THE PROPERTIES OF THE CIRCUMGALACTIC MEDIUM IN RED AND BLUE GALAXIES: RESULTS FROM THE COS-GASS+COS-HALOS SURVEYS

Galaxy growth is fundamentally connected to the cycle of accretion and ejection of matter into and out of galaxies. In the simplest picture, galaxies acquire gas that reaches the central regions via the circum-galactic medium (CGM). There, it condenses into neutral and then molecular gas, some of which is then converted into stars. Young stars, in turn, drive strong winds, outflows, and radiation that deposit mass, metals, energy, and momentum to the CGM, thus significantly influencing its properties (see review by Somerville & Davé 2015; Fielding et al. 2016, and references therein). These linked processes are commonly termed the baryon cycle. The CGM then lies at the heart of this cycle, as it is the interface between the stellar body of the galaxy and the intergalactic medium. It is the primary spatial pathway for the baryon cycle into and out of galaxies (Tumlinson et al. 2013; Brook et al. 2014; Shen et al. 2014; Borthakur et al. 2015; Mitra et al. 2015; Nielsen et al. 2015; Ford et al. 2016, and references therein).

The CGM is also a reservoir of low-density gas that may have as much mass as the stellar component of the galaxy (Tumlinson et al. 2013; Werk et al. 2013, 2014; Peeples et al. 2014; Richter et al. 2016). It extends out from the stellar disk out to the virial radius of the galaxy (Chen et al. 2001a; Borthakur et al. 2013; Stocke et al. 2013). However, due to its low surface-brightness, we have not yet been able to directly image this vast baryonic reservoir. On the other hand, absorption-line spectroscopy provides an avenue to probe the physical conditions in this low-density gaseous medium. Rest-frame ultra-violet (UV) spectroscopy enables us to use various absorption-line transitions, including hydrogen and metal-line species spanning a broad range of ionization states.

Mapping the CGM with the help of a large sample of sightlines probing a range of impact parameters is crucial for understanding its properties and its variations as a function of radius. The radial dependence in the properties of the neutral hydrogen in the CGM has been known for decades, based on observations of the Lyα absorption-line (Lanzetta et al. 1995; Chen et al. 1998; Tripp et al. 1998; Chen et al. 2001b; Bowen et al. 2002; Prochaska et al. 2011; Stocke et al. 2013; Tumlinson et al. 2013; Liang & Chen 2014; Borthakur et al. 2015, and references therein). However, only recently, with the installation of Cosmic Origins Spectrograph (COS) aboard the Hubble Space Telescope (HST), has it become feasible to undertake detailed probes of the CGM properties as a function of other global properties of the central galaxy.

One of the consequences of the accretion of gas passing through the CGM is that it provides the raw material to sustain the growth of the galaxy via star formation (e.g., Bouché et al. 2013). Not all galaxies produce stars at the same rate (Brinchmann et al. 2004; Daddi et al. 2007; Noeske et al. 2007; Salim et al. 2007; Rodighiero et al. 2011; Speagle et al. 2014; Snyder et al. 2015). In particular, galaxies show two distinct populations in terms of their star formation rate (SFR). Although most low-mass galaxies form stars at significant rates, most high-mass galaxies produce stars at negligible levels. This was termed the galaxy color bimodality, defined in terms of "blue" (star-forming) galaxies and "red" (quiescent) galaxies (e.g., Tully et al. 1982; Blanton et al. 2003; Kauffmann et al. 2003; Baldry et al. 2004; Brinchmann et al. 2004).

About a decade ago, cosmological hydrodynamical simulations revealed two distinct ways that galaxies accrete gas into their dark matter halo as a function of halo mass. The predominant mode of gas accretion for low-mass galaxies is believed to be the "cold" mode (Kereš et al. 2005, 2009; Dekel et al. 2009), where gas falls into galaxies as streams or lumps at temperatures much less than that of the virial temperature. For the higher-mass halos, the accretion process is expected to be in the "hot" mode (White & Frenk 1991; Fukugita & Peebles 2006), in which the incoming gas shock heats to the virial temperature. This broadly can explain why high-mass galaxies have little-to-no cold gas reservoirs to fuel star formation (see work on condensation in hydrodynamical simulations by Kaufmann et al. 2006, 2009; Sommer-Larsen 2006). Galaxies also recycle gas from previous generations of star formation that is stored in their CGM (Ford et al. 2013; Fraternali et al. 2015). However, the process of how gas gets into the disk from the CGM is fairly complex. Nonlinear perturbations in the filamentary flows may help the cool accreting gas condense and add cold gas to the disk (Kereš & Hernquist 2009; Joung et al. 2012a). These condensing clouds may contain as much as 25%–75% of the cold gas in the CGM (Fernández et al. 2012).

In addition to accretion, feedback driven by star formation may change the nature and properties of the gas in the CGM (Marasco et al. 2015; Nelson et al. 2015, 2016; Kauffmann et al. 2016; Liang et al. 2016). Massive young stars inject energy and/or momentum into outflows (Veilleux et al. 2005; Heckman et al. 2011; Borthakur et al. 2014; Heckman et al. 2015; Heckman & Borthakur 2016) that may travel into the CGM, enriching it with metals, shock-heating the cooler CGM clouds, and possibly even expelling/unbinding the CGM (Borthakur et al. 2013). Therefore, if feedback provided by massive stars plays a role in the observed bimodality, then we should see a change in the structure, ionization state, and/or kinematics of the CGM as a function of SFR.

To that end, we have selected a subsample of galaxies from the Galaxy Evolution Explorer (GALEX) Arecibo Sloan Digital Sky Survey (SDSS) Survey (GASS; Catinella et al. 2010, 2012, 2013) that have background UV-bright quasi-stellar objects (QSOs) located within a projected distance of 250 kpc in the rest-frame of the galaxy. This yielded the COS-GASS sample (Borthakur et al. 2015) whose members were observed with COS, using the G130M grating. This provided a spectral R = 20,000–24,000 (FWHM ∼ 12 to 15 km s⁻¹). We have multi-band data for these galaxies from the parent GASS survey: 21 cm H i spectroscopic data obtained with the Arecibo telescope, optical images and spectroscopy from the SDSS, UV imaging with the GALEX, molecular gas data from IRAM (COLD GASS; Saintonge et al. 2011), and long-slit optical spectroscopy (Moran et al. 2012) for a portion of the sample. Therefore, we have the stellar mass, SFR, gas-phase metallicity, stellar morphology, and atomic and molecular gas masses for all 45 galaxies from the COS-GASS sample.

Here, we present our study, which utilizes the combined COS-GASS (Borthakur et al. 2015) and COS-Halos (Tumlinson et al. 2013) samples. Detailed descriptions of our sample, the COS observations and data reduction are presented in Section 2. The results are presented in Section 3, and their implications are discussed in Section 4. Finally, we summarize our findings in Section 5. The cosmological parameters used in this study are ${H}_{0}=70\,\mathrm{km}\,{{\rm{s}}}^{-1}\,{\mathrm{Mpc}}^{-1}$ (in between the two recent measurements of $73.24\pm 1.74\,\mathrm{km}\,{{\rm{s}}}^{-1}\,{\mathrm{Mpc}}^{-1}$ (Riess et al. 2016) and ${67.6}_{-0.6}^{+0.7}\,\mathrm{km}\,{{\rm{s}}}^{-1}\,{\mathrm{Mpc}}^{-1}$ (Grieb et al. 2016)) , ${{\rm{\Omega }}}_{m}=0.3$, and ${{\rm{\Omega }}}_{{\rm{\Lambda }}}=0.7$. We note that varying the Hubble constant value from $65\,\mathrm{to}\,75\,\mathrm{km}\,{{\rm{s}}}^{-1}\,{\mathrm{Mpc}}^{-1}$ does not affect the conclusions in the paper.

2.1. Sample

We combined the COS-GASS sample (45 galaxies) with the COS-Halos sample (44 galaxies) to get our full sample. The COS-Halos sightlines cover the inner CGM (thoroughly up to ∼0.8 R_vir) whereas the COS-GASS sightlines extend the observations to the outer CGM (∼0.2–1.5 R_vir). The resulting combined sample contains a total of 89 sightlines probing the CGM from 17 to 231 kpc in the rest frame of the target galaxies.

The two programs probe similar stellar mass ranges. The COS-Halos program probes galaxies in the range ${10}^{9.6-11.5}\,{M}_{\odot }$ at $0.1\lt z\lt 0.2$, whereas the COS-GASS program probes galaxies in the range (${10}^{10.1-11.1}\,{M}_{\odot }$) at slightly lower redshifts of $0.02\lt z\lt 0.05$. A comparison of the stellar masses and virial radii (based on the prescription by Kravtsov et al. 2014 and Liang & Chen 2014) for both samples are provided in Figure 1. Recent gravitational-lensing-based results by Mandelbaum et al. (2016) show that blue galaxies of fixed stellar mass are found in lower mass halos than red galaxies of the same stellar mass. Based on their results, we add or subtract 0.15 dex to the halo masses for red or blue galaxies, respectively, with a given stellar mass. The dark matter halo masses of the combined sample range from 11.1 to 13.2  M_⊙ dex. We note that the redshift difference between the two samples is ∼0.1. However, the variation in CGM properties during this time is expected to be minimal (Chen 2012). Also, the COS-Halos sample was selected to be all centrals (with a couple of non-centrals Tumlinson et al. 2013, Section 2.5). This is not one of the criteria for COS-GASS, although the mass range of the galaxies ensured that most of the COS-GASS galaxies (34/45) are centrals (based on the group catalog by Yang et al. 2005, 2007). Thus, the combined sample is 85% centrals. We have retained the satellites in our analysis, but verified that they do not affect any of our conclusions.

Zoom In Zoom Out Reset image size

We identify galaxies with specific star formation rates (sSFR = SFR/M_⋆) > 10⁻¹¹ yr⁻¹ as blue (star-forming) galaxies, and those below this limit as red (quiescent) galaxies. sSFR values of <10⁻¹² yr⁻¹ can be considered as upper limit. A detailed description of our galaxy color assignment can be found in Borthakur et al. (2015).

For more information on the properties of the target galaxies, including their redshifts, stellar masses, SFRs¹³ , sSFR, galaxy colors, and impact parameter of the sightlines, we refer the reader to Table 1 presented in this paper (for the COS-GASS sample) and Table 2 from the published work by Tumlinson et al. (2013).

Table 1.  Description of Galaxy Properties for the COS-GASS Survey^a

Galaxy	GASS ID	R.A.	Decl.	zgal	M⋆	Mhalob	Rvirc	sSFR	Colord	vesc,Rvire
					(Log M⊙)	(Log M⊙)	(kpc)	(Log yr−1)		(km s−1)
J0159+1346	3936	29.941	13.781	0.0441	10.1	11.4	153	−9.5	Blue	114
J0808+0512	19852	122.068	5.216	0.0308	10.8	12.2	296	−12.0	Red	217
J0852+0309	8096	133.229	3.152	0.0345	10.3	11.5	166	−10.1	Blue	122
J0908+3234	22391	137.232	32.576	0.0490	10.5	11.9	232	−12.3	Red	174
J0914+0836	20042	138.684	8.601	0.0468	10.0	11.3	147	−9.6	Blue	110
J0930+2853	32907	142.538	28.898	0.0349	10.5	11.6	184	−10.7	Blue	136
J0931+2632	53269	142.817	26.550	0.0458	11.0	12.4	345	−12.6	Red	258
J0936+3204	33214	144.101	32.079	0.0269	10.3	11.8	217	−11.7	Red	158
J0937+1658	55745	144.292	16.977	0.0278	10.9	12.0	263	−10.3	Blue	192
J0951+3537	22822	147.937	35.622	0.0270	10.6	11.7	197	−10.4	Blue	143
J0958+3204	33737	149.714	32.073	0.0270	10.7	12.1	272	−12.7	Red	198
J1002+3238	33777	150.711	32.645	0.0477	10.1	11.7	191	−11.9	Red	143
J1013+0501	8634	153.352	5.025	0.0464	10.1	11.4	153	−10.8	Blue	115
J1032+2112	55541	158.196	21.216	0.0429	10.6	11.7	202	−10.1	Blue	150
J1051+1245	23419	162.827	12.757	0.0400	10.4	11.5	175	−10.0	Blue	130
J1059+0517	9109	164.811	5.292	0.0353	11.1	12.6	387	−11.9	Red	284
J1100+1210	23457	165.048	12.171	0.0354	10.1	11.4	154	−10.7	Blue	114
J1100+1043	23477	165.200	10.728	0.0360	11.1	12.3	313	−11.0	Blue	231
J1115+0241	5701	168.789	2.699	0.0442	10.7	11.8	218	−10.9	Blue	162
J1120+0410	12452	170.026	4.177	0.0492	10.8	12.2	296	−12.1	Red	222
J1122+0314	5872	170.642	3.244	0.0446	10.5	11.9	239	−12.0	Red	178
J1127+2657	48604	171.943	26.960	0.0334	10.6	11.7	201	−11.0	Blue	147
J1131+1553	29898	172.954	15.897	0.0364	10.2	11.7	199	−12.0	Red	147
J1132+1329	29871	173.052	13.492	0.0342	10.2	11.4	158	−9.7	Blue	116
J1142+3013	48994	175.575	30.230	0.0322	10.7	11.8	222	−10.4	Blue	163
J1155+2921	49433	178.903	29.351	0.0458	10.5	11.6	180	−10.3	Blue	135
J1241+2847	50550	190.367	28.791	0.0350	10.3	11.5	166	−10.0	Blue	123
J1251+0551	13074	192.894	5.864	0.0486	10.9	12.0	242	−10.4	Blue	182
J1305+0359	13159	196.356	3.992	0.0437	10.4	11.5	172	−10.8	Blue	128
J1315+1525	26936	198.855	15.423	0.0266	10.7	12.1	283	−12.3	Red	206
J1317+2629	51025	199.440	26.486	0.0450	10.3	11.4	162	−10.4	Blue	121
J1325+2714	51161	201.345	27.249	0.0345	10.1	11.4	156	−9.8	Blue	115
J1348+2453	38018	207.142	24.891	0.0297	10.1	11.3	153	−10.5	Blue	112
J1354+2433	44856	208.546	24.556	0.0286	10.1	11.6	191	−11.8	Red	139
J1404+3357	31172	211.122	33.953	0.0264	10.3	11.8	211	−12.3	Red	154
J1406+0154	7121	211.678	1.915	0.0472	10.2	11.7	202	−11.8	Red	152
J1427+2629	45940	216.954	26.484	0.0325	10.4	11.9	225	−12.0	Red	165
J1430+0323	9615	217.508	3.398	0.0333	10.2	11.7	197	−11.1	Red	145
J1431+2440	38198	217.894	24.682	0.0378	10.7	12.1	261	−12.7	Red	193
J1454+3050	42191	223.516	30.846	0.0320	10.1	11.4	155	−9.8	Blue	114
J1502+0649	41743	225.517	6.823	0.0462	10.5	11.6	180	−10.2	Blue	135
J1509+0704	41869	227.340	7.078	0.0414	10.1	11.4	155	−9.6	Blue	115
J1515+0701	42025	228.781	7.021	0.0367	10.9	12.3	314	−11.9	Red	231
J1541+2813	28365	235.344	28.230	0.0321	10.4	11.5	173	−9.6	Blue	127
J1544+2740	28317	236.034	27.673	0.0316	10.1	11.6	191	−12.1	Red	140

Notes.

^aDetails on the COS-Halos survey can be found in the published work by Tumlinson et al. (2013) and Werk et al. (2013). ^bUsing prescription from Kravtsov et al. (2014) and Mandelbaum et al. (2016). ^cUsing prescription from Liang & Chen (2014). ^dGalaxies with sSFR $\gt {10}^{-11}\,{\mathrm{yr}}^{-1}$ are defined as blue galaxies. Galaxies with sSFR below this value are defined as red galaxies. ^eEscape velocity at the virial radii probed by the QSO sightline assuming a NFW profile for the galaxy's dark matter distribution.

Download table as:  ASCIITypeset image

Table 2.  Description of QSO Sightlines and Absorption Line Measurements for the COS-GASS Survey

QSO	R.A.QSO	Decl.QSO	zQSO	ρ	$\rho /{R}_{\mathrm{vir}}$	Θa	WLyαb	${\rm{\Delta }}{V}_{\mathrm{Ly}\alpha }$ c	${v}_{\mathrm{Ly}\alpha }$ d	${b}_{\mathrm{Ly}\alpha }$ d	${W}_{\mathrm{Si}{\rm{III}}1206}$	${\rm{\Delta }}{V}_{\mathrm{Si}{\rm{III}}1206}$ c	${v}_{\mathrm{Si}{\rm{III}}}$ d	${b}_{\mathrm{Si}{\rm{III}}}$ d
				(kpc)			(Å)	(km s−1)	(km s−1)	(km s−1)	(Å)	(km s−1)	(km s−1)
J0159+1345	29.971	13.765	0.504	102	0.7	64	1.501 ± 0.023	−150–380	93, 358	85, 14	⋯	⋯	⋯	⋯
J0808+0514	122.162	5.244	0.361	215	0.7	7	⋯	⋯	⋯	⋯	<0.050	⋯	⋯	⋯
J0852+0313	133.247	3.222	0.297	178	1.1	67	0.113 ± 0.015	0–110	50	49	<0.051	⋯	⋯	⋯
J0909+3236	137.276	32.608	0.809	170	0.7	21	0.090 ± 0.014	50–200	101	71	<0.041	⋯	⋯	⋯
J0914+0837	138.632	8.629	0.649	189	1.3	69	0.104 ± 0.020	−50–80	30	45	<0.065	⋯	⋯	⋯
J0930+2848	142.508	28.816	0.487	214	1.2	42	<0.126	⋯	⋯	⋯	<0.108	⋯	⋯	⋯
J0931+2628	142.820	26.480	0.778	226	0.7	77	0.114 ± 0.013	150–250	200	55	<0.043	⋯	⋯	⋯
J0936+3207	144.016	32.119	1.150	160	0.7	0e	<0.113	⋯	⋯	⋯	<0.086	⋯	⋯	⋯
J0937+1700	144.279	17.006	0.506	63	0.2	64	0.135 ± 0.021	−300 to −150	−204	67	<0.056	⋯	⋯	⋯
J0951+3542	147.850	35.714	0.398	226	1.1	3	0.839 ± 0.014	−90–200	64	59	⋯	⋯	⋯	⋯
J0959+3203	149.812	32.066	0.564	162	0.6	0e	0.420 ± 0.016	−320 to −100	−238, −144	33,19	0.060 ± 0.013	−300 to −200	−256	12
J1002+3240	150.727	32.678	0.829	119	0.6	40	<0.063	⋯	⋯	⋯	<0.062	⋯	⋯	⋯
J1013+0500	153.325	5.009	0.266	102	0.7	0e	0.445 ± 0.024	−180–50	−72	49	0.088 ± 0.020	−120–0	−70	20
J1033+2112	158.270	21.204	0.315	214	1.1	45	0.437 ± 0.033	−100–95	−1	50	<0.065	⋯	⋯	⋯
J1051+1247	162.857	12.796	1.281	140	0.8	73	0.781 ± 0.022	−180–140	−37	75	⋯	⋯	⋯	⋯
J1059+0519	164.795	5.327	0.754	95	0.2	30	0.270 ± 0.022	−100–100	6	44	<0.062	⋯	⋯	⋯
J1059+1211	164.984	12.198	0.993	171	1.1	61	0.225 ± 0.014	−150–0	−61	26	<0.045	⋯	⋯	⋯
J1100+1046	165.199	10.770	0.422	108	0.3	0e	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯
J1115+0237	168.782	2.633	0.567	209	1.0	83	0.195 ± 0.020	−100–100	2	37	0.065 ± 0.012	−20–30	4	20
J1120+0413	170.021	4.223	0.545	162	0.5	78	0.830 ± 0.018	50–390	201, 370	73, 21	0.172 ± 0.015	150–290	194	26
J1122+0318	170.601	3.301	0.475	221	0.9	0e	0.110 ± 0.018	0–150	67	58	⋯	⋯	⋯	⋯
J1127+2654	171.902	26.914	0.379	140	0.7	26	0.705 ± 0.021	−250–100	−28, −178	54, 30	0.051 ± 0.017	−80–20	−34	12
J1131+1556	172.905	15.946	0.183	176	0.9	0e	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯
J1132+1335	173.044	13.586	0.201	230	1.5	5	0.319 ± 0.017	−40–130	57	55	<0.043	⋯	⋯	⋯
J1142+3016	175.551	30.270	0.481	104	0.5	50	0.886 ± 0.023	−200–170	3	67	0.227 ± 0.019	−150–50	−33	53
J1155+2922	178.970	29.377	0.520	208	1.2	1	0.742 ± 0.023	−200–230	72, −150	60,13	0.056 ± 0.015	40–160	92	66
J1241+2852	190.374	28.870	0.589	198	1.2	40	0.211 ± 0.020	−120–170	33	95	<0.040	⋯	⋯	⋯
J1251+0554	192.853	5.906	1.377	200	0.8	57	0.409 ± 0.021	−20–180	80	51	<0.063	⋯	⋯	⋯
J1305+0357	196.351	3.959	0.545	103	0.6	11	0.821 ± 0.016	−160–180	67, −43	48, 59	⋯	⋯	⋯	⋯
J1315+1525	198.938	15.432	0.448	155	0.5	17	0.405 ± 0.018	−50–170	64	48	0.131 ± 0.017	0–150	49	48
J1318+2628	199.508	26.475	1.234	198	1.2	86	0.184 ± 0.032	−120–120	0	50	⋯	⋯	⋯	⋯
J1325+2717	201.266	27.289	0.522	199	1.3	55	⋯	⋯	⋯	⋯	<0.095	⋯	⋯	⋯
J1348+2456	207.093	24.947	0.293	153	1.0	81	0.474 ± 0.035	−230 − 0	−97	76	<0.073	⋯	⋯	⋯
J1354+2430	208.604	24.502	1.878	155	0.8	78	0.545 ± 0.034	−170 − 50	−97, −5	37,26	<0.084	⋯	⋯	⋯
J1404+3353	211.118	33.895	0.549	111	0.5	57	0.749 ± 0.027	−150–150	−26	75	0.177 ± 0.024	0–150	37	45
J1406+0157	211.732	1.954	0.427	222	1.1	67	⋯	⋯	⋯	⋯	<0.061	⋯	⋯	⋯
J1427+2632	216.898	26.537	0.364	170	0.8	0e	⋯	⋯	⋯	⋯	<0.078	⋯	⋯	⋯
J1429+0321	217.420	3.357	0.253	231	1.2	0e	0.807 ± 0.027	−150–250	−39, 109	68, 104	0.052 ± 0.017	−50–50	−21	42
J1431+2442	217.858	24.706	0.407	110	0.4	18	0.569 ± 0.015	0–220	73,156	40,28	<0.048	⋯	⋯	⋯
J1454+3046	223.601	30.783	0.465	223	1.4	37	0.472 ± 0.035	−50–160	57	47	<0.079	⋯	⋯	⋯
J1502+0645	225.517	6.754	0.288	224	1.2	80	0.438 ± 0.013	−150–110	12, −55	38,72	<0.032	⋯	⋯	⋯
J1509+0702	227.368	7.043	0.418	130	0.8	62	0.956 ± 0.022	−275–130	49, −18, −215	29, 84, 30	0.137 ± 0.011	−275 to −195	−239	21
J1515+0657	228.781	6.952	0.268	180	0.6	14	0.270 ± 0.023	−500 to −300	−367	43	<0.061	⋯	⋯	⋯
J1541+2817	235.340	28.285	0.376	128	0.7	0e	0.864 ± 0.011	−520 to −250	−363	90	<0.039	⋯	⋯	⋯
J1544+2743	236.114	27.723	0.163	196	1.0	55	0.191 ± 0.022	50–210	126	69	<0.064	⋯	⋯	−

Notes.

^aOrientation of the QSO sightlines with respect to the disk of the galaxies. The values are based on SDSS r-band photometric measurements. ^bLimiting equivalent width denotes 3σ uncertainity. ^cFull width of the absorption feature in the rest-frame of the galaxy. ^dCentroid and b-value of the multiple components of the Lyα and Si iii absorption feature, as estimated via Voigt profile fit. These are printed in the order of each component's strength. Redshift of the absorber z_abs = z_gal + v_transition/c, where ${z}_{\mathrm{gal}}$ is the systemic redshift of the galaxy (from Table 1) and c is the speed of light in vacuum. ^eFace-on galaxies.

Download table as:  ASCIITypeset image

2.2. Observations and Data Reduction

The COS-GASS survey covered a wavelength range of ≈1150–1450 Å  for most galaxies. This includes the prominent transitions like H i Lyα λ1216, O i λ1302, C ii λ1334, Si ii λ1260, 1193, 1190, Si iiiλ1206, Si ivλλ 1393, 1402, and N vλ1239. H i Lyα and Si iii λ1206 are the strongest lines detected in our sample. In this paper, we will primarily focus on these, the two most sensitive probes. A paper presenting all the other metal-lines detected in COS-GASS survey is in preparation. Table 2 presents the measurements for Lyα and Si iii for each of the sightlines from the COS-GASS sample. In cases where we do not detect any absorption features, we quote a 3σ equivalent width as the upper limit. The detection limits for the COS-GASS sample are typically ∼50 mÅ, which corresponds to Log N(H i) = 12.96, Log N(Si ii) = 12.55, Log N(Si iii) = 12.37, Log N(Si iv) = 13.05, and Log N(C ii) = 13.39, respectively.

The measurements for COS-Halos sightlines can be found in the published work by Werk et al. (2013).

3.1. Overall Detection Rates

3.2. An Overview of the CGM Properties

We begin by summarizing the basic structural and kinematic properties of the CGM. Later, we will consider the dependence of these properties on the star-forming characteristics of the central galaxy.

3.2.1. Structure

The dark matter halo mass of the galaxy should influence the size and kinematic properties of the CGM (Hummels et al. 2012; Ford et al. 2016, and references therein). For example, it is expected that a galaxy with a larger halo mass could contain a more massive CGM and gravitationally bind it to larger radii (Chen et al. 2001b). In order to explore the radial profile of the CGM while accounting for the large range in halo mass, we use the variable ρ/R_vir, which we refer to as the normalized impact parameter (e.g., Stocke et al. 2013). This parameter scales the impact parameter (ρ) in terms of the size of the dark matter halo (R_vir). By doing so, we standardize the position of the sightlines for galaxies of different halo masses, and consequently, CGM sizes. Similar analyses have been performed on different data sets by Stocke et al. (2013) and Liang & Chen (2014), and on COS-Halos and COS-GASS by Tumlinson et al. (2013) and Borthakur et al. (2015), respectively.

We show the radial distribution of the equivalent width of Lyα normalized with respect to the virial radius of the galaxies in Figure 2. The distribution can be fit as a exponential with a scale-length of 1.1 R_vir, i.e., ${W}_{\mathrm{Ly}\alpha }=A\,{e}^{-\rho /1.1{R}_{\mathrm{vir}}}\,\mathring{\rm A}$, where the normalization factor, A, is equal to 0.9 Å. The fit was derived using the Buckley–James¹⁷ method (Buckley & James 1979) and Expectation-maximization algorithm as implemented in the survival analysis software package, ASURV (Feigelson & Nelson 1985). The equivalent width data presented here are the same as that of Figure 2 of Borthakur et al. (2015), however, the abscissa is different, because we have adopted the Kravtsov et al. (2014), Liang & Chen (2014) formalism with modifications based on the findings of Mandelbaum et al. (2016) for halo masses and virial radii. In addition, the fit presented here takes into account the censored data; hence, it has slightly different parameters.

Zoom In Zoom Out Reset image size

Similarly, the radial distribution of the equivalent width of Si iii (see right panel of Figure 2) can be fit as a exponential with a scale-length of 0.4 R_vir i.e., ${W}_{\mathrm{Si}{\rm{III}}}=0.4\,{e}^{-\rho /0.4{R}_{\mathrm{vir}}}\,\mathring{\rm A}$. Almost all of our Si iii absorbers were detected inward of 0.8 R_vir. The smaller characteristic size scale for the Si iii absorbers compared to Lyα and the lack of Si iii detections beyond about 0.8 R_vir are consistent with the conclusions drawn by Liang & Chen (2014).

Given the systematic radial decline in the strengths of both the Lyα and Si iii absorbers, we define a new parameter: the impact parameter corrected equivalent width (hereafter, the excess equivalent width, [Log W − $\overline{\mathrm{Log}\,W}{]}_{\mathrm{ion}}$). This refers to the offset in log ${W}_{\mathrm{ion}}$ in any individual sightline, with respect to the best-fit exponential for the entire sample. This parameter is then independent of impact parameter biases and allows us to compare all the absorbers in a uniform way.

3.2.2. Kinematics

We turn now to the kinematic properties of the CGM. To begin, a useful way of visualizing these properties is via the one-dimensional cross-correlation function. The cross-correlation function between galaxies and velocity centroids of Lyα absorbers is presented in the Figure 3. The cross-correlation function shows a strong signal for the presence of Lyα absorbers within 120 km s⁻¹ of the galaxy systemic velocity. The systemic velocity is defined as the velocity corresponding to the optical spectroscopic redshift from SDSS that traces the stars that form the bulk of the baryonic material in the central region of the galaxies.

Zoom In Zoom Out Reset image size

The uncertainties were derived using a Jackknife error estimator, and are indicated as the brown line. They are dominated by small number statistics although the random pairs were generated by using 10,000 Monte Carlo samples. The cross-correlation function also takes into account the same data analysis criteria as the observations. For example, we have taken into account the observational aspect of identifying features by designating absorbers within ±600 km s⁻¹ of the galaxy systemic as associated absorbers. For estimating the cross-correlation function, we randomly distributed the Lyα absorbers within ±600 km s⁻¹ of the galaxy systemic velocity. Therefore, it is not advisable to compare this analysis with blind galaxy-absorber cross-correlation functions, such as those published by Lanzetta et al. (1998), Chen et al. (2005), Ryan-Weber (2006), Wilman et al. (2007), Chen & Mulchaey (2009), Tejos et al. (2012), and others. We also considered each component of the Lyα absorption features as an individual absorber. This should be given due consideration when comparing our results to those from other studies that use spectrographs with significantly different velocity resolution from COS.

Similar conclusions can be drawn for the Si iii transition. The distribution of Si iii is also closely related to that of Lyα absorption, although not all Lyα absorbers have associated Si iii absorption. To illustrate the velocity distribution of Si iii with respect to Lyα, we calculated the cross-correlation function between the velocity offset between Lyα and Si iii. In doing so, we preserved the relationship between distribution of Lyα centroids w.r.t the galaxy systemic—which implies that Lyα is not randomly distributed w.r.t v = 0. The cross-correlation function is presented in Figure 4. The red line indicates the cross-correlation function with the uncertainties shown in brown. The uncertainties were calculated using the same procedure as described previously. The correlation is strongest within ±40 km s⁻¹ of the Lyα absorbers and drops gradually. The signal-to-noise is greater than 3.5σ up to 160 km s⁻¹.

Zoom In Zoom Out Reset image size

We now consider how the CGM kinematics depend upon the halo mass. Figure 5 shows the velocity offset of the absorbers relative to the systemic velocity of the galaxy ($| {v}_{\mathrm{Ly}\alpha }-{v}_{\mathrm{sys}}|$, hereafter ${\rm{\Delta }}v$) plotted as a function of dark matter halo mass. We find that the velocity distribution of the centroids of majority of the Lyα absorbers (as well as other detected metal absorbers) to typically lie within 200 km s⁻¹ of the systemic velocity of the galaxy. The escape velocities at impact parameters of 100, 200, 300 kpc, and ${R}_{\mathrm{vir}}$ are plotted as dashed and solid curves of different colors. The centroids of the strongest component are shown as the filled symbols and the other components are shown as open symbols. The colored vertical lines connecting the strongest component to the other components mark the full-width of the Lyα profiles. Nearly all the strongest Lyα components have velocity centroids within the escape velocity of their host galaxies. Therefore, we expect this material to be bound to the galaxies.

Zoom In Zoom Out Reset image size

Another way to characterize the kinematics of the CGM is to use the widths of the absorption features rather than their velocity displacement. In Figure 6, we plot the distribution of line-of-sight velocity dispersion of the strongest component of the Lyα feature (${\sigma }_{\mathrm{los}}=b/\sqrt{2}$, where b is the Doppler parameter in our fits). Several features are noteworthy in this pair of figures. First, the lines are generally very narrow (mean σ_los ∼ 30 km s⁻¹). Second, there is no trend for these widths to increase as the halo mass (virial velocity) increases (as was also the case for Δv).

Zoom In Zoom Out Reset image size

A compact representation of the information in the figures discussed above is given in Figure 7. We define the kinematic parameter $W={({\sigma }_{\mathrm{bins}}^{2}+{\sigma }_{\mathrm{avg}}^{2})}^{1/2}$. Here, ${\sigma }_{\mathrm{bins}}$ is the velocity dispersion of the distribution of the velocity differences between the Lyα absorber and the galaxy systemic velocities within a given bin in $\mathrm{log}\,{M}_{\mathrm{halo}}$. The term ${\sigma }_{\mathrm{avg}}$ is the average value of ${\sigma }_{\mathrm{los}}$ of the individual Lyα absorbers in this same bin in halo mass. Again, we see no dependence of CGM kinematics on halo mass. In particular, the low (sub-virial) velocity spread of CGM absorbers in the halos of massive red galaxies has been noted before (Zhu et al. 2014; Huang et al. 2016). We will discuss the possible implications of these results in Section 4.

Zoom In Zoom Out Reset image size

We can also examine the radial dependences of the CGM kinematics. In Figure 8, we plot the Lyα velocity distribution as a function of normalized impact parameter. This figure suggests that the velocity offset of the absorbers from systemic velocity (${\rm{\Delta }}v$) drops in the outer CGM. The differing kinematic properties of the inner versus the outer CGM are shown in histogram form in Figure 9. An F-test shows that the distributions differ at >99.99% (97.4%) confidence level for the blue (red) galaxies. In Figure 10, we show a similar plot of the radial dependence of the line-of-sight velocity dispersions (${\sigma }_{\mathrm{los}}$) of the Lyα absorbers. This shows that there is a statistically significant (99.6% confidence level) trend for ${\sigma }_{\mathrm{los}}$ to increase with increasing impact parameter.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We next define the dimensionless quantity ${\rm{\Delta }}v/{\sigma }_{\mathrm{los}}$, and plot it in Figure 11 as a function of normalized impact parameter. The most interesting result is that ${\rm{\Delta }}v/{\sigma }_{\mathrm{los}}$ has a mean value of about four interior to the virial radius. This figure also shows that the ratio declines in the outer CGM (a result significant at the 99.9% confidence level). This is due to the combined effects of the drop in the ${\rm{\Delta }}v$ in the outer CGM and the radial rise in ${\sigma }_{\mathrm{los}}$ that were described above.

3.3. The CGM in Blue versus Red Galaxies

We now compare the radial distributions of the absorbers in the CGM surrounding the blue versus the red galaxies. The radial Lyα profiles as a function of normalized impact parameter ($\rho /{R}_{\mathrm{vir}}$) are shown in Figure 12. One clear difference is the dispersion in the data between the two sub-samples. The blue galaxies show a fairly uniform radial distribution with a >95% detection rate of Lyα absorbers out to $\sim {R}_{\mathrm{vir}}$. On the other hand, the red galaxies show a much larger dispersion in the radial distribution (as indicated by weak absorption features, as well as non-detections with very good upper limits). It is worth noting that red galaxies do occasionally exhibit strong Lyα absorbers associated with their CGM, but their detection rate is not as large as in blue galaxies. This is particularly true for the inner CGM, and suggests that the warm CGM is more patchy in the red galaxies (has a smaller areal covering factor).

Zoom In Zoom Out Reset image size

The dashed black line is the fit to entire sample, whereas the blue and red solid lines are fits to blue and red galaxies respectively. The implied exponential scale-lengths for the blue and red galaxies are similar to one another (0.65 R_vir and 0.75 R_vir respectively). The difference in the normalization of the profiles of 0.45 dex (blue versus red) reflects the patchy nature of the absorbers in the CGM of the red galaxies.

Histogram representations of the excess Lyα equivalent widths for the blue and red galaxies are presented in Figure 13. The dispersion in the red galaxy sub-sample is much higher than in the blue galaxy sub-sample. Again, this signifies the difference in the covering fraction of neutral gas between the two populations. The distribution in the excess Lyα equivalent widths between the blue and red galaxies is significant enough that these two sub-samples can be considered as different populations with 99.9% confidence based on Logrank test statistics, using the software package ASURV.

Zoom In Zoom Out Reset image size

We find similar results for Si iii (Figure 14). The exponential scale lengths are similar for the blue and red galaxies (0.33 R_vir and 0.36 R_vir respectively). The normalization of the fit to the equivalent width radial distribution is 0.44 dex higher for the blue galaxies, which again reflects the patchy nature of the absorbers in the CGM of the red galaxies. The histogram showing the distributions of the excess Si iii equivalent widths is shown in Figure 15. The difference in the distribution of the excess Si iii equivalent width between the red and the blue galaxies is significant at the 99.8% confidence level.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

By and large, the kinematic properties of the CGM are quite similar between the blue and red galaxies. However, one notable difference is highlighted in Figure 16, which shows histograms of ${\rm{\Delta }}v$ for the individual Lyα absorbers. Although the majority of the values for ${\rm{\Delta }}v$ are less than 100 km s⁻¹ in both samples, the blue galaxy histogram has a pronounced tail extending out to ${\rm{\Delta }}v\sim 500$ km s⁻¹. An analysis of variance (ANOVA) F-test reveals that the blue and red samples differ at the 98.5% confidence level.

Zoom In Zoom Out Reset image size

3.4. Correlation between CGM Properties and SFR

In the discussion above, we have simply classified galaxies as star-forming (blue) or quiescent (red). In this section, we focus on quantitative measures of the SFR and the sSFR. In our recent study, we found a strong correlation between the neutral hydrogen content in the interstellar medium of galaxies (traced by the 21 cm hyperfine transition) and the Lyα equivalent width in the outer CGM probed by the COS-GASS sample (Borthakur et al. 2015). We also found correlations between Lyα strengths and both SFR and sSFR, but these were significantly weaker than those with the H i 21 cm mass or mass fraction. Here, we reinvestigate the correlations with SFR and sSFR, using a much larger combined sample that better covers the full radial range of the CGM.

As seen in Figure 17, we find a positive correlation between the equivalent width of the Lyα absorbers and the SFR at the 99.8% confidence level. We find an even stronger correlation with the excess equivalent width [Log W - $\overline{\mathrm{Log}\,W}{]}_{\mathrm{Ly}\alpha }$ and the galaxy sSFR (at the 99.99% confidence level.¹⁸ We also find similar correlations between SFR and the equivalent width of the Si iii and between the excess Si iii equivalent width ${[\mathrm{Log}W-\overline{\mathrm{Log}W}]}_{\mathrm{Si}{\rm{III}}1206}$ and sSFR (see Figure 18).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We want to highlight a number of the results above, and try to connect them together into a simple picture of the warm ionized phase of the CGM in both the blue (star-forming) and red (quiescent) galaxies.

We have found that the distribution of Si iii absorbers is more compact than that of the Lyα absorbers for both the red and blue galaxies (with exponential length scales of ∼0.35 versus 0.7 R_vir for the respective ions in both the blue and red galaxies). As a consequence, the detection fraction of Si iii for the full sample drops from 67% inside 0.7 R_vir to only 17% outside. This was also seen by Liang & Chen (2014), who interpreted it in terms of a boundary that represents the region that has been significantly enriched by metals expelled from the central galaxy (e.g., affected by stellar feedback at some point in the evolution of the galaxy and its CGM). We also see a change in the distribution of the velocity offsets between the Lyα lines and galaxy systemic velocities in the outer CGM. Although this could be related to feedback processes, it could also be due to line-of-sight projection effects, if the flow pattern in the CGM has a strong radial component (inward and/or outward).

Perhaps the most surprising result is that neither the red nor blue populations shows any trend for either the velocity offset of the Lyα absorbers with respect to the galaxy systemic velocity (${\rm{\Delta }}v$) or the line-of-sight velocity dispersions (${\sigma }_{\mathrm{los}}$) of the absorption lines to increase with increasing halo mass across a range of about 5 in implied virial velocity. The implied sub-virial velocities in the CGM around massive red galaxies have been noted before (Zhu et al. 2014; Huang et al. 2016), and are particularly mysterious. One possibility is that we are seeing the condensation of thermal instabilities out of a hot volume-filling phase (Voit et al. 2015). The hot gas is supported hydrostatically, and the denser cooler clouds that condense out have not (yet) been accelerated by gravity to the virial velocity of the halo. This would suggest that these clouds in massive red galaxies are transient (lifetimes less than a halo crossing time).

Another possibility is that the flow of the absorbing material in the warm CGM of these galaxies is significantly affected by drag forces associated with a more massive hot volume-filling phase. In any case, it appears that the CGM dynamics are not purely determined by gravitational forces alone. Because we expect that the absorption-line systems we see are imbedded in a multi-phase halo, their dynamics are likely to be influenced by processes such as drag forces, thermal instabilities, turbulent mixing, merger dynamics, and feedback-driven outflows (Maller & Bullock 2004; Santillan et al. 2007; Kwak & Shelton 2010; Kwak et al. 2011; Joung et al. 2012a, 2012b; Stinson et al. 2012; Ford et al. 2014; Suresh et al. 2015; Fielding et al. 2016).

The role of drag forces in reducing the velocities of the CGM clouds in massive halos is particularly interesting to consider. Following Bordoloi et al. (2016), it is straightforward to show that the terminal velocity for a CGM cloud that is significantly affected by drag is given by:

$$\begin{eqnarray}&&{v}_{\mathrm{term}}\sim {v}_{\mathrm{vir}}{({M}_{\mathrm{cs}}/{M}_{\mathrm{vf}})}^{1/2}\end{eqnarray} \tag{ 1 }$$

Here, v_vir is the virial velocity of the halo, and M_vf and M_cs are the total masses of the volume-filling gas and the system of clouds in the CGM. For drag forces to be important, the volume-filling phase needs to significantly exceed the total cloud mass. The amount of mass in a volume-filling phase is uncertain in typical Milky Way-like galaxies (e.g., Miller & Bregman 2013), but appears to be significant in the halos of more massive galaxies (Planck Collaboration et al. 2013; Greco et al. 2015).

We also find that the ratio ${\rm{\Delta }}v/{\sigma }_{\mathrm{los}}$ is typically large (mean value of ∼4, see Figure 11) interior to the virial radius. This is true in both the blue and red galaxies. This result is inconsistent with at least one simple, and otherwise plausible, model in which a single line of sight through the CGM intersects many clouds with distinct locations and a wide range of line-of-sight velocities (e.g., a sea of clouds orbiting in the halo potential). Instead it implies that a typical line of sight through the CGM is intersecting a coherent structure (a cloud, sheet, or filament). Current constraints on the size of these structures (which are based on models in which the gas is photoionized by the meta-galactic background) are rather weak, but characteristic path-lengths are of-order 1–10 kpc (Stocke et al. 2013; Werk et al. 2014).

Despite many of the similarities between the CGM in red and blue galaxies noted above, we do find significant differences. First, in terms of the radial distributions of the Lyα and Si iii equivalent widths, the red galaxies have lower normalizations for the exponential fits, reflecting a patchy distribution of absorbers. There is also a significant fraction of the blue galaxies showing large velocity differences between the radial velocity of the absorber and central galaxy (${\rm{\Delta }}v$ up to 500 km s⁻¹). Taken together with the COS-Halos results on the presence (absence) of highly ionized gas (seen as O vi absorbers Tumlinson et al. 2011) in blue (red) galaxies, there are therefore real differences between the CGM surrounding star-forming and quiescent galaxies. Unfortunately, as discussed in Section 3.1, the data from the COS-GASS do not cover O vi; thus, we are not able to do the same analysis as done by Tumlinson et al. (2011) for the combined sample.

The interpretation of these differences is not straightforward. The direction of the causal connection between the properties of the CGM and the galaxy could be in either direction (or both). The CGM properties could be influenced by feedback from the galaxy, and the amount and nature of this could be significantly different between the blue and red galaxies (e.g., feedback from massive stars and supernovae versus feedback associated with AGN-driven radio sources ("radio mode"). Alternatively (or in addition) the star-forming properties of the galaxies may be driven by the different observed properties of the CGM.

One the main motivations of this paper is to understand whether and how the properties of the CGM in star-forming and quiescent galaxies relate to the cessation of star formation in the latter. The results we have presented seem to imply rather subtle differences between the two types of galaxies. Here, we will argue that the apparently subtle differences could have significant implications.

Let us consider a simple model in which the star formation rate in a given galaxy is proportional to the total mass of the system of CGM clouds traced by the absorption-lines (M_cs) divided by the timescale for these clouds to be transported from the CGM to the galaxy (${t}_{\mathrm{in}}\propto {R}_{\mathrm{vir}}/{v}_{\mathrm{in}}$). The results in Figure 2 imply that ${M}_{\mathrm{cs}}\propto {f}_{c}{R}_{\mathrm{vir}}^{2}$, where f_c is the fraction of the cross-sectional area of the CGM ($\pi {R}_{\mathrm{vir}}^{2}$) covered by these clouds. Empirically, we see no dependence of the cloud kinematics on the mass of the dark matter halo. Hence, we will assume that v_in is likewise independent of M_halo. Taken together, this implies that the sSFR scales as $\mathrm{sSFR}\propto {f}_{c}{R}_{\mathrm{vir}}^{2}/{R}_{\mathrm{vir}}{M}_{* }\propto {f}_{c}{R}_{\mathrm{vir}}/{M}_{* }$. Using our adopted scaling between ${M}_{* }$ and R_vir, the predicted value of sSFR drops by a factor of ten, with increasing stellar mass over the range in ${M}_{* }={10}^{10.0}$ to ${10}^{11.5}$ ${M}_{\odot }$, even if f_c does not change. Given that f_c is a factor of 3 smaller in the red galaxies (Figures 12 and 14), this simple model then predicts a difference in the sSFR between the lowest-mass blue galaxies and highest-mass red galaxies in our sample of a factor of 30.

Of course, this model is purely phenomenological and does not explain the scaling of M_cs with R_vir or the invariance in v_in. These relations are simply based on the observations. One possible interpretation would be that the mass of the population of clouds relative to that of the hot diffuse volume-filling CGM phase drops as a function of M_halo. This reduces the normalized mass of the reservoir of clouds (${M}_{\mathrm{cs}}/{M}_{* }$) and increases the transport time of the clouds to the galaxy with increasing mass due to increasing drag forces. Such an idea is, at least qualitatively, consistent with the simple paradigm of a transition from predominantly cold accretion to a quasi-hydrostatic hot CGM as halo and galaxy mass increases.

We have presented the analysis of a comprehensive data set combining the COS-GASS and COS-Halos samples to probe the CGM of low-z galaxies spanning a stellar mass range of almost two orders-of-magnitude centered on the characteristic mass ($\sim {10}^{10.5}{M}_{\odot }$) at which the galaxy population transitions from mostly blue, star-forming galaxies to red, quiescent ones. These two surveys cover similar ranges in stellar masses and dark halo virial radii (R_vir). In addition, the combined sample uniformly samples a large range of radial distances from 0.02 to 1.3 R_vir. The COS-GASS survey primarily samples the outer CGM and COS-Halos survey primarily samples the inner CGM. We characterized the CGM properties, including its radial profile, its kinematics, and its correlation with the global properties of the galaxies. In particular, we have divided the sample into blue galaxies (with sSFR $\gt \,{10}^{-11}\,{\mathrm{yr}}^{-1}$) and red galaxies (with lower sSFR).

In this analysis, we discussed the Lyα λ1215 Å  and Si iii λ1206 Å  transitions tracing intermediate ionization gas. Si iii was chosen because it is the strongest metal transition detected in the combined data set. The typical detection limit for the COS-GASS sample is ∼50 mÅ  corresponding to 3σ uncertainty in the data. In the combined sample, the detection rates of Lyα and Si iii were 91% and ∼50%, respectively.

Based on the analysis of the combined sample, we conclude the following:

• 1.  
  The radial distribution of the equivalent width of Lyα as a function of normalized impact parameter ($\rho /{R}_{\mathrm{vir}}$) can be expressed as an exponential. The scale-lengths are similar for the red and blue galaxies (0.75 and 0.72 R_vir, respectively). The radial distribution of equivalent width of Si iii can also be expressed as an exponential with a scale-length of 0.36 (0.33) R_vir for the red (blue) galaxies. The detection rate of Si iii drops to almost zero beyond about 0.8 R_vir.
• 2.  
  The blue galaxies show a relatively uniform radial distribution of Lyα absorbers, implying an areal covering fraction of nearly 100% in the CGM. In contrast, the Lyα absorbers have a much less uniform radial distribution in the CGM of the red galaxies, suggesting a patchy distribution with smaller areal covering fractions. These differences are reflected in the overall normalization of the radial distribution of equivalent widths, which is higher for the blue galaxies (by 0.45 dex). Similar results were found for Si iii, but are restricted to the region interior to 0.8 R_vir (where Si iii is detected).
• 3.  
  We found a significant positive correlation between the equivalent width of Lyα and the star formation rate (at the 99.8% confidence level). The correlation is even more significant for normalized quantities: the impact- parameter-corrected equivalent with of Lyα (${[\mathrm{Log}W-\overline{\mathrm{Log}{\text{}}W}]}_{\mathrm{Ly}\alpha }$) and the specific SFR (SFR/M_⋆) were found to correlate at the 99.99% confidence level. Similar results were found for Si iii.
• 4.  
  We found the velocity distribution of the centroids of the majority of the Lyα and Si iii to generally lie within ∼150 km s⁻¹ of the systemic velocity of the galaxy. These velocities are smaller than the escape velocity, thus suggesting that the gas seen in absorption is the gravitationally bound within the halo. The metal-line transitions are also found mostly within ±100 km s⁻¹ of Lyα absorbers, although not all strong (>0.3 Å) Lyα absorbers showed associated Si iii.
• 5.  
  We find that the velocity offset between the Lyα centroid and the systemic velocity (${\rm{\Delta }}v$) is usually significantly larger than the line-of-sight velocity dispersion of the Lyα line (${\sigma }_{\mathrm{los}}$). The mean ratio ${\rm{\Delta }}v/{\sigma }_{\mathrm{los}}\sim 4$.
• 6.  
  We find no dependence of the kinematic properties of the CGM (${\rm{\Delta }}v$ or ${\sigma }_{\mathrm{los}}$) on the galaxy halo mass (virial velocity). This is surprising, because the sample spans ranges of about 10² in halo mass and ∼5 in v_vir.
• 7.  
  We found that the kinematic properties of the CGM are generally similar between the blue and red galaxies. However, although the majority of both the blue and red galaxies have Δv < 100 km s⁻¹, the distribution of ${\rm{\Delta }}v$ for the blue galaxies shows a pronounced tail out to values as high as 500 km s⁻¹.
• 8.  
  We found a significant change in the CGM kinematics at about a radius of 1.0 (0.7) R_vir for the blue (red) galaxies. In the outer CGM, ${\rm{\Delta }}v$ for the Lyα absorbers is always less than 150 km s⁻¹, whereas the distributions of ${\rm{\Delta }}v$ show tails out to values as high as 500 km s⁻¹ in the inner CGM. In addition, ${\sigma }_{\mathrm{los}}$ is higher, on average, in the outer CGM for both the blue and red galaxies. These two results lead to a corresponding decrease in ${\rm{\Delta }}v/{\sigma }_{\mathrm{los}}$ in the outer CGM.

The combined COS-GASS and COS-Halos sample has allowed us to conduct a comprehensive study of the connection of the properties of the CGM with those of the stellar body of the galaxy. We think that three of the specific results from above are particularly noteworthy. First, the differences in the radial distributions of the Lyα, versus the Si iii absorbers, suggest that the inner CGM is being (or has at some time been) affected by feedback associated with massive stars and supernovae. This feedback has chemically enriched the CGM. It is interesting that the kinematics of the inner and outer CGM also show differences, although it is not clear that these are related to feedback or to projection effects.

Second, the fact that the typical ratio of the velocity offset to the line-of-sight velocity dispersion for the Lyα absorption-lines is so large is an important clue as to the structure of the CGM. It implies that a line of sight through the CGM does not intersect a whole sea of many clouds orbiting in the halo, but is rather passing through a coherent structure (cloud, sheet, filament). Model-dependent estimates imply a path-length of order 1–10 kpc for these structures.

Third, the independence of the kinematic properties of the warm CGM on the halo mass is quite remarkable. This implies that, even though the observed absorption-line systems are mostly gravitationally bound to the halo, simple gravitational forces alone do not adequately explain the CGM dynamics. For the massive red galaxies, the "sub-virial" velocities could be understood if the absorbers represent material cooling and condensing out of (or suffering drag as they move through) a hot volume-filling phase that is supported hydrostatically against gravity.

A major motivation of this study is to understand how and why galaxies in this stellar mass regime exhibit the color bimodality stemming from a suppression/cessation of star formation in some of them. Because the CGM is the interface through which galaxies could exchange gas and energy that is required to form stars (or is expelled as a result of star formation), CGM properties could hold clues as to how this process of gas delivery may be disrupted leaving some galaxies deprived of fuel to form stars. We have explored a simple scenario in which the star formation rate in a galaxy is proportional to the total mass of CGM clouds divided by an inflow time. We then show that the empirical results on the independence of CGM kinematic properties on halo mass and the smaller covering factor in the CGM in the red galaxies would imply a drop in the sSFR by about a factor of 30 over the stellar mass range from 10¹⁰ to 10^11.5 M_⊙.

In any event, we believe the data we have presented here provide a valuable observational resource for ongoing and future numerical simulations that try to reproduce CGM properties such as Lyα and metal-line column density profiles, covering fraction, and dynamics, such as line-widths and velocity spreads (for example Hummels et al. 2012; Stinson et al. 2012; Ford et al. 2013, 2014, 2016; Shen et al. 2013; Fielding et al. 2016; Kauffmann et al. 2016; Liang et al. 2016, and references therein). The ultimate goal is understanding the role of the CGM in the evolution of galaxies.

We thank the referee for his/her useful comments. We thank Hsiao-Wen Chen, Cameron Hummels, Colin Norman, Josh Peek, Molly Peeples, Jason X. Prochaska, John Stocke, and Jessica Werk for helpful discussions. This work is based on observations with the NASA/ESA Hubble Space Telescope, which is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS5-26555. S.B. and T.H. were supported by grant HST GO 12603. B.C. gratefully acknowledges support from the Australian Research Council's Future Fellowship (FT120100660) and Discovery Project (DP150101734) funding schemes."

This project also made use of SDSS data. Funding for the SDSS and SDSS-II has been provided by the Alfred P. Sloan Foundation, the Participating Institutions, the National Science Foundation, the U.S. Department of Energy, the National Aeronautics and Space Administration, the Japanese Monbukagakusho, the Max Planck Society, and the Higher Education Funding Council for England. The SDSS Web Site is http://www.sdss.org/. The SDSS is managed by the Astrophysical Research Consortium for the Participating Institutions. The Participating Institutions are the American Museum of Natural History, Astrophysical Institute Potsdam, University of Basel, University of Cambridge, Case Western Reserve University, University of Chicago, Drexel University, Fermilab, the Institute for Advanced Study, the Japan Participation Group, Johns Hopkins University, the Joint Institute for Nuclear Astrophysics, the Kavli Institute for Particle Astrophysics and Cosmology, the Korean Scientist Group, the Chinese Academy of Sciences (LAMOST), Los Alamos National Laboratory, the Max-Planck-Institute for Astronomy (MPIA), the Max-Planck-Institute for Astrophysics (MPA), New Mexico State University, Ohio State University, University of Pittsburgh, University of Portsmouth, Princeton University, the United States Naval Observatory, and the University of Washington.

Facilities: Sloan - Sloan Digital Sky Survey Telescope, HST (COS). - Hubble Space Telescope satellite