2.1. SAMI Galaxy Survey

The SAMI-GS was carried out with the SAMI instrument mounted at the prime focus of the 3.9 m Anglo-Australian Telescope (AAT). The SAMI instrument uses 13 integral-field units (namely hexabundles; Bland-Hawthorn et al. Reference Bland-Hawthorn2011; Bryant et al. Reference Bryant, Bland-Hawthorn, Fogarty, Lawrence and Croom2014) composed of 61 fused 1.6 arcsec-diameter fibres. These hexabundles are fed into the AAOmega spectrograph (Sharp et al. Reference Sharp, McLean and Iye2006). As of SAMI Data Release 3 (DR3; Croom et al. Reference Croom2021), the SAMI-GS observed 3 068 unique galaxies sampled from the Galaxy and Mass Assembly Survey (GAMA; Driver et al. Reference Driver2011) equatorial regions (G09, G12, G15; 2151 galaxies) and eight cluster regions selected from the catalogue of De Propris et al. (Reference De Propris2002) in the 2-degree Field Galaxy Redshift Survey, and the Cluster Infall Regions in the Sloan Digital Sky Survey (CIRS; Rines & Diaferio Reference Rines and Diaferio2006), with halo masses of ${14.25\leq \log\left(\mathrm{M}_{200}/\mathrm{M}_\odot\right)\leq 15.19}$ (917 galaxies) described in Owers et al. (Reference Owers2017). The observed galaxies have stellar masses ranging between $8\lt$ ${\log\left(\mathrm{M}_{*}/\mathrm{M}_\odot\right)}$ $\lt 12$ .

The details for target selection are given in Bryant et al. (Reference Bryant2015) for GAMA regions and Owers et al. (Reference Owers2017) for the cluster sample. Briefly, the primary targets are selected with a step-like stellar mass function with an increasing lower limit as a function of redshift, where redshift refers to the flow-corrected value, $z_{\text{tonry}}$ , for the GAMA regions (Tonry et al. Reference Tonry, Blakeslee, Ajhar and Dressler2000; Baldry et al. Reference Baldry2012). To define primary targets for the cluster regions, only two steps are adopted as ${\log\left(\mathrm{M}_{*}/\mathrm{M}_\odot\right)}$ $\gt$ 9.5 if the cluster redshift, $ z_{cl},\lt 0.045$ , otherwise ${\log\left(\mathrm{M}_{*}/\mathrm{M}_\odot\right)}$ $\gt$ 10. Additionally, cluster targets are selected within ${R}_{200}$ and $|v_{\text{pec}}|/\sigma_{200}\leq3.5$ , where $v_{\text{pec}}$ is the peculiar velocity of galaxies with respect to the cluster redshift, $v_{\text{pec}} = c \times (z_{\text{gal}} - z_{\text{cl}}) / (1 + z_{cl}$ ) with c being the speed of light, and $\sigma_{200}$ is the velocity dispersion of the cluster measured within ${R}_{200}$ defined by Owers et al. (Reference Owers2017). Note that both $z_{\text{cl}}$ and $z_{\text{gal}}$ are not flow-corrected. This selection is less strict than the cluster membership allocation done by Owers et al. (Reference Owers2017), which was based on caustic analysis, and it therefore includes non-bona fide members with redshifts close to the cluster.

One of the main aims of this study is to understand the environmental impact on star formation activity. To achieve this, we need a sample from a non-cluster field as a reference/control. Therefore, we use the GAMA portion of the survey as a control sample, which will enable a direct comparison to assess environmental effects. GAMA galaxies are selected through a step-like stellar mass function for $ 0.005 \lt z_{\text{tonry}} \lt 0.095$ (Bryant et al. Reference Bryant2015), which is shown in Figure 2.

2.2. SAMI data products

2.2.2. Spectral classification maps

We adopt the spectral classification scheme described in Owers et al. (Reference Owers2019). The scheme incorporates the emission and absorption line features simultaneously. A spectrum is classified as ‘emission’ if a spaxel exhibits either H $\alpha$ or [NII] $\lambda$ 6584 (i.e. the primary lines) and an additional line from H $\beta$ , [OIII] $\lambda$ 5007 [SII] $\lambda6716$ , or [SII] $\lambda6731$ with SNR=(flux/ $\sigma_{\text{flux}})\gt 3$ , where $\sigma_{\text{flux}}$ is the associated flux uncertainty estimated through a least-square fitting procedure (Ho et al. Reference Ho2016). To avoid spurious detections due to template mismatches, the primary line must also have equivalent width, EW $\gt$ 1 Å. Otherwise, the spectrum is classified as ‘absorption’.

The emission line classification of Owers et al. (Reference Owers2019) is summarised as follows:

• • If H $\alpha$ , [NII] $\lambda$ 6584, and at least one of H $\beta$ or [OIII] $\lambda$ 5007 are detected, Baldwin, Phillips & Terlevich (BPT) diagnostic (Baldwin, Phillips, & Terlevich Reference Baldwin, Phillips and Terlevich1981) is applied to classify the ionisation source as ‘star-forming (SF)’, or ‘intermediate (INT)’, or ‘non-star-forming (NSF)’, through the demarcation lines defined by Kewley et al. (Reference Kewley, Dopita, Sutherland, Heisler and Trevena2001) (Ke01) and Kauffmann et al. (Reference Kauffmann2003) (Ka03).

• • If only H $\alpha$ and [NII] $\lambda$ 6584 are present, WHAN diagram of Cid Fernandes et al. (Reference Cid Fernandes2010) is used. SF, INT, and NSF classes are separated using $\log$ ([NII] $\lambda 6584/H\alpha$ ) = $-0.32$ and $\log$ ([NII] $\lambda6584/H\alpha$ ) = $-0.1$ values, which approximate the Ke01 and Ke03 demarcation lines, respectively.

• • In case of one of the primary lines missing, the classification is done as follows – star-forming (SF) if H $\alpha$ is present, and non-SF if only [NII] $\lambda$ 6584 is detected.

• • Additional sub-classes are defined based on the H $\alpha$ line strength, following the recipe given in Cid Fernandes et al. (Reference Cid Fernandes2010). If EW(H $\alpha$ ) $\lt$ 3 Å, the spectrum can be divided into ‘wSF’ or ‘rINT’ or ‘rNSF’. NSF spectra with EW(H $\alpha$ ) ranging between 3 and 6 Å are classified as ‘wNSF’, whereas those with EW(H $\alpha$ ) $\gt$ 6 Å are ‘sNSF’. Here, ‘r’, ‘w’, and ‘s’ stand for ‘retired’, ‘weak’, and ‘strong’ sub-classes defined by Cid Fernandes et al. (Reference Cid Fernandes2010).

Absorption line measurements are carried out on absorption spectra as well as on emission spectra classified as NSF spectra, whose emission is not powered by photoionisation. The spectra with strong Balmer absorption (i.e. EW $(\overline{\mathrm{H}_{\delta\gamma\beta}})\lt -3$ Å and $|\mathrm{SNR}(\mathrm{EW}(\overline{\mathrm{H}_{\delta\gamma\beta}}))|\gt 3$ ) are classified as ‘H $\delta$ -strong (HDS)’ or ‘non-star-forming H $\delta$ -strong (NSF-HDS)’. The absorption spectra not exhibiting strong Balmer absorption are categorised as ‘passive’.

Combining emission and absorption classification, the global spectroscopic classification is performed as follows: ‘passive galaxies (PASGs)’ are defined as with more than 90% of the spaxels classified as ‘passive’, ‘rNSF’, ‘rINT’, ‘wNSF’, ‘sNSF’, or ‘wSF’; ‘star-forming galaxies (SFGs)’ are those present with at least 10% of the spaxels as ‘SF’ or ‘INT’; galaxies, where ‘HDS’ or ‘NSF-HDS’ spaxels are making up more than 10% of the total, are classified as ‘H $\delta$ -strong galaxies (HDSGs)’. Please refer to Owers et al. (Reference Owers2019) for the complete procedure of spectroscopic classification.

Figure 1. A visualisation of the steps involved in producing the ionised gas map for an example galaxy, 9011900084. The two leftmost panels show the H $\alpha$ and [NII] $\lambda$ 6584 flux maps generated using the detection procedure described in Section 2.2.2, and the colour bar presents the flux values. The middle panel presents the final emission detection map, highlighting connected spaxels and coloured by the detected emission line, which is also shown in the colour bar. The two rightmost panels show the total emission map, derived from the final detection map using the same colour bar as the leftmost panels but applied to H $\alpha$ + [NII] $\lambda$ 6584 emission. The corresponding binary detection map is also shown, where ‘1’ and ‘0’ indicate spaxels with and without detected emission, respectively. The black and red contours displayed in all panels refer to the stellar continuum defined in Section 3.1, and the SAMI field of view, respectively. The orientation is as North towards up, and East towards the left.

2.2.4. SFR maps

One key element of this study is to understand how RPS influences star-formation activity. Using IFU observations from SAMI, we are able to conduct a spatially resolved examination of star-forming properties of cluster galaxies. In this section, we describe the procedure to generate resolved SFR maps estimated from H $\alpha$ fluxes.

We make use of the per-spaxel classification scheme of Owers et al. (Reference Owers2019) to identify star-forming spaxels. For these star-forming spaxels, we correct H $\alpha$ fluxes for intrinsic dust extinction and Galactic foreground extinction similar to the recipes given in Medling et al. (Reference Medling2018) and Mun et al. (Reference Mun2024). The dust extinction is corrected using the observed Balmer decrement (BD), $(\mathrm{H}\alpha/\mathrm{H}\beta)_\mathrm{{obs}}$ . Here, BD maps are smoothed by a truncated Gaussian kernel with FWHM of 1.6 spaxels in order to account for aliasing in the differential atmospheric refraction effect (refer to Medling et al. Reference Medling2018 for details). We assume an intrinsic BD, $(\mathrm{H}\alpha/\mathrm{H}\beta)_\mathrm{{int}}$ , value of 2.86 based on Case B recombination line (Osterbrock Reference Osterbrock1989; Calzetti et al. Reference Calzetti2000), and Milky Way extinction curve (i.e. $R_V \simeq 3.1$ ) from Cardelli et al. (Reference Cardelli, Clayton and Mathis1989). For spaxels without H $\beta$ detection, or with observed BD less than intrinsic value (i.e. 2.86), we fix the correction value and its error to 1 and 0, respectively, by following Medling et al. (Reference Medling2018). Additionally,Footnote ^a the spaxels with observed BD greater than 10, or H $\alpha$ flux $\gt 40\times 10^{-16}\ \mathrm{erg/s/cm^{-2}}$ are excluded, to eliminate spurious H $\alpha$ fits to the edges of the bundles. The foreground Galactic extinction is corrected for H $\alpha$ fluxes using dust maps produced by Schlegel et al. (Reference Schlegel, Finkbeiner and Davis1998), Schlafly & Finkbeiner (Reference Schlafly and Finkbeiner2011) for the given coordinates. We use the observed frame wavelengths to obtain the extinction coefficients.

Once extinction-corrected H $\alpha$ fluxes are obtained, we convert them to H $\alpha$ luminosities via luminosity distances for the adopted cosmology. The redshifts of cluster galaxies are fixed to the cluster redshift to minimise the scatter due to peculiar motions relative to the cosmological redshift of the cluster, while $z_{\text{tonry}}$ values (Baldry et al. Reference Baldry2012; Bryant et al. Reference Bryant2015) are used for the GAMA sample. We then determine SFRs by implementing the estimator given by Kennicutt (Reference Kennicutt1998) as follows:

(1) \begin{equation} SFR = \frac{7.92 \times 10^{-42} \times L_{H\alpha}}{1.53} \ \ [\mathrm{M_\odot \ yr^{-1}}]\end{equation}

Here, $L_{H\alpha}$ is H $\alpha$ luminosity, and 1.53 is the conversion factor between the Salpeter (Reference Salpeter1955) and Chabrier (Reference Chabrier2003) initial mass functions.

For the given redshift range, a spaxel covers a range of physical size in kpc. Therefore, to compare SFRs within the same physical size, we determine star formation surface densities, $\Sigma_{\mathrm{SFR}}$ , by normalising SFR with the physical area of each spaxel as follows:

(2) \begin{equation} \Sigma_{SFR} = \frac{SFR}{(0.5 \times PS(z) )^2}\ \ [\mathrm{M_\odot \ yr^{-1} \ kpc^{-2}}]\end{equation}

where 0.5 is the SAMI pixel scale in arcsecond/spaxel and PS(z) is the physical size in kpc/arcsecond for the adopted cosmology. Redshifts are treated the same as above. Uncertainties are estimated by propagating the formal errors of H $\alpha$ and H $\beta$ fluxes using the uncertainties Python package (Lebigot Reference Lebigot2024).

Figure 2. Left panel: Stellar mass and redshift ( $z_{\text{tonry}}$ for GAMA) distribution of the full SAMI sample (Croom et al. Reference Croom2021) with available spectroscopic classification from Owers et al. (Reference Owers2019) $(N=2\,908)$ . The cyan and red points represent the primary targets for GAMA and cluster regions, respectively, whereas the green pluses and black crosses show the secondary targets for the same samples. The black solid line shows the selection steps in stellar mass for primary GAMA targets as defined in Bryant et al. (Reference Bryant2015). The black dashed line marks the lower stellar mass limit we adopted in this study – ${\log\left(\mathrm{M}_{*}/\mathrm{M}_\odot\right)}=10$ . The blue hatched area encloses GAMA galaxies selected as a control sample. Middle panel: The distribution of detection fraction ( $f_{\text{det}}$ ) of cluster galaxies, as defined in Section 2.3, with $N_{\text{spaxel}}(Emission) \neq 0 (N=194)$ . The red, blue, and green step histograms show the distributions for passive galaxies (PASGs), star-forming galaxies (SFGs), and H $\delta$ -strong galaxies (HDSGs), respectively. The black vertical line marks the cut applied as $f_{\text{det}} \gt 0.2$ . Right panel: The stellar mass distributions of the final cluster sample defined in Section 3.2 $(N=81)$ and the selected star-forming GAMA sample ( $N=462$ , the cyan histogram). P-value from the KS test for the comparison between the final cluster and the GAMA samples is shown on the right.

2.2.5. Stellar Mass Maps

Stellar mass maps (i.e. per spaxel) are generated through the empirical relation given by Taylor et al. (Reference Taylor2011) and adjusted by Bryant et al. (Reference Bryant2015) and Owers et al. (Reference Owers2017) for the observed frame as follows

(3) \begin{equation}\begin{split} \log(M_\ast/M_\odot) & = -0.4i + 2\ \log(D_L(z)) -2 - \log(1 + z)\\ & \ \ + (1.2117 {-} 0.5893z) {+} (0.7106 {-} 0.1467z) {\times} (g - i)\end{split}\end{equation}

where i is the Galactic extinction-corrected i-band apparent magnitude, $(g-i)$ is the Galactic extinction-free colour, $D_L$ is the luminosity distance under the assumed cosmology, and $z$ refers to redshift. Here, the luminosity distances are determined using the same approach as in Section 2.2.4 for both cluster and GAMA regions. The remaining three terms use the redshifts as described in Section 2.1. Taylor et al. (Reference Taylor2011) produce this relation for SDSS photometry, which is missing for the southern sky portion of the SAMI cluster sample. Therefore, we exploit griz imaging from the Legacy Survey Data Release 10 (LS DR10; Dey et al. Reference Dey2019), for full SAMI cluster coverage and to generate maps homogeneously across the sample. We first retrieve cutout images matching the size and the pixel resolution to SAMI cubes (i.e. 25 $^{\prime\prime}$ and 0.5 $^{\prime\prime}$ /pixel, respectively) and apply $m = 22.5 - 2.5 \log_{10}(flux)$ for magnitude conversion. Unlike previous data releases, DR10 provides i-band data for the first time, incorporating additional DECam imaging, although not as complete as grz-bands in terms of spatial coverage. For any galaxies with missing i-band imaging, we generate mock i-band magnitudes/images via the relation, given below, for the available riz magnitudes, where we use the largest aperture (i.e. 14 $^{\prime\prime}$ ) and non-extinction corrected values from the Tractor catalogue.

(4) \begin{equation} i_{LS,\ {\text{mock}}} = 0.318 \times (r-z)_{LS} + 0.057 + z_{LS} \ \ \text{(rms)} = 0.361\end{equation}

We correct the magnitudes only for foreground Milky Way extinction, given that Equation (3) is not strongly impacted by the intrinsic dust attenuation (see Section 5.2.3 in Taylor et al. Reference Taylor2011). Using transformations given by Abbott et al. (Reference Abbott2021) for DES, we convert extinction-free Legacy magnitudes to SDSS magnitudes and then produce the stellar mass maps via Equation (3). Following the approach of Bryant et al. (Reference Bryant2015), only spaxels with $-0.2 \lt (g-i) \lt 2$ are included in the stellar mass maps. The empirical relation yields very precise mass measurements with an error ( $1\sigma$ ) of $\sim$ 0.1 dex independent of colour (Taylor et al. Reference Taylor2011).

Similar to SFRs, we also determine the spaxel-wise stellar mass surface densities, $\Sigma_\ast$ , as follows:

(5) \begin{equation} \Sigma_\ast = \frac{M_\ast}{(0.5 \times PS(z) )^2}\ [\mathrm{M_\odot \ kpc^{-2}}]\end{equation}

2.3. Sample selection

Our main sample is constructed from the SAMI-DR3 (Croom et al. Reference Croom2021) primary targets with spectroscopic classification from Owers et al. (Reference Owers2019) – 732 cluster members that are allocated by Owers et al. (Reference Owers2017) and 1 855 SAMI-GAMA galaxies ( ${N_{\text{TOT}}} = 2\,587$ ). Based on these galaxies, we select the sample by applying the following criteria.

In the original selection for cluster galaxies in the SAMI-GS, two different stellar mass limits were adopted. As shown in the left panel of Figure 2, for ${z_{\text{cl}}}\gt 0.045$ , galaxies with ${\log(M_\ast/M_\odot)}$ $\lt$ 10 were observed as secondaries. Therefore, to maintain and probe the same mass completeness across the sample, we include cluster members with ${\log(M_\ast/M_\odot)}$ $\gt$ 10, removing 143 galaxies. Moreover, we visually exclude 24 galaxies due to miscentring or nearby objects within the hexabundle, to avoid possible contamination. Finally, we apply a cut depending on the ionised gas detection. We define a detection fraction as follows

(6) \begin{equation} {f_{\text{det}}} = \frac{{N_{\text{emission}}}}{{N_{\text{detection}}}}\end{equation}

where ${N_{\text{detection}}}$ is the total number of spaxels with either emission or continuum detection (i.e. continuum emission). The right panel in Figure 2 highlights ${f_{\text{det}}}$ distribution as a function of spectral types. Here we only show galaxies with ${N_{\text{emission}}}\neq 0 (N=194)$ . Passive galaxies are the main occupants of the lower detection region, with their emission primarily originating from sources other than extended star formation, such as LINER/LIER emission from p-AGB stars, or a central AGN. To minimise this contamination, we exclude galaxies with ${f_{\text{det}}}\lt 0.2$ regardless of spectral type. Additionally, in the sample, some of the brightest cluster galaxies (BCGs) within each cluster are identified as having ionised gas emission larger than the threshold, ${f_{\text{det}}}\gt 0.2$ . Since their emission is not related to RPS, we exclude these BCGs. These cuts result in a cluster sample of 123 members.

As already mentioned in Section 2.1, we aim to compare cluster galaxies with their field counterparts. This is important since it will allow us to assess the environmental effects, through RPS, on galaxy properties, especially on star formation activity. For this purpose, as shown in the left panel of Figure 2, we select a subsample of star-forming GAMA galaxies with ${\log(M_\ast/M_\odot)}$ $\gtrsim10$ and ${0.02\lesssim z_{\text{tonry}} \lesssim0.065}$ to probe the similar redshift and stellar mass regions with the cluster sample, returning 462 galaxies. We compare how similar the stellar mass distributions of the final cluster sample defined in Section 3.2 $(N=81)$ and the GAMA galaxies selected as a control sample $(N=462)$ , using the two-sample Kolmogorov-Smirnov test (Smirnov Reference Smirnov1939). As shown in the right panel of Figure 2, this comparison returns a p-value of 0.58, quantifying the similarity.

3.1. Visual classification of ionised gas distribution

To understand how RPS alters the properties of cluster galaxies, it is essential to identify those undergoing RPS. To achieve this, we employ a visual classification scheme of the ionised gas morphology. Alongside optical imaging, visual classification has also been adapted by other wavelength regimes to examine different sub-components of galaxies, such as ionised gas and neutral gas components (Chung et al. Reference Chung, van Gorkom, Kenney, Crowl and Vollmer2009; Smith et al. Reference Smith2010; Poggianti et al. Reference Poggianti2016; Yoon et al. Reference Yoon, Chung, Smith and Jaffé2017; Jaffé et al. Reference Jaffé2018; Roberts et al. Reference Roberts2021b), which better trace and are used to identify ram-pressure affected galaxies in cluster environments. This is what motivates our approach of basing our visual classification scheme on the ionised gas distribution.

To do this, for the cluster galaxies, four visual classes are used to describe how the emission is distributed with respect to the stellar continuum. We define the continuum by taking the median value of the blue ( $6\,400 \lt \lambda \lt 6\,540$ Å) and red ( $6\,590 \lt \lambda \lt 6\,700$ Å) sidebands around the H $\alpha$ and [NII] $\lambda6584$ emission lines, considering only spaxels with continuum ${S/N} \gt 2$ . The description for each class is given below.

• • Ionised gas is evenly distributed across the galaxy (i.e. follows the continuum) without any clear truncation and/or asymmetry with respect to the continuum.

• • Relatively symmetric ionised emission distribution at the centre of the galaxy with marginal to clear truncation at sides.

• • Ionised gas exhibits clear asymmetric features such as one-sided tail, extra-planar emission, and truncation at the opposite side of these asymmetric features.

• • Irregular emission patterns and/or clear asymmetry which might not be explained by RPS or other features (e.g. spiral arms, etc.), or cases where other visual classes are not sufficient to describe.

Figure 3 shows examples of each of the visual classes defined above. The ionised gas maps were classified by four of us (OC, MSO, MP, and GQ) using the scheme above. Each person assessed the maps by giving a primary classification and a comment, such as ‘Asymmetric’ and ‘Mild Asymmetry’. Considering the primary classification, we assign a final classification for each galaxy as the visual class with the maximum count. If there is an equality between votes, such as $(2N - 2T)$ , we then take the comments into account. Here, we adopt a basic weighting scheme, shown below, as a tiebreaker:

\begin{equation*}{w =\begin{cases} 1, & \text{No comment}\\ 0.75, & \text{Mild Asymmetry}\\ 0.75, & \text{Mild Truncation}\\ 0.5, & \text{Another class (e.g. N, T, A, U)} \end{cases}} \end{equation*}

Figure 3. Examples of visual classes defined in Section 3.1. Galaxy IDs above each panel are coloured based on their classification stated below: unperturbed – grey, truncated – magenta, asymmetric – blue, unclear - green, aperture – black. The cividis colour map shows the total (i.e. H $\alpha$ +[NII] $\lambda$ 6584) emission, while the black and red contours represent the boundary of the stellar continuum at SNR = 2, and the SAMI field of view (i.e. $\sim$ 15 $^{\prime \prime}$ in diameter), respectively. The magenta and lime ellipses correspond to 0.5 and 1 ${R_e}$ , respectively. North is towards the up, and East is towards the left.

The minimum value we set for w is 0.5, indicating that the person voting is indecisive between two classes. If there is no comment, no correction is needed (i.e. $w=1$ ). The remaining comments are set to 0.75 as the mid-point. Accordingly, we multiply each count of given votes by the weights defined per comments (i.e. $1 \times w$ ). The class with the maximum weighted count is then assigned as the final class. Lastly, in case of no consensus (i.e. 1 vote per class), we classified them as ‘Unclear’.

With the final classifications established, we assessed the consistency between classifiers across all classes. For this purpose, we define a ‘Gold’ sample, in which at least three out of four classifiers agree on the same visual class (e.g., 3/4 or 4/4 agreement). This Gold sample includes $\sim$ 71% (87/123) of the galaxies, indicating that the majority of the votes are in agreement, as an initial check. For the classifier-based assessment, we treat the final classes from the Gold sample as the ‘ground truth’ and compare them with the votes per classifier. We use f1-scores and accuracy to quantify consistency. Three out of four classifiers achieve f1-scores over 0.80, 0.90 for ‘Unperturbed’ and ‘Asymmetric’ classes, respectively, and the minimum f1-score is 0.7. All f1-scores for ‘Truncated’ galaxies exceed 0.90. For the ‘Unclear’ class, f1-scores range between 0.36 and 0.8, which aligns with the definition of ‘unclear’. The minimum accuracy across classifiers is 0.86. Overall, we find good agreement across the classes, particularly for the ‘Truncated’ and ‘Asymmetric’ galaxies, as the primary focus of this study.

Note that the SAMI field of view corresponds to $\sim$ 7.5 arcsec in radius and $\sim$ 80% of primary targets have been imaged out to at least 1 $R_e$ (Croom et al. Reference Croom2012; Bryant et al. Reference Bryant2015). Still, there are galaxies having $R_e$ larger than the bundle radius, indicating that we only cover the central regions. However, since $R_e$ is measured using r-band imaging (Croom et al. Reference Croom2021), larger $R_e$ values may not necessarily mean that they are affected by aperture according to our visual classification. If the ionised gas of a galaxy fills the entire bundle, we visually classify that galaxy as ‘Aperture’, irrespective of its ${R_e}$ . An example of an aperture-affected case is shown in Figure 3 (i.e. ID=9016800425). Additionally, if a galaxy is partially covered by the bundle due to it is not well-centred, we also classify it as ‘Aperture’.

3.2. Populations

In the previous section, we classified galaxies based on their ionised gas morphologies. We are able to examine whether these classes differ in key properties, such as structural parameters, projected phase-space distribution, and resolved star formation activities. Identifying such differences may provide insights into transformation as a function of RPS activity. The first step towards investigating the impact of RPS is to understand the distribution of classes across the sample before examining their properties in detail.

Figure 4 shows the distribution of the visual classes. The numbers above each bin indicate the number of galaxies in each class. Nearly two-thirds of the sample are classified as either unperturbed or aperture-affected, comprising 36 galaxies ( $29.3^{+4.4}_{-3.7}$ %) and 37 galaxies ( $30.1^{+4.4}_{-3.8}$ %), respectively, out of 123. Truncated galaxies make up 30/123 ( $24.4^{+4.2}_{-3.4}$ %), asymmetric galaxies account for 15/123 ( $12.2^{+3.6}_{-2.3}$ %) of the sample, and the minority is represented by unclear galaxies with 5/123 ( $4.1^{+2.6}_{-1.1}$ %). This results in a total of $36.6^{+4.5}_{-4.1}$ % for RPS-affected galaxies, combining truncated and asymmetric classes (45/123). Associated uncertainties shown here and below are estimated following the approach given in Cameron (Reference Cameron2011).

Table 1. The breakdown of the visual classes as a function of spectral types described in Section 2.2.2.

Moreover, we examine the distribution of visual classes as a function of spectral types. Table 1 presents the breakdown of classes. Nearly all unperturbed galaxies (34/36; $94.4^{+1.8}_{-6.5}$ %) are star-forming. More than half of truncated galaxies ( $56.7^{+8.3}_{-9.1}$ %; 17/30) are also SFGs, while PASGs account for 11/30 ( $36.7^{+9.3}_{-7.7}$ %). The majority of HDSGs ( $70^{+10.0}_{-16.8}$ %) contribute to asymmetric galaxies, accounting for nearly the half of the class ( $46.7^{+12.4}_{-11.6}$ %; 7/15), followed by PASGs with $33.3^{+13.4}_{-9.5}$ % (5/15) and SFGs with 3/15 ( $20^{+13.6}_{-6.5}$ %). Aperture-affected sample only consists of star-forming galaxies, which is expected given that the majority of our sample (93/123) is star-forming.

Figure 4. The number of galaxies per visual class. The colours are the same as described in Section 3.1 – unperturbed – grey, truncated – magenta, asymmetric – blue, unclear – green, and aperture – black. The numbers above each bin show the number of galaxies per class.

To understand the aperture-affected cases more quantitatively, we compare the stellar masses and the half-light radii of the total H $\alpha$ +[NII] $\lambda$ 6584 flux, ${R_e}$ (H $\alpha$ + [NII] $\lambda$ 6584), which are measured using elliptical apertures. Figure 5 shows the distribution of mass and ${R_e}$ (H $\alpha$ + [NII] $\lambda$ 6584) across the samples. It is clearly seen that the majority of galaxies with larger ${R_e}$ (H $\alpha$ + [NII] $\lambda$ 6584) values are aperture-affected cases, while almost all unperturbed, truncated, and asymmetric galaxies exhibit lower ${R_e}$ (H $\alpha$ + [NII] $\lambda$ 6584) values, quantitatively agreeing that the ionised gas distribution of aperture cases is more extended. In addition, galaxies affected by the SAMI aperture appear to be more massive compared to their unperturbed counterparts. For the remainder of the paper, we exclude galaxies classified as either ‘Unclear’ or ‘Aperture’ from our analysis, which leaves us with 81 cluster galaxies as our final sample.

Figure 5. Distribution of stellar mass and half light radius of H $\alpha$ + [NII] $\lambda$ 6584 flux, measured within the SAMI bundle, for the visual classes defined in Section 3.1 – unperturbed – grey points, truncated – magenta squares, asymmetric – cyan stars, and aperture – black points. Here, we exclude ‘Unclear’ cases. The majority of galaxies with larger ${R_{e}}(\mathrm{H}\alpha \mathrm{+ [NII]})$ values are primarily aperture-affected galaxies, quantitatively supporting the reasoning behind them, which is the emission filling the bundle.

3.3. Visual classification on quantitative space

Regardless of how high the expertise across the classifiers is, any kind of visual classification is prone to being subjective. Therefore, quantitative analysis methods based on non-parametric measures of the stellar light distribution have been developed to understand galaxy morphologies in a coherent and unbiased manner, such as CAS statistics (Abraham et al. Reference Abraham1996; Bershady, Jangren, & Conselice Reference Bershady, Jangren and Conselice2000; Conselice, Bershady, & Jangren Reference Conselice, Bershady and Jangren2000; Conselice Reference Conselice2003) and Gini-M₂₀ (Lotz, Primack, & Madau Reference Lotz, Primack and Madau2004). These methods have also been applied to nebular emission (Nersesian et al. Reference Nersesian, Zibetti, D’Eugenio and Baes2023) and exploited to identify ram-pressure-affected galaxies through different wavelengths (McPartland et al. Reference McPartland, Ebeling, Roediger and Blumenthal2016; Roberts & Parker Reference Roberts and Parker2020; Roberts et al. Reference Roberts2021a; Bellhouse et al. Reference Bellhouse2022; Krabbe et al. Reference Krabbe2024). Therefore, we similarly aim to identify the ram-pressure-affected galaxies in our sample quantitatively using different non-parametric measures such as concentration, asymmetry, and the offset between the ionised gas and the stellar continuum, as described below.

Concentration (C): We adopt the concentration definition of Schaefer et al. (Reference Schaefer2017), given by the ratio of the half-light radius determined in emission and continuum flux maps using elliptical apertures, as follows

(7) \begin{equation} C = \frac{r_{50}(H\alpha + [NII])}{r_{50}(Continuum)}\end{equation}

This definition allows us to examine the emission relative to the continuum, aligning with the visual classification scheme. We estimate the concentration similar to the method outlined in Schaefer et al. (Reference Schaefer2017) and Owers et al. (Reference Owers2019), except we consider the total H $\alpha$ +[NII] $\lambda$ 6584 flux defined in Section 2.2.3. The interpretation of this parameter is that the higher the value, the more extended the ionised gas emission.

Asymmetry (A): We employ the standard asymmetry expression (Abraham et al. Reference Abraham1996; Conselice et al. Reference Conselice, Bershady and Jangren2000; Conselice Reference Conselice2003) as follows

(8) \begin{equation} A = \frac{\Sigma|I_0 - I_{180}|}{2\times\Sigma |I_0|} \end{equation}

with $I_0$ being the original emission line flux image and $I_{180}$ being the flux image rotated by 180 $^{\circ}$ . We fix the centre of rotation as the cube centre (i.e. galaxy centre). Due to its flux-weighted nature, the standard definition (hereafter ${A_{\text{flux}}}$ ) is not sensitive to the faint/low-surface-brightness features, for instance, possible gas tails in our case. In order to account for these features, Pawlik et al. (Reference Pawlik2016) introduced a new asymmetry measure, namely ‘shape asymmetry’ (hereafter ${A_{\text{shape}}}$ ). It uses the same formulation defined above, yet instead of a flux image, a ‘binary detection’ image, made of 1 and 0 based on the pixels included, is input into the calculation. In this way, pixels hosting the faint features, such as those found in ram-pressure stripped tails, are weighted equally with the rest. The following equation summarises the asymmetry definitions used here.

\[{A =\begin{cases} {A_{\text{flux}}}, & \text{if } {I_0} \text{ is flux image} \\{A_{\text{shape}}}, & \text{if } {I_0} \text{ is binary detection image}\end{cases}}\]

Offset (d): We compute the centroid offset similarly to Liu et al. (Reference Liu2021). As done in asymmetry, we determine two offsets between the cube centre, ${(x_c, y_c)}$ , and (i) emission line flux-weighted centre (FWC) and ii) emission line binary detection centre (BDC), defined as ${(x_i, y_i)}$ in the formula below, respectively.

(9) \begin{equation}{d_i} = \sqrt{{(x_{c}-x_{i})}^2 + {(y_{c}-y_{i})}^2} \ \ \ \mathrm{for} \ {i} = \mathrm{\{FWC, BDC\}}\end{equation}

Where $d_i$ returns the offset in pixels. Since a spaxel can cover a range of physical sizes in kpc, we multiply the offset by both the SAMI pixel scale and the physical size defined as in Section 2.2.4 to convert it to kpc.

Associated uncertainties are estimated via Monte Carlo realisations with ${N_{\text{iteration}}}=100$ . In each iteration, we perturb H $\alpha$ and [NII] $\lambda$ 6584 line maps with noise sampled from a normal distribution of $N(0, \sigma_{\text{flux}}$ ), where $\sigma_{\text{flux}}$ is the flux error per spaxel. We then run perturbed maps through the detection procedure described in Section 2.2.3 and measure the same quantities. The standard deviation of these measurements is assigned as the uncertainty per parameter.

Considering the combinations between flux and binary detection maps, we have five different non-parametric measurements in total. Figure 6 shows the distributions of these parameters as a function of the visual classes. Each visual class is coloured as described in Section 3.1. The top row presents concentration, flux-weighted asymmetry, and shape asymmetry parameters, while the offsets determined via flux-weighted and binary detection centroids are given in the bottom row.

Figure 6. The probability distribution functions (PDFs) of non-parametric measures as a function of visual class. The colour code is the same as defined in Section 3.1 – unperturbed - grey, truncated - magenta, asymmetric – blue. The vertical lines for all panels mark the $16{\mathrm{th}}, 50{\mathrm{th}}, \text{ and, }84{\mathrm{th}}$ percentiles, respectively. The medians (i.e. the $50{\mathrm{th}}$ percentile) are also given next to each distribution, and the associated uncertainties are the standard errors, defined as $1.253\times\sigma/\sqrt{N}$ . The stripes embedded within PDFs represent the individual values for each distribution. The top panel shows the distributions for concentration and asymmetry parameters (standard definition, ${A_{\text{Flux}}}$ , in the middle, shape asymmetry, ${A_{\text{shape}}}$ , in the right panel). Bottom panels show the offsets estimated via flux maps ( ${d_{\text{FWC}}}$ ; left) and binary detection maps ( ${d_{\text{BDC}}}$ ; right). These plots indicate that parameters, such as ${A_{\text{shape}}} \text{ and } {d_{\text{BDC}}}$ , are effective in separating visual classes quantitatively, especially those with asymmetric features. Both truncated and asymmetric galaxies have lower concentration values compared to their unperturbed counterparts. Moreover, especially the parameters defined using binary detection maps (e.g. ${A_{\text{shape}}}$ and ${d_{\text{BDC}}}$ ) are more robust at reproducing the visual classification.

Considering the concentration parameter distributions shown in the top left panel of Figure 6, the truncated and asymmetric galaxies both exhibit systematically lower concentration values than those of unperturbed galaxies. The median values are $-0.14$ , $-0.19$ , and $-0.10$ dex for truncated, asymmetric, and unperturbed galaxies, respectively. This similarity is primarily driven by the definition of the concentration parameter using fluxes. As already mentioned above, since asymmetric features, such as tails, have lower fluxes, this results in lower concentration values (i.e. more concentrated). Additionally, because the visual classification of asymmetric galaxies includes partially truncated emission, similar concentration values can be yielded. The distribution for unperturbed galaxies also extends towards low concentration values. Given that these unperturbed galaxies do not show any visual signatures of truncation, smaller values of concentration may be driven by extended emission in the presence of centrally concentrated flux, rather than physical truncation. We compare the concentration distributions pairwise employing the two-sample Anderson–Darling (AD) test (Anderson & Darling Reference Anderson and Darling1952), which evaluates whether two samples originate from the same parent sample or not by taking the tails of the distributions into account. P-values for the pairwise comparisons of each parameter are given in Table 2. The AD test yields ${p\gg0.05}$ only for the comparison of the concentration distributions between truncated and asymmetric galaxies, confirming the similarity quantitatively. In contrast, the test returns ${p\ll0.05}$ for comparisons involving unperturbed galaxies, indicating that unperturbed and RPS-affected galaxies are not likely sampled from the same parent distribution.

Table 2. p-values from the AD test of visual class pairwise comparisons of the shape parameters defined in Section 3.3. Each row lists the parameters, while each row notes the visual class pair. Here, red shaded cells indicate $p\lt 0.05$ .

Focusing on the upper middle and right panels of Figure 6, the flux weighted asymmetry, ${A_{\text{flux}}}$ , yields similar distributions across the visually identified classes, especially between truncated and asymmetric galaxies. On the other hand, the shape asymmetries, ${A_{\text{shape}}}$ , put forth more significant differences between classes, returning lower values for truncated and unperturbed galaxies, whereas asymmetric galaxies extend towards higher values. This is also quantitatively supported by the AD test. As shown in Table 2, only the asymmetric-unperturbed pair yields ${p \lt 0.05}$ for ${A_{\text{flux}}}$ , showing that the flux weighted approach is unable to differentiate the asymmetric and truncated samples. In contrast, the comparisons of the distributions of shape asymmetries return p-values less than 0.05 for all pairs, indicating that ${A_{\text{shape}}}$ is statistically more robust in capturing the asymmetric features compared to ${A_{\text{flux}}}$ .

As shown in the bottom panel of Figure 6, both ${d_{\text{FWC}}}$ and ${d_{\text{BDC}}}$ distributions are similar across visual classes. The majority of the truncated and unperturbed galaxies are clustered at the lower offset values ( $\sim$ 0.3 kpc) with few extending towards higher values ( $\gtrsim$ 1 kpc). Notably, the asymmetric sample exhibits systematically higher offset values as well as a broad distribution in the offset. As we compare the distributions, the AD test returns ${p \lt 0.05}$ for all pairwise comparisons of ${d_{\text{FWC}}}$ and ${d_{\text{BDC}}}$ , except for the comparison between truncated and unperturbed samples. This indicates that the asymmetric galaxies are not likely drawn from the sample parent distribution, compared to the truncated and unperturbed galaxies.

Comparisons shown above reveal that the ${A_{\text{shape}}}$ and ${d_{\text{BDC}}}$ parameters – derived from the binary detection map – and concentration are more robust in segregating the RPS candidates from their unperturbed counterparts. This means that we can use these three parameters to demarcate RPS-affected galaxies. Figure 7 shows a corner plot for two-dimensional joint distributions between concentration, ${A_{\text{shape}}}$ , and ${d_{\text{BDC}}}$ parameters. The grey points, magenta squares, and blue stars represent the visually classified unperturbed, truncated, and asymmetric galaxy samples, respectively. The black crosses, orange pluses, and cyan crosses show aperture-affected galaxies, and those exhibiting ‘mild truncation’ and ‘mild asymmetry’ based on the secondary comments, respectively. The red solid lines mark demarcation lines between visual classes, which are defined below.

Figure 7. Corner plot for concentration, ${A_{\text{shape}}}$ , and $d_{\text{{BDC}}}$ parameters. The grey points, magenta squares, and blue stars represent the galaxies classified as ‘unperturbed’, ‘truncated’, and ‘asymmetric’, respectively. The orange pluses and cyan crosses are the galaxies showing mild truncation and mild asymmetry, respectively, based on the secondary comments. The red solid lines mark the rough demarcation lines between visual classes, as defined in Section 3.3.

There are several noteworthy trends in the 2D parameter spaces shown in Figure 7. The majority of the visually classified truncated galaxies reside in the lower left corner of all three plots, that is, the quantitative morphologies align with them having concentrated ionised gas distributions that are symmetric about the centre of the galaxy. 5 out of 8 visually classified truncated galaxies that have ${A_{\text{shape}}}\gt 0.25$ and ${d_{\text{BDC}}} \gt 1$ were mostly also identified visually as having mild asymmetry, which gives further confidence in the quantitative measures. Furthermore, the majority of the visually identified asymmetric galaxies generally lie in regions of the three spaces that are distinct from the other two visual classes, for example, they generally have both larger ${A_{\text{shape}}}$ and ${d_{\text{BDC}}}$ when compared to the other galaxies. This is expected because these two parameters are designed to quantify asymmetry and, therefore, provide confidence that the visual and quantitative schemes are yielding reliable results.

We compare the 2D distributions shown in Figure 7 between the samples to understand whether they are drawn from the same parent sample or not. For this purpose, we use the non-parametric multivariate two-sample kernel density estimation (KDE) test developed by Duong et al. (Reference Duong, Goud and Schauer2012) and available within the ks library (Duong Reference Duong2007) in R. Table 3 summarises the p-values resulting from each pairwise comparison per parameter space. The KDE test returns ${p}\lt 0.05$ for the comparisons of ${\log(C_{\text{emission}}) - d_{\text{BDC}}}$ , and ${\log(C_{\text{emission}})} - {A_{\text{Shape}}}$ between the asymmetric and unperturbed galaxies, while it returns marginal significance, ${p = 0.057}$ , in ${\log(C_{\text{emission}})} - {d_{\text{BDC}}}$ comparison for the truncated and unperturbed galaxies.

After testing different combinations of morphological parameters, we define the following demarcation lines between each sample as shown in Figure 7: ${\log(C_{\text{emission}})} \sim -0.1$ roughly separates the unperturbed and RPS-affected galaxies; ${A_{\text{shape}}} \sim 0.2 \text{ and } {d_{\text{BDC}}} \sim 0.5$ kpc approximately divide the asymmetric sample from the truncated and unperturbed counterparts.

Table 3. p-values from the 2D KDE test per pairwise comparison of each parameter space. Each row lists the parameters compared, while each column corresponds to a visual class pair. Red shaded cells mark pairs with $p \lt 0.05$ .

3.4. Comparison of the projected phase-space distributions

Analyses presented in the previous sections showed that the ionised gas morphologies differ in unperturbed, truncated, and asymmetric galaxies both visually and quantitatively. As the primary goal of this paper, we aim to understand whether these differences arise as a function of RPS activity (i.e. their accretion times). For this purpose, in this section, we examine the projected phase-space (PPS) distribution of each sample. It is well-established that the PPS diagram determined from the normalised projected cluster-centric distances, ${R}/{R}_{200}$ , and the normalised line-of-sight (LOS) velocity, $v_{\text{pec}}/\sigma_{200}$ , is a useful tool to probe populations with different accretion times (Haines et al. Reference Haines2012; Haines et al. Reference Haines2015; Noble et al. Reference Noble2013; Oman, Hudson, & Behroozi Reference Oman, Hudson and Behroozi2013; Rhee et al. Reference Rhee2017; Owers et al. Reference Owers2019).

Figure 8 shows the distribution of each visual class in the PPS diagram, using absolute values of $v_{\text{pec}}/\sigma_{200}$ , as defined in Section 2.1. The grey points, magenta squares, and blue diamonds represent the unperturbed, truncated, and asymmetric galaxies, respectively. The same coloured contours are generated by ks library (Duong Reference Duong2007) in R, and show the $\mathrm{10{th}}, \text{ and } \mathrm{90{th}}$ percentiles of each distribution, respectively. The upper and right panels display 1D KDEs of the sample distributions for each axis, with the solid lines indicating the bi-weighted means per axis and sample, respectively. Furthermore, Table 4 presents the means and standard deviation of individual distributions of normalised cluster-centric distances and absolute LOS velocities per visual class, determined using a bi-weight estimator (Beers, Flynn, & Gebhardt Reference Beers, Flynn and Gebhardt1990). The uncertainties are estimated through bootstrapping as follows. We generate 100 mock distributions (i.e. ${N_{\text{iteration}}} = 100$ ) by randomly resampling the original parameter distribution and compute the mean and standard deviation in each iteration using the same procedure as for the real data. Uncertainties are then estimated using the 16th and 84th percentiles of the bootstrapped parameter distributions.

Figure 8 and Table 4 both show very clearly that the PPS distribution of asymmetric galaxies is strikingly different compared to those of truncated and unperturbed galaxies. The asymmetric galaxies are concentrated at lower cluster-centric distances ( ${R}/{R}_{200}$ $\lesssim$ 0.5) with the smallest standard deviation. In comparison, the truncated and unperturbed galaxies appear to be located at higher cluster-centric distances, yet exhibit almost uniform distributions across the plane. The 1D two-sample AD test returns ${p \ll 0.05}$ for the pairs with the asymmetric sample, and ${p = 0.154}$ for the unperturbed-truncated pair, confirming the difference in the radial distribution, quantitatively.

Table 4. The bi-weighted (BW) means and standard deviations (Beers et al. Reference Beers, Flynn and Gebhardt1990) of normalised cluster-centric distances and LOS velocities in absolute for each sample. Associated uncertainties are estimated through bootstrapping.

Figure 8. Projected phase-space distribution. The grey points, magenta squares, and blue diamonds represent the unperturbed, truncated, and asymmetric galaxies, respectively. The contours indicate 10th and 90th percentiles of the 2D distribution of each sample, generated by ks library (Duong Reference Duong2007) in R. The p-values from the 2D KDE test (Duong et al. Reference Duong, Goud and Schauer2012), comparing ${R}/{R_{200}}$ with $|v_{\text{pec}}/\sigma_{200}|$ for each pair, are given in the upper right. The top and right panels show the 1D KDEs of normalised cluster-centric distances and line-of-sight velocities, respectively, with solid lines indicating the bi-weigthed means for each class.

Considering the distribution of $|v_{\text{pec}}/\sigma_{200}|$ , both the truncated and unperturbed samples exhibit lower dispersions, as indicated by BW STDEV in Table 4, with respect to their asymmetric counterparts. The AD test returns p-values between 0.3 and 0.4 for $|v_{\text{pec}}/\sigma_{200}|$ comparison between pairs including unperturbed galaxies, whereas truncated – asymmetric pair yields a p-value of 0.045, indicating that they present a statistically significant difference in $|v_{\text{pec}}/\sigma_{200}|$ distribution. When non-absolute LOS velocities are compared (i.e. $v_{\text{pec}}/\sigma_{200}$ ), the same test returns p-values higher than 0.05 for all pair combinations, with a marginal significance of ${p = 0.064}$ for the asymmetric-unperturbed pair.

Lastly, to test the significance of the difference in PPS distributions across the samples, using the absolute LOS velocities (i.e. $|v_{\text{pec}}/\sigma_{200}|$ ), we employ the non-parametric multivariate two-sample KDE test, which has been adopted by several studies (Lopes, Ribeiro, & Rembold Reference Lopes, Ribeiro and Rembold2017; de Carvalho et al. Reference de Carvalho2017; Owers et al. Reference Owers2019; Costa et al. Reference Costa, Ribeiro, de Carvalho and Benavides2024; Çakır et al. Reference Çakır2025). As shown in the upper right of Figure 8, the KDE test yields $p = 0.033$ only for the asymmetric-unperturbed pair. When considering the non-absolute LOS velocities (i.e. $v_{\text{pec}}/\sigma_{200}$ ), p-values between 0.2 and 0.5 are returned. This leads us to the conclusion that the PPS distribution of the asymmetric sample is different to the unperturbed and truncated counterparts, although with marginal statistical significance.

3.5. Star formation activity

The difference in PPS found in Section 3.4 indicates that the asymmetric galaxies may be caught at a different accretion phase when compared with the unperturbed and truncated galaxies. The larger velocity and smaller cluster-centric distance, coupled with the asymmetric ionised gas morphology, further indicate that asymmetric galaxies may be observed at an evolutionary phase closer to peak ram-pressure compared to the truncated and unperturbed galaxies. To understand the impact of RPS on the star-forming properties, in this section, we investigate star formation activity across the asymmetric, truncated, and unperturbed samples. We first investigate the integrated star formation activity in Section 3.5.1. We then investigate the resolved star formation activity through two diagnostics: resolved star formation rate – mass surface density plane, $\Sigma_{\mathrm{SFR}} - \Sigma_\ast$ , and the specific star formation rates (sSFRs) as a function of the galactocentric distances ( ${r/r_{\text{eff}}}$ ).

3.5.1. Integrated star-forming properties

For the purpose of investigating global star-forming properties, we determine integrated SFRs by summing the SFR maps defined in Section 2.2.4, using only SF spaxels, following a method similar to previous studies (González Delgado et al. Reference González Delgado2016; Medling et al. Reference Medling2018; Sánchez et al. Reference Sánchez2019; Mun et al. Reference Mun2024). When summing the per-spaxel SFRs, we do not apply any aperture correction; this means that, as noted by Medling et al. (Reference Medling2018), the integrated SFRs given here are not true global SFRs. However, given the similar selection of the cluster and GAMA control samples, we expect that relative comparisons of the integrated star-forming properties should not be affected by the lack of aperture correction. We also note that the per-spaxel spectroscopic classification outlined in Section 2.2.2 differs from that adopted by Medling et al. (Reference Medling2018), Sánchez et al. (Reference Sánchez2019) in terms of diagnostics used and the SNR cut applied for the emission lines. Therefore, SFR estimates for the common galaxies may differ between these studies.

Figure 9 shows the integrated SFR-stellar mass distribution for our sample. The grey points, magenta squares, and cyan stars represent visually identified unperturbed, truncated and asymmetric galaxies, while the orange hexagons are the GAMA control galaxies. A small number of GAMA (4), asymmetric (4), and truncated (3) galaxies have no ‘star-forming’ spaxels, and we set their SFR to $10^{-5} \ \mathrm{M}_\odot\ \mathrm{yr^{-1}}$ . The solid lime line shows the best fit to the SFMS defined by Fraser-McKelvie et al. (Reference Fraser-McKelvie2021) and determined using SAMI and MANGA galaxies. The majority of the GAMA control sample (357/462) and the unperturbed cluster galaxies (19/36) lie within 1 $\sigma$ (0.76 dex) of the Fraser-McKelvie et al. (Reference Fraser-McKelvie2021) SFMS best fit. On the other hand, both the truncated and asymmetric galaxies are systematically below the SFMS fit of Fraser-McKelvie et al. (Reference Fraser-McKelvie2021), with 10/15 and 26/30 of the asymmetric and truncated galaxies, respectively, found more than 0.5 $\sigma$ below the SFMS. These findings indicate that global star formation in the asymmetric and truncated galaxies is suppressed relative to the unperturbed and GAMA control galaxies.

Figure 9. Integrated star formation rate – stellar mass plane. Integrated star formation rates are measured by summing star formation rates of star-forming spaxels within the SAMI aperture, while stellar mass is estimated through ( $g-i$ ) colours, as defined in Bryant et al. (Reference Bryant2015), Owers et al. (Reference Owers2017). The grey points, magenta squares, and cyan stars show the visually identified unperturbed, truncated, and asymmetric galaxies. The orange hexagons represent the GAMA sample; the darker the bins, the higher the number of galaxies in the bin. Here, we set log(SFR) values to $-5$ if no star-forming spaxels are detected. The solid lime curved line is the SFMS best fit defined by Fraser-McKelvie et al. (Reference Fraser-McKelvie2021) for the SAMI and MANGA galaxies.

3.5.2. Resolved star formation rate – stellar mass relation

While we found in Section 3.5.1 that the global SFR is suppressed for asymmetric and truncated galaxies, that analysis does not allow us to ascertain the impacts of ram-pressure on the local star-forming properties of galaxies in our sample. To that end, we now use the resolved star-formation – stellar-mass distribution to determine if star-formation is locally suppressed or enhanced, and how the resolved star-forming properties relate to those on a global scale.

Table 5. Best-fit coefficients (i.e. slope and intercept) of the spatially resolved $\Sigma_{\mathrm{SFR}}-\Sigma_\ast$ relation for each sample. ${N_{\text{GAL}}}$ and ${N_{\text{SF-spx}}}$ denote the number of galaxies and star-forming spaxels used to define the best-fit. RMS is the root mean square of residuals between best-fit and individual points.

We defined resolved star formation rate and stellar mass surface density maps using the $\Sigma_{\mathrm{SFR}}$ and $\Sigma_{\ast}$ defined in Sections 2.2.4 and 2.2.5. The left panel of Figure 10 shows the $\log\Sigma_{\mathrm{SFR}} - \log \Sigma_\ast$ relation for all ‘star-forming’ spaxels in the GAMA control (orange hexagons) and unperturbed (grey points) samples. The dark orange, black, magenta, and blue contours mark the loci enclosing 90% of 2D density, generated using the seaborn library (Waskom Reference Waskom2021), for the GAMA control, unperturbed, truncated, and asymmetric samples, respectively. The dark orange solid line shows the best-fit linear fit to the $\log\Sigma_{\mathrm{SFR}} - \log \Sigma_\ast$ relation of star-forming spaxels for the GAMA control sample, determined using ltsfit (Cappellari et al. Reference Cappellari2013). The dashed orange line shows the root-mean-square error on the fit. Table 5 presents the best-fit parameters for the GAMA control sample, as well as the parameters determined from fits to the truncated, asymmetric, and unperturbed samples.

Figure 10. The spatially resolved star formation rate – mass plane ( $\Sigma_{\mathrm{SFR}} - \Sigma_{\ast}$ ) for the whole sample (i.e. cluster + GAMA galaxies) using spaxels classified as ‘SF’ by Owers et al. (Reference Owers2019). Left panel: The orange hexagons represent the distribution for GAMA spaxels, while the grey points are spaxels from unperturbed galaxies. The dark orange, black, magenta and blue contours enclose 90% of the 2D density, estimated by seaborn using Scott’s rule of thumb, for the GAMA, unperturbed, truncated, and asymmetric galaxies, respectively. The dark orange solid line marks the best-fit estimated by ltsfit (Cappellari et al. Reference Cappellari2013) for the GAMA galaxies, with the dashed lines indicating the root mean square error of the fit. The best-fit parameters are shown in the upper left of the panel, and in Table 5. Right panel: The solid lines show the running medians for the same distributions shown in the left panel, with the same colour scheme. The medians are defined within 0.2 dex bins and shifted 0.05 dex (i.e. window size and shift) in $\Sigma_{\ast}$ . The error bars are the standard errors on the median (i.e. $1.253 \times \sigma_{\mathrm{bin}}/\sqrt{{N_{\text{bin}}}}$ ) and the dashed lines represent $\mathrm{16{th}} \text{ and } \mathrm{84{th}}$ percentiles of the distributions per bin. The dark orange solid line is the best-fit of the GAMA sample.

Despite the large scatter, the left panel of Figure 10 reveals that $\Sigma_{\mathrm{SFR}}$ and $\Sigma_\ast$ show a clear correlation in which spaxels with larger $\Sigma_\ast$ exhibit higher star formation rates. This holds independently of the sample to which the spaxels belong, although the truncated and asymmetric samples appear to host a steeper relation when compared with the GAMA control and unperturbed samples. These differences in slope are confirmed by the parameters from the best-fit relations shown in Table 5, where the truncated and asymmetric samples have slopes $\simeq 1.4-1.5$ , while the GAMA control and unperturbed samples have slopes $\simeq 0.8-0.9$ .

The right panel of Figure 10 shows the median log $\Sigma_{\mathrm{SFR}}$ values, estimated through a sliding window approach (i.e. running median), with a window of full width of 0.2 dex and a step size of 0.05 dex along $\mathrm{log } \ \Sigma_{\ast}$ . The uncertainties per bin are defined as the standard error on the median – $1.253 \times \sigma_{\mathrm{bin}}/\sqrt{{N_{\text{bin}}}}$ . The $\mathrm{16{th}} \text{ and } \mathrm{84{th}}$ percentiles of each bin are also shown with the dashed lines for each sample. The dark orange solid and dashed lines show the same best-fit and error as that shown in the left panel for the GAMA sample. For log $\Sigma_\ast\gt 8\ \mathrm{M}_\odot \ \mathrm{kpc}^{-2}$ , the median profiles show comparable trends for the GAMA, unperturbed, asymmetric, and truncated samples. In contrast, for lower masses, log $\Sigma_\ast \lt 8\ \mathrm{M}_\odot \ \mathrm{kpc}^{-2}$ , the asymmetric and truncated galaxies diverge from the unperturbed and GAMA samples, exhibiting lower $\Sigma_{\mathrm{SFR}}$ values, which drives the steeper slopes seen in Table 5. Furthermore, the truncated galaxies have systematically lower log $\Sigma_{\mathrm{SFR}}$ values than the asymmetric sample. This is most evident, at log $\Sigma_\ast \lesssim 8.5\ \mathrm{M}_\odot \ \mathrm{kpc}^{-2}$ , where the offset is $\sim$ 0.2 dex.

3.5.3. Radial trends in specific Star Formation Rate (sSFR)

In Section 3.5.2, we showed that there are differences between the resolved SFMS of unperturbed and ram-pressure-affected galaxies. As the next step, to rule out the mass dependency on the star-formation activity, we investigate the radial sSFR (i.e. SFR per unit mass) profiles. For this purpose, we first measure the distance of each spaxel with respect to the galaxy centre. The ellipticities and position angles from the input catalogue (Bryant et al. Reference Bryant2015; Owers et al. Reference Owers2017) are used to account for projection effects. We then normalise the distances by the effective radius, ${r_{\text{eff}}}$ , of each galaxy to enable like-for-like comparisons.

Figure 11 shows the radial sSFR profile of each sample. The orange, grey, magenta, and blue solid lines represent the medians for the GAMA, unperturbed, truncated, and asymmetric samples. Similar to that in Figure 10, a sliding window approach is used to determine the medians of log sSFR binned in normalised galactocentric distance, their associated uncertainties (error bars), and the $\mathrm{16{th}} \text{ and } \mathrm{84{th}}$ percentiles (dashed lines). The inset plot zooms in on the central ${r_{\text{eff}}}$ . The unperturbed galaxies show a comparable median radial profile to the GAMA control sample, with a small offset ( $\lesssim0.2$ dex). Both the unperturbed and GAMA samples show a slight decline in log sSFR, $\sim$ 0.25 dex, towards their centres. On the other hand, the asymmetric and truncated galaxies show the opposite trend, with a declining sSFR profile as the radial distance increases.

Figure 11. Radial profile of specific star formation rates for each sample. The medians are determined with the step size of 0.2 and shifted by steps of 0.05 in ${r/r_{\text{eff}}}$ . The colours are the same as Figure 10. The error bars and the dashed lines show the standard errors on medians, and the $\mathrm{16{th}} \text{ and } \mathrm{84{th}}$ percentiles of the distribution per bin. The inset zooms in on the central regions (i.e. ${r \lt r_{\text{eff}}}$ ). We overlay the radial sSFR profiles of the asymmetric galaxies derived by a jackknife approach on the main samples. Here, each colour from the nipy_spectral colour map (between black and light grey) represents the galaxy excluded for that iteration (i.e. the profile without that galaxy). While nearly all profiles show similarities with the main asymmetric profile, only the lime profile (without 9011900166) exhibits much lower sSFRs, comparable with the other samples.

Notably, the asymmetric galaxies show somewhat enhanced central star formation activity compared to the other samples. For ${r/r_{\text{eff}}} \lt 0.5$ , the difference in log sSFR between the asymmetric sample and the rest reaches $\sim 0.7$ dex. To assess whether the enhancement seen in the asymmetric galaxies is a population-wide trend or driven by a small number of galaxies, we use jackknife resampling to determine the median sSFR profiles for a subsample of asymmetric galaxies. In each iteration, we exclude a galaxy and then estimate the median. As shown with the lime profile in the inset plot, when 9011900166 is omitted, the level of central sSFRs becomes comparable with the remaining samples. Therefore, this enhancement is primarily due to this galaxy, instead of a representative trend for the asymmetric sample. Still, even after removing 9011900166, asymmetric galaxies maintain higher sSFRs in the outskirts (i.e. ${r/r_{\text{eff}}}\gt 0.5$ ) than their truncated counterparts, and this trend seems to disappear for ${r/r_{\text{eff}}} \gt 1.5$ , where data gets noisier for both samples.

In light of these findings, we can draw a conclusion that star formation in ram-pressure-affected galaxies is suppressed in the outskirts with respect to unperturbed counterparts (as well as the control sample), while central sSFRs remain comparable.