The MEGARA view of outflows in LINERs

Outflows are believed to be ubiquitous in all AGNs, although their presence in low luminosity AGNs, in particular, for LINERs, has only started to be explored. Their properties (geometry, mass and energetics) are still far from being properly characterised. We use integral field spectroscopic data from the MEGARA instrument, at GTC, to analyse a small sample of nine LINERs, candidates of hosting ionised gas outflows. We aim to study the main emission lines in the optical to identify their properties and physical origin. We obtained data cubes at the lowest (R$\sim$6000) and highest (R$\sim$20000) spectral resolution of MEGARA. We modelled and subtracted the stellar continuum to obtain the ionised gas contribution, and then fitted the emission lines to extract their kinematics (velocity and velocity dispersion). We identified outflows as a secondary component in the emission lines. The primary component of the emission lines was typically associated to gas in the galactic disc. For some objects, there is an enhanced-$\sigma$ region co-spatial with the secondary component. We associated it to turbulent gas produced due to the interaction with the outflows. We find signatures of outflows in six LINERs, with mass outflow rates ranging from 0.004 to 0.4 M$_{sun}$yr$^{-1}$ and energy rates from $\sim$10$^{38}$ to $\sim$10$^{40}$ erg s$^{-1}$. Their mean electronic density is 600cm$^{-3}$, extending to distances of $\sim$400 pc at an (absolute) velocity of $\sim$340 km s$^{-1}$ (on average). They tend to be compact and unresolved, although for some sources they are extended with a bubble-like morphology. Our results confirm the existence of outflows in the best LINER candidates identified using previous long-slit spectroscopic and imaging data. These outflows do not follow the scaling relations obtained for more luminous AGNs. For some objects we discuss jets as the main drivers of the outflows


Introduction
Outflows are believed to be ubiquitous in galaxies that host an active galactic nuclei (AGN) (e.g.Veilleux et al. 2005;Morganti 2017;Veilleux et al. 2020).They are thought to be powered either by the activity of the supermassive black hole (SMBH), or triggered by the kinetic energy injected by the jet, although in reality they could be a consequence of both processes acting together (Veilleux et al. 2005;Fabian 2012; Kormendy & Ho 2013;Heckman & Best 2014), or even some contribution from nuclear star formation.Outflows have been extensively studied in luminous AGNs, as they are more powerful and thus easy to identify (e.g.Fluetsch et al. 2019).However, they are also present in lower-luminosity AGNs such as Low-Ionisation Nuclear Emission-line Regions (LINERs) (Masegosa et al. 2011;Dopita et al. 2015;Cazzoli et al. 2018;Hermosa Muñoz et al. 2020;Cazzoli et al. 2022;Hermosa Muñoz et al. 2022).LINERs are the largest population of AGNs in the Local Universe (Heckman 1980;Ho et al. 1997;Márquez et al. 2017), and thus ideal targets for doing spatially resolved studies.
Hermosa Muñoz et al. (2022) (hereafter HM22) made the largest systematic study searching for outflows in LINERs by obtaining narrow band imaging data for 70 nearby (z<0.025)galaxies.The targets were confirmed to be genuine AGNs through the detection of a compact X-ray source in their centres (González-Martín et al. 2009;Márquez et al. 2017;Agostino et al. 2023).This imaging information allows to detect, or at least hint, the presence of outflows as extended ionised gas structures emerging from the nuclear region of the galaxies (see e.g.Pogge et al. 2000;Masegosa et al. 2011).However, the presence of extended, filamentary ionised gas may be related to processes other than outflows, such as past interactions or mergers (e.g.Raimundo 2021).To estimate the reliability of the ionised gas distribution as a tracer of outflows, HM22 complemented the images with all the kinematical information available in the literature for the galaxies.This is considered a more reliable proof for outflows,

Sample and data
We analysed a total of nine LINERs (see Table 1).All the objects were confirmed as true AGNs based on the detection of a hard Xray unresolved central source (González-Martín et al. 2009).The parent sample of our data set was extracted from Cazzoli et al. (2018) and Hermosa Muñoz et al. (2020), where the long-slit spectroscopic properties of 21 and nine type 1.9 and 2 LINERs were characterised, respectively; and from HM22, where narrow band imaging data for 70 LINERs were analysed.We selected targets for which Cazzoli et al. (2018); Hermosa Muñoz et al. (2020) and HM22 reported the detection of non rotational motions with the spectroscopic information and/or morphological signatures in the ionised gas that could be ascribed to outflows, presumably associated to the AGN.We gathered IFS data with the MEGARA instrument for nine targets, one of which was NGC 1052, analysed in detail in C22.The complete information of the galaxies is summarised in Table 1.The data and observations are described in Sect.2.1 and Table 2.We show as a reference the optical images of the host galaxies in Fig. 1.

Data gathering
We observed the galaxies with the MEGARA instrument (Gil de Paz et al. 2016, 2018;Carrasco et al. 2018), attached to the GTC telescope (10.4 m), at Roque de los Muchachos observatory in La Palma, Spain.We used MEGARA in its Large Compact Bundle Integral Field Unit (IFU) mode, that is composed by a total of 623 fibres (567 for the scientific target, and 56 for obtaining sky-background spectra), each fibre covers a spatial scale of 0.62 ′′ , that gives a total projected field-of-view (FoV) on sky of 12.5 ′′ ×11.3 ′′ .MEGARA uses several Volume Phase Holographics (VPHs) available at different wavelengths with spectral resolutions from R∼6000 to R∼20000.In this work we use three different VPHs at the lowest (LR-B, LR-V and LR-R) and one at the highest resolution (HR-R).The VPHs, following the MEGARA cookbook 1 , have the following characteristics: (i) LR-B: wavelength range of 4330-5200 Å with an average resolution R = 6060 (i.e.0.23 Å pix −1 ); (ii) LR-V: wavelength range of 5140-6170 Å with an average resolution R = 6080 (i.e.0.27 Å pix −1 ); (iii) LR-R: wavelength range of 6100-7300 Å with an average resolution R = 6100 (i.e.0.31 Å pix −1 ); and (iv) HR-R: wavelength range of 6440-6840 Å with an average resolution R = 18700 (i.e.0.09 Å pix −1 ) (for a detailed description, see Chamorro-Cazorla et al. 2023).At the redshift of our targets (z<0.02), in the wavelength range covered from LR-B to LR-R, we can detect the emission lines [O III]λλ4959,5007 ÅÅ, Hβλ4861 Å, [S II]λλ6716,6731 ÅÅ, [N II]λλ6548,6584 ÅÅ, Hαλ6563Å and [O I]λλ6300,6363 ÅÅ, and the absorption lines of NaIλλ5890,5896 ÅÅ.The data were gathered in a total of 10 nights (PIs: S. Cazzoli and A. Gil de Paz).We obtained two (three) individual exposures for VPH per galaxy, with typical exposure times of 1200 s (900 s) for the data obtained through the MEGADES (MEGARA Galaxy Disk Evolution Survey; see Chamorro-Cazorla et al. 2023) survey (open time).The average airmasses were 1.22 (LR-B), 1.25 (LR-V) and 1.3 (LR-R) (see Table 2).We note that for the galaxies in the MEGADES sample (see Table 2), they were observed with the LR-B, LR-V, LR-R and/or HR-R VPHs, as the whole wavelength range is needed to fulfil the scientific objectives of the survey (Chamorro-Cazorla et al. 2023).Nevertheless, the data obtained through our open time program of GTC (see Table 2) only uses LR-R and LR-V as these have typically higher sensitivity and signal-to-noise (S/N) per spaxel than that of LR-B (Chamorro-Cazorla et al. 2023).More importantly, the wavelength coverage of LR-V and LR-R is sufficient for obtaining a proper stellar modelling and a robust identification of the kinematical properties of the emission lines associated to the ionised gas, that may be related to outflows (see Sect. 3).

Data reduction and data handling
The data from MEGARA/GTC were reduced following the MEGARA Data Reduction Pipeline (v0.11Pascual et al. 2019Pascual et al. , 2021)).All the software and packages throughout the analysis are used in a python environment.The full process is described in Gil de Paz et al. (2018); Cazzoli et al. (2020) and Chamorro-Cazorla et al. (2023); here we briefly describe the main steps.The pipeline applies the standard data reduction procedures, as bias subtraction, flat-field correction and wavelength and flux calibration (with the spectrophotometric standard stars observed each night).The pipeline considers the information from all the fibres individually, to trace them over the FoV and apply for each one a specific illumination correction.For one object (NGC 4750) the last bundle (covering 20 fibres) of the LR-R cube was not properly traced due to a poor spectrograph focus and thus there are some missing fibres in the north part of the final cube (see Sect. 4.2 and Chamorro-Cazorla et al. 2023 for more details).Finally, we combine the individual exposures using a median to create the final RSS file2 for each VPH and each target.For two objects (namely NGC 0266 and NGC 0315), the different observing blocks (i.e.exposures) for a given VPH were not observed during the same night, so the individual exposures needed to be shifted in spatial position.In these cases, to obtain the final combined cube, we created a mosaic with the task hypercube under the megara-tools package (v.0.1.0;Gil de Paz et al. 2020).After that, for all targets we transformed the final RSS file to a data cube by applying a regularisation grid to obtain 0.4 ′′ square pixels (see Cazzoli et al. 2020).
Once the cubes were fully reduced, we combined the LR-V and LR-R data cubes for all the targets.The procedure was the same we used for NGC 1052 (see C22); the wavelength range corresponding to the LR-R VPH (see Sect. 2.1) lacked from sufficient stellar features to obtain a good modelling of the stellar component of the galaxies.Hence we decided to combine LR-R and LR-V data cubes to maximise the wavelength coverage and increase the quality of the stellar continuum fitting, specially in the LR-R range (see Fig. 2).Given the difference in resolution between both VPHs (0.27Å/pix vs 0.31Å/pix), we degraded the LR-V resolution to that of the LR-R.For two targets (namely NGC 4278 and NGC 4750) we additionally obtained LR-B data cubes (see Table 2).At their redshifts, the emission lines available are located at the reddermost part of the available wavelength range.This implies that these data cubes have sufficient stellar features to obtain a proper stellar modelling without combining with the other VPHs (i.e.LR-R & LR-V).Finally, NGC 4278 was observed with both the lowest (LR-B and LR-V) and highest resolution mode of MEGARA (HR-R instead of LR-R).The spectral coverage of HR-R is small (∆λ ∼400 Å; see Sect.2.1) and the continuum featureless, so the stellar subtraction was not performed, as for NGC 7469 in Cazzoli et al. (2020).
For two objects (namely NGC 0315 and NGC 4750) the LR-R and LR-V cubes were not aligned, as they were observed in different nights.To estimate the photometric centre of every cube, we created a broad band-like image by collapsing along the wavelength dimension and fitted them to ellipses.The central positions of the ellipses were taken as references for the photometric centre.When the positions differed significantly (>2 square-pixels, i.e. 0.8 ′′ ) from the LR-V data cube to the LR-R, we shifted the cubes to set their photometric centres at the same spatial positions.Consequently, their final maps cover a smaller field of view than the rest of the targets (see Sect. 4).The fluxes of both data cubes where scaled in the wavelength range common to both LR-R and LR-V cubes to match their continuum levels, using as reference the LR-R cube.
Finally, we complemented the spectroscopic data with optical images of the galaxies observed with ALFOSC, located at the Nordic Optical Telescope (NOT) or the Hubble Space Telescope (HST).The complete description of these images can be found in HM22.As mentioned in Sect. 1, we obtained Hα images of 70 targets among which we find all the nine galaxies of this paper.These images have a spatial sampling of 0.21 ′′ pix −1 and 0.05 ′′ pix −1 (ALFOSC/NOT and HST, respectively), and are useful to compare with possible substructures unresolved kinematically with the MEGARA data cubes (spatial sampling of 0.62 ′′ ).

Stellar continuum modelling
We used the Penalized PiXel-Fitting (pPXF) software by Cappellari & Copin (2003) (see Cappellari 2017 and references therein) to subtract the stellar contribution from the host galaxy.This is crucial for our LINERs, where the emission from the AGN is usually heavily diluted by the stellar continuum, especially in the Balmer lines (see e.g.Agostino et al. 2023).As mentioned in Sect.3, we combined the LR-V and LR-R cubes (hereafter we refer to the LR-VR cube) to increase the continuum range in the spectra and improve the overall fitting by including more stellar features.For the correct modelling of the stellar population, we masked all the emission, Na I  doublet absorption, telluric and atmospheric lines.We used the additive and multiplicative Legendre polynomials to adjust the continuum shape of the selected templates, with the lowest possible degree to obtain a proper modelling.We set the degree of the additive and multiplicative polynomials to be 8 and 4, respectively, for all the targets (C22).In the same way we fitted the stellar continuum of LR-B data cubes individually (see Sect. 3; additive and multiplicative polynomials 8 and 4, respectively).We used the synthetic stellar library by González Delgado et al. (2005), which provides a total spectral coverage from 4000 to 7000Å, enough to cover the spectral range of our data cubes with the same resolution.The 413 synthetic stellar spectra correspond to a metallicity of 0.02 Z ⊙ , and cover the whole range of other stellar parameters, such as the temperature or the gravity (see González Delgado et al. 2005;Martins et al. 2005, C22 for more details).
From the modelling we obtained the main kinematical properties on a pixel by pixel basis (see Figs. 2 and 3) and generated the final maps for each galaxy.The main results are found on the top panels of Figs. 4 to 11, where we show the stellar properties of velocity, velocity dispersion (σ) and the galactic continuum maps of each LINER.The stellar modelling for both the LR-VR combined cube and the LR-B cube of NGC 4750 is shown in Fig. 2 as an example.

Emission line modelling
Once the stellar component was subtracted, we modelled the emission lines using the Non-Linear Least-Squares Minimization and Curve-Fitting package, lmfit (Newville et al. 2014).We applied a similar technique as in our previous works (see Cazzoli et al. 2018;Hermosa Muñoz et al. 2020;Cazzoli et al. 2020, C22).The process is briefly summarised here.We first modelled the [S II] lines to use them as a reference for the remaining emission lines (i.e. the [N II] doublet, Hα and [O I] doublet when possible), fixing their central wavelength (associated to the gas velocity) and its width (associated to the velocity dispersion).We imposed the intensity ratios between the [O III] (LR-B), [O I], and [N II] doublets to be 2.99, 3.13 and 2.99, respectively (Osterbrock & Ferland 2006).The velocity dispersion was corrected from the instrumental broadening as follows: To ensure a proper modelling of the profiles, we fitted the reference lines (i.e.[S II], [O III] and Hβ, see below) when their S/N was larger than 5. Instead of using the [S II] as reference (the so-called S-method), we could also use the [O I] 6300Å emission line (the O-method; see more details in Cazzoli et al. 2018 andHermosa Muñoz et al. 2020).However the [O I] line is detected with enough S/N (>5) only in three galaxies from the sample (i.e.NGC 3226, NGC 4750 and NGC 5055), and exclusively in the inner regions (<2 ′′ from the centre).Thus, the O-method was discarded given the generally low S/N of the line.Additionally, we found that the kinematics derived from the [S II] lines reproduced better the profiles of the Residuals Fig. 3. Ionised gas modelling of the integrated LR-R spectrum on the PSF region of NGC 4438.In the top panel the green, pink and blue Gaussians represent the primary (P), secondary (S) and tertiary (T) kinematical components, respectively.The global fit is in red and the region where we estimated the ε from the continuum (ε c ) is marked in yellow (see Sect. 3.2).In the bottom panel the residuals of the fit are shown in black, with the horizontal, dashed, yellow lines indicating 3ε c .
only fixed parameter in the fit.
For the line modelling, we allowed up to three Gaussian components per emission line, plus an additional very broad component (full-width at half maximum: FWHM > 1000 km s −1 ) only within the point-spread function (PSF) region to reproduce the broad Balmer lines originated in the broad line region (BLR).
The PSF region is determined individually for each galaxy from the FWHM of brightness distribution of the standard star used for the data flux calibration.We refer to the three Gaussian and the BLR components as the primary (P), secondary (S), tertiary (T) and broad (B), respectively (see Sect. 3.4).They were ordered to ensure a continuity in the velocities and widths at the final maps, which implies that the primary component usually has smaller widths than the secondary.In order to evaluate the number of components preventing overfitting, we used the ε criterion (Cazzoli et al. 2018;Hermosa Muñoz et al. 2020;Cazzoli et al. 2020, C22).This parameter ε is defined as the standard deviation of the residuals under the lines when fitted with a Gaussian component, with respect to its value in the continuum (ε c ) on a 50Åwide region with no emission lines.We considered the inclusion of an additional component whenever ε > 2.5, as this means that the standard deviation of the remaining residuals could not be explained simply as continuum variations.We visually inspected the spectra for ε between 2 and 2.5.To test whether there was a very broad component in the Balmer lines for each galaxy, we produced a high S/N spectrum integrating all the individual spectra in the PSF region, so that the BLR component could be more easily detected.If a BLR component was required, we included it in the individual pixels corresponding to the PSF region.The ionised gas modelling of the integrated spectrum in the PSF region of NGC 4438 is shown in Fig. 3 as an example.
3.3.Estimation of the ionising source through line ratios By using the ratios between the fluxes of the emission lines of the individual components, we can derive their origin by using the BPT diagrams (Baldwin et al. 1981;Veilleux & Osterbrock 1987;Kauffmann et al. 2003;Kewley et al. 2006)

Association of the individual emission line components to a physical interpretation
Based on the emission line fitting, we can associate a physical interpretation to each kinematical component used in the modelling.The primary component of the gas is typically interpreted as the gaseous disc of the galaxy, which is expected to be correlated with the stellar component of regular galaxies3 .The secondary and tertiary components could be associated to rotational or non-rotational motions of the gas in the galaxies, depending on the properties.We identify possible outflows as typically done in the literature (see e.g.Harrison et al. 2014;Fiore et al. 2017;Cazzoli et al. 2018;Hermosa Muñoz et al. 2020;Venturi et al. 2021, and references therein), as a (normally) blueshifted and broad component, with very distinct properties to that of the primary (systemic) component of the gas.These differences do not only concern velocity and σ, but also the morphology of that component that could be filamentary or bubble-like (see e.g.Cazzoli et al. 2018;Fluetsch et al. 2019;Hermosa Muñoz et al. 2020, HM22).On the contrary, the ionised gas disc is typically continuous, with a rotating pattern all over the FoV.
As mentioned in Sect. 1, outflows may suppress the star formation of galaxies if they are able to remove the available gas.For determining the impact that the outflows may have, we need to estimate their main properties.In particular, these are the mass of the outflow (M OF,ion ), the mass outflow rate ( ṀOF,ion ), the kinetic energy (E OF,ion ) and the kinetic power ( ĖOF,ion ) of the ionised gas.The expressions used for estimating these parameters are taken from C22.The total mass of the ionised gas, M OF,ion , us-ing Hα is calculated as in Venturi et al. (2021) (see also Carniani et al. 2015;Fiore et al. 2017): This mass depends on the Hα luminosity and the electron density of the gas (n e − ), that can be estimated from the ratio between the lines of the [S II] doublet (Osterbrock & Ferland 2006).However, for estimating the total mass using Hβ or [O III] luminosities instead, there are several expressions assuming different conditions about the properties (i.e.geometry, density, etc) of the outflow (see e.g.Cano-Díaz et al. 2012;Harrison et al. 2014;Carniani et al. 2015;Husemann et al. 2016;Fiore et al. 2017;Kakkad et al. 2020;Speranza et al. 2021).In Carniani et al. (2015), they assume T e ∼10 4 K and n e − ∼ 500 cm −3 to derive the expression: where the constant C is defined as ⟨n e − ⟩ 2 /⟨n 2 e − ⟩ (see Fiore et al. 2017).Carniani et al. (2015) note that there is a difference in the mass determinations using Hβ or [O III] lines, such that Hβ always traces a higher mass than the [O III], which instead would give a lower limit to the outflow mass.Assuming a biconical geometry for the outflow, then the mass outflow rate, ṀOF,ion , is estimated as in Cresci et al. (2015) (see also Fiore et al. 2017): It depends on the outflow mass, the maximum velocity (V OF,ion ) and its spatial extent (R OF,ion ).However, if the outflow has instead a shell-like structure, the ṀOF,ion is estimated as in In this case, the mass outflow rate is measured at the position of the shell structure, not accounting for the overall structure.The V OF,ion is the maximum velocity within the shell, and ∆R is the shell-thickness.With the mass and the mass outflow rate, we can estimate the kinetic energy of the outflow, E OF,ion , as in Venturi et al. (2021) and the kinetic power of the outflow, ĖOF,ion , as in Rose et al. (2018) (see also Santoro et al. 2020): We note that there are other formulae in the literature that could be used to estimate the same parameters (e.g.Cano-Díaz et al. 2012;Liu et al. 2013;Harrison et al. 2014).However, the expressions mentioned here assume an electron temperature T∼10 4 K and an electronic density n e − ∼500 cm −3 , that provide a better match with what we actually found in our objects (see Sect. 4.3).

Results
In this section we describe the main results from the stellar (see Sect. 4.1) and ionised gas components of the galaxies (see Sect. 4.2).We show the kinematic and flux maps for both stars and ionised gas, accounting for the different emission lines and kinematical components (see Figs. 4 to 11).In all cases, the ionised gas kinematic maps are derived from the [S II] line (all lines share the same kinematics; see Sect.3.2); the Hβ and [O III] 5007Å maps are given when the LR-B cube is available (see Table 2).The derived parameters from the maps are given in Table 3.The line ratio maps are in Appendix B, and the BPT diagrams for NGC 4278 and NGC 4750 are in Fig. 13.
The following subsections refer to eight out of the nine targets (NGC 1052 was analysed in detail in C22; see Table 2).In this work we focus exclusively on the study of the ionised gas phase, thus on the emission lines analysis.
For each galaxy (see Sects.4.1 and 4.2.1 and Table 3), we refer to: the velocity semi amplitude (∆v) defined as half the difference between the maximum and minimum velocities; and the central sigma (σ PSF ⋆ and σ PSF gas ) as the average velocity dispersion measured within the PSF region.The velocities are estimated subtracting the systemic velocity of the galaxy calculated from the redshifts reported in Table 1.The position angles (PA ⋆ and PA gas ) were measured using the maximum and minimum velocities of the maps, using as reference the receding side of the velocity maps from north to east4 .If the components are spatially unresolved, we report exclusively the average kinematic properties.All the values for the ionised gas given per component are referred to as P, S or T (see Sect. 3.2).The reported line ratios are in logarithm scale (see Sect. 3.3).

Stellar component kinematics
The stellar component maps are shown in the upper panels of Figs. 4 to 11.The main properties are listed in Table 3, and the position-velocity (PV) and position-velocity dispersion (PS) diagrams are in Appendix C. Globally, the stars are always rotating with spider-regular rotational patterns, except for NGC 0315, where both velocity and velocity dispersion maps are homogeneous with no particular structures (see Fig. 5 and Table 3).For NGC 0266, the velocity map shows a rotating pattern (V/σ ∼ 1.8; see Fig. 4), with two regions with more extreme velocities than the rest of the map (see position-velocity diagram in Fig. C.1).They are located from the PSF region up to ∼4 ′′ in diameter in the major axis direction (PA∼121 • ).Contrarily, the velocity dispersion is rather homogeneous throughout the FoV, peaking at the photometric centre (σ = 205 km s −1 ), decreasing radially out to a radius of ∼1.2 ′′ down to ∼170 km s −1 , and then it gets uniform at larger radii (> 2.5 ′′ ).For NGC 3245 and NGC 5055, the Hα emission is highly diluted due to an important contribution from the stars in the Balmer spectral lines, specially in the nuclear region (for NGC 5055 see Dullo et al. 2021).Thus to ensure a proper modelling of their stellar component within the PSF region, we used higher additive polynomials (degree 10 vs 8; see Sect.3.1) than for the rest of the objects (see integrated PSF spectrum of NGC5055 in Fig. 12).For both NGC 3245 and NGC 5055, the velocity maps show a clear rotational pattern (V/σ of 1.6 and 1.4, respectively; see Sect.5.1) along a PA of 178 • and 99 • , respectively (see Table 3).The velocity dispersion is homogeneously distributed for NGC 5055 (median σ ⋆ ∼ 121 km s −1 ; see Table 3), whereas for NGC 3245 it is centrally peaked (σ PSF ⋆ = 245 km s −1 ; see Table 3).For NGC 3226 and NGC 4278 the stars rotate with a regular pattern and similar semi-amplitude velocities (106 km s −1 and 104 km s −1 along a PA of 49 • and 20 • , respectively; see Table 3).In both cases, the velocity dispersion is centrally peaked (σ PSF ⋆ ∼ 210 and 336 km s −1 , for NGC 3226 and NGC 4278 respectively; see Table 3) and gradually drops from the centre to the outskirts (see Fig. C.1).For NGC 4278, contrary to the previous galaxies, we obtained the stellar component maps by using the LR-B and LR-V data cubes individually and not the combined LR-VR cube, as we have the HR-R cube instead of the LR-R (see Table 2 and Sect. 2.1).The analysis with both spectral setups (LR-B and LR-V) gave consistent results, so in Fig. 8 we show only that of the LR-V cube.For NGC 4438 and NGC 4750, the velocity maps of the stars also show a rotational pattern with similar properties (∆v ∼ 111 and 97 km s −1 , respectively; see Table 3), but with a rather homogeneous velocity dispersion (σ PSF ⋆ ∼ 148 and 134 km s −1 , respectively; see Figs. 9 and 10).NGC 4438 is known to have an ionised gas bubble, traced in several works through the Hα emission (Kenney & Yale 2002;Masegosa et al. 2011).To map this feature completely with MEGARA, we centred the northwest bubble instead of the nucleus in the FoV of MEGARA (see Sect. 2).As a consequence, we detect better the redder velocities than the bluer velocities in the stellar velocity field, simply because is not centred in the FoV (see Fig. 9).We detected sightly larger values of σ (∼ 180 km s −1 ) on the location of the ionised gas bubble.

Main kinematics
We fitted exclusively one Gaussian component for the emission lines of two galaxies (NGC 0266 & NGC 5055), two components for five galaxies (NGC 0315,NGC 3226,NGC 3245,NGC 4278 and NGC 4750), and three components for NGC 4438 and NGC 1052 (see C22).We added a BLR component to the spectra in the PSF region for five galaxies (namely NGC 0315, NGC 1052, NGC 3226, NGC 4438 and NGC 4750), although for NGC 4438 the detection is not well constrained; these detections are discussed in Sect. 5.
NGC 0266 (Figs. 4 and B.1): We find that ∼40% of the spaxels have S/N>5 in the [S II] lines.The ionised gas is elongated in the direction of the stellar major axis (see Fig. 4).In general, the emission line profiles are narrow (median σ ∼ 116 km s −1 ) and thus were fitted with a single Gaussian component all over the FoV.The structure of the ionised gas velocity map clearly differs from that of the stars, with a large (i.e.∼6.5 ′′ ), non-rotating region in the inner parts (see bottom panel in Fig. 4) twisted towards the north direction.The outer parts of the velocity map (at radius >2.5 ′′ in the east-west direction) show a rotating-like pattern oriented along the stellar major axis, with a semi-amplitude velocity of ∼212 km s −1 .The Hα flux is concentrated on a spiral-like shape, spatially coincident with the twisted region of nearly flat rotation in the gas velocity map.The velocity dispersion map is centrally peaked (σ PSF gas,N = 173±28 km s −1 ; see Table 3) with the largest values corresponding to the spiral-like feature seen in flux and velocity, and then dropping radially outwards from that structure.
NGC 0315 (Figs. 5 and B.2): The emission corresponding to the ionised gas in this galaxy, measured with the S-method, is concentrated around the nucleus, and reaches only distances up to ∼2 ′′ in radius from the photometric centre.The emission lines show complex profiles, with various kinematical components difficult to disentangle.Most of the emission (∼70%) is located inside the PSF region, hence unresolved.We modelled the spectra with two kinematic components per forbidden line plus a very broad component in Hα.The primary component is barely resolved (total extension of ∼3.9 ′′ , i.e. ∼1.4 kpc; see Table 1), with a rotational pattern that differs from the velocity map of the stars (see Fig. 5), and homogeneous σ (see Table 3).The secondary component and the broad component in Hα are unresolved (total extension for the secondary of < 1.5 ′′ , i.e. < 500 pc; the broad component is within the PSF region).The BLR component has an average velocity of -53 km s −1 and σ gas,BLR = 1077 km s −1 .
NGC 3226 (Figs. 6 and B.3): We identified two kinematical components per emission line plus a spatially unresolved broad component in Hα (see Sect. 5).The velocity map and the Hα flux map of the primary component are oriented in the north-east direction, with a position angle differing in 30 • from that of the stars (PA gas = 19 • ; see Table 3)5 .Moreover, this component rotates with a semi-amplitude velocity twice larger than that of the stars (see Table 3).In general, this component is very narrow (σ gas,P ≤ 100 km s −1 ) although a nuclear structure with wider profiles (σ ∼ 200 km s −1 ) is visible oriented almost perpendicular to the major axis of the ionised gas (extended ∼4.7 ′′ , i.e. ∼400 pc at PA ∼107 • ).The secondary component is unresolved and embedded in the PSF region (∼2.2 ′′ , i.e. ∼190 pc).It is clearly blueshifted (average velocity of -91 km s −1 ) with larger velocity dispersion than the primary in the same spatial region (difference of ∼60 km s −1 ; see Table 3).We detect a very broad component (⟨σ gas,BLR ⟩ = 979±65 km s −1 ) with ⟨v⟩ = 269 ± 20 km s −1 by modelling the integrated PSF spectrum.
NGC 3245 (Figs. 7 and B.4): The ionised gas in the inner regions has a complex structure that was modelled using two kinematical components per emission line.The primary component shows a clear rotation pattern, larger than that of the stars (differences of ∼ 100 km s −1 ; see Table 3) despite its relatively low extension (∼10 ′′ in the north-south direction and ∼ 5 ′′ in the east-west direction).The velocity dispersion map has an enhanced region (σ > 140 km s −1 ) in the east-west direction resembling a bicone that does not correspond to any particular feature in the velocity or flux maps.The secondary component is unresolved (∼ 1.7 ′′ × 3.6 ′′ , i.e. ∼150 pc × 320 pc), and blueshifted (⟨v⟩ ∼ -200 km s −1 ).Its average value is consistent to that of the primary component in the same spatial region (see Table 3).As for NGC 3226, we tested the presence of a very broad component in the integrated PSF spectrum.To model the lines, we added a non-rotating (⟨v⟩ ∼ -30 km s −1 ), very broad component (σ > 700 km s −1 ), with a mild contribution to the total flux of the Hα+[N II] blend (<10%).
NGC 4278 (Figs. 8,13 and B.5): The analysis of the emission lines in the LR-B cube reveals the presence of two different kinematical components.The [O III] and Hβ lines show very similar σ for all the components (mean difference < 10 km s −1 ), but they are decoupled in velocity, with a difference of ∼50 km s (larger for Hβ), as it was already found in the case of NGC 1052 (see C22).Given that the S/N of [O III] is larger than that of Hβ (97% vs 78% spaxels with S/N > 5), we focus on the [O III] line for the LR-B analysis.The ISM and the stellar velocity maps have different PA (20 • vs 46 • ; see Table 3).The primary component of the gas is rotating (∆v/2 ∼ 240 km s −1 ), with a receding side that presents several regions with rest-frame velocities located north-east from the nucleus.The velocity dispersion is centrally peaked (σ PSF gas,P ∼ 146 km s −1 ), and has an enhanced region (σ ∼ 120 km s −1 ) west-east from the nucleus (PA ∼ 45 • ), extending along the whole FoV.The [O III] flux is distributed along the major axis of the primary gas component.The secondary component is resolved, expanding up to ∼6 ′′ , mainly in the north-east direction.It is oriented in a similar direction to the primary component, but with a shell-like structure.The velocities are more extreme than for the primary (∆v/2 ∼280 km s −1 ), specially in the receding side, while the velocity dispersion is smaller (⟨σ⟩ ∼ 50 km s −1 ).We have not detected a BLR component in Hβ.
The Hα and [S II] lines have a similar velocity map to that of [O III] (∆v/2 ∼ 222 km s −1 ) rotating along a similar PA (47±7 • ).The velocity dispersion is similar too, as we detect for Hα a region with enhanced values along the east-west direction.However, the different spectral resolutions (50 km s −1 for LR-R vs 15 km s −1 for HR-R) are reflected in the velocity dispersion, much smaller for the HR-R data cube (average σ ∼ 30 km s −1 ) than for the LR-B (average σ ∼ 70 km s −1 ).We detect an unresolved secondary component only in a few spaxels (∼1%), mainly in the PSF region, that is not spatially coincident with the secondary in [O III].This secondary component in Hα has rest-frame velocities and a velocity dispersion of ∼70 km s −1 .As we have not subtracted the stellar continuum for this cube given the small wavelength range (see Sect. 3.1), the detection of the secondary component could be affected.Moreover, this may also affect the detection of a BLR component, which presence we cannot constrain although there are hints of it in some individual spaxels.
NGC 4438 (Fig. 9 and B.6)): We modelled the ionised gas spectra of this galaxy with three different kinematical components per forbidden emission line.This is the only galaxy in this work for which all the kinematical components are resolved out of the PSF region.The primary, narrowest component (σ PSF gas,P = 99 ± 28 km s −1 ) is detected with S/N>5 in almost the whole FoV (∼ 75%).Its velocity field is very similar to that of the stellar component (PA gas = 28 • ), with ∆v/2 ∼ 154 km s −1 (see Table 3).The Hα flux is distributed along the disc, as the stellar continuum, tracing a regular rotating gas disc.We found an enhancement in the emission line fluxes in the same direction as the bubble (north-west).We visually inspected those spaxels finding hints of a secondary component, although not statistically significant to include them (ε < 2 in the [S II] lines).Thus the flux enhancement is probably due to the bubble being slightly more extended, but our data do not allow to further constraint its presence.The σ map of the primary component has an homogeneous structure except in the FoV limits (due to a lower S/N), with typical values σ ≤ 200 km s −1 .The Hα flux map of the secondary component has a bubble-like Fig. 4. Kinematic maps for the stellar (top panel) and ionised gas (bottom panel) components of NGC 0266.The stellar kinematics was calculated using the combined LR-VR data cube, whereas the ionised gas kinematics is extracted from the [S II] lines (all the lines share the same kinematics, see Sect.3.2).In all panels from left to right: velocity in km s −1 (corrected from the systemic velocity of the galaxy; see redshifts in Table 1), velocity dispersion in km s −1 and integrated flux per spaxel (stars) or integrated Hα flux (gas) in erg s −1 cm −2 .The black (grey) dashed line, when present, indicates the position angle of the stellar (primary ionised gas) component.The black cross indicates the photometric centre, and the PSF region is marked with a black circle.For all the maps, north is up and east to the left.structure and spatially matches the extension of the bubble detected in the images in previous works (∼ 10.5 ′′ i.e. ∼ 300 pc).In the velocity map (see Fig. 9) we distinguish a receding (north-west) and an approaching (south-east) side, oriented at the same PA of the bubble (PA gas,S ∼ 123 • ).Overall the σ is higher than for the primary component (σ < 200 km s −1 only in 10% of the spaxels), specially enhanced at the borders of the bubble and less intense in the inner parts (see Sect. 5.4).Finally, the tertiary component (not found in [O I]) is also present in the spaxels along the north-west direction from the nucleus.This component is generally redshifted and narrow (⟨σ⟩ ∼ 135 km s −1 ), being σ comparable to that of the primary component (see Table 3).However, it has a filamentary-like morphology, different to the other components, that reaches as far as the secondary component (∼ 4.8 ′′ , i.e. ∼ 130 pc).We discuss its possible origin in Sect. 5.4. NGC 4750 (Figs. 10,13 and B.7): In contrast to the stellar component, we found peculiar kinematic maps for the ionised gas.We modelled the LR-VR combined cube with two components per emission line plus an additional very broad component associated to the BLR in Hα (see Sect. 5).The primary component shows two blobs in the velocity map near the nucleus, with velocities differing from the rest of the map, barely rotating.These two regions located eastward (approaching side) and westward (receding side) from the nucleus do not correlate with any particular feature in the velocity dispersion map (average σ in the whole map of ∼ 120 km s −1 ), that has a S-like structure of high σ (i.e.σ > 200 km s −1 ).The position angle of the rotating region of the ISM velocity map differs by ∼ 45 • from that of the stellar component (PA ⋆ ∼ 175 • vs PA gas ∼ 210 • ; see Table 3).The flux is enhanced in a region located at ∼ 5 ′′ northeast from the centre, that corresponds to a low velocity dispersion region (σ ∼50 km s −1 ) of approximately 2 ′′ ×2 ′′ in size.The secondary component is barely resolved out of the PSF region, slightly extended to the west from the nucleus, up to 2.5 ′′ .It is mainly blueshifted and broader than the primary component (difference in σ of 150 km s −1 ; see Table 3).We found a very broad, BLR component in Hα with ⟨v⟩ = 156±18 km s −1 and ⟨σ⟩ = 929±127 km s −1 (see Sect. 5).
For the LR-B cube we could model both [O III] and Hβ lines with a single component.The S/N of the emission lines is > 5 only for 31% and 29% of the cube for [O III]λ5008Å and Hβ, respectively.Contrarily to the emission lines in the LR-R cube, we do not detect a secondary component for any of the lines in this cube, probably because of the lower S/N.The primary component has a similar behaviour to the LR-VR cube, with two blobs in the velocity map.There is a shift of ∼ 25 km s −1 between the velocity of the blobs of [O III] and Hα.The velocity dispersion is consistent within the errors to that  3).It is enhanced in the same direction of the Hα S-like structure, but with a different morphology (just elongated SE-NW for [O III]).We do not detect a broad component in Hβ (neither in individual spaxels nor in the integrated spectrum), which matches its classification as a type 1.9 LINER.
NGC 5055 (Figs.11 and B.8): In the ionised gas spectra, the [S II] lines have always S/N > 5, whereas the [O I] is not detected in the FoV.The line profiles are quite narrow (σ < 140 km s −1 ), so we could model them with a single kinematic component, including the PSF region.The median velocity dispersion of the primary component is σ = 40±20 km s −1 .The velocity map follows a rotating pattern with a semi-amplitude velocity of 126 km s −1 , larger than that of the stars along a similar direction (PA gas ∼ 107 • , see Table 3).The velocity dispersion map peaks at the photometric centre, but it becomes asymmetric in the north-east side with respect to the south-west region (see Fig. 11).Finally, the Hα flux map is centrally peaked, with various enhanced regions in the north-east and south-west direction (see Fig. 14).The blobs in the north-east direction are anticorrelated with the regions mentioned before where σ is more prominent.

Line ratios
We derived the log([S II]/Hα) and log([N II]/Hα) ratios for all galaxies and components, log([O I]/Hα) for five objects (namely NGC 0266, NGC 0315, NGC 3226, NGC 4438 and NGC 4750), and log([O III]/Hβ) for two objects (namely, NGC 4278 and NGC 4750), for which we could produce the BPT diagrams of their primary kinematical component (see Fig. 13).In general, all the line ratios for all the different components can be explained by AGN photoionisation (see our conservative limits in  4) at 6 We note that the line ratios in the discs could have a contribution from post-AGB stars, particularly in those objects with no detection of the [OI] line, as is the case for NGC 5055 (Ho et al. 1997;Sarzi et al. 2010;Singh et al. 2013;Ricci et al. 2015) the north-east and south limits of the FoV.These regions are correlated with the various blobs visible in the kinematical maps corresponding to the low-σ regions and the Hα flux map enhancements (see Sect.  4), specially in the direction of the bubble for the latter galaxy.Despite that the average values of their primary component line ratios are also consistent with AGN photoionisation (AGN/composite for NGC 3226), their maps show some substructures.On the one hand, for NGC 3226 The only two objects for which we could produce the BPT diagrams are NGC 4278 and NGC 4750 (see Table 4 and Fig. 13).In both cases the primary component is consistent with AGN/LINER photoionisation in the diagrams.Although in the case of NGC 4278 we detected a secondary component in all the lines, they are in different spatial positions, so we cannot represent them in the BPT diagrams.Nevertheless, the secondary component of both galaxies are also consistent with the AGN region (see Table 4 and Figs 4).We note that the spectra were not corrected from the stellar contribution from the host (see Sect. 3.1), so the ratios with Hα are upper limits.Nevertheless, considering the small contribution of the stars in the Hα line in Cazzoli et al. (2018), we do not expect a drastic difference in the final line ratios.For NGC 4750 the more distinctive kinematic feature in the primary component is a blob ∼ 5 ′′ north-east from the nucleus, that also has different line ratios from the rest of the maps

Estimation of the properties of the detected outflows
We have detected kinematical components consistent with an ionised outflow (i.e.mainly blueshifted secondary component) in five out of the nine galaxies in the sample (namely, NGC 1052, NGC 3226, NGC 3245, NGC 4278, NGC 4438 and NGC 4750; see Table 6).In this section we characterise their main properties using physical relations (see equations in Sect.3.4) that depend on observational parameters, such as the velocity, outflow extension, and fluxes, that we can estimate from the kinematical and flux maps.The results are summarised   ( †) indicates that the values correspond to the [O III] line; (2) major axis position angle of the stellar velocity maps using the receding sides as reference; (3) semi-amplitude velocity of the stellar maps; (4) central velocity dispersion in the PSF region of the stellar maps; (5) position angle for the major axis of the primary component of the ionised gas; ( 6) and ( 7) semi-amplitude velocity and central velocity dispersion (in the PSF region) of the primary component of the ionised gas; (8) and ( 9) average velocity and velocity dispersion integrated in the PSF region for the secondary component of the ionised gas; (10) indicates if there is or not a broad component associated to the BLR detected in the Balmer lines and ( 11) the FWHM of that component.For all columns, the reported errors are 1σ values.[] indicates that the BLR fit is not well constrained.
in Table 5.
To estimate the mass outflow rate, ṀOF,ion according to Venturi et al. (2021), we have to calculate the intrinsic luminosity of Hα, that depends on the ratio of the Hα/Hβ fluxes.We cannot estimate the ratio using MEGARA data, as we only have both Hα and Hβ for NGC 4278 and NGC 4750, but for the first the outflow is only detected in Hβ and for the latter in Hα; for the rest of the galaxies, only Hα is available.Thus we decided to measure the Hα/Hβ ratio uniformly in the ground-based long-slit spectra from Cazzoli et al. (2018).The long-slit nuclear spectrum of NGC 3245 was analysed in Hermosa Muñoz et al. (2020), although only in the Hα line range; thus for this galaxy we used the ratio Hα/Hβ = 4.76 reported by Ho et al. (1997).
As already mentioned, NGC 4278 is the only galaxy for which we detect the possible outflow exclusively in the wavelength range of the LR-B data cube.This implies that the equation 1 mentioned in Sect.3.4 cannot be used to estimate the mass of the  3), ( 4), ( 5) and ( 6) are the average line ratios in logarithm scale, considering the standard deviation as the 1σ error.outflow.We note though that there could be differences on the mass determinations depending on the line and on the assumptions of the gas temperature.For our determination of M OF,ion using Hβ, assuming a T ∼ 10 4 K and n e ∼ 500 cm −3 seem more appropriate (see Sect. 3.4), so in Table 5 we report the estimated parameters following Carniani et al. (2015).The average M OF,ion using Hα is 7.5 ± 1.2 × 10 4 M ⊙ , whereas for Hβ using the expressions of the various works from the literature (see Sect. 3.4), M Hβ OF,ion varies from 1.8 × 10 3 to 6.6 × 10 3 M ⊙ with typical errors from 50 to 160 M ⊙ .The values for [O III] are typically one order of magnitude less than those estimated with Hβ.As seen in Fig. 8 and Sect.4.3, the outflow in NGC 4278 is distributed with a shell-like morphology that has to be taken into account when estimating ṀOF,ion (see Eq. 4 in Sect.3.4).We identified two main shells at R ∼ 3.8 ′′ and R ∼5.4 ′′ in the north-east direction with respect to the photometric centre, with thicknesses (∆R) of 2.9 ′′ ±0.4 ′′ and 1.3 ′′ ±0.4 ′′ , respectively.Thus, we calculated ṀOF,ion , E OF,ion and ĖOF,ion in the two shells, finding similar results for both (we report in Table 5 the values for the closest shell to the nucleus).We note that these values do not represent the overall properties of the outflow given its morphology, but only lower limits.
Globally, we obtain for our ionised gas (Hα) outflows a mean outflow velocity ∼ 340 km s −1 , a mean extension of ∼ 400 pc, and a mean density of ∼ 600 cm −3 .The galaxy with the most powerful outflow is NGC 1052, being the most extended out-flow expanding at the largest velocities from the sample.On the contrary, the least powerful outflow is that of NGC 4438, with properties typically one order of magnitude smaller than for the other objects (see Table 5).The mean ṀOF,ion is 0.17 ± 0.06M ⊙ with a standard deviation of 0.13, which is translated to a mean ĖOF,ion of ∼ 2 × 10 40 erg s −1 , lower values than for other more luminous AGNs from the literature (see e.g.Fiore et al. 2017;Baron & Netzer 2019;Venturi et al. 2021).We discuss the implications of these results in Sect.5.3.

Discussion
The discussion is focused on the interpretation of the kinematical components used to model the emission lines, including NGC 1052 (see C22).All the targets, except for NGC 5055, were analysed in our previous imaging (Masegosa et al. 2011, HM22) and long-slit spectroscopic7 works (Cazzoli et al. 2018;Hermosa Muñoz et al. 2020).The interpretations given to the components in each work are summarised in Table 6.For NGC 0266 and NGC 4750 there were signatures of outflows with both long-slit spectroscopic and imaging data.For NGC 0315 and NGC 3226 there were hints of an outflow with the long-slit spectra but not with the imaging data.For NGC 3245 and NGC 4438, the kinematical components from the long-slit spectroscopic data were explained with rotational movements in Hermosa Muñoz et al. (2020) and Cazzoli et al. (2018), respectively, whereas both were classified as outflow candidates based on the imaging morphology ('Bubble' candidate in HM22).NGC 4278 was classified as candidate for an inflow in Cazzoli et al. (2018), and as 'Core-halo' in Masegosa et al. (2011).For NGC 5055, we have no prior kinematical information in our works, but we have a morphological classification of the Hα gas as 'Core-halo' in Masegosa et al. (2011).In Appendix A we report a more detailed comparison between all these works for each object individually.indicates that the M OF,ion was estimated with the Hβ luminosity instead of Hα.See Sect.4.3 for more details on the calculations.
In general, with the MEGARA data we find extended emission for all galaxies except for NGC 0315, which is mainly unresolved (see Sect. 4).We modelled with at least two different kinematical components the emission lines in seven out of the nine objects.The secondary component traces an unresolved kinematical component (i.e.contained in the PSF region) in three cases (namely NGC 0315, NGC 3226 and NGC 3245), whereas it is extended for four objects (namely NGC 1052, NGC 4278, NGC 4438 and NGC 4750).We needed a third component to model the emission lines only in two objects: unresolved in NGC 1052 (see C22) and extended in NGC 4438 (see Fig. 9 and Sect.5.4).We have interpreted the secondary component as a possible signature of outflows for NGC 1052 (extensively discussed in C22), NGC 3226, NGC 3245, NGC 4278, NGC 4438 and NGC 4750.Additionally, we modelled exclusively the Hα line with a very broad component associated to the BLR for four LINERs, namely NGC 0315, NGC 1052, NGC 3226 and NGC 4750, and we cannot constrain its presence for NGC 4438 (see Table 3).These results are consistent with the previous works by Cazzoli et al. (2018) and Hermosa Muñoz et al. (2020), using long-slit spectroscopic data.The nondetections with the ground-based data of a BLR component in Hα for type-1.9LINERs, as mentioned in Cazzoli et al. (2018), could be a consequence of the low luminosity of these AGNs, whose BLR can be diluted if seen with relatively low spatial resolution (due to the long-slit aperture and the spaxel size).As mentioned in Sect.4.2.1, we did not detect any BLR component for the Hβ line, when available (see Table 3), thus confirming the nature of the galaxies as either type 1.9 or type 2 LINERs.The flux maps of the primary component for all galaxies, except for NGC 0266, NGC 4438 (see Sect. 5.4) and NGC 4750, are centrally peaked and show no particular morphologies.There is, however, a particular feature that is common for five targets (NGC 1052, NGC 3226, NGC 3245, NGC 4278 and NGC 4750) in the ionised gas velocity dispersion maps: an enhanced σ region with a 'butterfly'-like shape.For NGC 1052 and NGC 4750 this feature is found along the stellar major axis; whereas for NGC 3226, NGC 3245 and NGC 4278, it is almost perpendicular to the major axis.Given that these galaxies have a secondary component, as NGC 1052 for which it was interpreted as an outflow, it is likely that we are in a similar case.Thus the high-σ region would be turbulent gas produced by an outflow, probed by the secondary component of the gas.We discuss this feature in more detail in Sect.5.2, and the properties and origin of the outflows in Sect.5.3.
There are five targets with radio emission detected in the sample (namely NGC 0266, NGC 0315, NGC 1052, NGC 4278 and NGC 4438; see Nagar et al. 2005;Giroletti et al. 2005;Hota et al. 2007;Baldi et al. 2018).There are three radio jets: in NGC 1052, already discussed in C22; in NGC 0315, for which the ionised gas emission is barely resolved in our data, so a possible correlation with the jet can not be explored; and in NGC 4278 (see Sect. 5.3); and two compact detections in radio continuum in NGC 0266 and NGC 4438.We discuss a possible connection between the radio and the ionised gas emission in Sects.5.2 and 5.4.

Dynamical support and galaxy morphologies
We can estimate the dynamical support of the galaxies through the V/σ ratio, estimated from the amplitude of the velocity maps and the mean value of σ in the stellar maps over the FoV of MEGARA.We note that none of the position-velocity diagrams (see Figs. C.1 and C.2) reach to the flattening region of the curves, so the values to estimate the velocity amplitude and the mean velocity dispersion correspond only to the inner parts of the stellar discs, except for NGC 1052, which was completely sampled with MUSE data (see C22).NGC 0266, NGC 3245, NGC 4438, NGC 4750 and NGC 5055 are classified as spiral galaxies, and thus we should expect for them a velocity map with rotational patterns and a centrally peaked velocity dispersion.In fact, the V/σ values for these objects are the largest (1.8, 1.6, 1.4, 1.5 and 1.4, respectively), consistent with being dynamically cold discs.For NGC 0266 there is a bar in the inner regions disturbing the velocity map in the FoV of MEGARA, with two kinematically distinct blobs in velocity and a depression in the velocity dispersion (see Figs. 4 and C.1) already reported in the literature in previous works (Cappellari et al. 2011;Foster et al. 2016).NGC 4438, NGC 4750 and NGC 5055 show a strong rotational pattern but σ is not centrally peaked (see Figs. 9,10,11,and C.2). NGC 3245 has one of the strongest rotational patterns and a centrally peaked velocity dispersion, consistent with an spiral object (see Figs. 7  and C.1). NGC 0315 is the only galaxy classified as elliptical that indeed shows a non-rotating map with V/σ of 0.7, thus dominated by random motions.However, as discussed in C22, NGC 1052 was classified as an elliptical galaxy (see Table 1), but the stellar kinematics suggest that is rather a lenticular galaxy.This is also the case for NGC 3226 and NGC 4278, classified as ellipticals but with a strong rotational pattern and centrally peaked σ (see Figs. 6, 8 and C.1).Their V/σ are 1.2 and 1.0, pretty similar to that found for NGC 1052 in the MUSE data (see C22).Moreover, the stellar and ionised gas components of NGC 3226 and NGC 4278 are misaligned by ∼30 • (see Table 3).In the case of NGC 3226, it is probably as a consequence of its interaction with the companion galaxy NGC 3227.Thus NGC 3226 and NGC 4278 have very similar characteristics to NGC 1052, which was interpreted in C22 as probably a lenticular galaxy given its strong rotational patterns.Nevertheless, some elliptical galaxies may have strong rotational patterns that would classify them as fast rotators (Cappellari 2016).We note though that for NGC 3226 and NGC 4278 we are mapping exclusively the inner regions, so other options such as the presence of a small nuclear disc could also be causing the rotational patterns.This is probably not the case for NGC 1052, where C22 mapped almost completely the whole galaxy, detecting rotation out to the FoV limits (∼80 ′′ ; i.e. 8.8 kpc).
The difference in amplitude between the radial velocities measured in the gas and the stars in NGC 3226, NGC 3245, and NGC 5055, which range between v gas − v ⋆ ∼ 60 km s −1 in NGC 5055 and ∼ 100 km s −1 in NGC 3245, is likely due to the stars not being fully supported by rotation in these regions, which leads to an (asymmetric) drift between the two components.This is supported by the relatively large V/σ ratios found, with line-of-sight velocity and velocity dispersion values measured for the stars in these objects of v ⋆ ∼ 70-100 km s −1 and σ ⋆ ∼ 100-200 km s −1 , respectively (see Figs. 6,7,and 11).In fact, adopting a slope for the Stromberg linear equation of 5.4, similar to that of the Solar neighbourhood8 , the expected drift would be a factor ×2-3 larger than the measured differences (see more on the calculations on Mario Chamorro-Cazorla's PhD thesis).

Feedback effects on the gas disc
As mentioned in Sect.4.3, we identified outflows as a secondary, blueshifted, broad component in the emission lines.
We find what may be a correlation between the existence of an outflow probed by the secondary component, and the presence of an enhanced σ region in the primary component.This region is found to be co-spatial with the secondary component, ascribed to an outflow for five objects (namely NGC In NGC 1052 (see C22) there is a radio jet oriented in the same direction of the ionised gas outflow (consistent with being triggered by the jet) and perpendicular to the high-σ region.There is a similar situation for NGC 4278, as there is a double-sided, parsec-scale jet detected with VLBA data (Giroletti et al. 2005;Pellegrini et al. 2012), perpendicular to the enhanced-σ region of the primary component and also to the secondary component in MEGARA.The BPT diagrams for NGC 4750 indicate LINER photoionisation for the secondary component and for the high-σ region, with similar ratios to NGC 1052 (see Sect. 4.2.2 and figures in Appendix B).The two sides of the radio jet have different PAs and the emission was detected in previous works as a milliarcsecond or even compact radio source (see e.g.Schilizzi et al. 1983;Nagar et al. 2005, and references therein).Pellegrini et al. (2012) suggested that the jet may be confined inside the galaxy due to both the low-power of the AGN, that cannot launch the jet to larger distances, and the interaction with the ISM in the inner regions.The enhanced-σ region reaches large distances in the galaxy out of the MEGARA FoV (> 50 ′′ , i.e. > 2 kpc; Sarzi et al. 2006;Pellegrini et al. 2012).Given that this galaxy is almost face-on (inclination of 16 • ; see Table 1), the difference with respect to the outflow direction in NGC 1052 may be due to inclination effects.There are no reported detections of radio jets for NGC 3226, NGC 3245 and NGC 4750.
Enhanced-σ regions in the ionised gas related to a radio jet have already been reported in works as Venturi et al. (2021) with data of the MAGNUM survey, modelling the [O III] line for four Seyfert galaxies.In all the cases, these enhanced-σ regions were perpendicular to low power radio jets (<10 44 erg s −1 ).The high-σ region in NGC 4750 is distributed in a spiral-like shape, bordering southward and northward the regions of lowest and highest velocities, respectively, and located at the base of the secondary component, interpreted as the outflow.This morphology of the primary component may have been produced by strong dust lanes from the spiral arms (see e.g.Carollo et al. 2002), and the enhanced-σ region most likely is a perturbation produced by the outflow, as for NGC 1052.Considering the line ratios, this region is consistent with both LINER photoionisation or shocks, thus a mixture of the AGN ionisation and the collision of the ISM with the outflowing gas may be triggering this enhanced-σ region.

Characterisation of the ionised outflows
We can estimate the main properties of the detected outflows following the expressions in Sect.3.4.We represent in Fig. 15 a comparison between the mass outflow rates of our sources compact to the work by Fiore et al. (2017) and Baron & Netzer (2019).Fiore et al. (2017) 2017), and the grey, dashed line is the linear fit they report in their paper; black triangles are the data from Baron & Netzer (2019); the filled circles with their respective errorbars represent the values obtained for the outflows in this work, except NGC 1052 (red circle), estimated in C22.The empty circles are the same objects, but assuming an electron density of 200 cm −3 , as Fiore et al. (2017) in their analysis.
gas phase, deriving various scaling relations between the outflow and the AGN properties, such as the mass outflow rate with the bolometric luminosity.This work misses the low luminosity end of the AGN family, and thus we have extrapolated the linear relation that they report down to the luminosity of our targets (L bol : 10 41 -10 43.5 erg s −1 , see Table 1).We have also added the targets from Baron & Netzer (2019), which are objects at lower luminosities than Fiore et al. (2017), but higher than our targets (L bol : 10 43.5 -10 45.5 erg s −1 ).Baron & Netzer (2019) measured the properties of ionised gas outflows with SDSS DR7 spectra of 234 type-II AGNs, mainly Seyferts, in a range of z = 0.005-0.15.They used the ionisation parameter to derive the densities, a method independent on the [S II] lines (see e.g.Davies et al. 2020;Revalski et al. 2022).They found electron densities of ∼ 10 4.5 cm −3 , two orders of magnitude larger than those in Fiore et al. (2017), that instead assumes a constant value of 200 cm −3 for all targets.We cannot compare the densities that we obtained with Baron & Netzer (2019), as the ionisation parameter depends on the [O III]/Hβ ratio, that we cannot estimate for the majority of our targets.We plotted our values in Fig 15 , and we can see that they tend to be located above the linear relation reported by Fiore et al. (2017) for ionised gas outflows.We note that if we consider the electron density to be 200 cm −3 as in Fiore et al. (2017), instead of our derived values (see Table 5), the ṀOF,ion increases, hence these points located even further away from the reported linear relation in that work (see Fig 15).Baron & Netzer (2019) also found that their data points do not follow the linear relation by Fiore et al. (2017), even accounting for the uncertainties on both the mass outflow rate (1 dex) and bolometric luminosities (0.3-0.4 dex).They state that in their case, different outflow ge-ometries and projection effects may play a role, given that they analyse type-II AGNs.If we were to assume densities similar to those reported in Baron & Netzer (2019) (i.e.n e − ∼ 10 4.5 cm −3 ), the mass outflow rates (and consequently the energy) would be two orders of magnitude lower than what we obtain, thus closer to the linear relation although still above it (mean ṀOF,ion would go from 0.17 to 0.003).We note that our ṀOF,ion values are not corrected for projection effects.If we consider the outflows to be perpendicular to the galactic plane, then the deprojected mass outflow rates would depend exclusively on the inclination of the host.Given the inclinations (see Table 1), we estimate that the reported values in Table 5 would diminish a 45% up to 75% in the galaxies inclined more than 45 • , whereas it would increase up to 63% (NGC 4750) and 250% (for NGC 4278) in those inclined less than 45 • .Despite this, the order of magnitude of the ṀOF,ion is maintained in all cases except for NGC 3245 (where it diminish from 0.11 ± 0.03 to 0.06 ± 0.02), NGC 4750 (from 0.08±0.02 to 0.13±0.10)and NGC 4278 (from 0.004±0.001to 0.014±0.004),and they would still be out of the linear relation in Fig 15 .Given that we cannot fully account for the projection effects in the outflows due to the uncertainty of their orientations, and that the galaxy inclination does not heavily impact the results, we prefer to use the projected values for the discussion.
Although we only have a few points, our results indicate that it may be a flattening in the curve towards lower luminosities, such that the detected outflows are not as weak as expected.We note that these results have to be considered carefully as every work assumes different conditions, specially with respect to the electron density; thus a unified treatment, also accounting for projection effects, would be needed to extract robust conclusions.
The mass outflow rates and kinetic powers of our sources are equivalent to those found by Kukreti et al. (2023) with optical and radio data for 129 radio-AGNs in a slightly larger redshift range (z<0.25;mean 0.12) and luminosities than our sources (14% detection of outflows; ṀOF,ion = 0.09 -0.41 M ⊙ yr −1 and E OF,ion = 0.1-1.8× 10 41 erg s −1 ).These reported mass outflow rates and those in our work are small to efficiently quench the galaxies.In fact, the kinetic coupling efficiency ( ĖOF,ion /L bol ; see Table 5) for our sources is always < 0.01.This means that there is little energy transferred from the outflow to the medium, and so these outflows alone are not expected to heavily impact and/or quench their host galaxies (Harrison et al. 2018).However, if the turbulence of the medium is high as a consequence of the outflow and/or due to an interaction with the jet (specially if it is young), then the gas could still be unable to form stars (Kukreti et al. 2023).Further exploration of the quenching phenomena is out of the scope of this paper.

The bubble in NGC 4438 vs the prototypical LINER NGC 1052
NGC 4438 together with NGC 1052 (see C22) are the only galaxies for which we detected three different kinematical components in all the analysed emission lines, but spatially resolved only for NGC 4438.Both are lenticular galaxies with different kinematical properties (see Sect. 5.1).The primary component of NGC 4438 is likely tracing the gas disc.The velocity map is rotating regularly along the same position angle and with slightly larger velocities (difference in the semi-amplitude velocity of ∼40 km s −1 ) to the stellar component (see Table 3).This is in contrast to the case of NGC 1052, for which the primary component was rotating along the minor axis of the stars, and it was not consistent with a regular gas disc, showing spiral-like flux maps with a non-regular morphology.∼340 vs ∼220 km s −1 ; see Fig. 9).These kinematical properties and the overall morphology of this component are consistent with an outflow.The previous spectroscopic analysis of this object with long-slit spectroscopy did not find strong evidences of an outflow (Cazzoli et al. 2018).This can be understood since the slit was oriented perpendicular to the outflow, so it could not be traced (PA slit = 39 • vs PA Bubble ∼ 141 • ).The secondary component detected in NGC 1052 had more extreme kinematics than the primary component, being broader and aligned with the radio jet.It was also interpreted as an outflow, in this case consistent with being triggered by the radio jet (see C22).Given that the tertiary component in the MEGARA data is oriented in the direction of the outflow (PA∼127 • ), its origin may be connected to it.In fact, it is coincident with the north-western side of the bubble, traced with high-σ by the secondary component.The tertiary component is mainly redshifted (v MEGARA gas,T >50 km s −1 ), broader than the primary but narrower than the secondary (average σ MEGARA gas,T of 135 km s −1 ; see Table 3).This galaxy has a double-lobed radio continuum source oriented in a similar direction to that of the bubble (PA ∼ 125 • ; Hota et al. 2007).The extension of this doublelobed, asymmetric radio emission was estimated to be ∼250 pc west and ∼730 pc east from the nucleus (Hota et al. 2007).This is consistent with the estimated size of the outflow in the west direction (∼300 pc), as extended as the tertiary component in MEGARA data.In fact Kenney & Yale (2002) confirmed the co-spatiality between this radio emission and the outer ends of the Hα bubble, initially associating the borders of the ionised gas bubble to shocks, which Hota et al. (2007) associated to the interaction of the radio emission with the ISM.For NGC 1052 we also reported the existence of perturbed emission in the ionised gas due to the ISM-jet interaction, through the turbulence in the enhanced-σ region, as discussed in Sect.5.2.However, for NGC 1052 the tertiary component was unresolved, and thus no further interpretations were given.For NGC 4438, considering the kinematics, line ratios and spatial extension of the tertiary component derived with MEGARA, it may be tracing the optical counter-part of the radio emission.Summarising, the two galaxies are prototypical LINERs inside lenticular galaxies, with a detected outflow in ionised gas consistent with being powered by a radio jet.Both show a complex kinematical structure that needed to be modelled with multiple components, highlighting the importance of spatially resolved studies for obtaining a complete characterisation of the physics modelling the ISM in these objects.Although our sample is small, the results support the idea that the jet-mode might be play a very important role in driving outflows for low luminosity AGNs.

Summary and conclusions
In this work, we analyse the optical IFS data of nine LINERs observed with MEGARA/GTC with its low and high resolution modes (R∼6000 and R∼20000) in different wavelength bands (LR-R, LR-V, LR-B and HR-R).Prior to the study of the ionised gas properties, a subtraction of the stellar component of the galaxies is required.For that we applied a pPXF modelling (Cappellari & Copin 2003;Cappellari 2017) to the combined data cubes of LR-R and LR-V when both were available (seven LINERs, see Sect.3).After the stellar subtraction we obtained the spectra of the ISM and applied a multicomponent fitting of the optical ionised emission lines available in the wavelength range of each VPH.We used the [S II] lines as reference for modelling the rest of the emission lines as in previous works (Cazzoli et al. 2018;Hermosa Muñoz et al. 2020), and modelled [O III] and Hβ independently given that they have different kinematics, as was found for NGC 1052 (C22).These are the main conclusions of this work: -Stellar kinematics: The stars are generally rotating following a regular pattern, with a centrally peaked or homogeneous σ map, except for NGC 0315, that shows no rotation.Morphologically, four galaxies were ellipticals, two spirals and two lenticulars (see Table 1), but our results indicate that all the ellipticals except for NGC 0315 show a rotational pattern.-Ionised gas kinematics: In contrast to the stars, the ionised gas has a complex behaviour, with at least two kinematical components needed to model the emission lines in the majority of the LINERs (7), three components for two cases, and an additional very broad component in Hα associated to the BLR in four of them.The two objects modelled with a single component in the emission lines are NGC 0266 and NGC 5055.For NGC 0266, the observed disturbed kinematics can be explained by the large-scale bar, whereas for NGC 5055 the regular velocity and σ maps are tracing the ionised gas disc, with distinguishable star forming regions.-Enhanced-σ regions: For five objects, namely NGC 1052 (see C22), NGC 3226, NGC 3245, NGC 4278 and NGC 4750, the primary component has a rotational pattern with a peculiar σ.They have an enhanced-σ region perpendicular (parallel) to the stellar major axis in NGC 3226, NGC 3245 and NGC 4278 (NGC 1052 and NGC 4750) with a 'Butterfly' -like shape.This region is probably triggered by a mixture of turbulence caused by the interaction of the secondary component and/or the radio emission.The secondary component is co-spatial with the enhanced-σ region but with differing kinematics, typically blueshifted and broad (⟨σ S ⟩ ∼ 200 km s −1 ), explained here as an outflow, compact and unresolved for NGC 3226 and NGC 3245.-Outflow detections: All the galaxies in our sample were candidates of hosting an outflow from our previous analyses using both imaging and long-slit spectroscopic data (see Cazzoli et al. 2018;Hermosa Muñoz et al. 2020, HM22, C22).The MEGARA data has higher spectral resolution than our previous analysis of the optical spectra, and traces spatially the central regions of our targets.This new analysis allowed us to confirm the presence of outflows for the majority of the sample (six objects, namely NGC 1052, NGC 3226, NGC 3245, NGC 4278, NGC 4438 and NGC 4750).For NGC 4438 and NGC 4750, the outflow is detected with a bubble-like shape emerging north-west and west from the centre, respectively, and for NGC 4278 it has a shell-like structure, for all cases traced by the secondary component.For the remaining objects, the outflow is resolved only for NGC 1052 (see C22). -Energetic of the outflows: The detected outflows in Hα show mass outflow rates ranging from 0.004 to 0.4 M ⊙ yr −1 (average 0.17±0.06M ⊙ yr −1 ) and energy rates from ∼ 10 38 to ∼ 10 40 erg s −1 (average 2 × 10 40 erg s −1 ).On average, the outflows are extended to a distance of ∼ 400 pc at an (absolute) velocity of ∼ 340 km s −1 , with an electronic density of 600 cm −3 .-Tertiary kinematical component: For NGC 1052 and NGC 4438 a tertiary component is needed to model the emission lines.For NGC 1052 this component is unresolved, whereas for NGC 4438 it is extended, and co-spatial with both the outflow and a double-sided radio continuum emission.Thus it is most likely that these emissions are related, and we suggest that it could be the optical counterpart of the radio emission.-Line ratios and BPT diagrams: The average line ratios for the primary components of the galaxies indicate star-forming or composite photoionisation for NGC 3245, NGC 4438 and NGC 5055, and AGN photoionisation for NGC 0266, NGC 0315, NGC 3226, NGC 4278 and NGC 4750.The nuclear regions of all the galaxies are consistent with AGN photionisation.We were able to distinguish several star forming blobs in galaxies such as NGC 4750 and NGC 5055.The secondary and tertiary components of the emission lines are always consistent with AGN-photoionisation.We cannot exclude a possible contribution from shocks, specially regarding the [O I] ratios, that tend to be more affected by these processes.-Broad Line Region component: We detected a BLR component exclusively in Hα for five galaxies (namely NGC 0315, NGC 1052, NGC 3226, NGC 4438 and NGC 4750), although for NGC 4438 it is not well constrained.We did not detect it in Hβ, which confirms the nature of the galaxies as either type 1.9 or type 2 LINERs.
Our results have allowed us to characterise and unify all the results coming from various techniques, finding the existence of outflows for the majority of the LINERs.This reinforces the results from HM22 about outflows being common also in LINERs, and that although they can be detected with less time consuming techniques than IFS, only spatially resolved spectroscopy allows us to fully understand the complex kinematics of the ionised gas in these galaxies.
in the velocity dispersion (consistent only within 4σ).These differences could be produced due to the use of the [O I] line to model the remaining emission lines for the long-slit spectrum instead of the [S II] lines as we did with MEGARA data (as for NGC 0266), given that they fell in the limits of the wavelength range (see Fig. B7 in Cazzoli et al. 2018).Within the MEGARA data, the outflow may be probed by the unresolved, blueshifted secondary component.
NGC 3245: The Hα emission in Masegosa et al. (2011) is extended in the north-south direction, aligned with the slit used with the Palomar spectroscopic data in Hermosa Muñoz et al. (2020).The orientation and extension of the ionised gas in the imaging data is consistent with the primary component derived in the MEGARA data.Hermosa Muñoz et al. (2020) detected a secondary component in the long-slit spectroscopic data, rather narrow (σ HM20 gas,S of 136 ± 21 km s −1 ), and redshifted (94 ± 17 km s −1 ).However, in the MEGARA data this secondary component is barely resolved out of the PSF region, being generally blueshifted (−214 ± 142 km s −1 ) and with a width consistent with that derived for the long-slit spectrum (198 ± 172 km s −1 ).In general, the kinematics of the primary and secondary component are very similar to that of NGC 3226 (see Sect. 5).
NGC 4278: The ionised gas in the imaging data is compact with no particular signatures.This is in contrast to the Hα or [O III] flux distributions in MEGARA (see Fig. 8), that are slightly tilted towards the gas major axis (see Table 3).The spectrum studied in Cazzoli et al. (2018) was modelled with two components using simultaneously [S II] and [O I] lines, although the latter were not well constrained.The primary component with [S II] was almost at restframe (6±7 km s −1 ), with σ C18 gas,N = 177 ± 35 km s −1 ([O I] modelling showed similar results) consistent with the MEGARA data for the same component (∼ 150 km s −1 ; see Table 3).The secondary component however was redshifted (77 ± 34 km s −1 ), and wider σ C18 gas,S = 240 ± 4 km s −1 ([O I] not well constrained with values of 669 ± 67 km s −1 ).
NGC 4750: The Hα imaging data from HM22 has a similar morphology than the primary component in MEGARA data (see Fig. 14).Both works detect an enhanced region ∼4 ′′ north-east from the centre, that given the morphologies and kinematics, most likely corresponds to a star forming region that is spatially resolved in the HST image (see Fig. 14).The spectroscopic detection in Cazzoli et al. (2018) is based on the kinematics derived from the [O I] line.They found two kinematical components in the emission lines, with the primary being explained with rotational movements, and the secondary, blueshifted (v C18 gas,S = −296± 21 km s −1 ), broad (σ C18 gas,S = 380 ± 42 km s −1 ) component as a possible outflow.For our MEGARA data, the secondary component is barely resolved out of the PSF region in a bubble-like morphology emerging westwards from the centre up to ∼ 2.5 ′′ .It is blueshifted (⟨v MEGARA gas,S ⟩ = −92± 50 km s −1 ) and ⟩σ⟨ 297 ± 41 km s −1 (see Table 3), consistent with the derived value in Cazzoli et al. (2018).Considering the morphological and the kinematical signatures, this secondary component is likely tracing the emission from a compact, ionised outflow.The primary component show a low-σ region north-east from the nucleus, consistent with star-forming regions considering the line ratios of the emission lines (see Figs. 14 and B.7).
NGC 5055: The Hα ionised gas in the imaging data (see Fig. 2 in Masegosa et al. 2011) is concentrated around the nucleus, where we also see a centrally peaked flux with MEGARA.This galaxy is a spiral system with both spiral arms and star forming regions falling within the MEGARA and HST FoV.The apparent asymmetry in the flux and velocity dispersion maps may be caused by the strong dust features produced by the spiral arms, that are more prominent to the south as already seen in Masegosa et al. (2011), partially obscuring the nuclear region.The star forming regions are proved by the enhanced Hα-flux regions at ∼2.5 ′′ north-east and ∼ 3 ′′ south-west from the centre in the MEGARA flux map, corresponding to low σ in the corresponding map (σ < 50 km s −1 ; see Fig. 11).These were reported as an extended emission at a PA∼30 • in Masegosa et al. (2011).In general, considering the velocity map of the primary component, the detected ionised gas emission is more likely associated to a regular gas disc.

Fig. 1 .
Fig. 1.Optical images of the nine LINERs observed with MEGARA.From left to right, top to bottom: NGC 0266, NGC 0315, NGC 1052, NGC 3226, NGC 3245, NGC 4278, NGC 4438, NGC 4750 and NGC 5055.All images have been obtained from the Sloan Digital Sky Survey DR9, except for NGC 4750, that is from the Digitalized Sky Survey.The physical scale is indicated for all images; north is up and east to the left.

Fig. 2 .
Fig. 2. Stellar modelling of the LR-B (top) and LR-VR (bottom) integrated spectra in the PSF region of NGC 4750.The red line shows the best fitting using pPXF, the green line shows the residuals of the fit, and the blue line is the final subtracted spectrum.The main emission and absorption lines are indicated with dashed vertical lines.Grey regions were masked during the fitting.
[N II]-Hα blend (i.e. with lower residuals) than the [O I].Only the S-method can be applied to model the HR-R cubes, given that this cube lacks the [O I] lines due to the wavelength coverage.For the LR-B cube, we fitted the [O III] and Hβ lines independently of the emission lines in the LR-R/HR-R cubes.Similarly to the kinematics of [O I] and Hα lines, we found that [O III] and Hβ have different kinematics (see Sect. 4.2.1) as already reported for NGC 1052 using MUSE data in C22.Thus the emission lines of these data cubes were modelled separately, with the intensity ratio between the two [O III] lines being the

Fig. 5 .
Fig. 5. Kinematic maps for the stellar (top panel) and ionised gas components of NGC 0315.The spectra of this galaxy was modelled with a primary (second panel) and an unresolved, secondary (third panel) kinematical components per emission line.A full description is provided in Fig. 4.

Fig. 6 .
Fig.6.Kinematic maps for the stellar (top panel) and ionised gas components of NGC 3226.The spectra of this galaxy was modelled with a primary (second panel) and an unresolved secondary (third panel) kinematical components per emission line.A full description is provided in Fig.4.
4.2.1).All the components used to model the spectra of NGC 0266 and NGC 0315 (see Figs. B.1 and B.2) have homogeneous line ratios, with mean values consistent with AGN ionisation (see Table 4 and Sect.3.3).The line ratios of the secondary kinematical component of NGC 3226 (see Fig. B.3) and the secondary and tertiary (not detected in [O I]) components of NGC 4438 (see Fig.B.6) are consistent with AGN photoionisation (see Sect. 3.3 and Table

Fig. 7 .
Fig. 7. Kinematic maps for the stellar (top panel) and ionised gas components of NGC 3245.The spectra of this galaxy was modelled with a primary (second panel) and a secondary (third panel) kinematical components per emission line.A full description is provided in Fig. 4.
. B.5 and B.7).For NGC 4278 in general the line ratios of the primary component are quite homogeneous (standard deviation ∼ 0.1; see Fig. B.5), except slightly higher values near the limits of the FoV probably due to S/N effects, and in the region of the secondary component for the log([O III]/Hβ) ratio (0.15±0.10; see Table

Fig. 8 .
Fig. 8. Kinematic maps for the stellar component of NGC 4278.For this galaxy we modelled the stellar component using only the LR-V cube (see Sect. 3).The second and third (fourth and fifth) rows of panels correspond to the primary and secondary component of the [O III] (Hα) maps, respectively.A full description is provided in Fig. 4.

Fig. 9 .
Fig. 9. Kinematic maps for the stellar (top panel) and ionised gas components of NGC 4438.The emission lines were modelled with a primary (second panel), secondary (third panel) and tertiary (bottom panel) kinematical components.A full description in Fig. 4.

Fig. 10 .
Fig. 10.Kinematic maps for the stellar (top panel) and ionised gas components of NGC 4750.The spectra of this galaxy was modelled with a primary (second panel) and a secondary (third panel) kinematical components per emission line.The kinematic maps for the primary component of the [O III] line are in the fourth panel.A full description is provided in Fig. 4.

Fig. 11 .
Fig. 11.Kinematic maps for the stellar (top panel) and ionised gas (bottom panel) components of NGC 5055.The spectra of this galaxy was modelled with a single narrow component on the emission lines.A full description is provided in Fig. 4.

Fig. 12 .
Fig. 12. Stellar component modelling of the integrated PSF region for NGC 5055.Same description as in Fig. 2.

LFig. 13 .
Fig. 13.BPT diagrams of the primary kinematical component of the ionised gas for NGC 4750 (upper panels) and NGC 4278 (bottom panels), with the dividing lines fromKauffmann et al. (2003) andKewley et al. (2006).For NGC 4278 the [O I] lines are not available in the wavelength range covered by the HR-R data cube (see Sect. 2).

Fig. 14 .
Fig. 14.Hα flux derived from the MEGARA data superimposed with the contours of the narrow band HST images of the Hα emission from Masegosa et al. (2011) and HM22 for NGC 4438 (total flux accounting for all components), NGC 4750 and NGC 5055 (only primary component of the emission lines).The inner black circle and the black crosses indicate the PSF region and the photometric centre.The contours for NGC 4750 and NGC 4750 represent the 3σ, 5σ and 10σ (black, grey and white) flux levels of the narrow band image, with an extra contour at 100σ for NGC 4438 (black, black, grey, white).

Fig
Fig. C.1.Position-velocity and position-velocity dispersion of the stellar component of NGC 0266, NGC 3226, NGC 3245, NGC 4278 and NGC 4438.The x-axis represents the distance with respect to the kinematic centre.The velocities are corrected from the systemic velocity of the galaxy.

Table 1 .
General properties for all the 9 LINERs analysed in this work.

Table 3 .
Main kinematical parameters derived from the stellar modelling and for the individual components detected in the gas modelling of the Hα line for all the cubes.

Table 4 .
Average line ratios of the ionised gas for all the derived components.

Table 6 .
Interpretation of the secondary component fitted to the ionised gas emission lines for each galaxy in this and previous works.Candidate' indicates those candidates for non-rotational motions.'Bubble' refers to bubble-like, filamentary emission; 'Core-halo' to compact emission in the circumnuclear region; and 'Dusty' when dust lanes covered the central parts of the galaxy.A complete definition of the latter three categories can be found in HM22.
The enhanced-σ region of NGC 3226 and NGC 3245 are very similar (see Figs.6 and 7), both in size (∼ 500 pc 9 ), in width (σ > 150 km s −1 ; see Sect.4.2.1) and in velocity (almost flatrotation).As already mentioned, they are both coincident with the secondary component, which is unresolved, blueshifted (average values for NGC 3226 of −91 km s −1 and NGC 3245 of −214 km s −1 ; see Table3) and broader than the primary component of the emission lines in the same spatial region.
put together data from many different works to estimate the properties of outflows depending on the Mass outflow rate versus bolometric luminosity for ionised gas outflows from different objects.Black crosses represent data obtained fromFiore et al. ( (Kenney & Yale 2002)1)n of NGC 4438 is relatively homogeneous σ, enhanced in the borders of the bubble (see Sect. 4.2.1, whereas for NGC 1052 we detected an enhanced-σ region within the MEGARA FoV as already mentioned (see C22 for more details).The secondary and tertiary components of NGC 4438 are most likely related to the Hα bubble detected in this galaxy, as hinted by its morphology and kinematics.Kenney & Yale (2002)studied HST Hα narrow band images of this LINER, as so did afterwardsMasegosa et al. (2011), characterising a prominent ionised gas bubble emerging north-west from the nucleus.Its projected length was estimated to be 1 kpc and, despite it was classified as a bipolar outflow, one bubble is more visible north-west of the nucleus, probably due to intrinsic differences in the ISM in the two sides of the galaxy(Kenney & Yale 2002).The Hα emission derived from the MEGARA data clearly maps the bubble emerging from the nucleus (see Sect. 4.2.1 and Fig.14).The velocity map of the secondary component indicates that we are looking at the receding (i.e.redshifted) part of the bubble, whereas the approaching (i.e.blueshifted) part is on the south-east direction.The velocity dispersion in the borders of the bubble reaches larger values than for the inner parts (av-