Arp 65 interaction debris: massive HI displacement and star formation

Context: Pre-merger interactions between galaxies can induce significant changes in the morphologies and kinematics of the stellar and ISM components. Large amounts of gas and stars are often found to be disturbed or displaced as tidal debris. This debris then evolves, sometimes forming stars and occasionally tidal dwarf galaxies. Here we present results from our HI study of Arp 65, an interacting pair hosting extended HI tidal debris. Aims: In an effort to understand the evolution of tidal debris produced by interacting pairs of galaxies, including in situ star and tidal dwarf galaxy formation, we are mapping HI in a sample of interacting galaxy pairs. The Arp 65 pair is one of them. Methods: Our resolved HI 21 cm line survey is being carried out using the Giant Metrewave Radio Telescope (GMRT). We used our HI survey data as well as available SDSS optical, Spitzer infra-red and GALEX UV data to study the evolution of the tidal debris and the correlation of HI with the star-forming regions within it. Results: In Arp 65 we see a high impact pre-merger interaction involving a pair of massive galaxies (NGC 90 and NGC 93) that have a stellar mass ratio of ~ 1:3. The interaction, which probably occurred ~ 1.0 -- 2.5 $\times$ 10$^8$ yr ago, appears to have displaced a large fraction of the HI in NGC 90 (including the highest column density HI) beyond its optical disk. We also find extended ongoing star formation in the outer disk of NGC 90. In the major star-forming regions, we find the HI column densities to be ~ 4.7 $\times$ 10$^{20}$ cm$^{-2}$ or lower. But no signature of star formation was found in the highest column density HI debris, SE of NGC 90. This indicates conditions within the highest column density HI debris remain hostile to star formation and it reaffirms that high HI column densities may be a necessary but not sufficient criterion for star formation.


Introduction
Tidal interactions between pairs of galaxies before their merger are known to induce significant changes in the morphologies and kinematics of the stellar and interstellar medium (ISM) components on both large and small scales (Wielen 1990;Struck et al. 2005;Chien et al. 2007). Observations of pre-merger interacting pairs of galaxies provide the opportunity to investigate the way in which major tidal interactions affect the evolution of galaxies, including the impact on the rates and location of star formation (SF) as well as the morphological and kinematic responses of their stellar and ISM components. The 'Tails and Bridges' sample from Smith et al. (2007), drawn from the Arp atlas of interacting galaxies, contains a sample of nearby relatively isolated interacting pairs of galaxies. The pairs in the Smith sample are observed at various stages before merging. In our resolved 21 cm H i line studies of Arp pairs with the Giant Metrewave Radio telescope (GMRT) we are focusing on the impacts of these interactions on the H i morphology and kinematics of late-type galaxies (spirals), paying special attention to star formation in the low-density tidal debris and tidal dwarf galaxy (TDG) formation (Sengupta et al. 2012(Sengupta et al. , 2013(Sengupta et al. , 2014. The Arp 65 pair ( Fig. 1left) consists of two large spirals, NGC 90 and NGC 93. NGC 90 is a grand design barred spiral (SABc) with impressively enhanced spiral arms and long tidal tails to the north-west (NW) and south-east (SE). NGC 93 (spiral) is the more massive of the pair in terms of stellar mass and is projected about 3.0 ′ (64 kpc) to the east of NGC 90. The galaxies are catalogued as members of the SRGb063 galaxy group, (Fig. 1 -right) n (number of members) = 45, V opt = 5771 km s −1 (Mahdavi & Geller 2004). NGC 90 is projected ∼ 3.3 ′ (70 kpc) to the east of the group's intra-group medium (IGM) centroid based on X-ray observations by Mahdavi & Geller (2004). Further basic properties of the pair are given in Table  1.
We present here results from our GMRT H i observations of the Arp 65 system. In addition to the H i data from the GMRT, we have used the Sloan Digital Sky Survey (SDSS), Spitzer and Galaxy Evolution Explorer (GALEX) public archive data and images to understand the relation between the gaseous and stellar components of the system. The paper is set out as follows. Sect. 2 gives details of the GMRT observations, with observational results in Sect. 3. A discussion follows in Sect. 4 with concluding remarks in Sect. 5. The optical heliocentric radial velocities of NGC 90 and NGC 93 are respectively 5197 km s −1 and 5643 km s −1 (Hyperleda 1 ). Using the average radial velocity (5420 km s −1 ) and assuming H 0 to be 75 km s −1 Mpc −1 , we estimate the distance to Arp 65 pair to be 73 Mpc. Based on this distance the spatial scale is ∼ 21.2 kpc per arcmin.

Observations
H i 21 cm line observations of Arp 65 were carried out using the GMRT on 15 March, 2009. Further details of the observations are given in Table 2. The baseband bandwidth used was 16 MHz for the H i observations giving a velocity resolution of ∼27 km s −1 . The Astronomical Image Processing System (AIPS) software package was used for data reduction. Bad data from malfunctioning antennas and antennas with low gain and/or radio frequency interference were flagged. The flux densities are on the scale of Baars et al. (1977), with flux density uncertainties of ∼5%. After calibration, continuum subtraction in the uv 1 http://atlas.obs-hp.fr/hyperleda/

H i content and distribution
The top panel of Fig. 2 shows the low-resolution integrated H i contours (in white) overlaid on an SDSS r-band image for Arp 65. The blue contours in the figure show the far-ultraviolet (FUV) emission from GALEX. The bulk of the Hi detected in this system is associated with the SABc galaxy NGC 90. The other member of the pair, NGC 93 appears to be a more evolved spiral with a redder SDSS g-i colour of 1.58 compared to NGC 90 (SDSS g-i =1.4). H i is only marginally detected in NGC 93. The most striking feature in the low-resolution map is that the Hi intensity maximum and the bulk of the H i is projected ∼ 1 arcmin (21 kpc) to the south-east of the NGC 90 optical centre and beyond its optical disk.
The integrated H i flux density obtained by Springob et al. (2005) from single-dish observations for NGC 90 was 5.4 Jy km s −1 as compared to 4.9 Jy km s −1 from our GMRT observations, meaning that the GMRT recovered ∼ 90% of single-dish H i flux density. Conversion of the single dish flux density to an H i mass gives M HI = 6.8 × 10 9 M ⊙ . We use this H i mass derived from the single-dish observations for all our calculations requiring the NGC 90 H i mass. The single-dish H i flux density for NGC 93 was 1.46 Jy km s −1 (Springob et al. 2005). However the marginal Hi detection for NGC 93 in our GMRT map (Fig. 2,top panel), the Arecibo beam size (∼ 3.5 arcmin) and the coincidence in velocity of flux maxima in the Arecibo spectrum of NGC 93 (in red) with the high-velocity horn of the GMRT spectrum (in black) for NGC 90 (Fig.2, middle right panel) leads us to conclude that most of Hi flux in the NGC 93 Arecibo singledish spectrum is in fact H i spatially associated with NGC 90.
The middle right panel of Fig. 2 also shows H i spectra for NGC 90 from both the GMRT (in black) and the single-dish (in blue) observations. The single-dish spectrum has V HI = 5333 km s −1 and W 20 = 477 km s −1 (Springob et al. 2005). This W 20 value is ∼ 45 km s −1 higher than an estimate of W 20 based on the M* values in section 4.1 and equation 19 in Torres-Flores et al. (2011). The GMRT resolved images and channel maps provide insights into the enhancement of the line width. The middle left panel of Fig. 2 shows the velocity field for H i for NGC 90 (colour scale) overlaid with the FUV (GALEX) contours (in black). The red contour line shows the Hi extent (3 σ) from the GMRT low-resolution integrated H i map. West of the NGC 90 optical centre the velocity field indicates a more or less regular velocity gradient. But the eastern part of the disk shows an anomalous velocity behaviour. The separation between highand low-velocity components in the channel maps ( Fig. 5) is indicated by the lack of H i detections between 5347 km s −1 and 5402 km s −1 . A gas blob appears SE of NGC 90 at velocities > 5000 km s −1 and moves NW across the optical disk with a velocity increasing up to ∼ 5320 km s −1 . At around 5480 km s −1 another blob appears SE of the NGC 90 optical disk and remains detected there at velocities up to 5650 km s −1 . Based on the channel maps, we divided the high-and low-velocity components of NGC 90 at 5350 km s −1 . The bottom left panel of Fig 2 shows the velocity field of the low-velocity component, overplotted with the total intensity contours (in blue) of the high-velocity component and the FUV (GALEX) contours (in black) defining the edge of the optical galaxy. The H i velocity field for low-velocity component H i in NGC 90 presents as a disk with fairly regular rotation and no significant asymmetry. This velocity field shows that the H i velocity at the optical centre of NGC 90 is ∼ 5188 km s −1 (V optical = 5197±25) and its rotation velocity is ∼ 100 km s −1 , before adjustment for inclination. No sign of regular rotation is found in the velocity field for the high-velocity component, referred to as the "SE high-velocity H i mass" from now on. This separation of high-and low-velocity gas is also clearly seen in the position-velocity (pv) diagram (Fig. 3). The signal-to-noise ratio (S/N) is poor in the pv diagram, but it successfully shows the gradient in the disk of NGC 90 and the lack of a velocity gradient in the SE high-veloctiy gas mass. About 50% of the total H i mass of 6.8 × 10 9 M ⊙ can be attributed to the low-velocity component (NGC 90) implying ∼ 3.4 ×10 9 M ⊙ for the SE highvelocity H i mass. We conclude that both the abnormally large H i line width and the perturbed H i morphology of NGC 90 are largely the result of the superposition of the SE high-velocity H i mass on the gas disk of NGC 90 that rotates with a lower velocity.

H i and star formation beyond the NGC 90 optical disk
As the top panel of Fig. 2 shows, NGC 90 hosts several star forming zones beyond its optical disk along the western tidal tail and east of the disk. The most prominent, labelled as SF1, SF2 and SF3, are visible in both the SDSS r-band and the FUV (GALEX) images. Hi is detected at the positions of these starforming zones, but without any evidence of local enhancement in Table 3. GMRT H i observation results H i mass (total) in the Arp 65 system 6.8 × 10 9 M ⊙ ± 0.7 × 10 9 M ⊙ H i mass of the NGC 90 low-velocity disk 3.4 × 10 9 M ⊙ ± 0.4 × 10 9 M ⊙ H i mass of the SE high-velocity H i mass 3.4 × 10 9 M ⊙ ± 0.4 × 10 9 M ⊙ H i velocity of NGC 90 5188 ± 27 km s −1 H i velocity of the SE debris 5531 ± 27 km s −1 H i rotation velocity of NGC 90 100 ± 27 km s −1 H i flux density. The Hi column densities 2 in these star-forming zones are ∼ 1.2 -4.7 × 10 20 cm −2 . The high-and low-resolution H i maps both show the column density maxima projected at ∼ 10 kpc east of the NGC 90 optical disk edge (within the SE high-velocity H i mass). The Hi column density maxima from the high-resolution map ( Fig. 2 There is no evidence of ongoing star formation at or near the position of Hi column density maxima from the optical, UV or Spitzer 24 µm images.

Discussion
In this discussion we consider what the observations to date, including our GMRT H i data, reveal about the interaction between the pair and SF in the tidal debris.

Evidence of an NGC 90/NGC 93 interaction
The estimated stellar masses (M*) for NGC 90 and NGC 93 are ∼ 4.9 × 10 10 M ⊙ and 1.4 × 10 11 M ⊙ respectively, based on the parameters from Bell et al. (2003) and Blanton et al. (2003) for their SDSS r-band magnitudes and r-i colours. The M* values for NGC 90 (including the tails) and NGC 93 derived from the formula in Eskew et al. (2012) using Spitzer 3.6µm and 4.5µm flux densities from Smith et al. (2007) are 4.4 × 10 10 M ⊙ and 1.4 × 10 11 M ⊙ respectively. Both methods produce M* values that agree well with each other and give a stellar mass ratio of ∼ 1:3.
Interacting galaxy models and observations (e.g. Mihos 2001;Oh et al. 2008;Casteels et al. 2013) indicate enhanced stellar spiral arms as well as transitory tidal bridges and tails can result from prograde high-impact, low-velocity flyby encounters. Such interactions are expected to produce a long optical tidal tail on the far side and a bridge joining the pair on the near side (Mihos 2001). Strong arm enhancement is expected in bulge-dominated galaxies and a bar in disk-dominated galaxies. A stellar bridge is not visible between the NGC 90/93 pair, but a bar, prominent stellar arms and a far side tidal tail are observed in NGC 90 ( Fig.  1, left panel). Oh et al. (2008) used models to study the enhancement by a perturber of the optical spiral arms, tidal bridges and tidal tails (together with their visibility time-scales) in a disk galaxy. The baryonic mass of NGC 90 (∼ 5.4 × 10 10 M ⊙ ) is close to that of the Oh model disk galaxy (∼ 5.2 × 10 10 M ⊙ ), therefore it is reasonable to assume that the galaxy properties, including the dynamical times-cale, are sufficiently similar to make valid comparisons between the two, accepting that the orbital parameters remain poorly constrained compared to the Oh models. The Oh models use a tidal strength parameter, and investigate interactions with S < 0.3. For NGC 90 we can say the ratio of perturber to galaxy mass ( M p M g ) is ∼ 3. But we have no direct constraint on the more important ratio of the peri-centre to galaxy radius ( R g R peri ) variable or ∆T T , the angular speed of the perturber relative to stars at the galaxy edge. In the Oh et al. (2008) models the tidal tails dissipate rapidly after reaching their most prominent phase 1.4 to 2.5 × 10 8 yr after the interaction. But for S > 0.3 the tail dissipation time-scale may extend to ∼ 1 Gyr and in this case the tail may fragment to form TDGs (Barnes 1992;Oh et al. 2008). However four pieces of evidence support an interaction within the times-cales derived from the Oh et al. (2008) models: (i) H i morphological perturbation signatures from a full merger of galaxies with total baryonic masses of the same order as NGC 90, only remain identifiable for a maximum of ∼ 4 ×10 8 yr to 7 × 10 8 yr (Holwerda et al. 2011). It seems likely that H i perturbations from a full merger would be of the same order as the pre-merger interaction we observe in NGC 90. But because the H i morphology perturbations in NGC 90 are extreme compared to the Holwerda H i merger signature thresholds, it is reasonable to conclude the NGC 90 perturbation occurred well within the upper limits of the (Holwerda et al. 2011) times-cale; (ii) the projected distance between NGC 90 and NGC 93 is ∼ 64 kpc. Assuming the separation velocity is the SRGb0063 group velocity dispersion of ∼ 336 km s −1 , we estimate the shortest possible time since their closest approach to be ∼ 1.9 × 10 8 years ago; (iii) the time for a single rotation 3 of NGC 90 is ∼ 0.7 × 10 9 yr, based on the ∼ 180 km s −1 H i rotational velocity (inclination adjusted) of the lowvelocity component and data from Tab. 2. A de-projected SDSS g -band image 4 of NGC 90 indicates that the truncated bridge remnant is now offset ∼ 60 • counter-clockwise from a line joining the optical centres of NGC 90 and NGC 93. If we assume that the NGC 90/NGC 93 bridge formed along an axis joining the two galaxies soon after their closest approach, as interaction models predict, then the current ∼ 60 • offset (between a line joining the two galaxies and the bridge remnant) reflects the approximate fraction the time for single NGC 90 disk rotation since the bridge was formed. This then implies a time-scale since the interaction of ∼ 1.2 × 10 8 yr 5 ; and (iv) the NGC 90 optical morphology agrees well with that of the Oh et al. (2008) model, their figure 1, at t = 0.1 and 0.2 Gyr. While the uncertainties for each of these time-scales are large and difficult to quantify, taken together they indicate that the interaction probably occurred within the last ∼ 1.0 -2.5 × 10 8 yr. The presence of a prominent tidal tail in this time frame indicates that the Oh et al. (2008) models apply to the interaction and implies an interaction with S < 0.3. The ∼ 1.0 -2.5 × 10 8 yr interaction time-scale indicates NGC 93 is the most probable perturber of NGC 90 which would rule out interactions with other similar sized SRGb063 galaxy group members (see Fig. 1, right panel) projected at greater distances. Tidal interactions between pairs of galaxies are known to be capable of displacing substantial fractions of their H i, but it is quite unusual for the the highest density H i to be observed beyond the optical disks. Some other examples are NGC 3921 (Hibbard & van Gorkom 1996), Arp 105 (Duc et al. 1997) and Arp 181 (Sengupta et al. 2013).
H i masses for galaxies of a size and morphological type similar to NGC 90 and NGC 93 are expected to be ∼ 1.0 × 10 10 3 T rot [Gyr]= 6.1478 r/ V rot , where r = the optical radius [kpc] and V rot = 0.5 ∆V [km s −1 ]/ sin(i). 4 based on PA = 113.4 • and inclination = 34.5 • from HyperLeda 5 60 • /360 • × 0.7 × 10 9 yr = 1.2 × 10 8 yr M ⊙ and ∼ 4.3 × 10 9 M ⊙ respectively, with an error of ∼50% (Haynes & Giovanelli 1984). If we assume that it is only the low velocity H i component (see section 3.1) that remains viralised within the NGG 90 potential, then its H i mass is ∼ 3.4 × 10 9 M ⊙ , with a higher than usual uncertainty (≥ 10%) because this mass is derived from an approximate separation of the NGC 90 spectrum into high-and low-velocity components. These complications together with the uncertainty in the expected H i mass values (Haynes & Giovanelli 1984) prevents us from making a conclusive statement on the Hi deficiency of NGC 90, although it is possible that its apparent H i deficiency is real. Based on the marginal detection of H i in NGC 93 from our GMRT map we conclude that NGC 93 has no significant H i content. Again the expected Hi mass estimate of NGC 93 is also highly uncertain as its morphological type is not clearly known (Table 1). But marginal GMRT detection of Hi in NGC 93 strengthens the case for it being Hi deficient.

Possible explanations
While it remains unclear whether the Arp 65 pair, individually or together, is Hi deficient, we observed a massive Hi cloud without detected optical, UV or Spitzer 24 µm counterparts, projected beyond the pair's optical disks. Additionally, the prominent tidal arms or tails at optical wavelengths are a clear signature of a recent tidal interaction (∼ 1.0 -2.5 × 10 8 yr). We discuss three possible scenarios below which could explain these features and reach a conclusion about which is the most probable scenario.

Intra-group medium and ram pressure
NGC 90 is projected ∼ 3.3 arcmin (70 kpc) east of the intragroup medium (IGM) X-ray emission centroid of the SRGb0063 group mapped by Mahdavi et al. (2000). In their ROSAT map, NGC 90 is projected at the easternmost edge of the 3 σ contour. The SRGb0063 group has an L X (0.5) = 3.63 x 10 42 ergs s −1 (Mahdavi & Geller 2004) and a velocity dispersion of 336 ±36 km s −1 . NGC 90 has a V opt of 5197 km s −1 , which is 574 km s −1 lower than the group V opt = 5771 ±48 km s −1 (Mahdavi & Geller 2004). The strong ram pressure that can arise in galaxy clusters because of their dense ICM and high velocity dispersions is capable of stripping and displacing large quantities of Hi. But ram pressure is predicted to be much weaker in galaxy groups. Comparing the L X and velocity dispersion of the SRGb063 group with that of galaxy clusters, we see the SRGb063 L X is about two orders of magnitude lower and the velocity dispersion is a factor of 2 to 3 times lower than for galaxy clusters. We would therefore expect ram pressure in the SRGb063 group to be at least an order of magnitude or two lower than in clusters. While there is evidence that ram-pressure stripping has produced H i deficiencies in galaxy groups with X-ray emitting IGM (e.g. Davis et al. 1997;Sengupta & Balasubramanyam 2006), several studies concur that for galaxies with M* similar to NGC 90 (whether in galaxy groups or galaxy clusters), large offsets of their highest density H i beyond optical disk cannot be attributed to ram pressure alone (Bravo-Alfaro et al. 2000;Chung et al. 2009;Scott et al. 2010). Still, we also note that the extent of the NGC 90 H i disk (assuming it is traced by the low-velocity H i component) is similar to that of its optical disk. Taken in isolation this could suggest ram-pressure stripping, but in this case the recent tidal interaction seems a more likely cause for the truncation of low-velocity H i disk. We conclude that while there may be low-level ram-pressure stripping, it is highly unlikely to be cause of the displacement of the high-density H i gas to the SE of NGC 90.

SE high-velocity H i mass belongs to a dwarf galaxy
An alternative scenario is that the entire SE high-velocity H i mass belongs to a small dwarf galaxy. However a dwarf galaxy with an H i mass of ∼ 3.4 × 10 9 M ⊙ is expected to have a stellar mass of least a few times 10 8 M ⊙ and a surface brightness of ∼ 24-25 magnitude arcsec −2 , which would be visible in the SDSS and UV images. But in this case there is no visible optical or UV counterpart at these wavelengths. Moreover the velocity field of the SE high-velocity H i mass is irregular, more resembling debris than a galaxy.

Gas-rich NGC 90 interacts with gas-poor NGC 93
In this scenario the SE high-velocity H i mass that lies between NGC 90 and NGC 93 in velocity space and offset ∼ 25 kpc in projection SE of the NGC 90 disk (see Fig.2, bottom left panel) is a result of the tidal interaction between NGC 90 and NGC 93 which also produced the optical interaction signatures in NGC 90. The reason for the marginal GMRT H i detection in NGC 93 is that it was already essentially devoid of H i prior to the interaction, which means almost all of the H i detected with the GMRT belonged to NGC 90 before the interaction. Had NGC 93 contained a significant mass of H i before the interaction, we would expect the bulk of it to have remained bound to NGC 93 because of its deeper potential well compared to NGC 90. In this scenario the pair interaction caused a massive displacement of about half of the H i in NGC 90 (the SE high-velocity H i mass) ∼ 1.0 -2.5 × 10 8 yr ago. This an effect on H i similar to the one that caused the massive H i displacement beyond the optical disks observed in another interacting pair, Arp 181 (Sengupta et al. 2013). But, unlike Arp 181 (M* ratio 1:1.6), in Arp 65 the displacement is primarily in the velocity plane, although with some offset in the plane of the sky. The displaced H i was tidally stripped from NGC 90 during the interaction and accelerated toward the NGC 93 velocity during the closest approach, attracted by the deeper gravitational potential of NGC 93 (see Fig. 2, middle right panel). The ∼ 3.4 × 10 9 M ⊙ H i that is now observed within the NGG 90 optical disk (the low-velocity NGC 90 H i component) rapidly re-virialised in the gravitational potential of the galaxy and now presents as a rotating H i disk. In contrast, despite having a similar H i mass and greater maximum H i column density than the low-velocity NGC 90 component, the SE H i mass remains as evolving H i debris that currently lacks H i rotation and SF signatures. Assuming NGC 93 orbit takes it farther from NGC 90 before their next encounter it seems likely that the near-term fate of the SE H i mass will be to fall back into the NGC 90 potential. While it would take tailored modelling to prove this scenario, it appears to provide the most reasonable explanation amongst the alternatives considered for observations.

Star formation in the tidal debris
The FUV (GALEX) image of the Arp 65 pair reveals several interaction-induced star-forming zones in the tidal arms and outer disk of NGC 90. At the base of the tidal tail projected north-west of the NGC 90 optical disk, a particularly bright hinge clump was reported by Hancock et al. (2009) (labelled as SF1 in Fig. 2). This clump shows strong emission in GALEX FUV and NUV images and is very bright in the Spitzer 8µm image ( Fig. 1 in Hancock et al. (2009)). In addition, beads of star formation are also visible all along the NGC 90 tail north-west of the hinge clump. There are several star-forming zones to the east and south-east of NGC 90, and several of them are beyond the optical disk (the prominent ones are labelled as SF2, SF3 in Fig. 2, top panel). We derived the stellar masses of these clumps using online Spitzer IRAC 3.6µm and 4.5µm photometry data (Smith et al. 2007) and the recipe from Eskew et al. (2012). The hinge clump (SF1) is the most massive star-forming zone with an estimated stellar mass ∼ 4.5 × 10 8 M ⊙ . The stellar mass estimates for SF2 and SF3 are 1.5 × 10 8 M ⊙ and 7.2 × 10 7 M ⊙ respectively. Uncertainties for these masses are ∼ 10%. In the Spitzer images SF3 consists of at least two clumps while SF2 has a bright main clump and a diffuse area of star formation around it. The star formation rates (SFR) of each clump were estimated using Spitzer MIPS 24 µm data using the calibration reported by Rieke et al. (2009). As expected, the stellar clump with the highest mass, SF1, shows the highest SFR amongst the three main star-forming zones. The SFRs of SF1, SF2 and SF3 were estimated to be 0.025, 0.003 and 0.007 M ⊙ yr −1 , respectively.
The Hi image of NGC 90 (Fig. 2, top panel) reveals that all three star formation zones are embedded in Hi tails and debris. However, the H i column densities at the positions of the SF zones vary between 1.2 -4.7 × 10 20 cm −2 . None of the three SF zones coincide with the local maxima of the Hi column density. The highest Hi column density in these regions is ∼ 4.7 × 10 20 cm −2 , where the synthesised beam samples a 16.6 kpc × 13.2 kpc area. The high-resolution H i maps were made using a taper on the u,v data and also yielded similar results. The bottom right panel of Fig. 2 shows the last three highest column density contours (in black), between 1.5 -8.3 × 10 20 cm −2 , overlaid on the low-resolution Hi contours (in white) and the SDSS optical image of NGC 90. Only the high column density contours are plotted to highlight the Hi local maxima. Here the synthesised beam samples an area of 8.0 kpc × 5.2 kpc. The distribution appears to be more clumpy and fragmented than its low-resolution counterpart, however none of the Hi clumps coincide with the known star-forming zones. While there is no consensus in the literature about the threshold Hi column density at which star clusters form in tidal debris and outer disk regions, Maybhate et al. (2007) found an increase in the number of star clusters above N(Hi) ∼ 4 × 10 20 cm −2 . They suggested that a Hi column density threshold may be a necessary but not a sufficient condition for star cluster formation. An important point to be noted here is that the synthesised beam size plays a crucial role and the largest spatial scale sampled in Maybhate et al. (2007) was 7.5 kpc. Our high-and low-resolution GMRT maps sample between 7 kpc to 16 kpc respectively. Considering this, we find that in NGC 90, most of the star-forming zones more or less satisfy the threshold of Maybhate et al. (2007). Moreover, as suggested in Maybhate et al. (2007), in Arp 65 we also find that this threshold is indeed not a sufficient criterion for star formation. The highest column density H i in Arp 65 (∼ 8.0 × 10 20 cm −2 , in our highresolution maps) was detected within SE high-velocity H i mass.
A star formation map of NGC 90 (Fig. 4) was created by co-adding Spitzer 24 µm and GALEX images in units of M ⊙ yr −1 pix −1 where the size of each pixel is 2.4". We used skybackground-subtracted archive image and converted the observed flux to star formation rates using the calibrations provided by Rieke et al. (2009) and Kennicutt (1998) for 24 µm and FUV emission, respectively. The figure shows no sign of any star-forming activity above the sky-noise limit within a radius of 20" (∼ 7 kpc) of the highest H i column density of the SE highvelocity H i mass. Sky background subtraction for the Spitzer 24 µm image was difficult. The archival Spitzer 24 µm images are not cleaned properly and the background due to CCD reflection is severe. The observed typical rms across the map is ∼2 × 10 −5 M ⊙ yr −1 per pixel and can be considered as a reasonable upper limit. Using this, the integrated upper limit SFR within a 20 ′′ radius of the Hi column density maximum in the SE high-velocity H i mass is 2.2 M ⊙ yr −1 . No evidence of star formation associated with this high H i column density was found in the SDSS, Spitzer or GALEX data.

Concluding remarks
We have mapped the H i in Arp 65 (NGC 90 and NGC 93) with the GMRT and found that approximately half of the detected H i is in a high-velocity component projected SE of the NGC 90 optical disk, which was probably removed from NGC 90 during the interaction ∼ 1.0 -2.5 × 10 8 yr ago with its more massive gas-poor partner NGC 93. The balance of the H i is detected in a disk rotating at a lower velocity that is associated with NGC 90. The high-velocity H i component contains the Hi column density maxima ∼ 1.2 arcmin (25 kpc) E of the NGC 90 optical centre. The tidal interaction that produced the impressive optical tidal tails in NGC 90 was probably also responsible for removal of H i from the NGC 90 disk. This means that the asymmetric H i morphology of NGC 90 disk is the result of the superposition of gas removed from NGC 90 during its tidal interaction with NGC 93 on the remaining NGC 90 H i disk. Our observations of Arp 65 and Arp 181 indicate that high-impact pre-merger tidal interactions involving large spirals, with M* ratios in the range ∼ 1:1.6 to 1:3, have the ability to displace large factions of the H i beyond the optical disks of the participants. Arp 65 belongs to a small number of galaxy pairs where the highest density H i is observed beyond the optical disks of the pair following an interaction (e.g. Hibbard & van Gorkom 1996;Sengupta et al. 2013). We found that NGC 90 has extended on-going star formation in the outer optical disk, most likely triggered by interaction. In the three main star-forming clumps in the tidal debris and arm, we found that the star formation is occurring in regions with H i column densities ∼ 4.7 × 10 20 cm −2 or lower. But in contrast to Arp 181, which contains a star-forming tidal dwarf candidate, no signature of star formation was found near the highest column density H i located in the high-velocity H i debris south-east of the NGC 90 optical disk. This debris hosts no star formation signatures and reaffirms that high Hi column densities are a necessary, but not sufficient criterion for star formation.  The Hi column density levels are 10 20 cm −2 × (0.6, 1.2, 1.8, 2.9, 4.7, 5.9, 7.2) Middle left: The H i velocity field for NGC 90, in colour. The black contour shows the GALEX FUV extent NGC 90. The red contour is the Hi 0.6 × 10 20 cm −2 contour from the top panel. Middle right: Hi spectra for NGC 90: GMRT (black) and single dish (blue) from Springob et al. (2005). NGC 93: single dish spectrum (red) Springob et al. (2005)