A blind ATCA H\scaleto$I$1.2ex survey of the Fornax galaxy cluster

It is known that the evolution of galaxies is faster in denser environments (Diaferio et al., 2001) and that, as a consequence, the relative abundance of red early-type galaxies and blue late-type galaxies changes with environment density (Hubble & Humason, 1931; Oemler, 1974; Dressler, 1980). In this context, galaxy clusters are the most extreme environments within which galaxies evolve. They are characterised by a high number density of galaxies and by the presence of a dense intra-cluster medium (ICM), which galaxies move through. Thus, in clusters, both hydrodynamical and gravitational interactions such as ram-pressure stripping and tidal interactions, respectively, are likely to happen (Gunn & Gott 1972; Toomre & Toomre 1972), as well as mergers between galaxies (at least in the cluster outskirts - Sheen et al. 2012; Oh et al. 2018). These interactions deplete the cold gas reservoirs of galaxies and, therefore, affect their star formation activity. The balance between these types of interactions depends both on galaxy properties – including their orbits – and on cluster properties, such as the number density of galaxies and the ICM density. Therefore, we expect galaxy evolution to proceed differently in different clusters.

The aim of this work is to study galaxy evolution in the Fornax cluster, which is a low mass cluster (7$\times 10^{13}$ M_⊙ within a radius of 1.4 Mpc - twice the virial radius, $R_{\mathrm{vir}}$; Drinkwater et al. 2001a), located in the southern sky at a distance of 20 Mpc (Blakeslee et al., 2009). Fornax is thus the second closest galaxy cluster to us after Virgo. Fornax has roughly 200 spectroscopically confirmed galaxies within $R_{\mathrm{vir}}\sim$700 kpc (Maddox et al., 2019). NGC 1399 is the central dominant galaxy, which coincides with the peak of the X-ray emission (Paolillo et al., 2002), and the majority of the massive galaxies in the cluster central region are of an early morphological type. Thus, Fornax shows a more dynamically evolved state than Virgo (Grillmair et al., 1994; Jordán et al., 2007), but its growth is not over at all.

The Fornax region includes two main substructures (Drinkwater et al., 2001a). The first is the cluster itself centred on NGC 1399, whose interaction with NGC 1404 is revealed by a perturbed ICM distribution (e.g. Sheardown et al. 2018). Here, three well defined groups of galaxies with different light and colour distributions, kinematics, and stellar populations were found by combining the Fornax Deep Survey imaging (Iodice et al., 2019b) with Fornax 3D spectroscopy (Iodice et al., 2019a) in the two-dimensional projected phase space: the core, the north-south clump and the infalling galaxies (see Fig. 7 in Iodice et al. 2019a). The core is still dominated by NGC 1399, which is one of only two slow-rotators inside the virial radius (NGC 1427 is the other one, located on the east side of the cluster). The NS-clump is located in the high-density region of the cluster (within $0.4R_{\mathrm{vir}}\sim 0.3$ Mpc in projection), where the X-ray emission is still bright. It hosts the reddest and most metal-rich galaxies, all of them fast-rotating early-type galaxies. The bulk of the gravitational interactions between Fornax galaxies as well as most of the intra-cluster baryons (i.e. diffuse light, globular clusters and planetary nebulae) (Cantiello et al., 2018; Spiniello et al., 2018; Iodice et al., 2019b) are found in this NS clump, where galaxy growth is still ongoing through the accretion of mass onto galaxies’ outer regions (Spavone et al., 2020). The third group of objects in the cluster includes the infalling galaxies, which are distributed nearly symmetrically around the core in the low-density region outside $\sim 0.4R_{\mathrm{vir}}\sim 0.3$ Mpc in projection. The majority of these galaxies are late-types with ongoing star formation. Most of them exhibit signs of interaction with the environment and/or minor mergers in the form of tidal tails and disturbed molecular gas (Zabel et al., 2019; Raj et al., 2019). Previous works had also shown that numerous dwarf galaxies are currently falling into the centre of the cluster (Drinkwater et al., 2001b; Schröder et al., 2001; Waugh et al., 2002). Finally, the other major structure in the Fornax volume is the infalling group centred on NGC 1316 (Fornax A), located $\sim$ 1.5 Mpc south-west of the cluster centre and experiencing ongoing interactions between its members (Schweizer, 1980; Horellou et al., 2001; Iodice et al., 2017; Serra et al., 2019).

The observation of galaxies’ atomic hydrogen through the 21 cm wavelength emission line (H\scaleto$I$1.2ex) gives us information on the evolutionary state of galaxies as well as a global picture of the cluster. The evolution of galaxies depends on the evolution of their atomic hydrogen reservoir, which is the primary reservoir of fuel for star formation. Since it typically extends to the outskirts of galaxies, H\scaleto$I$1.2ex gas is the first component that is affected by tidal interactions, ram-pressure stripping and mergers. Indeed, cluster galaxies are usually deficient in atomic hydrogen with respect to non-cluster galaxies (Giovanelli & Haynes, 1983; Haynes & Giovanelli, 1986; Boselli & Gavazzi, 2006). H\scaleto$I$1.2ex is, therefore, a crucial observable for understanding galaxy evolution in dense environments (Hughes & Cortese, 2009; Chung et al., 2009).

H\scaleto$I$1.2ex emission in the Fornax cluster was studied in several works. Bureau et al. (1996) used the Parkes radio telescope to measure the amount of atomic hydrogen in 21 undisturbed galaxies with morphologies S0$/$a or later, located within 6 deg from NGC 1399 and with cz $\leq$ 2520  km s^-1. The eight galaxies within $R_{\mathrm{vir}}$ did not show any peculiar value in the $M_{\text{H\scaleto{$I$}{1.2ex}}}$/I$-$band infrared luminosity ratio with respect to the rest of the sample. The same ratio evaluated for Ursa Major galaxies, a lower density environment than Fornax, led them to conclude that the observed Fornax galaxies are not H\scaleto$I$1.2ex deficient.

The first blind H\scaleto$I$1.2ex survey of Fornax was carried out by Barnes et al. (1997), also with the Parkes telescope. They detected H\scaleto$I$1.2ex in eight galaxies within an area of 8 $\times$ 8 deg². Of those, two galaxies are within the cluster $R_{\mathrm{vir}}$: ESO 358G-063 and NGC 1365. Thanks to the survey sensitivity, they excluded the existence of a significant population of optically undetected H\scaleto$I$1.2ex clouds with H\scaleto$I$1.2ex mass greater than 10⁸ M_⊙.

The number of H\scaleto$I$1.2ex detections within the Fornax central region increased with the targeted survey by Schröder et al. (2001). They found H\scaleto$I$1.2ex in 37 out of 66 galaxies. Of those, 14 are within $R_{\mathrm{vir}}$, while the rest lies within 5 deg from the centre of the cluster. They found a lack of H\scaleto$I$1.2ex in the cluster centre, and measured the H\scaleto$I$1.2ex deficiency parameter to be $0.38\pm 0.09$ (as defined in Solanes et al. 1996). This shows a modest H\scaleto$I$1.2ex depletion in Fornax (usually, galaxies with H\scaleto$I$1.2ex deficiency parameter ¿ 0.3 are considered to be H\scaleto$I$1.2ex deficient, e.g. Dressler 1986; Solanes et al. 2001). Furthermore, the $M_{\text{H\scaleto{$I$}{1.2ex}}}$$-$to$-$blue light ratio of Fornax galaxies - mean value ($0.68\pm 0.15$) M${}_{\odot}/$L_⊙ - is a factor 1.7 lower than in the field. They also showed that the velocity dispersion of the sample of H\scaleto$I$1.2ex deficient galaxies is lower than that of the remaining H\scaleto$I$1.2ex detected galaxies. This difference in velocity dispersion agrees with the deficient galaxies having more radial orbits (as shown in Dressler 1986), which makes them good candidates for ram-pressure stripping (as discussed in Solanes et al. 2001).

Waugh et al. (2002) presented a blind survey of Fornax based on the H\scaleto$I$1.2ex Parkes All Sky Survey (HIPASS $-$ Barnes et al. 2001) data with a $M_{\text{H\scaleto{$I$}{1.2ex}}}$ limit of 1.4 $\times$ 10⁸ M_⊙. They detected 110 galaxies within an area of $\sim$620 deg² around the cluster. Of those, nine H\scaleto$I$1.2ex detections are within $R_{\mathrm{vir}}$. The authors confirmed a H\scaleto$I$1.2ex depletion in the H\scaleto$I$1.2ex detections (all late types) near the centre of the cluster, and suggested that H\scaleto$I$1.2ex-rich galaxies detected in the outer parts of the cluster are infalling towards the cluster for the first time. H\scaleto$I$1.2ex detections are arranged in a large scale sheet-like structure with a negative velocity gradient from south-east to north-west.

Waugh (2005) presented the result of the deepest blind H\scaleto$I$1.2ex survey of the cluster so far: the Basketweave survey of Fornax carried out with the Parkes telescope. This survey used a new scanning technique, which improved the sampling and the noise level compared to HIPASS. H\scaleto$I$1.2ex was detected in 53 galaxies within 100 deg² down to a detection limit of 10⁸ M_⊙. Of those, 15 were new detections, which confirmed the results of Waugh et al. (2002). In addition, the author presented higher resolution H\scaleto$I$1.2ex observations (2 arcmin, 3.3  km s^-1) of 28 individual Fornax galaxies carried out with the Australia Telescope Compact Array (ATCA). Of those, six are within $R_{\mathrm{vir}}$. Within this region, ESO LV-3580611 was marked as having an intriguing H\scaleto$I$1.2ex morphology with an elongation to the north-east of the system. As suggested in Schröder et al. (2001), also Waugh (2005) pointed out that this galaxy may be moving towards us.

The only other Fornax galaxy within $R_{\mathrm{vir}}$ with resolved H\scaleto$I$1.2ex imaging is NGC 1365 (van der Hulst et al., 1983; Jorsater & van Moorsel, 1995). The latter study observed an elongated H\scaleto$I$1.2ex distribution to the west of the system. They suggested that it may be caused by the interaction with the ICM.

All previous blind H\scaleto$I$1.2ex surveys of Fornax were carried out with the Parkes telescope. Their angular resolution of 15 arcmin was insufficient to study the H\scaleto$I$1.2ex morphology of the detected galaxies, which is a powerful tracer of environmental effects. Better angular resolution was achieved with the interferometric observation of a few selected galaxies in Fornax (see references above), but those observations covered only a small portion of the cluster volume. Here we present the first, blind, interferometric H\scaleto$I$1.2ex survey of the Fornax cluster.

Our survey was carried out with the ATCA and covers an area of 15 deg² centred on NGC 1399 with a spatial and velocity resolution of 67″ $\times$ 95″($\sim 6\leavevmode\nobreak\ \times\leavevmode\nobreak\ 9$ kpc at a distance of 20 Mpc) and 6.6  km s^-1, respectively. The average column density sensitivity within the survey area is 2$\times$10¹⁹ cm^-2 (3$\sigma$ over 25  km s^-1) and the $M_{\text{H\scaleto{$I$}{1.2ex}}}$ sensitivity is 2$\times 10^{7}$ M_⊙ (3$\sigma$ over 100  km s^-1). Initially, a case study by Lee-Waddell et al. (2018) revealed the tidal origin of NGC 1427A using spatially resolved H\scaleto$I$1.2ex images from a subregion of our ATCA mosaic. Here we present the results of the full survey. In Sect.2 we describe observations and data reduction. In Sect.3 we present the H\scaleto$I$1.2ex detections, their H\scaleto$I$1.2ex images and spectra. We also compare their H\scaleto$I$1.2ex mass, H₂\scaleto1.2ex0pt mass and SFR relative to non-cluster control samples, and compare their spatial and velocity distribution with those of the general Fornax population. In Sect. 4 we discuss our results, which we then summarise in Sect.5. We include supplementary material on each galaxy in Appendix A.

Our blind Fornax survey covers an area of 15 deg² (defined at a sensitivity level 3 $\times$ higher than in the best region of the H\scaleto$I$1.2ex cube), spanning from the centre of the cluster to a distance slightly further than $R_{\mathrm{vir}}$. The observations were carried out with the ATCA in the 750B configuration, from December 2013 to January 2014 (project code C2894)¹¹1Data available on https://atoa.atnf.csiro.au/query.jsp. The cluster was observed for 336 hrs using 756 different pointings with a spacing of 8.6 arcmin (1/4 of the primary beam FWHM at 1.4 GHz)

The 64 MHz bandwidth, centred at 1396 MHz, was divided into 2048 channels, providing a velocity resolution of 6.6  km s^-1. We reduced the data using the MIRIAD software (Sault et al., 1995). PKS B1934-638 and PKS 0332-403 were chosen as the bandpass calibrator and the phase calibrator, respectively. The latter was observed at 1.5 hr intervals between on-source scans. We flagged strong radio frequency interference based on Stokes V visibilities. After flagging and calibration we further processed a restricted frequency range 1407.2 MHz - 1419.6 (which corresponds to the velocity range of 166 - 2783  km s^-1), which includes all spectroscopically confirmed Fornax galaxies (Maddox et al., 2019). Within this range we used the UVLIN task of MIRIAD to fit and subtract continuum emission using 2nd-order polynomials.

We obtained the dirty cube with the INVERT task using natural weights to maximise surface-brightness sensitivity. We used MOSMEM and RESTOR to clean and restore H\scaleto$I$1.2ex emission, respectively. The restoring Gaussian PSF has a major and minor axis FWHM of 95 and 67 arcsec, respectively, and a position angle of 0.4 deg. The root mean square (RMS) noise level of the final cube goes down to 2.0 mJy beam^-1 in the most sensitive region. Within the survey area the RMS noise is $\leq 6.0$ mJy beam^-1 and, on average, 2.8 mJy beam^-1. This corresponds to a $3\sigma$ H\scaleto$I$1.2ex column density sensitivity of $N_{\text{H\scaleto{$I$}{1.2ex}}}$ $\sim 2\times$10¹⁹ cm^-2 assuming a line width of 25  km s^-1 and a $3\sigma$ $M_{\text{H\scaleto{$I$}{1.2ex}}}$ sensitivity of  $\sim$2$\times$10⁷ M_⊙ over a linewidth of 100  km s^-1. We searched for H\scaleto$I$1.2ex sources with the SoFiA source-finding package (Serra et al., 2015) within the survey footprint (see the green outline in Fig. 1). By smoothing and clipping, we convolve the input cube with a set of kernels and detect emission above 3.5$\sigma$ of the local noise level of each kernel. Reliable detections are identified based on the reliability algorithm presented in (Serra et al., 2012), which assumes that true sources have positive total flux and that the noise is symmetric around 0. For one faint source, NGC 1436, visual inspection was necessary to improve the SoFiA detection mask used to estimate galaxy parameters. Lastly, our final list of H\scaleto$I$1.2ex detections includes a faint source which did not pass the reliability test, but which we consider a genuine detection given its spatial correspondence with the known optical source FCC 323.

The population of H\scaleto$I$1.2ex detected Fornax galaxies exhibits several interesting features. Despite the limited resolution of our data, half of all detections reveal H\scaleto$I$1.2ex asymmetries and offsets relative to the stellar body (Fig. 2). Furthermore, the H\scaleto$I$1.2ex sample as a whole is gas-poorer and is forming stars at a lower rate than samples of non$-$cluster galaxies in the same $M_{\star}$ range (Fig. 6 and Fig. 10), and half of the galaxies with $M_{\star}$$>$10⁹ M_⊙ have an anomalous $M_{\text{H}_{2}}$/$M_{\text{H\scaleto{$I$}{1.2ex}}}$ ratio. Finally, the H\scaleto$I$1.2ex detections are distributed in a noticeably different way with respect to the majority of cluster galaxies both on the sky and in projected phase space (Fig. 12 and Fig. 13). This body of evidence suggests that the Fornax environment is influencing the evolution of these galaxies, which may be the most recent arrivals in the cluster. In particular, the difference between the 3D distribution of H\scaleto$I$1.2ex detected galaxies compared with that of spectroscopically confirmed Fornax members (Maddox et al. 2019), indicates that H\scaleto$I$1.2ex is a crucial observable to test the volume of the cluster where H\scaleto$I$1.2ex rich galaxies become H\scaleto$I$1.2ex deficient, before a complete H\scaleto$I$1.2ex removal in the inner part of the cluster.

An outstanding question is how long it takes for a galaxy to lose its H\scaleto$I$1.2ex – the dominant component of the interstellar medium – as it falls into a cluster. In the case of Fornax we can gain some insight through a joint analysis of Fig. 11 and Fig. 13. If we assume that H\scaleto$I$1.2ex is being actively removed from within galaxies in Fornax (as suggested by the frequently disturbed H\scaleto$I$1.2ex morphologies - see Fig. 2), the lack of H\scaleto$I$1.2ex detections in the virialised region of Fig. 13 implies that H\scaleto$I$1.2ex removal happens on a time scale $\tau_{\mathrm{HI,loss}}$ shorter than the cluster’s crossing time: $\tau_{\mathrm{HI,loss}}\leq\tau_{\mathrm{cross}}\leavevmode\nobreak\ \sim\leavevmode\nobreak\ R_{\mathrm{vir}}/\sigma_{\mathrm{cl}}\leavevmode\nobreak\ \sim\leavevmode\nobreak\ 2$ Gyr (see Sect.  3.4; this $\tau_{\mathrm{cross}}$ is similar to that estimated from simulations, 1.2$\leavevmode\nobreak\ \pm\leavevmode\nobreak\$0.5 Gyr independent of cluster mass; Rhee et al. 2017.) Hence, we expect that our 16 detections will have lost most of their H\scaleto$I$1.2ex by the time they reach the pericentre.

On the other hand, Fig. 11 suggests that, so far, H\scaleto$I$1.2ex has been lost slowly in our H\scaleto$I$1.2ex detections (having been removed and/or consumed, and not replenished). In that figure, our Fornax H\scaleto$I$1.2ex detections are distributed along the same correlation defined by non-cluster galaxies. Thus, for those Fornax galaxies, the SFR has so far had sufficient time to respond to a variation in H\scaleto$I$1.2ex mass – just like outside clusters. Since the transition from H\scaleto$I$1.2ex to new stars happens through the intermediate phase of H₂\scaleto1.2ex0pt, this ‘equilibrium’ between H\scaleto$I$1.2ex and SFR implies that H₂\scaleto1.2ex0pt is depleted faster than H\scaleto$I$1.2ex is lost in galaxies currently at the cluster’s outskirts, giving the entire cycle of H\scaleto$I$1.2ex-to-H₂\scaleto1.2ex0pt-to-SFR enough time to ‘see’ the varying H\scaleto$I$1.2ex content. Thus, while for Fornax as whole $\tau_{\mathrm{HI,loss}}\leq\tau_{\mathrm{cross}}\sim 2$ Gyr (see above), for our H\scaleto$I$1.2ex detections in the outer regions of Fornax $\tau_{\mathrm{HI,loss}}$ has so far been $\geq\leavevmode\nobreak\ \tau_{\mathrm{H_{2},depl}}\leavevmode\nobreak\ \sim\leavevmode\nobreak\ 1\text{-}2$ Gyr (Bigiel et al., 2008). (The H₂\scaleto1.2ex0pt depletion time $\tau_{\mathrm{H_{2},depl}}$ is only marginally shorter in Fornax, in particular at its outskirts, and anyway with large variations from galaxy to galaxy; Zabel et al. 2020.)

It is likely that, further inside Fornax, H\scaleto$I$1.2ex is removed faster than what we estimated above for galaxies at the cluster’s outskirts. Indeed, simulations indicate that for a galaxy within $0.5\leavevmode\nobreak\ R_{\mathrm{vir}}$, $\tau_{\mathrm{HI,loss}}\leavevmode\nobreak\ <\leavevmode\nobreak\ 0.5$ Gyr (Marasco et al., 2016). Such a rapid H\scaleto$I$1.2ex removal may not leave sufficient time for the SFR to ‘track’ the decrease in H\scaleto$I$1.2ex mass in Fig. 11, resulting in galaxies moving to the left of the non-cluster sample. A confirmation of this effect may come from the eight H₂\scaleto1.2ex0pt-detected galaxies (most of them H₂\scaleto1.2ex0pt-deficient; Zabel et al. 2019) where H\scaleto$I$1.2ex has already been removed at least down to the ATCA $M_{\text{H\scaleto{$I$}{1.2ex}}}$ sensitivity (left-pointing arrows in Fig. 11), and possibly by some H\scaleto$I$1.2ex detections closer to the cluster centre (e.g. NGC 1436). Indeed, these galaxies occupy a region to the left of the comparison sample, showing that their SFR is still significant despite their low H\scaleto$I$1.2ex content. For these galaxies, H\scaleto$I$1.2ex is likely to have been removed faster than H₂\scaleto1.2ex0pt is depleted: $\tau_{\mathrm{HI,loss}}\leq\tau_{\mathrm{H_{2},depl}}$. This conclusion is supported by the anomalously high $M_{\text{H}_{2}}$/$M_{\text{H\scaleto{$I$}{1.2ex}}}$ ratio of these galaxies in Fig. 9 (in the $M_{\star}$ range where we have a reliable comparison).

Within the picture discussed above, H\scaleto$I$1.2ex-detected galaxies in the outer regions of Fornax are thus first infallers, which are starting to interact with the Fornax environment and will lose most of their H\scaleto$I$1.2ex by the time they reach the pericentre. Even at the current early stage of infall, they already show a relatively large diversity in H\scaleto$I$1.2ex morphology (Fig. 2) and mass (Fig. 6). More specifically, we found two morphologically undisturbed H\scaleto$I$1.2ex rich galaxies ($12\%$ of the H\scaleto$I$1.2ex detections); four morphologically disturbed H\scaleto$I$1.2ex rich galaxies (25$\%$ of the H\scaleto$I$1.2ex detections); four morphologically disturbed H\scaleto$I$1.2ex deficient galaxies (25$\%$ of the H\scaleto$I$1.2ex detections); six morphologically undisturbed - within the ATCA resolution - H\scaleto$I$1.2ex deficient galaxies (38$\%$ of the H\scaleto$I$1.2ex detections).

The morphologically undisturbed H\scaleto$I$1.2ex-rich galaxies are ESO 358-G060 and NGC 1437A. They both reside outside the caustic curves in projection (Fig. 13) and, given their low $M_{\star}$, they should be easily perturbed by the Fornax environment. Thus, we speculate that they are recent Fornax members which have not yet had enough time to be significantly affected by the cluster. One possible caveat is the low resolution of our images, which may hide H\scaleto$I$1.2ex disturbances in particular in the case of the optically peculiar galaxy NGC 1437A. Another possibility is that they are outside the cluster volume or in a region with lower ICM density. ESO 358-G060 is also an outlier in Fig. 11, where the relatively high $M_{\text{H\scaleto{$I$}{1.2ex}}}$/$M_{\star}$ does not correspond to a high SFR. In this galaxy no SFR was detectable from WISE W3/W4 bands (after subtracting the old stellar population light), and the NUV contribution to star formation is low. Several physical processes might account for the low SFR with respect to the large H\scaleto$I$1.2ex reservoir: for example, an inefficient H\scaleto$I$1.2ex to H₂\scaleto1.2ex0pt conversion due to a small amount of dust, a large angular momentum which prevents the H\scaleto$I$1.2ex from collapsing, and/or an H\scaleto$I$1.2ex external origin (e.g. see Geréb et al. 2016, 2018).

The morphologically disturbed H\scaleto$I$1.2ex-rich galaxies are ESO 358-G063, NGC 1351A, NGC 1427A, NGC 1365. We discuss the last galaxy when we focus our attention on the H\scaleto$I$1.2ex detected subgroup of interacting galaxies. ESO 358-G063 and NGC 1351A, are both disc galaxies north of NGC 1399. They are both quite isolated from all other Fornax Galaxies in RA, Dec and velocity in projection (see Fig. 12 and Fig. 13). Thus, their slightly H\scaleto$I$1.2ex disturbed morphologies described in Sect.3 might be due to the interaction with the ICM of the Fornax cluster. Furthermore, the agreement of $M_{\text{H}_{2}}$/$M_{\text{H\scaleto{$I$}{1.2ex}}}$ of ESO 358-G063 and NGC 1351A with the xGASS scaling relation in Fig. 9 suggests that they are recent Fornax members, thus the cluster environment has not had enough time to significantly deplete their H\scaleto$I$1.2ex reservoirs. The last morphologically disturbed H\scaleto$I$1.2ex rich galaxy is NGC 1427A. This is the galaxy with the second-highest H\scaleto$I$1.2ex mass in our sample of Fornax H\scaleto$I$1.2ex detections. In Fig. 12 and Fig. 13, we see that NGC 1427A is the closest H\scaleto$I$1.2ex detected galaxy to the centre of the cluster in projection, but it has also the highest velocity. Thus, it may be a new Fornax member which is infalling from the foreground. As already mentioned, Lee-Waddell et al. (2018) studied the origin of the H\scaleto$I$1.2ex tail using the same data we present here. They concluded that, unlike previously suggestions, ram-pressure is unlikely to be the main process shaping the galaxy’s optical appearance. Instead, NGC 1427A is most likely a recent merger remnant, thus shaped by tidal forces. The recent merger might be the cause of the low molecular column density and/or its low metallicity, resulting currently undetected by ALMA (Zabel et al. 2019; it is the upper limit with the highest $M_{\star}$ in Fig. 9). NGC 1427A is also the second H\scaleto$I$1.2ex rich outlier with low SFR in Fig. 11. The H\scaleto$I$1.2ex-to-H₂\scaleto1.2ex0pt conversion might be inefficient to have a SFR consistent with its $M_{\text{H\scaleto{$I$}{1.2ex}}}$/$M_{\star}$ ratio.

The only H\scaleto$I$1.2ex deficient and morphologically disturbed galaxies which do not belong to the NGC 1365 subgroup are: ESO-LV 3580611 and NGC 1437B. They are very close to one another on the sky but have significantly different velocities. The former is outside the caustic curve in projection (Fig. 13), which supports the hypothesis of an new infalling Fornax member made by Schröder et al. (2001). The latter has a velocity similar to the recessional velocity of the cluster. H\scaleto$I$1.2ex and molecular morphologies (the latter detected by Zabel et al. 2019) are elongated in the same direction. This suggests that both gas phases are experiencing the same environmental interaction. Raj et al., 2019 detected a tidal tail in NGC 1437B, which may be due to a recent fly-by of another galaxy. As mentioned in Sect.3.4, although the evidence is not strong, NGC 1437B might be part of a subgroup of interacting galaxies which includes FCC 323. Thus, FCC 323 might be the fly-by galaxy that NGC 1437B has interacted with.

It is difficult to comment on the morphologically undisturbed H\scaleto$I$1.2ex deficient galaxies, since their symmetric H\scaleto$I$1.2ex distribution may be a consequence of the ATCA resolution. However, a very peculiar case is the truncated H\scaleto$I$1.2ex disc of NGC 1436. It is the closest spiral galaxy to the centre of the cluster detected in H\scaleto$I$1.2ex. Raj et al. (2019) observed an ongoing morphological transition into lenticular: the spiral structure is found only in its inner region, while the outer disc has the smooth appearance typically found in S0 galaxies.

The inner part of NGC 1436 appears regular also in H₂\scaleto1.2ex0pt, and the galaxy is just moderately H₂\scaleto1.2ex0pt deficient (Zabel et al., 2019). It is also the only galaxy detected both in H\scaleto$I$1.2ex and H₂\scaleto1.2ex0pt which shows a high $M_{\text{H}_{2}}$/$M_{\text{H\scaleto{$I$}{1.2ex}}}$ ratio in Fig. 9. These results and the evidence of morphological distortions only in the outer part of the galaxy suggest that this galaxy may have gone through a quick interaction with the cluster environment, which did not affect the inner spiral structure yet. This idea is corroborated by the fact that NGC 1436 lies on the upper edge on the comparison sample in Fig. 11, which means that despite its H\scaleto$I$1.2ex deficiency it is still forming stars at a significant rate.

Finally, we focus on the NGC 1365 subgroup. We discussed all the H\scaleto$I$1.2ex morphologies and the 3D distribution of the subgroup members in Sect.3.1 and Sect.3.4, respectively. As mentioned earlier, NGC 1365 is the only H\scaleto$I$1.2ex rich galaxy of the subgroup (Fig. 6) with a $M_{\text{H\scaleto{$I$}{1.2ex}}}$ at least two orders of magnitude larger than the other members. The H\scaleto$I$1.2ex distribution both in NGC 1365 and in FCC 102 is elongated to the north, while it is elongated to the south in FCC 090. ESO 358-G016 is the only galaxy with a regular H\scaleto$I$1.2ex morphology. It is also the only galaxy located north of NGC 1365. We propose two scenarios in order to explain the properties of the subgroup and its members: the former is a case of interaction between galaxies and the cluster environment which is responsible of the high $M_{\text{H\scaleto{$I$}{1.2ex}}}$/$M_{\star}$ ratio of the low mass members. In contrast, due to to its deep gravitational potential, NGC 1365 has been able to retain its H\scaleto$I$1.2ex, although some of it has been perturbed. The latter scenario we propose, it is a case of preprocessing in a group of galaxies where the local environment of the subgroup was able to affect the $M_{\text{H\scaleto{$I$}{1.2ex}}}$/$M_{\star}$ ratio of the low mass members before the group began to interact with the cluster environment.

In general, although we found a large variety of H\scaleto$I$1.2ex properties in our sample of H\scaleto$I$1.2ex detections, we detected an overall trend towards H\scaleto$I$1.2ex disturbances and deficiency in Fornax. Fig. 6 makes evident an already evolved state of Fornax H\scaleto$I$1.2ex galaxies where $\sim$2/3 of the galaxies are H\scaleto$I$1.2ex deficient. Fig. 6 also shows that the Fornax environment is more effective in altering the gas content of galaxies with $M_{\star}$ $<$ 3 $\times$ 10⁹ M_⊙) (see Sect.3.2).

Zabel et al. (2019) presented a similar study, where they compare Fornax and field galaxies based on their molecular gas properties. They found some molecular deficiency in all their detections except NGC 1365. However, galaxies with $M_{\star}$ $<$ 3$\times$10⁹ M_⊙ are both more H₂\scaleto1.2ex0pt-deficient and morphologically disturbed with respect to more massive galaxies, whose molecular gas morphology is always regular. We do not observe such a clear difference in H\scaleto$I$1.2ex. Indeed, we observe disturbed H\scaleto$I$1.2ex morphologies across our entire $M_{\star}$ range, confirming that atomic hydrogen is the best tracer of early interactions. Despite this difference between H\scaleto$I$1.2ex and H₂\scaleto1.2ex0pt morphologies, we also note some similarities between our results and those in Zabel et al. (2019). Indeed, the mass range of molecular disturbed galaxies is also characterised by a stronger H\scaleto$I$1.2ex depletion. Conversely, the H\scaleto$I$1.2ex depletion is weaker in the mass range in which galaxies show regular molecular-gas morphologies ($M_{\star}$ $>$ 3$\times$10⁹ M_⊙; Sect.3.2). Thus, both gas phases show that the Fornax environment is more effective in altering the gas content of low-mass galaxies compared to high-mass galaxies.

Finally, our results are in agreement with the FDS$\&$F3D results (Iodice et al., 2019a, b). Indeed, almost all galaxies with a disturbed H\scaleto$I$1.2ex morphology are of late type and belong to the group of the infalling galaxies in (Iodice et al., 2019a), which are symmetrically distributed around the cluster’s central region. These galaxies have active star formation and are located in the low-density region of the cluster, where the X-ray emission is faint or absent. Our results show that they are interacting with the cluster environment. Deeper into the cluster, the lack of H\scaleto$I$1.2ex detections is consistent with the result that this region is dominated by evolved early-type galaxies. Some of these galaxies have been able to retain part of their H₂\scaleto1.2ex0pt reservoirs (Zabel et al., 2019), but not their H\scaleto$I$1.2ex.

The blind ATCA H\scaleto$I$1.2ex survey of the Fornax galaxy cluster covers a field of 15 deg² out to a distance of $\sim R_{\mathrm{vir}}$ from the cluster centre. It has a spatial and velocity resolution of 67″ $\times$ 95″and 6.6  km s^-1, respectively, and a 3$\sigma$ $N_{\text{H\scaleto{$I$}{1.2ex}}}$ and $M_{\text{H\scaleto{$I$}{1.2ex}}}$ sensitivity of  $\sim$2$\leavevmode\nobreak\ \times$ 10¹⁹cm^-2 and $\sim$2$\leavevmode\nobreak\ \times\leavevmode\nobreak\$ 10⁷M_⊙, respectively. The survey revealed H\scaleto$I$1.2ex emission from 16 Fornax galaxies covering a mass range of about three orders of magnitude, from $8\times 10^{6}$ to $1.5\times 10^{10}$ M_⊙. These galaxies exhibit a variety of disturbances of the H\scaleto$I$1.2ex morphology, including asymmetries, tails, offsets between H\scaleto$I$1.2ex and optical centres and a case of a truncated H\scaleto$I$1.2ex disc (Fig. 2). This suggests environmental interactions within or on their way to Fornax (whether with other galaxies, the large-scale potential or the intergalactic medium), supported by the offset of Fornax galaxies towards low $M_{\text{H\scaleto{$I$}{1.2ex}}}$/$M_{\star}$ ratios with respect to the xGASS $M_{\star}$-$M_{\text{H\scaleto{$I$}{1.2ex}}}$/$M_{\star}$ scaling relation (Fig. 6), and resulting in H\scaleto$I$1.2ex deficiencies similar to those observed in the Virgo cluster (Fig. 8). The H\scaleto$I$1.2ex sample of Fornax galaxies is also forming stars at a lower rate than samples of non-cluster galaxies at fixed $M_{\star}$ (Fig. 10). This deficit of SFR is consistent with the deficit of H\scaleto$I$1.2ex when compared to non-cluster galaxies (Fig. 11).

Our 16 detections reside outside the virialised region of the cluster – where the distribution of the general population of Fornax galaxies is clustered – both on the sky and in the projected phase space diagram (Fig. 12 and Fig. 13). This result implies that H\scaleto$I$1.2ex is lost down to the ATCA sensitivity within a crossing time ($\tau_{\mathrm{HI,loss}}\leavevmode\nobreak\ \leq\leavevmode\nobreak\ \tau_{\mathrm{cross}}\leavevmode\nobreak\ \sim 2$ Gyr), and that our H\scaleto$I$1.2ex detections are recent arrivals in the cluster. They still reside at the outskirts of Fornax, where their H\scaleto$I$1.2ex and SFR properties suggest that H\scaleto$I$1.2ex has so far been lost on a time scale longer than the H₂\scaleto1.2ex0pt depletion time ($\tau_{\mathrm{HI,loss}}\leavevmode\nobreak\ \geq\leavevmode\nobreak\ \tau_{\mathrm{H_{2},depl}}\leavevmode\nobreak\ \sim 1\mathrm{-}2$ Gyr). In the cluster’s central regions H\scaleto$I$1.2ex removal is likely to proceed faster ($\tau_{\mathrm{HI,loss}}\leavevmode\nobreak\ <\leavevmode\nobreak\ \tau_{\mathrm{H_{2},depl}}$). This is supported by the relatively high SFR of H\scaleto$I$1.2ex-undetected, H₂\scaleto1.2ex0pt-detected galaxies and by the anomalously high $M_{\text{H}_{2}}$/$M_{\text{H\scaleto{$I$}{1.2ex}}}$ ratios of galaxies in those regions (Fig. 9, Fig. 12 and Fig. 13). These are galaxies where SFR is likely to be proceeding relatively unperturbed after rapid removal of the H\scaleto$I$1.2ex.

This picture is enriched by the new detection of the NGC 1365 subgroup – where both pre-processing and early interaction with the cluster environment are plausible scenarios to account for the H\scaleto$I$1.2ex properties of its members – and by the detection of several galaxies with peculiar ISM properties, such as some H\scaleto$I$1.2ex-rich but H₂\scaleto1.2ex0pt-poor and low-SFR galaxies (NGC 1427A, ESO 358-G060). The future MeerKAT Fornax Survey (Serra et al., 2016) will observe this cluster with a better resolution and sensitivity than those of our ATCA survey, enabling a further step forward in the study of the evolution of Fornax galaxies.