b for details), and (3) multiplying by the square root of the total number of beams represented in the moment0 aperture.

To calculate the integrated line flux for each galaxy, we summed the channels shaded in green in Figure 4 and multiplied by the velocity width of each channel, listed in Table 2. The line flux rms was then calculated by multiplying the rms per channel by the channel velocity width and the square root of the total number of channels determined to contain line emission. In the case of the two SPOGs observed by both the IRAM 30 m and CARMA, the CO fluxes agreed to within the standard 20% errors.

Figure 3 show the three-color gri images from SDSS of all 52 CO-SPOGs. The thumbnails show a large number of galaxies that include signs of interaction, though only a handful appear to be major mergers in the coalescence phase (Dopita et al. 2002). A significant fraction of CO-SPOGs have tidal features prominent enough to be seen in SDSS images. The remaining sample of CO-SPOGs mainly consists of early-type spirals, lenticular galaxies with bars, and objects with peaked nuclei consistent with AGNs. Our selection did not include morphology, so it is interesting that the number of tidally disrupted objects represented in CO-SPOGs is larger than that in the general SPOG sample.

In order to quantify whether the CO-SPOG galaxies are disrupted, six team members (Alatalo, Appleton, Cales, Lanz, Lisenfeld, Nyland) independently visually inspected the gri thumbnails of these galaxies (Figure 3). Based on the (potential) presence of tidally features, they classified the galaxies as (possibly) disrupted or not. There was complete consensus of the classification of 47 galaxies, indicated by the presence or absence of a check mark in Table 1, as well as a ($\infty$) symbol in Figure 3. The other five galaxies had at least two classifications differing from the others and are therefore marked as ambiguous, represented as a question mark in Table 1 and ($\sim$) in Figure 3. Based on these classifications,¹²  19–24 (37%–46% ± 7%) of our CO-SPOGs are classified as disrupted. A detailed analysis of the morphologies of SPOGs will be presented in a future paper.

Figure 4 shows the CO(1–0) spectra of all CO-SPOGs, color-coded to identify which facility was used to make the observation (turquoise for the IRAM 30 m and green for CARMA). A few double-horned spectral profiles, consistent with molecular gas rotating in a disk that extends out to the flat part of the rotation curve, are present but are the minority of detections (although this could in part be due to sensitivity). In many of the strong molecular gas detections, multiple components and peaks are present in the molecular gas, consistent with the morphological disruption present in gri images of the galaxies.

Table 2 lists the derived properties for CARMA-observed CO-SPOGs, including the rms noise in the channel maps, the spatial extent of the molecular gas (determined by summing the total number of unmasked pixels in the moment map; see Alatalo et al. 2013 for details of moment map construction), channel widths, and total on-source hours. Derived molecular gas properties are listed in Table 3. CO luminosities are derived using the equation in Solomon & Vanden Bout (2005):

$$\begin{eqnarray}&&{L}_{{\rm{CO}}}=1.20\times {10}^{-1}\,\displaystyle \frac{{S}_{{\rm{CO}}}{\rm{\Delta }}v\,{D}_{{\rm{L}}}^{2}}{1+{v}_{{\rm{sys}}}/c}\,{L}_{\odot },\end{eqnarray} \tag{ 1 }$$

where ${S}_{{\rm{CO}}}{\rm{\Delta }}v$ is the CO(1–0) flux (in Jy km s⁻¹), ${D}_{{\rm{L}}}$ is the luminosity distance (in Mpc), ${v}_{{\rm{sys}}}$ is the optically defined systemic velocity (in km s⁻¹), and c is the speed of light (in km s⁻¹). Converting the CO(1–0) luminosity into molecular gas masses requires the assumption of a ${L}_{{\rm{CO}}}\mbox{--}{M}_{{{\rm{H}}}_{2}}$ conversion factor (Bolatto et al. 2013), which is known to be dependent on the state of the molecular gas. Interacting galaxies and ultraluminous infrared galaxies (ULIRGs) are known to have ${L}_{{\rm{CO}}}\mbox{--}{M}_{{{\rm{H}}}_{2}}$ conversions that are lower than normal, star-forming galaxies by a factor of about 5 (Downes & Solomon 1998), though Narayanan et al. (2011) have shown that there is a large scatter above and below the standard value, even within merging systems. Sandstrom et al. (2013) showed that the nuclei of normal star-forming galaxies can exhibit CO–to–H₂ conversion up to an order of magnitude below the Milky Way value. Determining the proper ${L}_{{\rm{CO}}}\mbox{--}{M}_{{{\rm{H}}}_{2}}$ conversion is therefore essential to deriving the molecular mass of a galaxy.

Table 3.  SPOG CO(1–0) Values

	SPOG	Vel. Range	rms	${{\rm{F}}}_{{\rm{CO}}}$	SNR${}_{{\rm{CO}}}$	${{\rm{L}}}_{{\rm{CO}}}$	log(${M}_{{{\rm{H}}}_{2}}$)	${F}_{{\rm{gas}}}$
	Name	(km s−1)	(mJy)	(Jy km s−1)		(10${}^{4}\,{L}_{\odot }$)	(${M}_{\odot }$)
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
1	J0003+0048	41070–41580	7.20	4.19 ± 0.78	5.4	18.99 ± 3.53	10.22 ± 9.5	0.24
4	J0011−0054	14150–14350	8.80	3.05 ± 0.58	5.3	1.58 ± 0.30	9.14 ± 8.4	0.06
7	J0029+1433	42800–43200	9.05	$\lt 7.80$	—	$\lt 37.58$	$\lt 10.52$	0.40
13	J0037+0024	24100–24400	7.45	$\lt 5.45$	—	$\lt 8.13$	$\lt 9.85$	0.32
24	J0119+1334	57030–57480	8.70	3.94 ± 0.92	4.3	34.41 ± 7.99	10.48 ± 9.8	0.25
77	J0803+2530	40250–40620	7.80	2.86 ± 0.73	3.9	12.28 ± 3.11	10.03 ± 9.4	0.11
81	J0807+2006	19590–20190	7.05	3.79 ± 0.80	4.7	3.79 ± 0.80	9.52 ± 8.8	0.12
98	J0816+1936	33670–33980	6.15	2.01 ± 0.52	3.9	5.98 ± 1.54	9.72 ± 9.1	0.17
142	J0845+2006	36960–37530	10.05	4.79 ± 1.16	4.1	16.95 ± 4.09	10.17 ± 9.6	0.27
157	J0853+0310	38660–39080	7.80	5.42 ± 0.77	7.1	21.15 ± 2.99	10.27 ± 9.4	0.20
169	J0859+1006	16240–16670	7.65	4.15 ± 0.74	5.6	2.80 ± 0.50	9.39 ± 8.6	0.07
186	J0914+3753	21340–21780	1.40	7.48 ± 0.13	57.1	8.83 ± 0.15	9.89 ± 8.1	0.28
191	J0918+4200	12010–12560	1.04	1.70 ± 0.17	9.9	0.64 ± 0.06	8.75 ± 7.8	0.03
200	J0925+0623	22450–22750	1.42	3.30 ± 0.11	29.9	4.30 ± 0.14	9.58 ± 8.1	0.10
209	J0928+0741	31410–31680	6.20	2.47 ± 0.49	5.1	6.32 ± 1.24	9.74 ± 9.0	0.30
224	J0938+1819	26380–26780	1.88	10.40 ± 0.17	61.9	18.78 ± 0.30	10.22 ± 8.4	0.27
253	J0957−0012	9710–9970	1.30	3.75 ± 0.09	39.9	0.92 ± 0.02	8.90 ± 7.3	0.07
267	J1008+0936	8050–8250	1.17	1.88 ± 0.07	25.5	0.31 ± 0.01	8.43 ± 7.0	0.03
268	J1008+1916	54750–55270	6.60	3.21 ± 0.74	4.3	25.31 ± 5.84	10.35 ± 9.7	0.20
270	J1008+5123	46650–47170	6.60	3.84 ± 0.73	5.3	22.21 ± 4.22	10.29 ± 9.6	0.33
293	J1018+1536	32920–33320	5.05	4.12 ± 0.48	8.6	11.70 ± 1.36	10.01 ± 9.1	0.15
305	J1026+4340	31509–31690	1.47	2.51 ± 0.09	28.5	6.44 ± 0.23	9.75 ± 8.3	0.24
308	J1028+5736	21120–21880	10.95	7.32 ± 1.40	5.2	8.93 ± 1.71	9.89 ± 9.2	0.34
322	J1031+0540	48490–48980	9.55	4.66 ± 1.02	4.5	29.46 ± 6.49	10.41 ± 9.8	0.33
349	J1046+2804	38300–38740	8.45	$\lt 2.55$	—	$\lt 9.75$	$\lt 9.93$	0.24
365	J1057+0554	16299–16560	11.05	4.04 ± 0.83	4.9	2.70 ± 0.56	9.37 ± 8.7	0.17
437	J1126+1913	30760–31160	1.67	11.69 ± 0.15	78.1	28.69 ± 0.37	10.40 ± 8.5	0.45
439	J1127+1256	45610–45890	2.73	1.06 ± 0.29	3.7	5.81 ± 1.58	9.71 ± 9.1	0.06
462	J1136+2453	9600–9900	4.71	8.40 ± 0.37	23.0	2.01 ± 0.09	9.25 ± 7.9	0.12
470	J1139+4631	51710–52130	0.77	3.09 ± 0.07	43.9	22.10 ± 0.50	10.29 ± 8.6	0.15
498	J1153+0930	41220–41860	0.60	5.43 ± 0.07	80.3	24.61 ± 0.31	10.33 ± 8.4	0.28
533	J1211+2936	31970–32200	5.50	2.32 ± 0.40	5.9	6.12 ± 1.04	9.73 ± 9.0	0.12
547	J1216+1904	22510–22900	6.30	4.93 ± 0.59	8.4	6.35 ± 0.75	9.74 ± 8.8	0.07
578	J1229+3224	51500–52140	0.88	6.94 ± 0.10	69.3	49.38 ± 0.71	10.64 ± 8.8	0.26
619	J1248+5514	24610–25090	5.20	6.25 ± 0.54	11.7	9.82 ± 0.84	9.93 ± 8.9	0.13
658	J1313+0207	8920–9360	10.45	5.50 ± 1.01	5.5	1.13 ± 0.21	8.99 ± 8.3	0.04
662	J1314+2106	13520–13980	1.37	9.97 ± 0.13	75.9	4.71 ± 0.06	9.62 ± 7.7	0.12
663	J1315+2437	3700–4059	15.14	76.82 ± 1.28	59.8	2.88 ± 0.05	9.40 ± 7.6	0.16
689	J1326+1922	52110–52430	0.88	2.34 ± 0.10	23.5	16.89 ± 0.72	10.17 ± 8.8	0.16
704	J1336+3008	7530–7950	1.35	3.97 ± 0.15	26.3	0.59 ± 0.02	8.72 ± 7.3	0.08
711	J1339+4422	18570–18970	2.23	12.16 ± 0.20	60.9	10.90 ± 0.18	9.98 ± 8.2	0.24
745	J1356+2816	39340–40140	0.77	11.37 ± 0.10	117.4	46.77 ± 0.40	10.61 ± 8.5	0.40
767	J1409+1016	28390–28809	6.85	6.14 ± 0.66	9.3	13.11 ± 1.41	10.06 ± 9.1	0.13
859	J1505+5847	43630–44220	8.15	4.04 ± 0.96	4.2	20.06 ± 4.77	10.24 ± 9.6	0.22
862	J1506+0806	11770–12010	8.75	4.29 ± 0.63	6.8	1.53 ± 0.22	9.13 ± 8.3	0.07
908	J1529+0601	31669–32130	7.40	2.79 ± 0.75	3.7	7.22 ± 1.94	9.80 ± 9.2	0.11
909	J1529+0913	37900–38140	8.45	2.52 ± 0.63	4.0	9.42 ± 2.34	9.92 ± 9.3	0.21
955	J1555+2955	20610–21010	11.65	5.10 ± 1.08	4.7	5.68 ± 1.20	9.70 ± 9.0	0.19
980	J1611+0840	49700–50000	11.60	$\lt 8.75$	—	$\lt 57.40$	$\lt 10.70$	0.56
1014	J1645+3048	17620–17850	14.45	3.35 ± 1.02	3.3	2.63 ± 0.80	9.36 ± 8.8	0.19
1057	J2245+1232	27400–27750	9.35	$\lt 2.46$	—	$\lt 4.88$	$\lt 9.63$	0.08
1062	J2326−0114	58470–59340	8.70	4.51 ± 1.27	3.6	42.07 ± 11.8	10.57 ± 10.0	0.35

Note. Column (1): SPOG name. Column (2): velocity range (optically defined, local standard of rest). Column (3): spectral rms. Column (4): CO(1–0) line flux. Column (5): the S/N of the CO detections. Column (6): CO(1–0) luminosity. Column (7): observed mass of H₂ (using the conversion factor of Bolatto et al. 2013). Column (8): molecular gas fraction, ${F}_{{\rm{gas}}}=\tfrac{{M}_{{{\rm{H}}}_{2}}}{{M}_{{{\rm{H}}}_{2}}+{M}_{{\rm{star}}}}$.

Download table as:  ASCIITypeset image

We calculate ${M}_{{{\rm{H}}}_{2}}$ with the following equation:

$$\begin{eqnarray}&&{M}_{{{\rm{H}}}_{2}}=1.05\times {10}^{4}\displaystyle \frac{{S}_{{\rm{CO}}}{\rm{\Delta }}{{vD}}_{{\rm{L}}}^{2}}{1+{v}_{{\rm{sys}}}/c}\,{M}_{\odot }\end{eqnarray} \tag{ 2 }$$

assuming ${X}_{{\rm{CO}}}=2\times {10}^{20}$ cm⁻² (K km s⁻¹)⁻¹, the mean conversion factor presented in Bolatto et al. (2013), the derived H₂ masses for our CO-SPOGs range between ${10}^{8.4-10.6}\,{M}_{\odot }$.

Figure 5 shows the molecular gas fraction distribution for different subsets of galaxies that contain molecular gas. Normal, star-forming galaxies from the CO Legacy Database for the GALEX Arecibo SDSS Survey (COLD GASS; top; Saintonge et al. 2011) have an average molecular gas fraction $\langle \mathrm{log}({f}_{{\rm{mol,SF}}})\rangle =-1.29$. The classical poststarbursts (middle; French et al. 2015) have $\langle \mathrm{log}({f}_{{\rm{mol,PSB}}})\rangle =-1.19$, and CO-SPOGs (bottom) have $\langle \mathrm{log}({f}_{{\rm{mol,SPOG}}})\rangle =-0.83$. SPOGs tend to have higher gas fractions than both normal galaxies and poststarbursts. The molecular gas fractions in relation to the stellar masses range from 10^−1.56 to ${10}^{-0.34}$, with those SPOGs at the high end of the molecular gas fraction range being comparable to interactions (in which log(${M}_{{{\rm{H}}}_{2}}/{M}_{{\rm{star}}}+{M}_{{{\rm{H}}}_{2}}$) ∼ −0.3; Combes et al. 2013; Kaneko et al. 2014). We ran a Mann–Whitney U-test¹³ (which is used in cases of small numbers) and were able to show that ${f}_{{\rm{mol}}}$ in SPOGs is distinct from both star-forming galaxies (p ≈ 0) and poststarbursts (p = 0.00189), where p is the probability of a null hypothesis.

Zoom In Zoom Out Reset image size

The fraction of our CO-SPOGs classified as having disrupted morphologies is large (37%–46% ± 7%). This optical disruption suggests that the molecular gas should also be disrupted to some extent. Downes & Solomon (1998) showed that in many interacting systems, the molecular gas mass predicted from the CO(1–0) flux was larger than the dynamical mass of the system, usually by about a factor of 5, though this is quite uncertain (Yao et al. 2003; Bolatto et al. 2013). They conclude that this is due to the nature of the molecular gas in interacting systems being more diffuse and warm (Aalto et al. 1995; Rangwala et al. 2011), creating a more continuous coverage of CO-emitting gas rather than distributed in discrete giant molecular clouds (which would also span a smaller velocity range). If the molecular gas in SPOGs is indeed disrupted and therefore not confined to a disk as in the Milky Way, we could be overestimating the molecular gas mass by ∼5. If CO-SPOGs are found on the extreme end of the conversions found by Sandstrom et al. (2013) for nuclear regions of star-forming galaxies, then the conversion could be off by as much as a factor of 7. If all CO-SPOGs require one of these reduced conversion factors, then the molecular-to-stellar mass ratio would shift significantly, to an average ratio of ${10}^{-1.68}$, consistent with the COLD GASS sample (Saintonge et al. 2011), and less than the poststarburst sample (French et al. 2015).

Figure 5 also overlays the molecular gas fraction of a sample of Hickson Compact Group (HCG) galaxies, with stellar masses calculated from U. Lisenfeld et al. (2016) in preparation and H₂ masses calculated from CO observations (Leon et al. 1998; Verdes-Montenegro et al. 1998; Martinez-Badenes et al. 2012; Lisenfeld et al. 2014), using X_CO = 2 × 10²⁰ cm⁻² (K km s⁻¹)⁻¹. HCG galaxies span the range of ${f}_{{\rm{mol}}}$ values, suggestive of their large range of properties and environments (Bitsakis et al. 2010, 2011, 2014; Zucker et al. 2016). Many HCGs contain warm H₂-luminous galaxies, known to host shocks (Cluver et al. 2013), which may be a better comparison sample with CO-SPOGs than interactions. Given that the SPOG criteria aimed to select objects that host shocks, the warm H₂-bright galaxies serve as a reasonable comparison. Alatalo et al. (2015a) used CARMA to study the molecular gas in turbulent HCG galaxies and found that the molecular gas-to-dust ratios within these systems were consistent with the Milky Way value and predicted gas masses that were consistent with a normal CO-to-H₂ conversion. Despite the fact that both types of systems host shocks, the molecular-to-stellar mass ratio for warm H₂-bright HCG galaxies tends to be lower than that for SPOGs (${10}^{-1.21};$ Martinez-Badenes et al. 2012; Lisenfeld et al. 2014).

It is unclear which ${L}_{{\rm{CO}}}\mbox{--}{M}_{{{\rm{H}}}_{2}}$ conversion factor is more relevant to our population of CO-SPOGs. Alatalo et al. (2014a, 2016) suggested that SPOGs were at an earlier stage of quenching than other poststarburst galaxies and thus could still contain larger reservoirs of molecular gas as they undergo quenching. Observations of denser gas tracers, such as CS and HCN, will be necessary to better infer the mass of dense gas and better define the ${X}_{{\rm{CO}}}$ factor.

The well-known connection between mid-IR emission and star formation (Calzetti et al. 2007) is caused by hot dust re-radiating UV photons from the young stars. Therefore, it is likely that a relation exists between the 22 μm and CO(1–0) fluxes, given that the 22 μm emission traces the star formation and the CO(1–0) traces the star-forming fuel (Rosenberg et al. 2015). The left panel of Figure 6 puts this relation to the test by comparing the 22 μm and CO(1–0) fluxes of different samples of galaxies, including star-forming galaxies (Lisenfeld et al. 2011; Saintonge et al. 2011; Bauermeister et al. 2013), early-type galaxies (Young et al. 2011; Alatalo et al. 2013; Davis et al. 2014), radio galaxies (Evans et al. 2005), and poststarburst galaxies (French et al. 2015). The 22 μm emission was obtained through a cross-matching with the ALLWISE catalog (Wright et al. 2010). In cases in which objects were flagged as extended in the Two Micron All Sky Survey (2MASS; Skrutskie et al. 2006), the extended source flux was used (w4gmag); otherwise, the profile fit flux (w4mpro) was used.

Zoom In Zoom Out Reset image size

The star-forming objects of the COLD GASS survey (Saintonge et al. 2011), Evolution of molecular Gas in Normal Galaxies (Bauermeister et al. 2013), and Analysis of the interstellar Medium of Isolated GAlaxies (Lisenfeld et al. 2011) samples (black and gray dots) in Figure 6 do indeed trace a reliable relation, allowing us to predict the CO(1–0) flux from the 22 μm flux. We used the scaling of Calzetti et al. (2007) of SFR ∝ ${L}_{24\,\mu {\rm{m}}}^{0.885}$, as well as that of Carilli & Walter (2013) of ${L}_{{\rm{FIR}}}$ ∝ ${L}_{{\rm{CO}}}^{1.4}$ (as well as ${L}_{{\rm{FIR}}}$ being linearly related to the SFR; Kennicutt 1998), to derive the expected slope of the relation between F(22 μm) and ${S}_{{\rm{CO}}}{\rm{\Delta }}v$, setting it to ${S}_{{\rm{CO}}}{\rm{\Delta }}v$ ∝ F(22 μm)${}^{0.632}$. Using a 4σ-clipped subset of the star-forming objects and the relation noted above, we find the following relationship between ${S}_{{\rm{CO}}}{\rm{\Delta }}v$ and ${F}_{22\mu {\rm{m}}}$:

$$\begin{eqnarray}&&\displaystyle \frac{{S}_{{\rm{CO}}}{\rm{\Delta }}v}{(\mathrm{Jy}\,\mathrm{km}\,{{\rm{s}}}^{-1})}={2.402}_{-0.122}^{+0.128}\,{\left(\displaystyle \frac{{F}_{22\mu {\rm{m}}}}{{\rm{mJy}}}\right)}^{0.623},\end{eqnarray} \tag{ 3 }$$

which corresponds to the bisecting line shown in Figure 6. For regular star-forming galaxies, this relation can predict a rough CO(1–0) flux, though the non-star-forming samples show that this breaks down with the presence of other dominant contributors to the 22 μm emission.

The early-type galaxies from the ATLAS${}^{{\rm{3D}}}$ sample show a much larger scatter than the star-forming galaxies on this relation, also noted by Davis et al. (2014), with a slight mid-IR enhancement (${\epsilon }_{{\rm{MIR}}};$ defined as the ratio of the 22 μm emission in the source to the expected 22 μm at a given ${S}_{{\rm{CO}}}{\rm{\Delta }}v$ based on the star-forming objects defined line), $\langle {\epsilon }_{{\rm{MIR,ETG}}}\rangle ={4.22}_{-0.74}^{+0.90}$. This scatter is likely due to the fact that there are other sources of 22 μm emission in early-type galaxies, including the aged stellar population and the hard ultraviolet field created by post-AGB stars (responsible for the UV upturn phenomenon; O'Connell 1999; Davis et al. 2014).

A subset of the radio galaxies and quasars (Evans et al. 2005) tend to diverge the most from this relation, as their 22 μm emission overpredicts the CO(1–0) flux with $\langle {\epsilon }_{{\rm{MIR,AGN}}}\rangle ={5.58}_{-1.75}^{+2.53}$. Many of the galaxies that sit near the relation are known starbursting AGN hosts (Lanz et al. 2016). The enhancement therefore is unsurprising, given that AGNs contribute substantially to the hot dust component of a galaxy's spectral energy distribution (SED) in the mid-infrared (Ward et al. 1987; Sanders et al. 1989; Elvis et al. 1994). Thus, a quasar's divergence from this relation is due to the fact that star formation is not the primary contributor to the 22 μm emission.

The poststarburst galaxies from French et al. (2015) appear to be strongly MIR-enhanced, sitting below the ${S}_{{\rm{CO}}}{\rm{\Delta }}v$–${F}_{22\mu {\rm{m}}}$ relation by a factor of $\langle {\epsilon }_{{\rm{MIR,PSB}}}\rangle ={10.51}_{-1.60}^{+1.89}$. An additional heating mechanism (that would only heat the dust, but not increase the CO flux by injecting turbulence) must be present in poststarburst galaxies to cause this enhancement, such as the radiation from the intermediate-age stellar population or deeply buried AGNs.

CO-SPOGs are also a divergent population, albeit not to the extent of the poststarbursts, with all objects lying below the ${S}_{{\rm{CO}}}{\rm{\Delta }}v$–${F}_{22\mu {\rm{m}}}$ relation set by the star-forming galaxies with $\langle {\epsilon }_{{\rm{MIR,}}{\rm{SPOG}}}\rangle ={4.91}_{-0.39}^{+0.42}$. In fact, if we limit ourselves only to strong (S/N $\geqslant$ 5) detections, $\langle {\epsilon }_{{\rm{MIR,}}{\rm{SPOG}}}\rangle$ increases to ${9.75}_{-1.60}^{+1.91}$. This suggests that there could be AGNs in our CO-SPOGs, but that the bolometric luminosities of these AGNs are not as strong as the extreme objects in the Evans et al. (2005) sample. The selection criteria we applied to create the CO-SPOG sample (intermediate-age stars, a lack of star formation from the line diagnostic diagram, and detectable 22 μm emission) favor AGNs; thus, the ${S}_{{\rm{CO}}}{\rm{\Delta }}v$–${F}_{22\mu {\rm{m}}}$ relation indicating the presence of AGNs is expected. How CO-SPOGs and poststarbursts compare will be presented in detail in Section 5.4.

Their positions in Figure 6 predict that SPOGs have AGN luminosities that are intermediate between the star-forming population and the quasar/radio galaxy population. Given the prevalence of intermediate-age stars found in quasars (Canalizo & Stockton 2013), it is possible that some SPOGs in our sample may migrate to more luminous AGN phases as the galaxy transition proceeds.

The right panels of Figure 6 have taken the relation and broken CO-SPOGs down into subpopulations based on their Na i D properties (top), radio detections (middle), and morphological properties (bottom) to determine whether any of these properties influence ${\epsilon }_{{\rm{MIR}}}$. Our CO-SPOGs have $\langle {\epsilon }_{{\rm{MIR,}}{\rm{SPOG}}}\rangle ={4.91}_{-0.39}^{+0.42}$, falling off the star-forming relation by a factor of ≈5. Objects with either radio emission or an Na i D enhancement (detailed in Section 5.5) appear to fall slightly farther from the relation with $\langle {\epsilon }_{{\rm{MIR,radio}}}\rangle ={4.95}_{-0.55}^{+0.62}$ and $\langle {\epsilon }_{\mathrm{MIR},\mathrm{Na}{\rm{I}}{\rm{D}}}\rangle ={5.40}_{-0.73}^{+0.85}$, respectively, but agree within the uncertainty. CO-SPOGs that were classified as morphologically disrupted were found to have smaller deviations from the relation, $\langle {\epsilon }_{{\rm{MIR,disrupted}}}\rangle ={4.37}_{-0.43}^{+0.47}$. In all of these cases, the overall ${\epsilon }_{{\rm{MIR}}}$ was not sufficiently deviant from the overall CO-SPOG population to be distinct, and thus we cannot conclude whether the presence of any of these properties enhances or suppresses the mid-IR emission as compared to the CO flux (relative to star-forming galaxies).

In CO-SPOGs, shocks can dominate the ionized gas emission (Allen et al. 2008; Rich et al. 2011), including [O ii] and Hα. UV emission suffers from large uncertainties due to extinction and contamination due to non-star-formation-dominated phenomena, such as heating by intermediate-age stars. Finally, as Figure 6 laid out, the 22 μm emission can be significantly contaminated by the presence of an AGN. In NGC 1266, the 22 μm emission overestimates the SFR by a factor of 10 (Alatalo et al. 2011), despite star formation being suppressed by a factor of >50 (Alatalo et al. 2015c). While centimeter-wave radio emission is a sensitive tracer of recent star formation (Condon 1992), the continuum emission at 1.4 GHz can also be contaminated by the synchrotron emission from an AGN (Zensus 1997; Laor & Behar 2008).

Constructing an SED that includes far-infrared emission is by far the most reliable tracer of star formation in all systems except those with the most deeply buried AGNs, because the cold dust nearly unambiguously traces star formation originating in imbedded clouds. It is possible that a single data point on the Rayleigh–Jeans tail of the dust continuum blackbody could reduce the uncertainty of SFRs in ambiguous systems, such as SPOGs. For instance, an observation at 850 μm would be able to anchor the blackbody. If we assume that the dust temperatures do not vary much away from T_d ≈ 25 K (Scoville et al. 2014), we should be able to use the 22 and 850 μm points to interpolate the modified blackbody for the dust continuum across, inferring a star formation rate. While this method is not as precise as fitting a full SED and dust continuum (with far-IR data near the peak of the blackbody), it could significantly improve SFR estimates for SPOGs.

Alatalo et al. (2016) argue that the SPOG criterion might identify objects at an earlier stage of transition than the classical poststarburst criterion. French et al. (2015) showed that 53% of poststarburst galaxies contain non-negligible reservoirs of molecular gas, modifying the standard picture of galaxy evolution in which a galaxy expels its interstellar medium before transitioning (Hopkins et al. 2006). We have also shown that compared to their molecular gas content, the poststarbursts of French et al. (2015) are significantly 22 μm enhanced.

The French et al. (2015) poststarburst sample has reservoirs of molecular gas ranging from ${10}^{8.6}$ to 10${}^{9.8}$ ${M}_{\odot }$ in mass, and molecular-to-stellar mass fractions between 10⁻² and 10${}^{-0.5}$. In that sample, the majority of the poststarburst galaxies contain ionized gas line ratios that are consistent with LINERs, and a substantial fraction show disrupted optical morphologies. Of the 17 CO-detected poststarbursts from the French et al. (2015) sample, 15 were detected in the 22 μm WISE band with an S/N > 3; thus, all but two French et al. (2015) poststarbursts would have surpassed the 22 μm criterion of the CO-SPOGs. There is no difference in average ${f}_{{\rm{mol}}}$ in the poststarburst sample when we remove the two 22 μm nondetected poststarbursts from consideration.

CO-SPOGs compare well to the French et al. (2015) poststarbursts in ionized gas ratios and 22 μm properties, and morphologies, including both sets presenting substantial LINER-like ionized gas line diagnostics, similar detection rates in WISE 22 μm emission, and a similar fraction of objects present morphological disruptions. Due to a selection against strong [O iii] emission, the French et al. (2015) CO-detected poststarbursts have an average EW[O iii] that is an order of magnitude smaller $(\langle \mathrm{EW}{[{\rm{O}}{\rm{III}}]}_{{\rm{PSB}}}\rangle =0.60)$ than what is found for CO-SPOGs $(\langle \mathrm{EW}{[{\rm{O}}{\rm{III}}]}_{{\rm{COSPOG}}}\rangle =5.74)$. The poststarbursts of French et al. (2015) show low CO fluxes compared to their 22 μm fluxes, and CO-SPOGs seem to contain larger reservoirs of molecular gas (Figure 7; assuming equivalent values for ${X}_{{\rm{CO}}}$). We also see discrepancies between SPOGs and poststarbursts in the molecular-to-stellar mass ratio (Figure 5), where SPOGs exhibit much higher gas fractions.¹⁴

Zoom In Zoom Out Reset image size

The fact that SPOGs contain more molecular gas than French et al. (2015) poststarbursts could suggest one of the following. Either (1) the SPOG criteria allowed for galaxies that contained more substantial reservoirs of molecular gas in the first place, due to not strictly excluding Hα or [O ii] emission, or (2) SPOGs are in an earlier phase of their transition; possibly both of these effects are at play. It is also possible that in some CO-SPOGs, the bulk of star formation is taking place behind an optically thick screen, although IFU studies of objects at different merger stages have shown that such "skin effect" phenomena are rare (Rich et al. 2015) and do not explain the excess 22 μm enhancement present in the poststarbursts. Optical IFU studies of CO-SPOGs would help elucidate whether this effect is present in any of our objects.

Figure 9 in Alatalo et al. (2016) indicates that SPOGs are bluer than poststarburst galaxies (Goto 2007; Yesuf et al. 2014; Rowlands et al. 2015) and lie on the blueward side of the green valley, and Figure 2 confirms that CO-SPOGs follow the colors of the general SPOG population. SPOG optical colors may be bluer than poststarbursts because they have truncated their star formation more recently and thus have a larger population of blue stars (to contribute to the galaxy colors). A large-scale stellar population analysis of these two populations would be able to better pinpoint how SPOGs and poststarbursts are related, but the larger reservoir of molecular gas in SPOGs could suggest that they are at an earlier phase of evolution, if depletion of the molecular reservoir is related to the time since the start of a galaxy's transition.

If galaxy transition does correlate with the depletion of molecular material, then the larger gas reservoirs and gas fractions are suggestive of a timescale effect, which correlates with the phase of transition. Alatalo et al. (2015a) used interferometric CO measurements to show that in some warm H₂-bright HCG galaxies, the expulsion of the molecular reservoir was not required for a galaxy to transform. Instead, these authors argue that the inhibition of the molecular gas to form stars was a more important factor in predicting a galaxy's color. Whether CO-SPOGs or the poststarbursts of French et al. (2015) show this impact too will require sufficiently resolved molecular gas and accurate star formation rates to study whether a star formation suppression phase globally manifests in transforming galaxies or whether it is a special phenomenon seen in the compact group environment.

Alatalo et al. (2016) showed that, as a population, SPOGs have a higher fraction of sources with Na i D properties that require an interstellar component compared to the (star-formation-dominated) ELG sample. Figure 8 shows the underlying distribution from the ELG sample (in grayscale), along with the SPOG distribution (green contours), where we see that there is an Na i D-enhanced wing. Overplotted are CO-SPOGs (stars), labeled based on whether they were detected at 1.4 GHz by FIRST (dark blue) or not (light blue), and whether they are disrupted (magenta outline). We can see that the CO-SPOGs are even more Na i D enhanced than the underlying SPOG population. We ran the Mann–Whitney U-test to test whether the Na i D properties of CO-SPOGs could be randomly drawn from the SPOG sample, and found that the null hypothesis was ruled out (p ≈ 0). A total of 19 (37% ± 7%) of the CO-SPOGs contain Na i D emission above the 3σ boundary defined by the Na i D-Mg b relation of the ELG sample (Equation (10) in Alatalo et al. 2016). This excess suggests the possibility that many CO-SPOGs host interstellar winds (Rupke et al. 2005; Veilleux et al. 2005; Murray et al. 2007; Park et al. 2015).

Zoom In Zoom Out Reset image size

Figure 8 indicates that a fraction of CO-SPOGs host Na i D properties as extreme as what is seen in AGN-driven outflow hosts (Rupke & Veilleux 2011, 2015), including NGC 1266 (Alatalo et al. 2011; Davis et al. 2012; Nyland et al. 2013). The extreme Na i D widths of these sources suggest that they may host an AGN-driven molecular outflow like NGC 1266. If so, it would indeed shed light on how such outflows relate to galaxy transformation. However, it is also possible that Na i D enhancements are due to unsettled neutral gas along the line of sight instead.

We have investigated how Na i D absorption changes across the ELG sample. Table 4 shows the median values and 10th and 90th percentile values of the ELG, SPOG, and CO-SPOG subsamples.¹⁵ Overall, objects that have been detected in the 22 μm band of WISE appear to show more enhanced Na i D in both the ELG and the SPOG samples. In the case of galaxies within the ELG (and thus star-formation-dominant) sample, this Na i D enhancement might be able to be explained by neutral winds being launched by the starburst (Murray et al. 2007; Park et al. 2015; Sarzi et al. 2016), which would correlate with the 22 μm hot dust emission (and therefore the star formation rate). The SPOG selection criteria eliminated strong star-formers (based on ionized gas emission-line ratios; Baldwin et al. 1981; Veilleux & Osterbrock 1987; Kewley et al. 2006), so the Na i D excess and 22 μm emission are possibly from an alternative source.

Table 4.  Na i D Properties of SPOGs

Type	${N}_{{\rm{obj}}}$	$\widetilde{{\epsilon }_{{\rm{NaD}}}}$	${\epsilon }_{{\rm{NaD,}}{\rm{10}}}$	${\epsilon }_{{\rm{NaD,}}{\rm{90}}}$
(1)	(2)	(3)	(4)	(5)
ELG	159,387	−0.073	−1.142	0.833
ELG 22 μm > 3σ	71,301	−0.009	−0.966	1.023
SPOGs	1,067	−0.043	−1.111	1.656
SPOG 22 μm > 3σ	491	0.115	−0.856	2.105
CO-SPOGs	52	1.421	−0.485	3.273
CO-SPOGs, S/N > 5	34	1.452	−0.341	3.463
CO-SPOGs, radio	30	1.984	−0.072	3.463
CO-SPOGs, morph.	24	1.495	−0.641	3.686

Note. Column (1): galaxy type. Column (2): total number of objects in each class. Column (3): median Na i D enhancement of each sample. Column (4): the 10th percentile Na i D enhancement. Column (5): the 90th percentile Na i D enhancement.

Download table as:  ASCIITypeset image

As discussed in Section 2, the ramifications of this selection are that our CO-SPOGs likely favor the presence of AGNs. SPOGs also follow the trend that objects hosting radio emission and 22 μm emission show a significant Na i D enhancement when compared to other samples, including radio or 22 μm nondetected SPOGs. Radio-detected SPOGs exhibit a median Na i D enhancement (defined as the deviation from the mean relation of the ELG of Na i D = (0.685 Mg b+0.8) from Alatalo et al. 2016) of $\langle {\epsilon }_{{\rm{NaD,radio}}}\rangle =1.421$, and 13/19 (68% ± 11%) objects found with 3σ Na i D enhancements were radio detected (although 57% ± 7% of the CO-SPOGs have been detected in radio). Objects that were not detected in FIRST have a median Na i D enhancement of $\langle {\epsilon }_{\mathrm{NaD},\mathrm{non} \mbox{-} \mathrm{radio}}\rangle =0.547$. CO-SPOGs with radio emission have higher Na i D enhancements than objects that do not. This is further supported by the results of the Mann–Whitney U-test, which was able to rule out that the Na i D properties of radio nondetected CO-SPOGs were drawn from the same distribution of radio-detected CO-SPOGs with the probability of a null hypothesis of p < 0.04. Given that AGN activity is a probable origin of the radio emission, we suggest that AGN-hosting SPOGs are the most likely to contain enhanced Na i D.

We also tested whether the morphologies of the SPOGs had a strong influence on the Na i D enhancement, given that many objects that host strong neutral winds are starbursts in ULIRGs, which are mostly major mergers (Veilleux et al. 2005). The overall effect of morphological disruption seems to have a slight impact on the Na i D enhancement, with median $\langle {\epsilon }_{{\rm{NaD,}}{\rm{disrupted}}}\rangle =1.495$, slightly larger than the median for galaxies that were not classified as being disrupted of $\langle {\epsilon }_{{\rm{NaD,}}{\rm{undisturbed}}}\rangle =1.206$. Therefore, morphological disruption appears to have a smaller influence on the Na i D enhancement of CO-SPOGs than the presence of radio emission.

Sarzi et al. (2016) showed that in a sample of 456 nearby galaxies, sources with Na i D enhancements attributable to interstellar winds (as opposed to [Na/Fe] overabundances as is seen in the most massive galaxies; Jeong et al. 2013) were not observed in objects detected with very long baseline interferometry (Deller & Middelberg 2014). These authors concluded that the majority of Na i D-enhanced objects were prolifically star-forming galaxies with neutral winds that were star formation driven, and that Na i D winds originate most commonly in star-forming galaxies due to star formation driving. In CO-SPOGs, it appears that there are (radio-detected) AGNs and Na i D enhancements in the same objects, consistent with the results of Lehnert et al. (2011) that 33% of radio galaxies within SDSS also contain Na i D enhancements. The lack of this type of object within the Sarzi et al. (2016) sample suggests that they could be quite rare or are possibly a short and violent phase in galaxy evolution, in which the star formation has quenched and an AGN is driving a neutral wind that is stirring up the remaining interstellar medium. The SPOG criterion selected against star-forming objects (which is able to rule out strong starbursts that are needed to drive winds; Alatalo et al. 2016) and thus includes many more galaxies with LINER emission, as opposed to the Sarzi et al. (2016) sample.

Further studies are needed, including observations with higher spectral resolution to investigate the shape and structure of the Na i D line, to look for outflows (Rupke et al. 2005; Veilleux et al. 2005; Rupke & Veilleux 2011). Integral field spectrographs will be able to determine the extent of the Na i D absorption, provide high-resolution kinematics to determine the neutral mass flux, and compare the Na i D kinematics to the stellar kinematics. Deeper CO observations might detect broad molecular wings, resulting in the measurement of a range of the mass outflow (and mass escape) rates observed in transitioning objects. In-depth studies of these objects' star formation rates and molecular gas distributions will shed light on how often star formation is suppressed, leading to conserving and extending the lifetime of the molecular gas as galaxies undergo transformations from late types to early types. Deep X-ray observations will put limits on the AGN luminosity, helping to determine the range in energy budgets these systems might exhibit, and two-dimensional stellar population studies might be able to provide a range of timescales over which triggering mechanisms started the process of star formation quenching in these systems.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size