The NuSTAR, XMM-Newton, and Suzaku view of A3395 at the intercluster filament interface

NuSTAR observed the northwestern region of A3395 at the junction of cluster and intercluster filament (Fig. 1) for $\sim$130 ks during 2020 with the focal plane modules (FPMA and FPMB) with an offset pointing from the central cluster emission (Table 1).

In order to filter the data, standard standard pipeline processing is carried out using HEASoft (v. 6.28) and NuSTARDAS (v. 2.0.0.) tools. Since the NuSTAR observation of A3395 was performed after 2020 March 16, the use of versions of HEASoft earlier than v. 6.27.1 and of NuSTARDAS version earlier than v. 1.9.2 results in errors during pipeline processing⁴⁴4http://nustarsoc.caltech.edu/NuSTAR_Public/NuSTAROperationSite/software_calibration.php/. To clean the event files, the stages 1 and 2 of the NuSTARDAS pipeline processing script nupipeline are used. Regarding the cleaning of the event files for the passages through the South Atlantic Anomaly (SAA) and a tentacle-like region of higher activity near part of the SAA, instead of using SAAMODE=STRICT and TENTACLE=yes calls, we have created light curves and applied different filters to create good time intervals (GTIs) manually without fully discarding the passage intervals.

The new set of GTIs are reprocessed with nupipeline stages 1 and 2, and images are generated at different energy bands with XSELECT. nuexpomap is used to create exposure maps. To produce the corresponding spectra for the regions of interest as well as the corresponding response matrix files (RMFs) and ancillary response files (ARFs), stage 3 of nuproducts pipeline are used.

The background assessment of NuSTAR is particularly challenging due the open mast between the the focal plane modules and the optics assembly. The main components of the background are instrument Compton-scattered continuum emission, instrument activation and emission lines, cosmic X-ray background from the sky leaking past the aperture stops, reflected solar X-rays, and focused and ghost-ray cosmic X-ray background. Modeling the background where there is a lack of cluster emission regions is tricky since the ICM emission becomes an additional component in the background fitting procedure.

To apply these models for the cluster background assessment, a set of IDL routines called ${\tt nuskybgd}$, which defines the background spatially and spectrally, is utilized (Wik et al., 2014). The procedure starts by selecting regions in the FOV where the cluster emission is inherently present yet not the most dominant. To account for the ICM emission, an apec model is included in the full set of models, and jointly fitted with the background (Fig. 12 in Section A). The global background model is used to create background images, which are then subtracted from the cleaned images and are corrected by the corresponding exposure maps. Background-subtracted, exposure-corrected images at different energy bands are presented in Fig. 2.

Once the background is defined for any region in the FOV both spatially and spectrally, the next steps are to select regions of interest, to extract spectra and the corresponding files, and to generate the specific background model, followed by spectral fitting to evaluate the physical properties of the ICM.

XMM-Newton observations lasted for $\sim$30 ks during 2007 (Table 1). The FOV of this observation is focused on the central region of A3395, which extends to $\sim$r₅₀₀ (Fig. 1). The three EPIC cameras MOS1, MOS2 (Turner et al., 2001), and PN (Strüder et al., 2001) were operated in full frame mode with the Thin1 filter.

We used standard procedures from the Science Analysis System (SAS) software version 16.1.0, along with the Extended Source Analysis Software (ESAS⁵⁵5https://heasarc.gsfc.nasa.gov/docs/xmm/esas/cookbook/xmm-esas.html) package. Calibrated photon event files were produced using the epchain and emchain tasks. For obtaining the good time intervals, we used mos-filter and pn-filter tasks. Point sources were detected by the cheese routine.

ESAS routines mos_back and pn_back embedded in SAS create model quiescent particle background (QPB) spectra and images for selected regions from the intermediate files produced from the mos-spectra and pn-spectra. To create background-subtracted, exposure-corrected images, first a spectrum for the source and the QPB was extracted from the full FOV. This spectrum was fitted with the absorbed apec model for the cluster emission plus the models for the background. The background model components are the cosmic diffuse X-ray background (CXB), solar wind charge exchange (SWCX), particle and residual soft proton contamination (K$\alpha$ SP).

CXB is modeled with an unabsorbed thermal component of about 0.1 keV, an absorbed thermal component of about 0.25 keV, and the extragalactic power law with a spectral index of 1.46. The particle background is Gaussian lines at 1.496 keV and 1.75 keV representing the Al K$\alpha$  and Si K$\alpha$  lines in the MOS, lines at 1.496 and near 8 keV representing the Al K$\alpha$  and Cu fluorescent lines in the PN. Possible SWCX produces two more lines at 0.56 and 0.65 keV, and the residual SP is represented by a broken power law and fitted with a diagonal matrix supplied in the XMM-ESAS CalDB release, mos1-diag.rsp.gz, mos2-diag.rsp.gz, and pn-diag.rsp.gz. To find the solid angle on each detector for the fit proton_ scale was used, which were used as a constant model between the cameras in the fit model. The routine proton takes the fitted parameters from the fit of the SP model, then creates SP background images later to be subtracted as a background from the photon image.

This process is repeated for an annulus selected in an outer region of the FOV to exclude the bright cluster emission at the center of the FOV in order to produce a better fit for the SP contamination. The resulting SP normalization from the annulus fit is then renormalized to the full FOV using proton_scale. The resulting adaptively smoothed, background-subtracted, exposure-corrected images are obtained with tasks comb and adapt respectively, as shown in Fig. 3.

In addition to the ESAS background, the total background model used to fit the spectra contains the following components: gaussian + gaussian + gaussian + gaussian + gaussian + gaussian + gaussian + constant$\times$constant$\times$(gaussian + gaussian + apec + (apec + apec + powerlaw$\times$wabs), where the second constant corresponds to the cross calibration of instruments MOS1, MOS2 and PN. This background modeling is used for the rest of the spectral analysis in this paper.

Suzaku (Mitsuda et al., 2007) also covered a region enclosing the NuSTAR FOV (Fig. 1) for $\sim$47 ks in 2013. The X-ray Imaging Spectrometer (XIS: Koyama et al., 2007)) of Suzaku is one of the best instruments to investigate shallow cluster emission in the outskirts owing to their low and stable background. The basic data analysis and results are presented in Sugawara et al. (2017). Here we briefly explain the data reduction and spectral analysis approach.
We followed the approach presented in Sugawara et al. (2017). We used the latest calibration file (20160607) and performed event screening with cutoff rigidity greater than 8 GV. The area damaged by micrometeoroids in the XIS0 instrument was excluded in the following analysis (http://www.astro.isas.jaxa.jp/suzaku/doc/suzakumemo/suzakumemo-2010-01.pdf). An additional screening is applied for XIS1 to mitigate the increase in non-X-ray background (NXB) due to an increase in the amount of charge injection (http://www.astro.isas.jaxa.jp/suzaku/analysis/xis/xis1_ci_6_nxb). The cleaned exposure time is about 33 ks for each instrument. In order to characterize the ICM emission, all background components need to be constrained well. For the estimation of the NXB, we used a database constructed from observations of Earth at night using the ftool ${\tt xisnxbgen}$ (Tawa et al., 2008). For the sky background consisting of cosmic-ray background, local hot bubble, and Milky Way halo, we used same model presented in Table 2 of Sugawara et al. (2017).
RMFs and ARFs are generated by the ftools ${\tt xisrmfgen}$ and ${\tt xissimarfgen}$ (Ishisaki et al., 2007). For the ARFs, we assumed uniform emission from a circular region with 20$\arcmin$ radius as an input image. Throughout the fitting procedure, the normalization between BI and FI instruments is kept free.

We start the analysis with the characterization of the global emission at the NuSTAR FOV that is pointed at the northwestern region of A3395. We selected a square 12$\arcmin$ $\times$ 12$\arcmin$ region from which we extracted a spectrum and applied a single-temperature ${\tt apec}$ model (Smith et al., 2001) using ${\tt XSPEC}$ (v. 12.11.1; Arnaud, 1996). Since NuSTAR is not sensitive to emission below 3 keV, it is also not affected by foreground absorption by the Galactic hydrogen column density, and any changes in the $N_{H}$ value do not affect the thermodynamical values obtained from the fit. We visually inspected the images for the point sources, and excluded a 1$\arcmin$ (comparable to the half-power diameter of NuSTAR’s point spread function) circular region from the location of the point source that is visible in Fig. 2.

The spectral fit and the corresponding values are presented in Fig. 4 and Table 2, respectively. It is evident from the spectra and the C-stat values that a single-temperature plasma does not fully describe the cluster. This is expected since the NuSTAR observation is pointed at a region where there are multiple substructures, which may involve emission features from different sources. To assess the possibility of multiple temperatures, we have added another ${\tt apec}$ model for a secondary temperature structure, yet the higher-temperature component was not constrained.

One of these emission features can also be due to a nonthermal component, namely inverse Compton (IC) scattering, which is well represented by a power-law distribution and corresponds to the model ${\tt powerlaw}$ in ${\tt XSPEC}$. The addition of a ${\tt powerlaw}$ model to the original ${\tt apec}$ seems to describe the overall cluster emission well, but the reason for this extra component may be completely unrelated to cluster physics, namely scattered light caused by the two main subclusters lying just outside the FOV. This additional component also seems to suppress the ICM to an unrealistic low temperature with respect to what is found in the literature (Markevitch et al., 1998; Lakhchaura et al., 2011; Alvarez et al., 2018), further suggesting a scattered light origin.

Since A3395 has many substructures within the NuSTAR FOV, a typical assumption of spherical symmetry cannot be used for region selection. For the characterization of the regions, we have created thermodynamical maps of a 12$\arcmin$ $\times$ 12$\arcmin$ square region, encompassing a 6 $\times$ 6 grid system as shown in Fig. 5. This method is used to aide our understanding of the cluster ICM in detail, as done by Gastaldello et al. (2015) for the NuSTAR observation of the Coma cluster. We use these results to select regions with similar thermodynamical properties to define larger regions with higher signal-to-noise ration (S/N).

For this analysis, we calculated projected maps, since the main goal is to achieve a comparative study, rather than to study the specific density, entropy, and pressure properties. To assess the ICM properties of box 7, we excluded the point source in that region. Although there will also be scattered light contamination in these regions, we expect the scattered light to change mildly in the adjacent regions due to the small region size we select. This means that, even if a most complete assessment of ICM properties may not be managed, we aim to still have a good proxy for the definition of regions of interest.

The grid system analysis is achieved with NuSTAR data, where we extracted spectra from each grid then fitted them using a single-temperature ${\tt apec}$ model. The abundances for the fits are fixed to Z=0.3 Z_⊙, because they were difficult to constrain due to low S/N. For the selection of regions with similar thermodynamical properties for the whole FOV, maps of NuSTAR temperature, density, entropy, and pressure were created. The normalization of the ${\tt XSPEC}$ model ${\tt apec}$ is defined as $\left[10^{-14}/4\pi\left[D_{A}(1+z)\right]^{2}\right]\int n_{e}n_{H}dV$, where the integrand is the emission measure (EM), n_e and n_H are electron and hydrogen densities in cm^-3, and the angular diameter distance to A3395 is D_A $\simeq$ 6.20 $\times$ 10²⁶ cm. By using the normalization of ${\tt apec}$ model, we estimated the EM, and the corresponding pseudo-pressure and pseudo-entropy maps using P = kT $\times$ EM^1/2 and S = kT/EM^1/3 (e.g. Rossetti et al., 2007). The projected density is calculated as the square root of the normalization parameter of the ${\tt apec}$ model.

Following the results of our grid analysis and with the guidance of the XMM-Newton photon image, we defined 6 regions of interest on the NuSTAR FOV Fig. 7. Region A represents the location of the hot spot detected by XMM-Newton  (Lakhchaura et al., 2011), Region B is the region extending to the intercluster filament, Region E is the bridge between the subclusters of A3395, and Region F is the tail of the southwestern subcluster (or the W clump). Region C is isolated since it shows higher-density regions than Regions A and B from our NuSTAR grid analysis (Fig. 6), and also shows excess emission in the eROSITA image, probably of filamentary origin (Fig. 1 and Fig. 7, middle panel). We note that Regions A, B, and C are mostly enclosed in the region isolated by Lakhchaura et al. (2011) (indicated by NW), which they suggest that connects the cluster to the intercluster filament.

NuSTAR is affected by scattered light contamination (also known as Ghost Rays; Madsen et al., 2017) due to X-ray photons that undergo only a single reflection off of either the primary (upper) or secondary (lower) mirror, as opposed to a properly focused double reflection. Due to the lack of precollimators, photons from bright sources outside the FOV can reach the focal plane without double reflection.

The upper single reflection is due to the photons that strike the upper mirrors at angles steeper than the nominal focusing graze angle; it fades away when the angle becomes too steep so that the adjacent shell shadows it (Madsen et al., 2017). The aperture stop also helps to block some of these photons. The lower single reflection, on the other hand, is caused by photons reaching the mirrors at angles that are shallower than the nominal graze angle (Madsen et al., 2017). In addition, back reflections can occur, in which photons hit the back side of the upper mirror of the adjacent shell first, then reflect off the front side of the mirror shell. The scattered light contamination appears as early as 2$\arcmin$ off-axis but only becomes significant above 3$\arcmin$, out to $\sim$1$\arcdeg$ (Madsen et al., 2017). The off-axis and energy dependence of scattered light is understood for point sources (Madsen et al., 2017), but no data analysis tool yet exists to model its effect in the case of extended emission. Ray-trace simulators, which trace out the paths of photons through the optics and onto the detectors, have been shown to reproduce observed scattered light patterns (Westergaard, 2011; Madsen et al., 2017).

We analyzed the effect of the scattered light for the NuSTAR observation, in an attempt to explain the excess power-law emission in the global spectra. Also, it is already expected since the two main subclusters, namely NE and SW (Fig. 7), lie near the FOV of the NuSTAR pointing. Our FOV covers distances of 3$\arcmin$-20$\arcmin$ from either of the subcluster centers, where the scattered light contamination is known to be significant (Madsen et al., 2017).

We did not study the effect of the back reflection contamination, since this component appear at very high fluxes and becomes effective at $\sim$15$\arcmin$, whereas the largest distance covered by our NuSTAR FOV from either of the subcluster centers is $\sim$20$\arcmin$. Also, the relative geometric area, i.e., fraction of back reflection to on-axis geometric area, is below 1% between 15$\arcmin$ and 20$\arcmin$ (Madsen et al., 2017).

To model the scattered light, we used the ray-tracing code for X-ray telescopes, MT_RAYOR (version 4.6.9) (Westergaard, 2011) written in Yorick interpreted language (Munro & Dubois, 1995), to simulate the properly focused photons (good photons) and photons that undergo single reflection (scattered light). MT_RAYOR takes a count rate image in the 0.5-2.5 keV energy band and a temperature map constructed from other X-ray missions—XMM-Newton in this case—guided by literature (Lakhchaura et al., 2011), as an input for the distribution of photons in the cluster, divides the cluster into predefined pixels, and predicts the expected double and single bounce photons from a NuSTAR observation, mainly treating the extended ICM emission as a collection of point sources that have different spectral properties as well as photon distribution.

These MT_RAYOR simulations provide a spatial distribution of good photons as well as the scattered light. We then extracted spectra from the 6 regions from both the good photons and scattered light simulation results (Fig. 8). Good photons describing the cluster emission were then fit by a single ${\tt apec}$ model, whereas we applied different models to fit the scattered light to find the best model that describes this contamination. We used ${\tt apec}$, ${\tt powerlaw}$, and ${\tt bknpower}$ models and find that ${\tt powerlaw}$ is the best model describing the scattered light for all regions. The model ${\tt powerlaw}$ has two parameters, photon index ($\Gamma$) and model normalization ($\kappa$). The resulting photon indices are presented in Table 3. The spectral fits of the simulated regions are shown in Fig. 13 in Section B.

We also created hardness ratio maps with a selection of different energy bands to investigate the energy dependence of the spatial distribution of the ray-traced scattered light, yet no apparent gradient was observed.

Once we calculated the 3-15 keV flux from both the good photons and scattered light for all six regions of interest, we deduced the ratio of the scattered photons to the total photons as presented in Table 3. The flux ratios as well as simulated images showed that the main region of interest, i.e. Region A, showed very little contamination from the scattered light. Since the number of ray-trace-simulated photons may differ from the real data, direct addition of the ray-traced scattered light spectra to the background spectrum may result in overestimation or underestimation of scattered light. Therefore, we calculated the fluxes from all six regions of the observational data, then with the 3-15 keV flux ratios provided from the aforementioned analysis, we used the ${\tt fakeit}$ function of ${\tt XSPEC}$, to simulate the scattered light with the previously obtained fitted parameter $\Gamma$, and reset the model norms to match the flux-estimated ray-trace flux ratios. Our recipe is summarized in the flow chart in Fig. 9.

We fitted the ${\tt apec}$ model to the XMM-Newton spectrum extracted from the six ROI. In this analysis, the abundance was a free parameter, and we present best-fit results where the N_H is fixed to the literature value of N_H = 6.30 $\times$ $10^{20}$ cm^-2 (Lakhchaura et al., 2011), in Table 4.

The faked spectra of each region were added to the corresponding background model spectra using ${\tt addspec}$ script by HEAsoft ${\tt ftools}$. After combining the scattered light and ${\tt nuskybgd}$ background, we fit the NuSTAR observational data with an ${\tt apec}$ model for the six regions, and the results are shown in Table 4.

For comparison and to see the effect of scattered light on model parameters, we also repeated the analysis without including our scattered light analysis. The results for this secondary analysis are shown in Table 4.

We also repeated the spectral analysis for the same six ROI using Suzaku data. The result of the fits are presented in Table 4.

As the next step, we fitted NuSTAR and XMM-Newton spectra jointly for all 6 ROI, where the abundance values were fixed to the values obtained from XMM-Newton spectral fits. The results are shown in Table 4.

We also jointly fitted the NuSTAR and XMM-Newton spectra from the regions of interest using an additional ${\tt apec}$ or ${\tt powerlaw}$ component to the original ${\tt apec}$ model. From these regions, the emission coming from Regions A and E is found to be better described with an additional spectral model.

Region A hints at two-temperature structure, the high-temperature component being kT = 16.2 keV, for which only a lower limit of 8.5 keV is obtained. The lower temperature is found to be kT = 2.90${}^{+0.47}_{-0.29}$ keV. This ${\tt apec}$ + ${\tt apec}$ model, improved the statistics by $\Delta$C/$\Delta\nu$ = 12.47/2. Applying ${\tt apec}$ + ${\tt powerlaw}$ model also improved the statistics ($\Delta$C/$\Delta\nu$ = 12.36/2), giving a photon index of $\Gamma$ = 1.70${}^{+0.23}_{-0.62}$.

An additional ${\tt apec}$ component improved the statistics of the Region E spectrum with respect to the single ${\tt apec}$ model by $\Delta$C/$\Delta\nu$ = 19.42/2. This two-temperature model resulted in a high-temperature component of kT = 15.91${}^{+15.92}_{-4.70}$ keV and a low-temperature component of kT = 3.78${}^{+0.37}_{-0.32}$ keV. Instead, when the ${\tt apec}$ + ${\tt powerlaw}$ model was used, the statistics were improved by $\Delta$C/$\Delta\nu$ = 24.41/2. This combined model resulted in a power-law emission with a photon index of $\Gamma$ = 1.80${}^{+0.12}_{-0.17}$, and the thermal component was kT = 4.28${}^{+0.57}_{-0.27}$ keV.

Since we did not find a constrained, strong high-temperature component in Region A, we restricted our spectral analysis to a smaller region with r = 1$\arcmin$.5 centered at the X shown in Fig. 7.

We fitted an ${\tt apec}$ model to the joint NuSTAR and XMM-Newton spectra and the XMM-Newton spectrum alone. For these analyses, we used both fixed and free abundance and N_H parameters, having eight different spectral fit. The abundance was fixed to Z = 0.23 Z_⊙ as obtained from the XMM-Newton spectral fit for Region A. We present the results of this analysis in Section C.

We also created deprojected thermodynamical maps using the ${\tt XSPEC}$ model ${\tt apec}$, which is defined as $\left[10^{-14}/4\pi\left[D_{A}(1+z)\right]^{2}\right]\int n_{e}n_{H}dV$, to obtain electron density n_e, with the assumption of a fully ionized plasma, n_e $\simeq$ 1.2 n_H. For the volume, we used the area of the ROI obtained using SAOImageDS9⁶⁶6https://sites.google.com/cfa.harvard.edu/saoimageds9 multiplied by the depth, and for the depth we assumed 1 Mpc for the line of sight (LOS), therefore the density values are scaled by (LOS/1 Mpc)^-1/2 similar to the approach adopted by Akamatsu et al. (2017). For entropy and electron pressure, we used the commonly adopted definitions, S = kT $\times$ n${}^{-2/3}_{e}$ (scaled by (LOS/1 Mpc)^1/3) and P_e = n_e $\times$ kT ((LOS/1 Mpc)^-1/2), respectively (Gitti et al., 2010). Gas pressure then becomes P = n $\times$ kT, where we assume n = 2n_e. The resulting maps are shown in Fig. 10, and the corresponding errors are given in Table 5.

We made two more assumptions for LOS for our thermodynamical values to be easily compared with future work. Firstly, we adopted r₁₈₀ = 34$\arcmin$.6 (Markevitch et al., 1998) and r₁₈₀ = r_vir, and estimated the distance of the center of each region from the midpoint of subcluster centers of A3395. The distances (x) we find for regions A, B, C, D, E, and F are 637 kpc ($\sim$0.32 r_vir), 876 kpc ($\sim$0.44 r_vir), 776 kpc ($\sim$0.39 r_vir), 438 kpc ($\sim$0.22 r_vir), 338 kpc ($\sim$0.17 r_vir), and 458 kpc ($\sim$0.23 r_vir), respectively.

These estimated distances are used for a secondary estimation of LOS. Along the center of the cluster where x=0 and $\theta$=0, the depth is 2r₅₀₀ $\sim$2 Mpc, and then it scales with varying [x,$\theta$]. The given distances (x) for these regions can be used as sin($\theta$), where the LOS is the 2cos($\theta$) assuming a perfect sphere with r = r_vir. With these assumptions, LOS for regions A, B, C, D, E, and F becomes $\sim$1.33 Mpc, $\sim$0.62 Mpc, $\sim$1.02 Mpc, $\sim$1.64 Mpc, $\sim$1.73 Mpc, and $\sim$1.62 Mpc, respectively.

A third estimation for LOS is achieved by using r_vir instead of r₅₀₀. This time, along the center of the cluster where x=0 and $\theta$=0, the depth is 2r_vir $\sim$4 Mpc, and then it again scales with varying [x,$\theta$]. We now assume r = r_vir. With these assumptions, LOS for regions A, B, C, D, E, and F becomes $\sim$3.76 Mpc, $\sim$3.56 Mpc, $\sim$3.68 Mpc, $\sim$3.90 Mpc, $\sim$3.92 Mpc, and $\sim$3.89 Mpc, respectively.

Using these physical distances along with the scaling factors used for density ((LOS/1 Mpc)^-1/2), entropy ((LOS/1 Mpc)^1/3) and pressure ((LOS/1 Mpc)^-1/2), results presented in Table 5 can easily be used with future work depending on the choice of radius for the cluster emission.

We also calculated the luminosity of the cluster from the XMM-Newton spectra within r = 796$\arcsec$ = 0.83 R₅₀₀ adopting R₅₀₀ = 930 kpc = 954$\arcsec$ (Alvarez et al., 2018). This selected region covers the whole FOV of the XMM-Newton observation. We find the X-ray bolometric (0.01-100 keV) luminosity to be L_X = 2.342${}^{+0.015}_{-0.011}$ 10⁴⁴ erg s^-1, and within 0.5-2.0 keV we find L_X = 1.204${}^{+0.041}_{-0.071}$ 10⁴⁴ erg s^-1 using ${\tt XSPEC}$ convolution model ${\tt clumin}$. Within this region, we find kT = 4.37$\pm{0.04}$ keV for the 0.5-7.0 keV band using a single-temperature model. In Fig. 11, we added the core X-ray properties of A3395 (black dot) to the bolometric luminosity vs. X-ray temperature scaling relation presented in Gaspari et al. (2014). We note that since the error bars are small, they are contained within the black dot, which is enlarged to facilitate the differentiation from the rest of the sample.

We also analyzed the spectral properties of the point source residing inside Region C previously cataloged as 2MASX J06261214-5417071 at z = 0.05050 (SIMBAD, Donnelly et al. (2001); Paturel et al. (2003); Jones et al. (2009); Marziani et al. (2017)). We extracted a circular region of r = 1$\arcmin$.2 centered on the point source. We find that the best-fit model that describes the emission from the source in the 3-20 keV band is an ${\tt apec}$ + ${\tt powerlaw}$ model with a photon index of $\Gamma$ = 1.48${}^{+0.56}_{-0.83}$, and C/$\nu$ = 412.10/457. The plasma temperature was found to be kT = 1.81${}^{+0.95}_{-0.59}$ keV, with Z = 0.43${}^{+2.47}_{-0.31}$ Z_⊙. The luminosity of the ${\tt powerlaw}$ component is found to be L_X = 1.434${}^{+0.153}_{-0.166}$ 10⁴² erg s^-1 within 3.0-20.0 keV and L_X = 5.962${}^{+0.706}_{-0.676}$ 10⁴⁰ erg s^-1 within 0.5-2.0 keV. For the ${\tt apec}$ component, we found 3.0-20.0 keV luminosity to be L_X = 7.330${}^{+1.008}_{-0.977}$ 10⁴¹ erg s^-1 and the 0.5-2.0 keV luminosity is L_X = 3.333${}^{+0.458}_{-0.444}$ 10⁴¹ erg s^-1.

This two-component model improved the statistics by with respect to a single-temperature model by $\Delta$C/$\Delta\nu$ = 14.15/2. The single-temperature model resulted in kT = 6.44${}^{+0.86}_{-0.74}$ keV plasma with an unconstrained abundance value. When we freed the redshift, we found z = 0.061${}^{+0.025}_{-0.028}$, which agrees with both the cluster and source (2MASX J06261214-5417071, Jones et al. (2009)) redshifts within 1$\sigma$.

We extracted a spectrum using the same region from XMM-Newton data. We grouped the XMM-Newton spectrum by 3 as well and applied ${\tt cstat}$ for direct comparison with NuSTAR results. A single-temperature model indicates kT = 4.64${}^{+0.60}_{-0.37}$ keV plasma with Z = 0.58${}^{+0.25}_{-0.30}$ Z_⊙. An addition of ${\tt powerlaw}$ component did not improve the fit and the photon index was not constrained. However, a single ${\tt powerlaw}$ model fit without an ${\tt apec}$ component gives a photon index of $\Gamma$ = 1.90${}^{+0.05}_{-0.09}$. Since these two emission models can be degenerate within XMM-Newton operating energy band given the similar statistics, we extracted a spectrum in the vicinity of the point source with the same area, and characterized the ICM within that region. A single-temperature model in this region showed a plasma with kT = 3.04${}^{+1.83}_{-0.76}$ keV and Z = 0.36${}^{+0.87}_{-0.25}$ Z_⊙. The parameters obtained from the ${\tt apec}$ fit of the region in the vicinity of the point source were then inserted and fixed during the fitting procedure of the point source, where a combined ${\tt apec}$ + ${\tt powerlaw}$ model was implemented. We then found a photon index of $\Gamma$ = 1.86${}^{+0.10}_{-0.11}$ for the ${\tt powerlaw}$ component. Within 0.5-2.0 keV, we estimate a luminosity of L_X = 3.555${}^{+0.308}_{-0.279}$ 10⁴¹ erg s^-1 again for the ${\tt powerlaw}$ component.

For the NuSTAR data of A3395, we first considered the global spectrum of the part of the cluster within the FOV, which was fit to a single temperature thermal plasma model (apec). We found an average temperature of $kT=5.59\pm 0.11$ keV. Although this is the global result form NuSTAR FOV, the pointing only covers the northwestern part of the cluster, and therefore this is not the global temperature of the cluster itself. However, the lack of a good fit suggests a single-temperature model insufficiently describes the data. As this is a merging cluster, we cannot assume isothermality across the FOV, and we expect the FOV to be contaminated by scattered light due to the presence of the subcluster centers in the vicinity of the pointing. We therefore fit this spectrum with two more models, with either an additional ${\tt apec}$ or ${\tt powerlaw}$ component.

We find that the ${\tt apec}$+${\tt powerlaw}$ model better describes the global spectrum with kT = 2.06${}^{+0.31}_{-0.22}$ keV and $\Gamma$ = 1.82${}^{+0.18}_{-0.29}$. However, this $\simeq$ 2 keV temperature is much smaller than expected based on the results of Markevitch et al. (1998), Donnelly et al. (2001) and Lakhchaura et al. (2011). In addition, the single-temperature model result of $\simeq$ 5.6 keV, is then higher than the global temperatures found for the cluster in the literature as found by Markevitch et al. (1998); Donnelly et al. (2001); Lakhchaura et al. (2011); Alvarez et al. (2018). The global temperatures they report account for the hot plasma at the cluster center and the NuSTAR FOV does not cover the central region of A3395. The cluster temperatures are expected to drop with respect to increasing radius assuming there are no shocks. The $\simeq$ 5.6 keV result we find for NuSTAR FOV should be even lower than their reported values and not higher. Although Lakhchaura et al. (2011) points to high-temperature regions enclosed in our NuSTAR FOV, which gave a motivation to study these regions in detail with NuSTAR  we recall that they also state a 60% error on these values.

Further investigating the impact of scattered light, we found that its contribution could also be modeled with a powerlaw model. This fact, combined with the presence of multi-temperature gas, suggests that the ${\tt apec}$+${\tt powerlaw}$ model is sufficiently flexible to capture these more extensive components and that the ${\tt powerlaw}$ component should not be interpreted to have physical meaning. Therefore, taking advantage of NuSTAR’s imaging capability, we continued with a grid analysis as described in Section 3.2. This analysis and the XMM-Newton photon image (upper panel of Fig.7) showed that there are six regions in the NuSTAR FOV with similar thermodynamical properties and substructures.

We also studied the XMM-Newton FOV that covers the central $\sim$0.83 R₅₀₀ of the cluster to obtain the global temperature and luminosity of the cluster. Our temperature result kT = 4.37$\pm{0.04}$ keV, which is in agreement with the literature (Markevitch et al., 1998; Donnelly et al., 2001) within 1$\sigma$ errors. We found a luminosity of L_X = 1.204${}^{+0.041}_{-0.071}$ 10⁴⁴ erg s^-1 within 0.5-2.0 keV, which is in agreement with the luminosity estimation of De Grandi et al. (1999) within 1$\sigma$. We found the X-ray bolometric (0.01-100 keV) luminosity to be L_X = 2.342${}^{+0.015}_{-0.011}$ 10⁴⁴ erg s^-1, as plotted in the L_X-T_X scaling relation in Fig. 11. Cool-core systems tend to reside in the upper envelope of the scaling relation due to a higher L_X per given T_X (or M_total since T_X is a tight proxy for cluster mass), whereas merging clusters whose cool cores have been disrupted reside in the lower envelope (Gaspari et al., 2014). A3395 lies at the lower boundary, in agreement with the typical behavior of the population of non-cool core clusters, which do not have the inner cool region of the ICM. This suggests that A3395 might have evacuated significant gas mass (moving toward the bottom), heated the gas (moving toward the right), or this cluster has assembled in a poor gas environment. Having a hot ($\sim$6 keV) intracluster filament in between its subclusters (Region E in our analysis, Region 2 in Markevitch et al. (1998), Region 3 in Donnelly et al. (2001), Region F in Lakhchaura et al. (2011)), and lacking a cool core, A3395 appears to be still in an early stage of merger, since it lacks a major overheating.

Using ray-trace simulations with MT_RAYOR, we assessed the scattered light contamination in our observation. We find that the scattered light is best modeled with a ${\tt powerlaw}$, and we provide a quantitative description of this contamination. This is an empirical model and due to the nonuniformity of the scattered light contamination, its complete behavior needs to be assessed with further studies of the off-axis angle and source energy dependence, as well as the position on the detector. This is a multidimensional problem, and our method is the only known approach to study NuSTAR data that has scattered light contamination, to the best of our knowledge. The method is described in detail in Section 3.3, and the method is summarized in flow chart shown in Fig. 9.

We find that in Regions A and B, where A3395 connects with an intercluster filament, the scattered light contamination is below 5% of the total flux. Regions C, D, and F suffer from scattered light at the $\sim$15% level, and this contamination rises to $\sim$25% for Region E. This is expected since Region E is near both of the bright subcluster centers, namely the NE and SW regions denoted in the upper panel of Fig. 7.

A quick comparison of $C$-stat values of NuSTAR  spectral fits with and without SL treatment in Table 4 shows that the fits improved for all ROIs when the scattered light component is included, with $\Delta C$ ranging from $\sim$2 to $\sim$27. Thus, the inclusion of our scattered light as an additional background component results in a better assessment of the source emission. However, we note that for all ROIs, the NuSTAR temperature results from the spectra with and without scattered light treatment agree within 1$\sigma$. Our temperature results from the NuSTAR data are in agreement to those from XMM-Newton  and joint NuSTAR and XMM-Newton analyses for a region with $\sim$25% scattered light contamination within 1$\sigma$. Our results show that, for regions of interest where scattered light contamination is above 10% (Regions C, D, E, and F), the temperature values are in agreement within 1.6$\sigma$ for NuSTAR, XMM-Newton, Suzaku, and joint XMM-Newton and NuSTAR spectral fits, which validates our approach to tackling the NuSTAR scattered light contamination for this observation.

We claim that, although the effect of scattered light contamination depends on the flux and emission features of structures, at a level of up to $\sim$25%, we seem to be safe within 1$\sigma$ errors of the face value of temperature. However, we also note that further investigation is needed to fully understand the effect of scattered light at various plasma temperatures. Our technique can be used for future NuSTAR observation proposals to estimate the possible scattered light contamination.

NuSTAR temperature results from all regions of interest seem to be lower than what is found with Suzaku (Table 4). This trend is also seen in regions A, B, C, and D for NuSTAR versus XMM-Newton. Cross-calibration studies similar to XMM-Newton and Chandra by Schellenberger et al. (2015) are required for NuSTAR, XMM-Newton, and Suzaku to understand this behaviour.

In our detailed analysis of the possible connection region of the cluster with the intercluster filament, Region A, we cannot confirm the existence of a strong high-temperature component in any of our NuSTAR, XMM-Newton, Suzaku, and joint NuSTAR and XMM-Newton fits, as previously reported in the temperature map of Lakhchaura et al. (2011). We consistently find temperature values around 4–5 keV in all datasets considered. This region is also present in a temperature map based on ASCA observations (Markevitch et al., 1998), and our temperature results from NuSTAR, XMM-Newton, Suzaku, and joint NuSTAR and XMM-Newton fits all agree with their results within 1$\sigma$. However, Lakhchaura et al. (2011) do note 60% errors on their map for regions lying at the edge of the FOV of the XMM-Newton pointing due to the low S/N.

Although we find a high-temperature component ($\sim$16 keV) for Region A with NuSTAR analysis, where two-temperature plasma model was applied to the spectrum, only a lower bound ($\sim$8 keV) for this higher-temperature component is found; the ICM is mainly dominated by the cooler component ($\sim$4 keV).

To better isolate a possible hot spot lying in Region A, we extracted spectra in a smaller region ($r=1.5\arcmin$) from both NuSTAR, and XMM-Newton, centered on the hottest region in the XMM-Newton temperature map reported by Lakhchaura et al. (2011). During these fitting procedures, we investigated the effect that column density and abundance may have on the temperature measurements, due to the possible bias against XMM-Newton  temperatures caused by the uncertainties in the effective area at soft energy (Schellenberger et al., 2015). While keeping the N_H frozen, we first kept the Z parameter fixed to the value found from the XMM-Newton analysis of Region A. Then we fit the spectra again by allowing Z parameter to be free. We repeated the same process by freeing the N_H.

The highest temperature we find is $kT=5.13^{+1.80}_{-1.48}$ keV with our XMM-Newton analysis (Table 6). We found lower temperatures with a joint XMM-Newton and NuSTAR analysis than with the XMM-Newton analysis alone (Table 7). In addition, all temperature values from these procedures agree within 1$\sigma$. Since background dominates at the edge of the XMM-Newton FOV, we hypothesize that the discrepancy between our results and the literature at the hot spot location may be due to how the background was treated; for example, the background model of Lakhchaura et al. (2011) might underestimate the true background at those locations.

Region B is also located near the connection region of A3395 and the intercluster filament. This region was also partially covered by the analysis of Markevitch et al. (1998), whose temperature results agree with ours at the 1$\sigma$ level. Regions C and D are at least partly included in the NW region studied by Lakhchaura et al. (2011), where our XMM-Newton temperature result agrees with their findings within 1$\sigma$.

The bridge emission between the two subclusters of A3395 enclosed by our Region E shows high temperature, entropy, and pressure with respect to the surrounding ICM. Moreover, this region was well fit by a two-temperature plasma model. This bridge is thought to be ram pressure-stripped from the northern subcluster in the A3395 merging system. Lakhchaura et al. (2011) finds a higher temperature for this intracluster filament than we do, yet our temperature value is within 1.5$\sigma$ of their value.

In addition, we searched for a nonthermal X-ray counterpart of the faint extended radio source to the west of A3395, lying in our Region F (Reiprich et al., 2021). They argue they this source denoted as S2/S3 in their work, may be a radio relic or may due to reaccelerated relativistic plasma. We find no significant nonthermal emission in this region, possibly because the hot ICM dominates the emission.

The electron entropy is closely related to the thermodynamical history of the clusters (Voit et al., 2005). In particular, the entropy of the ICM decreases in the process of radiative cooling and increases when heating energy is introduced into the ICM, e.g., via merging and feedback processes (Gaspari 2015). And at the interface regions of cluster outskirts and WHIM filaments is the zone where the entropy flattening is observed (Alvarez et al., 2018). To assess the merger history of the cluster and possible interaction of the filament and the cluster, we estimated the entropy for our ROIs. Since it is difficult to create radial profiles of a sample of non-cool-core systems due to asymmetrical morphology as well as nonthermal processes caused by mergers, we compared the entropy values with the 13 nearby cooling-flow cluster entropy profiles studied by Piffaretti et al. (2005, Figure 5) in the following paragraph.

In order make this comparison, we used the distance estimations of regions explained in Section 3.4. For Region A at $\sim$0.32 r_vir, we find the entropy to lie below the fitting curve yet within the scatter of the sample, and in addition, within 1$\sigma$ of galaxy cluster 2A 0335+096. Region B, at $\sim$0.44 r_vir, again is at the lower boundary of the scatter, in agreement with the entropy of Sersic 159-3 at the same distance from the core. Region C (at $\sim$0.39 r_vir), has a lower entropy than the whole sample range. The entropy of Region D at $\sim$0.22 r_vir is within the sample entropy values, still lying below the mean. The entropy of Region E ($\sim$0.17 r_vir), lies above the fitted curve. Finally, the entropy value of Region F ($\sim$0.23 r_vir), seems to lie above the fitted curve as well. All regions except for Region C have similar entropy with cooling-flow cluster entropy profiles, where Regions A, B, and D lie below the fitted curve, and Regions E and F lie above.

The high entropy and the high temperature of Region E (the bridge) indicate a heating process that may be caused by the gravitational pull of the ICM from subclusters that are at a pre-merger stage.

In addition to X-ray studies, the ICM of A3395 has been studied through the SZ effect with Planck. Planck Collaboration et al. (2013) report temperature and pressure values for the subclusters in A3395. Our Region D is enclosed in their A3395E region, and Region F is enclosed in Planck A3395SW region. Their GNFW2 pressure profile model results in kT = 5.0 keV and P = 0.40 $\times$ 10^-2 keV/cm³ for A3395E (Region D), and kT = 4.8 keV and P = 0.40 $\times$ 10^-2 keV/cm³ for A3395SW (Region F). With our NuSTAR analysis we find 4.01${}^{+0.44}_{-0.37}$ and P = 0.33$\pm{0.04}$ $\times$ 10^-2 keV/cm³ (5.35${}^{+0.68}_{-0.58}$ $\times$ 10^-12 dynes cm^-2) for Region D, and 5.27${}^{+0.39}_{-0.34}$ keV and P = 0.48${}^{+0.04}_{-0.03}$ $\times$ 10^-2 keV/cm³ (7.77${}^{+0.64}_{-0.57}$ $\times$ 10^-12 dynes cm^-2) for Region F. They do not report uncertainties for these specific regions (being model-dependent on global fits).

Although it is difficult to make a direct comparison between NuSTAR and Planck analyses since the derivation of pressure is based on the different methods, it is important to state that our result reaches the Planck value within 1$\sigma$ errors. This is also a good validation of our scattered light treatment, as well as the assumptions used for the deprojection of thermodynamical maps, since the estimation of these parameters is the end product of multiple treatments, assumptions, and analyses.

We visually detected a point source in both NuSTAR and XMM-Newton images, and extracted a circular region with r = 1$\arcmin$.2 region to study the source in detail from both NuSTAR and XMM-Newton data. We found that a ${\tt powerlaw}$ component with $\Gamma$ = 1.48${}^{+0.56}_{-0.83}$ and a thermal plasma with kT = 1.81${}^{+0.95}_{-0.59}$ keV using NuSTAR data best describe the emission. For XMM-Newton analysis, we selected a region with the same area in the vicinity of this source due to the degeneracy of the ${\tt powerlaw}$ and ${\tt apec}$ components within XMM-Newton bandpass. This analysis helped in estimating the plasma properties in the vicinity, and resulting parameters were adopted and fixed in the fit of the point source region, which resulted in a ${\tt powerlaw}$ component with $\Gamma$ = 1.86${}^{+0.10}_{-0.11}$, which is better constrained than the NuSTAR  photon index, since all apec parameters were frozen during the XMM-Newton fit. We also tried using an additional ${\tt apec}$ model instead of ${\tt powerlaw}$, yet then the secondary ${\tt apec}$ temperature was around 5 keV, and the statistics were comparable, i.e., C/$\nu$ = 570.35/702 for ${\tt apec}$ and C/$\nu$ = 569.03/702 for ${\tt powerlaw}$. This $\sim$5 keV value is higher than what we find in Region C, where the point source resides, with NuSTAR, XMM-Newton, and Suzaku, as well as joint NuSTAR and XMM-Newton results, pointing to the degeneracy of a kT $\sim$ 5 keV thermal emission and $\Gamma$ $\sim$1.86 power-law emission for the XMM-Newton analysis. Since NuSTAR covers a wider energy range, we claim that the true emission includes a power-law emission as well as a thermal emission within that region. The photon indices obtained from the two analyses agree within 1$\sigma$, therefore we cannot claim a statistically significant variability between NuSTAR and XMM-Newton observations. When the XMM-Newton and NuSTAR spectra were simultaneously fit using a powerlaw + apec model, the apec component dominated the spectra, as expected due to the inclusion of more lower energy XMM-Newton photons in the spectra.

In addition, results from NuSTAR spectral analysis for the temperature and the luminosity of the gas confined within this r = 1$\arcmin$.2 suggest that the point source may be a thermal corona embedded in a hot environment (Sun et al., 2007; Tümer et al., 2019). The photon indeces of $\Gamma$ = 1.86${}^{+0.10}_{-0.11}$ for XMM-Newton and $\Gamma$ = 1.48${}^{+0.56}_{-0.83}$ favor an AGN emission rather than X-ray binaries ($\Gamma$ $\leq$ 1.4) (see, e.g., Tozzi et al., 2006). Furthermore, ${\tt powerlaw}$ component accounts for the $\sim$66% of the total luminosity within 3.0-20.0 keV, and $\sim$15% within 0.5-2.0 keV based on NuSTAR analysis.

Guided by the literature (Tittley & Henriksen, 2001; Lakhchaura et al., 2011; Planck Collaboration et al., 2013; Bourdin et al., 2020; Reiprich et al., 2021), we studied Regions A, B, and C in detail with the assumption that these regions may represent an interface of the A3395 ICM and the intercluster filament. Alvarez et al. (2018) find a global temperature of $kT=4.45^{+0.89}_{-0.55}$ keV and density $n_{e}=1.08^{+0.06}_{-0.05}\times 10^{-4}$ cm^-3 for the intercluster filament. The temperature results obtained from NuSTAR, XMM-Newton, Suzaku, and the joint NuSTAR and XMM-Newton spectral analysis of Regions A and B, are in agreement with Alvarez et al. (2018) for the filament within 1$\sigma$. However, for Region C, the temperature results from NuSTAR analysis show a cooler plasma than what is found for the filament, yet in agreement within 1.3$\sigma$ of Alvarez et al. (2018), and within 1$\sigma$ with our XMM-Newton, Suzaku, and the joint NuSTAR and XMM-Newton analyses. In addition, the density of the filament found by Sugawara et al. (2017) and Alvarez et al. (2018) is much smaller ($\sim 1/4$) than the density we find for Regions A, B, and C.

The entropy is expected to rise to values higher than 1000 keV cm² outside 0.5 R₂₀₀ (Pratt et al., 2006), due to heating by accretion shocks, and Lakhchaura et al. (2011) report high entropy and high temperature values for these regions. However, we do not observe such high entropy or temperature in Regions A, B, and C, which are expected to have higher entropy than the regions close to the center of the cluster (Piffaretti et al., 2005). In addition, the entropy we find for these regions are even lower than what is found at that similar radii for cool-core clusters Piffaretti et al. (2005).

These results, when studied in conjunction with the low temperature, low pressure, and high density values, suggest an excess of radiative cooling, which points to a flow of ICM into the filament, in contrast to Reiprich et al. (2021), who find high-temperature gas in the interface region and suggest heating by shocks via the ongoing merger activity. Such an offset cooling process is analogous to the more vigorous multiphase condensation ‘weather’ occurring in the dense cluster cores (Gaspari et al. 2018). Indeed, mergers drive significant amount of turbulent motion, which can locally enhance density (Gaspari & Churazov 2013) and thus lead to localized enhanced filamentary cooling (Wittor & Gaspari 2020). Such detections of cooling in merger systems have become more frequent in recent years (e.g., Somboonpanyakul et al. 2021). Our results are also in line with Alvarez et al. (2018), who suggest that the ICM gas in the outskirts may be tidally moved into the filament during the interaction, as a part of the merging processes of A3395 and A3391. Such tidal motions can be seen as another form of large-scale turbulence, with the related eddies locally enhancing density.