H i Observations of Major-merger Pairs at z = 0: Atomic Gas and Star Formation

It has been well documented that galaxy–galaxy interaction can induce enhanced star formation (Larson & Tinsley 1978; Keel et al. 1985; Kennicutt et al. 1987; Bushouse et al. 1988; Telesco et al. 1988; Sulentic 1989; Xu & Sulentic 1991; Barton et al. 2000; Lambas et al. 2003; Alonso et al. 2004; Nikolic et al. 2004; Ellison et al. 2008; Li et al. 2008; Xu et al. 2010). More recently, Scudder et al. (2012) and Patton et al. (2013) found that wide galaxy pairs with separation as large as ∼80 kpc still show significant star formation rate (SFR) enhancement at the ∼40% level. Early studies (Hummel 1981; Haynes & Herter 1988; Bergvall et al. 2003) that failed to detect SFR enhancement in interacting galaxies may have suffered from biases in selecting the interacting galaxy sample and the control sample (c.f. Ellison et al. 2010; Xu et al. 2010).

Of particular interest are close (separation ≤20 h⁻¹ kpc) major mergers of galaxies of nearly equal mass (primary-to-secondary mass ratio ≲3). Most extreme starbursts, such as ultra-luminous infrared galaxies (ULIRGs), are close major mergers (Sanders & Mirabel 1996; Dasyra et al. 2006). For a sample of K-band-selected close major-merger pairs, IR observations carried out using Spitzer (Xu et al. 2010) and Herschel (Cao et al. 2016) found that in these pairs the average specific SFR (sSFR = SFR/M_star) in spirals is a factor of ≳2 higher than that of their counterparts in the control sample. Furthermore, spirals in spiral–spiral pairs (S+S pairs) are strongly enhanced, with a mean sSFR ≳ 3 times higher than that of control galaxies, but spirals in mixed spiral–elliptical pairs (S+E pairs) do not show any significant SFR enhancement compared to the control galaxies. Using WISE and Herschel data, Domingue et al. (2016) also found that spirals in S+S pairs exhibit significant enhancements in the interstellar radiation field and dust temperature, while spirals in S+E pairs do not.

Why is the sSFR enhancement of spirals in mixed S+E pairs different from that of spirals in S+S pairs? If the enhancement is purely due to gravitational tidal effects, then the spirals in S+E pairs should behave similarly to spirals in S+S pairs, unless the former have systematically less cold gas (i.e., the fuel for star formation) than the latter. This possibility has been tested by Cao et al. (2016). Using fluxes in six Herschel bands (70, 100, 160, 250, 350, and 500 μm), they estimated dust mass (M_dust), and assuming a constant dust-to-gas mass ratio, total gas mass (M_gas) for the paired spirals. They found only marginal evidence for spirals in S+E pairs having slightly lower gas content than those in S+S pairs (δlog(M_gas/M_star) = −0.14 ± 0.10). It appears that the difference between the sSFR enhancements of spirals in mixed S+E pairs and of those in S+S pairs is mainly due to their different star formation efficiency (SFE = SFR/M_gas), suggesting significant roles for non-tidal effects (e.g., collision between gas in two galaxies) in the interaction-induced star formation.

In this paper, we present a study of the H i gas content of pairs. The main science goal is to constrain the relation between H i gas content and SFR enhancement, and check the consistency with the relation between gas content (estimated using the dust mass) and SFR enhancement obtained in the Herschel study (Cao et al. 2016). In Sections 2–4, we describe the sample, GBT observations, and data reduction. Literature data are described in Section 5. The main results are presented in Section 6. Section 7 is devoted to discussions. Section 8 provides a summary. Throughout this paper, we adopt the Λ-cosmology with Ω_m = 0.3 and Ω_Λ = 0.7, and H₀ =70 (km s⁻¹ Mpc⁻¹).

The KPAIR is an unbiased and large sample of 170 close major-merger galaxy pairs selected in the K-band, from cross-matches between the Two Micron All Sky Survey (2MASS) and the Sloan Digital Sky Survey (SDSS)-DR5 galaxies (Domingue et al. 2009). The parent sample includes 77,451 galaxies of Ks ≤ 13.5 mag, with a sky coverage of 5800 deg² and redshift completeness of 86%. The selection criteria are: (1) the Ks magnitude of the primary is not fainter than 12.5; (2) at least one component has a measured redshift; (3) if both components have measured redshifts, the velocity difference is not larger than 1000 km s⁻¹; (4) the Ks difference between the two galaxies is not larger than 1 mag; and (5) the projected separation is in the range of 5 h⁻¹ kpc ≤ r ≤ 20 h⁻¹ kpc. When only one component has a measured redshift, the separation is calculated according to that redshift and the angular separation of the components. Visual inspections, complemented by the results of an automatic algorithm, classified 62 pairs as S+S, 56 as S+E, and 52 as E+E.

The H-KPAIR sample (Cao et al. 2016) includes all S+S and S+E pairs in the original KPAIR sample that have (1) measured redshifts for both components, (2) relative velocity <500 km s⁻¹, and (3) pair recession velocity <2000 km s⁻¹. It contains 88 pairs (44 S+S and 44 S+E).

For 67 pairs, the 21 cm H i fine-structure line observations were carried out using the National Radio Astronomy Observatory (NRAO)⁹ Robert C. Byrd Green Bank 110 m Telescope (GBT)¹⁰ Spectrometer in the L-band (1.15–1.73 GHz) between 2012 August and 2013 January. For each object, the data were collected in ∼2 hr on/off source pairs with 12.5 MHz bandwidth. The two spectral windows were centered at the same frequency (1420.4058 MHz). Using 9-level sampling and two IFs, the observations provide 1.5 kHz (0.3 km s⁻¹) spectral resolution for the dual polarization L-band system. The beam size is 9' × 9'. GBT has a well-calibrated structure and a stable gain at the 21 cm wavelength. We observed 3C 286 as the primary flux calibrator to monitor the instrumental performance. This observation of a bright calibration source verified the stability of the telescope gain factor. As a test, we also observed 5 nearby normal galaxies: NGC895, NGC2718, NGC3027, UGC10014, and NGC6140. Comparisons with the literature show a systematic difference of ∼15% between our measurements and data in the literature (see Appendix A), suggesting a minor deviation in the calibration. This shall not affect our main conclusions significantly.

The H i spectra were reduced using GBTIDL (Marganian et al. 2006). We used the Jy/K calibration to convert the H i line fluxes to units of Jy, applying an atmospheric opacity of 0.008 and aperture efficiency of 0.71. The scans and channels with Radio Frequency Interferences (RFIs) were flagged. For each polarization, the data were accumulated and averaged together. A polynomial of the order of 3–8 was used to fit the baseline over a range of ≈4500 channels for every pair. The Hanning-smoothed and then decimated spectra were used to subtract the baseline. The velocity resolution is ∼30 km s⁻¹ per channel after boxcar smoothing. The two polarizations were then averaged together to produce the full intensity spectra shown in Figure 1.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

View figure set (5 panels)

Tarball of figure set images

Some observations were significantly affected by RFIs. For 9 pairs (J0913+4742, J0926+0447, J1010+5440, J1020+4831, J1137+4728, J1148+3547, J1205+0135, J1505+3427, J1628+4109), the RFIs are so severe that no informative signals could be extracted from the data. The pairs are excluded from our analysis. Forty-six targets are detected with the H i 21 cm emission peak >3σ. Their H i masses were calculated using the relationship of M_{H i} = 2.36 × 10⁵D²(SΔV)M_⊙ (Condon & Ransom 2016), where D is the luminosity distance in Mpc and SΔV is the velocity-integrated H i flux density in Jy km s⁻¹. The H i velocity range was visually decided, with the constraint that the center is within ±400 km s⁻¹ from the system velocity (optical) of the target. The H i masses of the 12 undetected pairs were calculated based on the 3σ upper limits of the spectral line with an assumed line width of ∼800 km s⁻¹. The error was estimated by the quadratic sum of the 10% systematic uncertainty (mostly due to the calibration and the baseline subtraction error) and the measured rms noise. We also measured the mean H i velocity (weighted by signal) and the W20 of the spectral line for detected pairs.

For the 46 detected pairs, we visually inspected their SDSS images and found that 22 have neighboring spiral galaxies with redshifts inside the bandwidth of the H i observation and locations within 10' from the pair center (Figure 8 in Appendix B). We performed our search down to the limit of 17.71 mag (i-band), 18.35 mag (g-band) and 17.78 mag (r-band). All spiral galaxies in the search radius and brighter than these limits had observed redshifts. As the beam of GBT is 9', contamination due to blending could be significant for these sources, therefore a correction was carried out based on the algorithm developed by Zhang et al. (2009). Details about the contamination correction are presented in Appendix B. Most of the neighboring galaxies cause minor corrections. The average and the range of the factor by which M_{H i} was changed due to the correction are 1.26 and 1.03–1.84, respectively. Also, as a test for the algorithm, we found that for paired galaxies in our sample, the ratio between estimated and observed H i mass is consistent with unity (0.8 ± 0.2). The H i mass after this correction is listed as M_{H i,c} in column (7) in Table 1.

The 21 pairs in H-KPAIR that we did not observe with GBT were covered by previous H i emission-line observations. However, detailed inspections showed that among them, 9 pairs (J0915+4419, J1015+0657, J1150+1444, J1211+4039, J1219+1201, J1429+3534, J1506+0346, J1514+0403, J1608+2529) are in galaxy groups and the H i observations were not pointed toward the pairs but toward neighboring galaxies in the same group, as shown in Figure 8. Therefore, their H i masses are too uncertain and they are excluded from our analysis. For each of the remaining 12 pairs, the H i data collected from the literature are listed in Table 2. No neighboring spiral galaxies that can cause significant H i contaminations were found for these pairs (Figure 8 in Appendix B).

Table 2.  Pairs from the Literature

(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)
Pair ID	R.A.	Decl.	VH i	MH i	100 × Mdust	Mstar	SFR	Type	References	Beam size
(H-KPAIR)	(J2000)	(J2000)	(km s−1)	(109M⊙)	(109M⊙)	(109M⊙)	(M⊙ yr−1)
J0020+0049	00:20:27.4	+00:49:59	5498	3.30 ± 0.49	1.90	31.24	1.39	SE	1	33
J0823+2120	08:23:32.6	+21:20:16	5400	12.12 ± 1.17	3.96	33.79	3.98	SS	2	100
J0829+5531	08:29:15.0	+55:31:21	7758	25.59 ± 2.65	21.48	85.47	3.98	SS	1	100
J1043+0645	10:43:51.9	+06:46:00	8238	6.70 ± 0.27	12.34	59.98	5.01	SS	3	35 × 38
J1308+0422	13:08:28.3	+04:22:01	7251	8.26 ± 0.26	4.93	23.58	0.66	SS	3	35 × 38
J1315+6207	13:15:34.6	+62:07:28	9100	5.00 ± 0.83	12.77	103.90	62.89	SS	2	100
J1406+5043	14:06:21.7	+50:43:29	1860	1.63 ± 0.17	2.09	13.92	0.92	SE	1	100
J1423+3400	14:23:42.5	+34:00:30	3865	1.03 ± 0.10	2.90	25.92	1.23	SS	1	33
J1425+0313	14:25:05.5	+03:13:59	10680	1.50 ± 0.35	<1.02	23.11	<0.10	SE	4	35
J1444+1207	14:44:20.7	+12:07:55	8895	5.93 ± 0.64	12.84	186.60	4.64	SS	1	33
J1608+2328	16:08:22.5	+23:28:46	12121	14.87 ± 1.50	17.42	71.43	12.19	SS	1	33
J2047+0019	20:47:19.0	+00:19:17	4204	22.44 ± 3.45	15.14	123.10	1.84	SE	1	33

Note. Column descriptions are as follows. (1) Pair ID. The designations are "H-KPAIR J0020+0049," etc. (2) R.A. (h:m:s, J2000). (3) Decl. (d:m:s, J2000). (4) H i velocity. (5) H i mass (10⁹M_⊙). (6) Dust mass multiplied by 100 (10⁹M_⊙). The 100 × M_dust of S+S pairs with only one detection equals that of the detected component. The 100 × M_dust of S+E pairs include only the mass of the spiral component. (7) Stellar mass (10⁹M_⊙). The M_star of S+E pairs includes only the mass of the spiral component. (8) Star formation rate (M_⊙ yr⁻¹). The SFR of S+S pairs with only one detection equals that of the detected component. The SFR of S+E pairs includes only that of the spiral component. (9) Type of major-merger pair. (10) References: 1 Springob et al. (2005), 2 Huchtmeier & Richter (1989), 3 Haynes et al. (2011), 4 Catinella et al. (2010).

Download table as:  ASCIITypeset image

Our final sample includes 70 pairs (34 S+S pairs, 36 S+E pairs) whose H i mass M_{H i} can be found in Table 1 and Table 2. For S+S pairs, since the GBT beam cannot resolve them into individual galaxies, we treated each pair (including both components) as a single source. In these tables, we also listed 100 × M_dust as an estimate of total gas mass, stellar mass M_star, and SFR, all taken from Cao et al. (2016). The corresponding M_star, 100 × M_dust and SFR for each S+S pair are sums of the two components. In cases in which one of the two (100 × M_dust or SFR) is undetected by Herschel, the true value for the pair should be limited between the detection and the sum of the detection plus the upper limit of the undetected component. A test has shown that, for our statistical results (Table 3), the difference between calculations adopting either of these two limits is negligibly small (0.01–0.02 dex). We chose to take the detection as the value of the pair. For each S+E pair, we assumed that the H i mass is associated only with the S component and the contribution from the E component is negligible. We tested this assumption using the gas mass derived from the dust mass (Cao et al. 2016). We calculated the mean and error of the ratio of M_gas(E)/M_gas(S) using the Kaplan–Meier (K–M) estimator (Kaplan & Meier 1958), which exploits the information in the upper limits of M_gas(E). The result is 0.11 ± 0.01. This indicates that E galaxies contribute only 10% of the gas mass of S+E pairs, which is indeed negligible. Other variables, including M_star, 100 × M_dust and SFR, are also only for the S component.

In Table 3, means and errors of log[M_{H i}/(100 × M_dust)], log(M_{H i}/M_star), and log(SFE_{H i}) are presented for the total sample of pairs and for the following three sets of contrasting sub-samples: (1) S+S versus S+E; (2) log(M_star/M_⊙) < 10.7 versus log(M_star/M_⊙) > 10.7 (for an S+S pair, M_star is the mean of the two components); (3) "JUS" versus "INT & MER" pairs. According to Cao et al. (2016), "JUS" pairs are those without clear signs of interaction, and "INT" and "MER" are pairs with signs of interaction and merging. In order to exploit information in the upper limits, the calculations were carried out using the maximum likelihood K–M estimator (Kaplan & Meier 1958). We derived the survival curves of these ratios through VOStat (VOStat Development Group 2013). The means and errors of statistical calculations are based on the integrated areas of the curves. The log[M_{H i}/(100 × M_dust)] analysis is confined to sources (55) with M_dust detections, and the log[SFR/M_{H i}] analysis is confined to sources (58) with M_{H i} detections.

In Figure 2 we compare M_{H i} (contamination corrected) with 100 × M_dust. There is a good correlation between the two values, both probing the cold gas content in these pairs. The linear correlation coefficient is 0.53. For the Spearman's rank correlation, the coefficient is 0.61 and the significance is 2.60 × 10⁻⁸. The strong correlation suggests that a significant fraction of dust resides in H i gas, or the ratio between M_{H i} and ${M}_{{{\rm{H}}}_{2}}$ is relatively constant. For the total sample, the mean log[M_{H i}/(100 × M_dust)] = −0.06 ± 0.06. Cao et al. (2016) adopted 100 × M_dust as an estimate of the total gas mass M_gas. Taken at face value, our result indicates that the contribution of H i gas to the total gas mass is 87(±12)%. Draine et al. (2007) found an average dust-to-gas mass ratio of 0.007 for nearby spiral galaxies, corresponding to M_gas/M_dust = 143. Assuming this gas-to-dust mass ratio for our pairs, the contribution of H i gas to the total gas mass would be 61(±9)%).

Zoom In Zoom Out Reset image size

In Figure 3, we plot log(M_{H i}/M_star) against log(M_star) for S+S and S+E pairs using different symbols. Since each S+S pair has two spirals, while an S+E pair has only one, both M_{H i} and M_star of the former are divided by 2. The mean log(M_{H i}/M_star) (= −1.12 ± 0.06) for the total pair sample corresponds to a H i gas fraction of f_{H i} = 7.6 (±1.1)%. More massive paired galaxies (M_star > 10^10.7 M_⊙) have a slightly lower average (mean log(M_{H i}/M_star) = −1.13 ± 0.07, corresponding to f_{H i} = 7.4 (±1.2)%), compared with less massive paired galaxies of M_star < 10^10.7 M_⊙, which have mean log(M_{H i}/M_star) = −1.06 ± 0.11 (corresponding to f_{H i} = 8.7 (±2.2)%). This is consistent with the results of Catinella et al. (2010), who showed that for a large sample of ∼1000 SFGs of 10¹⁰ < M_star <10^11.5 M_⊙, there is a significant trend in which the H i gas fraction decreases with increasing M_star, with an overall average of f_{H i} ∼ 10% (Figure 3). We found no significant difference (<1σ) between the means of log(M_{H i}/M_star) of S+S and of S+E pairs (Table 3). This is different from Cao et al. (2016), who found, although with only marginal significance, that spirals in S+E pairs have on average lower total gas mass (estimated using the dust mass) to stellar mass ratios than those in S+S pairs.

Zoom In Zoom Out Reset image size

In Figure 4, we present a log(SFR) versus log(M_{H i}) plot for S+S and S+E pairs. Here again, the M_{H i} and M_star of S+S pairs are divided by 2. There is a sub-population of very active star-forming galaxies (with SFR ≳ 10 M_⊙ yr⁻¹) in S+S pairs. These galaxies are largely missing in S+E pairs.

Zoom In Zoom Out Reset image size

In Figure 5, log(SFR) is plotted against H i gas fraction M_{H i}/M_star. This shows again that the SFR of S+S pairs is systematically higher than that of S+E pairs. There is no clear dependence of SFR on H i gas fraction.

Zoom In Zoom Out Reset image size

The ratio SFR/M_{H i} measures the star formation rate per unit H i gas mass, which will hereafter be referred to as SFE_{H i}. In Figure 6, log(SFE_{H i}) is plotted against the H i gas fraction M_{H i}/M_star. It appears that for a given M_{H i}/M_star, the SFE_{H i} of S+E pairs is systematically lower than that of S+S pairs, and the difference is not sensitive to the H i fraction. As listed in Table 3, the mean SFE_{H i} of S+S pairs (10^{−9.26±0.11} yr⁻¹) is ∼4.6× higher than that of S+E pairs (10^{−9.92±0.11} yr⁻¹). There is a small but insignificant difference between the mean SFE_{H i} of pairs with signs of merger/interaction (10^{−9.45±0.14} yr⁻¹) and that of pairs without (10^{−9.60±0.09} yr⁻¹). The mean SFE_{H i} of the whole pair sample is 10^{−9.55±0.09} yr⁻¹, corresponding to a H i consumption time of 3.5 ± 0.7 Gyr.

Zoom In Zoom Out Reset image size

Given the large beam (FWHM = 9') of the GBT observations, the H i detected for each pair includes both gas inside the disks and the stripped gas in tidal features and debris. Because of the exclusion of mergers with component separations less than 5 kpc and nearby pairs with recession velocities less than 2000 km s⁻¹, the H-KPAIR sample preferentially selects massive (≳10¹⁰ M_⊙) early stage merger systems (before the final coalescence). Our observations show that these systems have similar H i gas fractions compared to normal spiral galaxies (Figure 3). For gas-rich dwarf galaxy pairs M_star < 10^9.7 M_⊙, Stierwalt et al. (2015) reached similar results. On the other hand, spiral galaxies in compact groups are found to be H i deficient (Verdes-Montenegro et al. 2001; Borthakur et al. 2015; Walker et al. 2016).

A major science goal of this study is to address the puzzling result of the significant difference between the sSFR enhancement of spirals in S+E pairs and in S+S pairs (Xu et al. 2010; Cao et al. 2016). Because all pairs were selected using the same criteria regardless of morphological type (Domingue et al. 2009; Cao et al. 2016), this difference cannot be due to any selection bias. One possibility is that the sSFR in a paired galaxy is influenced by the immediate surrounding environment. This hypothesis is in agreement with the correlation between sSFRs of the primaries and secondaries in major-merger S+S pairs (i.e., the "Holmberg effect" Kennicutt et al. 1987; Xu et al. 2010). On the other hand, Xu et al. (2010, 2012) did not find any significant difference between the local densities around S+E pairs and S+S pairs within a projected radius of 2 Mpc. Therefore, the linear scale of the environment effect must be less than 2 Mpc. Xu et al. (2010, 2012) speculated that the IGM in the dark matter halo (DMH) shared by both galaxies of a pair may play a significant role here. For example, when a DMH has strong (weak) cold flows (Dekel et al. 2009; Kereš et al. 2009), galaxies inside it may have abundant (scarce) cold gas supply to fuel active star formation. A prediction of this hypothesis is that spiral galaxies in S+E pairs have systematically lower gas content than those in S+S pairs.

However, this is not supported by our result, which shows no significant difference between the mean log(M_{H i}/M_star) of S+S and that of S+E pairs. It appears that the higher sSFR and SFE_{H i} of S+S pairs are mainly due to a sub-population of very active SFGs, which are missing in S+E pairs (Figure 4). It will be very interesting to find out how the high star formation enhancement in these pairs is triggered, and why it is not happening in S+E pairs.

Some insights can be gained from the examples studied by Hibbard et al. (2001) using high-resolution VLA H i maps for galaxies in the "merger sequence." Three of the five systems in their sample are early stage mergers with active star formation. They all show very extended tidal features in the H i gas distribution. However, most active star formation is confined to the central region where high density molecular gas and bright H_α emission are found. Dynamical simulations of Olson & Kwan (1990) demonstrated that, in the central ∼2 kpc of merging galaxies, interaction-induced collisions between gas clouds may play very important roles in triggering enhanced star formation, and the effect is stronger in S+S systems than in S+E systems.

Scudder et al. (2015) carried out VLA observations (beam FWHM = 14'') of the H i 21 cm line emission for 34 galaxies in 17 nearby S+S pairs, and obtained 17 detections. Compared to a control sample of galaxies, they found marginal evidence (at the ∼2σ level) for a positive correlation between the H i fraction and the SFR enhancement. On the other hand, Cao et al. (2016) did not see any significant correlation between SFR/M_gas enhancement and gas fraction in H-KPAIR. Our results in Figure 5 (Figure 6) show also that the difference between SFR (SFE_{H i}) of spirals in S+S and S+E pairs does not depend on H i gas fraction.

In this paper we present a study of the H i gas content of a large K-band-selected sample of galaxy pairs (H-KPAIRs). Among 88 pairs (44 S+S pairs, 44 S+E pairs), we observed 67 pairs using GBT for the 21 cm H i fine-structure emission. Except for 9 pairs that have severe RFIs and thus no informative signals could be extracted from the data, we derived H i mass from the spectral line. The results include detections (46 pairs) and upper limits (12 pairs). In addition, the H i masses of the other 12 pairs were collected from the literature. Compared with the Herschel data of the same sample, the relations between M_{H i} and 100 × M_dust, M_{H i} and SFR, M_{H i} and M_star, and SFE_{H i} and H i fraction are studied. The means and errors of log[M_{H i}/(100 × M_dust)], log(M_{H i}/M_star), and log(SFE_{H i}) are derived and analyzed for the total sample and for three sets of contrasting sub-samples. The primary results are as follows.

• 1.  
  Both linear and Spearman rank correlation analyses show a significant correlation between M_{H i} and 100 × M_dust. For the total sample, the mean log[M_{H i}/(100 × M_dust)] = −0.06 ± 0.06, corresponding to a H i-to-total gas ratio of 87(±12)%) if the gas-to-dust mass ratio is assumed as 100.
• 2.  
  The mean H i-to-stellar mass ratio of spirals in these pairs is 0.076 ± 0.011, consistent with the average H i gas fraction of spiral galaxies in general. There is no significant difference (<1σ) between the means of log(M_{H i}/M_star) of S+S and of S+E pairs (Table 3).
• 3.  
  The mean SFE_{H i} of S+S pairs (10^{−9.26±0.11} yr⁻¹) is ∼4.6× higher than that of S+E pairs (10^{−9.92±0.11} yr⁻¹), and the difference is not sensitive to the H i fraction. A sub-population of very active star-forming galaxies in S+S pairs is largely missing in S+E pairs.
• 4.  
  The difference between the mean SFE_{H i} of pairs with signs of merger/interaction (10^{−9.45±0.14} yr⁻¹) and that of pairs without (10^{−9.60±0.09} yr⁻¹) is insignificant (<1σ).
• 5.  
  The mean SFE_{H i} of the whole pair sample is 10^{−9.55±0.09} yr⁻¹, corresponding to a H i consumption time of 3.5 ± 0.7 Gyr.

This work is supported by National Key R&D Program of China grant No. 2017YFA0402600, the Open Project Program of the Key Laboratory of FAST, NAOC, Chinese Academy of Sciences, the National Natural Science Foundation of China grant No. 11643003 and No. 11373038, and the International Partnership Program of Chinese Academy of Sciences grant No. 114A11KYSB20160008. This work is sponsored in part by the Chinese Academy of Sciences (CAS), through a grant to the CAS South America Center for Astronomy (CASSACA) in Santiago, Chile. U.L. acknowledges support by the research projects AYA2014-53506-P from the Spanish Ministerio de Economia y Competitividad, from the European Regional Development Funds (FEDER) and the Junta de Andalucía (Spain) grant FQM108. D.L. acknowledges support from the "CAS Interdisciplinary Innovation Team" program. C.C. is supported by NSFC-11503013, and NSFC-11420101002.

In Figure 7, we present the full intensity spectra of 5 nearby normal galaxies NGC895, NGC2718, NGC3027, UGC10014, and NGC6140. The H i observations and data reduction of these galaxies are the same as those for the paired galaxies (Sections 3 and 4). The results are compared with literature data in Table 4. When there is more than one previous H i observation for a given galaxy, its literature data is chosen according to the following order of preference: (1) the latest GBT observation, (2) the latest NRAO 91 m observation, (3) the latest Arecibo observation, (4) the latest observation by other telescopes. The comparison shows a systematic difference on a ∼15% level, possibly due to a minor deviation in the calibration. This shall not significantly affect our main conclusions.

Zoom In Zoom Out Reset image size

In this appendix, we present the algorithm of the correction for contaminations due to neighboring galaxies, and the postage stamp images (taken from SDSS-DR14) used in the correction (Figure 8).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

View figure set (15 panels)

Tarball of figure set images

In order to correct for the contamination, we estimated the H i mass of neighboring spiral galaxies using the following algorithm. First the H i-gas-to-stellar-mass ratio G_{H i}/S is estimated by log₁₀(G_{H i}/S) = −1.732 38(g − r) + 0.215 182μ_i −4.084 51 (Zhang et al. 2009), where μ_i is the i-band surface brightness and (g − r) is the optical color derived from the g-band and r-band Petrosian magnitudes. The surface brightness used here is defined as ${\mu }_{i}={m}_{i}+2.5\mathrm{log}(2\pi {R}_{50}^{2})$, where m_i is the apparent Petrosian i-band magnitude and R₅₀ is the radius (in units of arcsecond) enclosing 50 percent of the total Petrosian i-band flux. Then, the stellar mass was estimated from the i-band luminosity and g − r color using the formula log(M_*) =log(L_i) − 0.222 + 0.864(g − r) (Bell et al. 2003). The estimated H i mass was then multiplied by the GBT beam response function at the distance of the galaxy, assuming the beam is a Gaussian with FWHM = 9'. Finally, for each pair, the contamination due to the H i mass of neighboring spiral galaxies as estimated was subtracted from its observed H i mass.

In Figure 8, postage stamp images taken from SDSS-DR14 are presented for individual pairs. In each image, paired galaxies are marked with yellow letters and neighboring galaxies with redshifts inside the bandwidth of the H i observation are marked with green letters. For pairs with H i detections, the white circle represents the beam and the red circle denotes the searching circle (r = 10'). For pairs observed but undetected by GBT, only the beam circle is plotted. For pairs with bad data, neither circle is plotted. In addition, there are nine pairs in galaxy groups and their H i observations were not pointed toward the pairs, which are excluded from our analysis.