Discovery of a Shock-compressed Magnetic Field in the Northwestern Rim of the Young Supernova Remnant RX J1713.7–3946 with X-Ray Polarimetry

Supernova remnants (SNRs) provide insights into cosmic-ray acceleration and magnetic field dynamics at shock fronts. Recent X-ray polarimetric measurements by the Imaging X-ray Polarimetry Explorer (IXPE) have revealed radial magnetic fields near particle acceleration sites in young SNRs, including Cassiopeia A, Tycho, and SN 1006. We present here the spatially resolved IXPE X-ray polarimetric observation of the northwestern rim of SNR RX J1713.7–3946. For the first time, our analysis shows that the magnetic field in the particle acceleration sites of this SNR is oriented tangentially with respect to the shock front. Because of the lack of precise Faraday rotation measurements in the radio band, this was not possible before. The average measured polarization degree (PD) of the synchrotron emission is 12.5% ± 3.3%, lower than the one measured by IXPE in SN 1006, comparable to the Tycho one, but notably higher than the one in Cassiopeia A. On subparsec scales, localized patches within RX J1713.7–3946 display a PD of up to 41.5% ± 9.5%. These results are compatible with a shock-compressed magnetic field. However, in order to explain the observed PD, either the presence of a radial net magnetic field upstream of the shock or partial reisotropization of the turbulence downstream by radial magnetohydrodynamical instabilities can be invoked. From comparison of PD and magnetic field distribution with γ-rays and 12CO data, our results provide new inputs in favor of a leptonic origin of the γ-ray emission.

1. INTRODUCTION Supernova remnants (SNRs) are thought to play an essential role in accelerating Galactic cosmic rays (CRs), with diffusive shock acceleration (DSA) generally accepted as the leading mechanism (see e.g.Malkov & Drury 2001).This mechanism is capable of accelerating CRs up to hundreds of TeV energies (see e.g., Ginzburg & Syrovatskii 1964;Amato 2014) thanks to an amplified magnetic field.Evidence of its presence in the SNR is provided by non-thermal X-ray emission generated by relativistic gyrating electrons, known as synchrotron radiation, first seen in radio waves (see, e.g., van der Laan 1962), and then in the X-ray band (Koyama et al. 1995).The discovery of X-ray synchrotron emission in SNRs provided crucial information on acceleration of electrons up to TeV energies.However, the exact relationship between magnetic fields influencing CR acceleration, and CRs modifying magnetic fields is still an open question.Indeed, the presence of synchrotron alone cannot be considered a definitive proof of CR presence.CRs are mainly protons and, in order to be sure that a source is accelerating protons, detection of hadronic γ-ray emission due to neutral pion decay is needed (Ackermann et al. 2013;Amato 2014;Cristofari 2021).Polarimetry provides information on the magnetic field turbulence level, and indeed radio band polarimetry revealed that young (up to a few thousands of years old) SNRs show low polarization fractions of 5−10% (Dickel & Jones 1990, e.g.), indicative of high turbulence, and a preferential radial direction of the magnetic field.On the other hand, older (∼10,000 years old or more) SNRs show a more ordered and preferentially tangential magnetic-field orientation (Milne & Dickel 1975;Fürst & Reich 2004).While the latter can be easily interpreted as shock compression of the ambient magnetic field which enhances the tangential component, the radial magnetic-field orientation is less understood, with most theories invoking radial stretching induced by the onset of hydrodynamical instabilities, such as Rayleigh-Taylor (RT) at the SNR contact discontinuity (CD) (Gull 1973;Jun & Norman 1996), or Richtmeyer-Meshkov (RM) instabilities near the forward shock (Inoue et al. 2013a).As an alternative, the radial magnetic field orientation may appear as an artifact of preferential (i.e., more efficient) electron acceleration, where the magnetic field is parallel to the shock normal, biasing polarization measurements towards regions with a radial magnetic-field direction (West et al. 2017).Once relegated only to the radio and, in some cases, in the infrared band (e.g., Jones et al. 2003), the polarimetric study of the magnetic-field topology in young SNRs can now also be explored in X-rays thanks to the launch of the NASA/ASI Imaging X-ray Polarimetric Explorer (IXPE, Weisskopf et al. 2022), whose imaging capabilities are perfectly suited for probing magnetic field distribution and turbulence in high-energy extended sources, such as SNRs.X-ray band polarimetry has many advantages with respect to that in the radio band: TeV-energy electrons responsible for the X-ray emission have a shorter lifetime compared to the radio emitting electrons (Vink & Zhou 2018).For this reason they are confined in regions closer to particle acceleration sites.Moreover, in the X-ray band the synchrotron spectra of SNRs tend to be steeper than in the radio one, and because the synchrotron polarization degree depends on the photon spectral index (Ginzburg & Syrovatskii 1965), this allows for higher theoretical maximum polarization values.Finally, X-ray radiation is not affected by Faraday rotation.Consequently, X-ray polarimetry provides the perfect opportunity to investigate magnetic field geometry at particle acceleration sites.IXPE previously detected significant polarization in three younger SNRs: Cas A (Vink et al. 2022), Tycho (Ferrazzoli et al. 2023), and SN 1006 (Zhou et al. 2023), all showing radially oriented magnetic fields.This demonstrates that the processes responsible for radial magnetic fields observed in young SNRs are already at work at the scales where X-rays are emitted.The lowest polarization degree (PD) of the non-thermal emission was reported for Cas A (Vink et al. 2022), PD∼ 5%, followed by Tycho (Ferrazzoli et al. 2023), PD∼ 12% and finally by SN 1006 (Zhou et al. 2023), PD∼ 22%.Comparing the X-ray polarization measurements of these three remnants, the remarkable similarity in the radially oriented magnetic field is in contrast with the very different PD, likely reflecting a decreasing level of magnetic field turbulence.The influence of different environmental densities has been put forward as a possible explanation (Zhou et al. 2023).Here, we present a significant IXPE detection of a tangential magnetic field in the northwestern rim of the SNR RX J1713.7-3946(from now on RX J1713 for brevity).Also known as G347.3-0.5, RX J1713 is a very large (≈ 1 degree diameter) shell-type SNR residing in the Galactic plane, discovered in the ROSAT all-sky survey (Pfeffermann & Aschenbach 1996).Its distance, inferred from CO-observations and from investigation of X-ray absorbing material in the remnant, is ∼ 1 kpc (Fukui et al. 2003;Uchiyama et al. 2003;Cassam-Chenaï et al. 2004).This remnant is thought to be the result of a type Ib/c supernova explosion (also referred to as "stripped core-collapse supernova") of a relatively low-mass star in a close binary system (Katsuda et al. 2015), where the progenitor massive hydrogen envelope was removed either in a binary interaction or through a stellar wind.RX J1713 is commonly associated with SN 393 (Wang et al. 1997;Tsuji & Uchiyama 2016;Acero et al. 2017), making it the oldest remnant for which an X-ray polarization signal is reported so far, with an age of 1631 years.Its hard X-ray emission has been discovered to be purely non-thermal (Koyama et al. 1997;Slane et al. 1999), representing the second detection of synchrotron X-ray radiation from a SNR shell after SN 1006.Lazendic et al. (2004) analyzed Chandra X-ray and ATCA 1.4 GHz radio observations of the north western region of RX J1713 resolving bright filaments of about 20−40 arcsec of inferred width, corresponding to linear sizes of 0.1−0.2pc at a 1 kpc distance.Their analysis showed a one-to-one correspondence between X-ray and radio morphological structures, providing strong evidence that, in this part of the remnant, the same population of electrons is responsible for synchrotron emission in both the bands.Lazendic et al. (2004) also detected linear radio polarization from the SNR at the northwestern part of the shell, although Faraday rotation made it impossible to determine the magnetic field direction from the polarization vectors.RX J1713 is also the best studied young SNR in the γ-ray band: it has been detected at TeV energies by HESS (Aharonian et al. 2004(Aharonian et al. , 2007;;H. E. S. S. Collaboration et al. 2018) and Fermi (Abdo et al. 2011;Federici et al. 2015).The finding was that the RX J1713 γ-ray emission at TeV energies closely follows the Xray morphology.However, despite the detailed morphological and spectral studies, the nature of the mechanism responsible for the production of GeV and TeV γ-rays in RX J1713 has not clearly been established yet.There are different mechanisms that can be responsible for efficient γ-ray production: leptonic emission due to inverse-Compton scattering of soft -CMB, IR, or optical -photons by relativistic electrons, hadronic emission due to two-photon decay of neutral pions produced by relativistic protons interacting with the ambient medium, or a mixture of both.RX J1713 is one of the most debated sources in the discussion on γ-ray from the aforementioned radiation mechanisms.Acero et al. (2009); H. E. S. S. Collaboration et al. (2018) argue that the γ-ray emission morphology, as well as the hard-ness of the spectrum, favors a leptonic origin, although it cannot be explained by a single electron population (Finke & Dermer 2012).Self-consistent hydrodynamical modeling, that includes the effects of efficient DSA, supports this interpretation (Ellison et al. 2010(Ellison et al. , 2012)).The main argument there is that the density at the shock seems too low for a sufficiently high pion production rate from hadronic cosmic-rays.On the other hand, the hadronic mechanism could also fit data under the assumption that dense clumps have survived the shock passage.The most energetic hadronic cosmic rays could penetrate those cloudlets, producing a relatively hard hadronic γ-ray spectrum (Fukui et al. 2012;Gabici & Aharonian 2014;Federici et al. 2015;Celli et al. 2019;Fukui et al. 2021).A very recent hypothesis is that the emission could be also produced by the sum of reaccelerated CR electrons emission and a subdominant contribution from freshly accelerated protons (Cristofari 2021).X-ray polarimetry is important as it traces the level of magnetic-field turbulence, which is one of the key ingredients of particle acceleration through DSA.In this paper, we describe the IXPE observation of RX J1713 and data reduction in Section 2, present and discuss our results in Sections 3 and 4, respectively, and finally present our conclusions in Section 5.

IXPE data
As extensively outlined in Weisskopf et al. (2022) and its cited sources, the IXPE observatory is composed of three identical X-ray telescopes, each including an Xray mirror module assembly (provided by NASA) and a polarization-sensitive Gas Pixel Detector (Costa et al. 2001;Bellazzini et al. 2006;Soffitta et al. 2021;Baldini et al. 2021, GPD, provided by ASI).This configuration enables X-ray imaging spectropolarimetry with ∼ 30 arcsecond angular resolution within the 2−8 keV energy band.IXPE observed the north west part of the shell of RX J1713 (see Figure 1) three times in 2023: August 24-27, August 28 to September the 1 st , and September 24 to October 5 for a total exposure time of ∼ 841 ks (obs.id.02001499).The pointing of the spacecraft was dithered with an amplitude of 0.8 arcmin to further average-out any residual instrumental effects.Because RX J1713 is among the faintest SNRs observed so far by IXPE, the background contribution needs to be carefully taken into account.This can be divided into two categories: the diffuse X-ray background of astrophysical origin and the particle-induced instrumental background.In order to mitigate the latter, we conducted an energydependent background rejection by removing the most probable particle-induced events through the algorithm  of Di Marco et al. (2023).This allows one to remove up to 40% of the instrumental background events.After rejection, the 2−8 keV background level of previous IXPE observations is in the range (0.9 − 1.2) × 10 −3 counts s −1 arcmin −2 (Di Marco et al. 2023).Because the remaining background after rejection is still non-negligible, we select two source-free regions as seen in the right panel of Figure 1.In these regions the emission consists of the remaining instrumental background, the Galactic Ridge X-ray diffuse background, and the local sky background.The 2−8 keV total background counting rate from both the two regions combined is found to be 2.2×10 −3 counts s −1 arcmin −2 .We verified that the background is compatible with being unpolarized with PD < 7.5% in the 2−5 keV energy range and PD < 6.9% in the 2−8 keV energy range at 99% confidence level.The details of the background properties are described in the Appendix.
A small fraction of the observation was affected by high solar activity, which caused background spikes in the data light curves.In order to remove the time intervals affected by this background contribution, we produced a light curve by binning the 2−8 keV photons from the whole dataset every 15 seconds, we then fitted the his-togram of the count distribution with a Gaussian function and filtered out all the time intervals with count rates more than 3σ above the mean of the distribution.Consequently, we removed ∼ 10 ks from the data, leaving an exposure time of ∼ 830 ks.
Because of uncertainties on the photon absorption point reconstruction in the GPD, spurious polarization patterns on scales comparable to the IXPE angular resolution may potentially contaminate Stokes maps in sources exhibiting significant gradients of the brightness distribution (that is a second derivative).Through estimates based on a Mueller matrix characterization of the IXPE response (Bucciantini et al. 2023), we verified that the impact of this effect is negligible (≪ 1% polarization) in all the regions of interest considered in this work.The alignment of the three DUs was good, and no manual alignment was needed.

Data analysis
Based on the well-tested procedure of previous IXPE analysis of SNRs (Vink et al. 2022;Ferrazzoli et al. 2023;Zhou et al. 2023), we adopt an approach that starts from a small-scale search of signal in binned and smoothed polarization maps, and then goes on with a large scale search in regions of interest, using both model-independent and model-dependent analysis techniques.In this paper, we produce polarization maps and perform model-independent analysis of IXPE data with the publicly available software package ixpeobssim (version 30.6.3, Baldini et al. 2022), using the version 12 of instrument response functions.ixpeobssim is a simulation and analysis toolkit developed by the IXPE collaboration including both a tool for generating realistic Monte Carlo simulations of IXPE observations and methods for data selection and binning to produce Stokes maps and spectra.We used the algorithm PMAPCUBE of the ixpeobssim binning tool xpbin to produce images of the Stokes parameters I, Q, and U, and then to compute polarization degree (PD) and polarization angle (PA) maps.We selected the 2−5 keV energy band to minimize the contamination from both instrumental and diffuse background (see Figure 7 in the Appendix).For in-depth analysis, we chose a region delimited by the contour enclosing the shell with surface brightness of at least 0.90 counts per arcsecond 2 in the 2−5 keV energy range, that is twice the background surface brightness of ∼ 0.45 counts per arcsecond 2 .We further divided the selected region in North (N), South (S) and West (W) sub-regions, and defined three elliptical sub-regions called P1, P2, and P3, identified in the smoothed polarization maps (see right panel of Figure 2).The largescale search for polarization in the regions of interest was carried out with the PCUBE algorithm of the xpbin tool: this allows us to extract the Stokes parameters of the events collected in each region and to calculate polarization properties.We also extracted the Stokes I, Q, and U spectra of these regions through the PHA1, PHA1Q, and PHA1U algorithms in xpbin, in order to perform a model-dependent analysis using XSPEC version 12.12.1 (Arnaud 1996).In the spectropolarimetric analysis, we used weighted analysis and response files (Di Marco et al. 2022) that allow one to improve the detection sensitivity by weighting the photoelectron tracks through their elongation.We grouped the Stokes I spectra to contain at least 50 counts per bin, while we applied a constant 0.2 keV energy binning to the Stokes Q and U spectra.

Other data
RX J1713 is a very well studied object from radio to γ-ray energy band.In Figure 1 (a) we show a composite three-color image of the whole RX J1713 in radio (red, ATCA 20cm), X-ray (green, Chandra 0.5−7 keV), and γ-ray (blue, HESS >2 TeV) bands: the IXPE field of view covers a region that is bright in all these three wavelengths.Consequently, it is important to present the X-ray polarimetric results in a multi-wavelength context.For this purpose, we use in our analysis several collected data sets.For the radio band, we obtain data from the Australia Telescope Compact Array (ATCA), namely the image and polarization fraction map at 20 cm (Lazendic et al. 2004).We also employed high-resolution X-ray data as an exposure-corrected Chandra broadband (0.5−7 keV) image obtained from a mosaic of the obs.id.736, 737, 5560, 5561, 6370, 10090, 10091, 10092, 10697, 12671, and 21339 processed with the CIAO toolkit (version 4.15, Fruscione et al. 2006).We then retrieve the HESS >2 TeV image described in H. E. S. S. Collaboration et al. (2018) with improved angular and spectral resolution.Finally, we also use data from the Mopra Southern Galactic plane CO survey data release 3 (Braiding et al. 2018) in order to trace the distribution of the molecular gas clouds associated with RX J1713.We produce maps of the J = 1 − 0 transitions of 12 CO with 0.6 arcmin spatial resolution and 0.1 km s −1 velocity resolution integrated in the −6 to −16 km s −1 velocity range, associated with the distance to RX J1713 (Fukui et al. 2003;Cassam-Chenaï et al. 2004) using the CARTA software (Comrie et al. 2021).

Polarization maps
The IXPE telescopes have an angular resolution (halfpower diameter) of ∼ 30 arcseconds, allowing us to produce spatially-resolved maps of the X-ray polarization properties.In Figure 2 (left) we show the polarization map of the NW rim of RX J1713 obtained by using the PMAPCUBE algorithm, binned with a pixel size of 2 arcmin (that is, about four IXPE point spread functions), with the vectors indicating the direction of the magnetic field, orthogonal to the observed PA.Binned maps require a trade-off in determining the pixel size: if this is too small there will be not enough statistics to achieve a significant measurement, if too big the polarization may be diluted.For this reason, we also applied a smoothing technique to the polarization maps with an over-sampled pixel-size of 10.5 arcseconds, and we increased the signal-to-noise ratio by smoothing the images with a Gaussian kernel of 6 pixels (corresponding to about ∼ 1 arcmin).We show this smoothed map in Figure 2 (right).In both maps, the significance of the magnetic field directions is colorcoded as > 2σ in cyan, and > 3σ in green.In the binned map we find four pixels with significance higher than 3σ.In the smoothed map, instead, we identify three ellip-tical regions of high-significance that we called P1, P2, and P3, and are broadly consistent with the positions of the most significant pixels in the binned map. Figure 3 (a), shows the Mopra 12 CO image of RX J1713 integrated in the −6 to −16 km −1 velocity range overlapped by the magnetic field inferred from the IXPE observation.The other panels of Figure 3 show the ATCA 20 cm radio image (b), the high-resolution broadband Chandra image (c), and the HESS > 2 TeV excess counts image (d).In the radio image we also show the radio PD as grey contours of 2% steps.In the 12 CO image, the highsignificance polarization is measured in regions that appear to have lower gas density, while in the 20 cm radio image there is a weak correlation between radio emission/polarization and X-ray polarization.On the other hand, the IXPE results overlayed to the high-resolution Chandra image show the magnetic field lines following the shape of the curvature of the shock.Because of the low spatial resolution of the HESS γ-ray image, no new information can be derived at the present time.

Model-independent polarization results
For each region of interest, we show the observed PD and PA as polar plots in Figure 4, with their 50%, 90%, 99%, and 99.9% confidence level contours, obtained from the PCUBE analysis.Their values are tabulated in Table 1 togheter with the values calculated after background subtraction.The polarization signal from the the region "All" (made from the combination of regions N, S, and W shown in the right panel of Figure 1) is detected at a significance of 5.2σ, with an average PD of 13.0% ± 3.5%, after background subtraction.In the combined North and South sub-regions of the shell, we have a detection at 4.8σ with PD of 12.8% ± 3.7%, after background subtraction.Individually, we detect X-ray polarization in the North and South parts of the shell at 3.7 σ and 4.0 σ, respectively, with an intrinsic PD of 12.4%±4.2%and 18.3%±5.3%.In each region, we find the PAs to be consistent with being orthogonal to the shock direction.We do not obtain statistically significant detection for the West part of the Shell.The three smaller regions P1, P2, and P3 selected from the polarization maps show a significant (3.9σ, 5.6σ, 4.8σ) PD of 25.2% ± 8.0%, 45.1% ± 8.3%, 46.0% ± 9.8%, after background subtraction, that are higher than the average shell value, and with PAs again consistent with being perpendicular to the local shock, in agreement with those of the entire shell.
In contrast to what was previously done in other IXPE observation of SNRs (e.g. in Vink et al. 2022;Ferrazzoli et al. 2023), and similarly to the results obtained on SN 1006 (Zhou et al. 2023), in RX J1713 a realignment of the Stokes parameters according to a radial or tangential magnetic field model, using the xpstokesalign tool, yields no conclusive evidence of a circular symmetry over a uniform distribution.This is because of the large angular size of the remnant and of the difficulty in determining the reference frame center for the Stokes parameters realignment.Hence we test the hypothesis of a constant magnetic field direction in the region encompassing the P1, P2, and P3 regions, using the teststatistic, where the weighted mean P A w is 105 degrees and PAn and PAn ERR , with n from 1 to 3, are the measured polarization angles and their uncertainties for the regions P1, P2, and P3, respectively.The estimated test-statistic value is χ 2 = 29.9 with 2 degrees of freedom.Thus, the probability that the PA is constant is only 3.26 × 10 −7 .The major axes of the elliptical regions, P1, P2, and P3, on the right panel in Figure 1 follow the bright arc in the IXPE count map that in turn is along the shock front.These major axes are normal to the directions of PA in the P1, P2, and P3 regions, respectively, within statistical errors.This suggests that the magnetic field is tangential to the shock front.

Spectropolarimetric analysis
A model-dependent approach for calculating the polarization of the five regions is to use the I, Q, and U spectra.Firstly, we extracted source and background spectra from the previously defined regions.For each region, we then jointly fit the background-subtracted I, Q, and U spectra with a model obtained by combining foreground absorption, power-law spectrum, and PD and PA constant in the 2−5 keV energy band (tbabs * powerlaw * constpol) in XSPEC.We fixed N H in the IXPE spectropolarimetric analysis to 8.15 × 10 21 cm −2 based on Tsuji et al. (2021), and left the powerlaw index, normalization, and polarization values as free parameters.The results of the fit are tabulated in Table 1, and show agreement with the model-independent results, both for the PD and PA.Also the best fit value of the photon spectral index Γ is in agreement with the values ∼ 2.0 − 2.4 for the region observed by IXPE as reported in the literature (Koyama et al. 1997;Slane et al. 1999;Uchiyama et al. 2003;Lazendic et al. 2004;Cassam-Chenaï et al. 2004;Hiraga et al. 2005;Sano et al. 2015;Tsuji et al. 2019).In Figure 5 we present the spectropolarimetric fit for the "All" region as an example.

Behavior of the magnetic field
The discovery of a tangential magnetic field in RX J1713 differs from previous IXPE findings that revealed radial magnetic fields for other young SNRs, Cas A, Tycho, and SN1006 (Vink et al. 2022;Ferrazzoli et al. 2023;Zhou et al. 2023).In these three cases, it was argued that hydrodynamical instabilities responsible for the on-set of radial field at longer wavelengths appear to be already active at sub-parsec scales, close to particle acceleration sites.The tangential magnetic field in RX J1713 is instead compatible with a model of anisotropic turbulence produced by the shock compression of an upstream isotropic turbulence (Bykov et al. 2020).In this model, as the turbulence is swept across the shock upon compression, show the observed polarization degree (PD) and angles (PA) in the the radial and position angle coordinates, respectively.The black dots mark the measured PD and PA values.We represent the 50%, 90%, 99%, and 99.9% confidence levels (based upon χ 2 with 2 degrees of freedom) as red, orange, green, and blue contours, respectively.the magnetic field becomes predominantly tangential to the shock front, resulting in predominantly radial synchrotron polarization.Interestingly, as suggested by the polarization map and the Chandra X-ray image (Fig. 3 c) the highest PDs we reported are very close to the shock front, with offsets from the peak polarization up to 1 arcmin.There may be shifts between IXPE and Chandra -due to proper motions and/or alignment issues -but based on the measured expansion rate (Tsuji & Uchiyama 2016), the shock would have shifted by 8 arcsec at most over the 10 year baseline, much less than the IXPE angular resolution.Moreover, we found no misalignment between the IXPE and Chandra images.So, if IXPE itself shows that the X-ray emission peak is not co-spatial with the polarization peak, this could imply that the turbulence level of the magnetic field is increasing further downstream of the shock.Under our interpretation, the PD is determined by the ratio of chaotic-to-ordered magnetic fields in the X-ray emitting regions.The PD values are observed to differ considerably between the four SNRs studied to-date, and differences in the ambient conditions may play a role.However, the distinctly different magnetic field geometry inferred for RX J1713 which reminds of the already known dichotomy observed in the radio band, suggests fundamental differences in the development of the ordered field component, making it difficult to draw firm conclusions on the chaotic magnetic field compo-nent.The reason of the dichotomy is not fully understood, because it involves many aspect of SNR physicsmagnetic field amplification due to CR acceleration, instability at the shock front dependent on ambient conditions, instability within the SNR shell due to its internal structure, and ultimately the SN explosion dynamicsthat no simple description can be taken as fully reliable.The general idea is that the dichotomy is due to the presence or absence of RT instability at the CD separating the SN ejecta from the shocked interstellar matter (ISM).This strongly depends on the acceleration to which the SNR shell is subject and this is stronger in younger remnants than in older ones (where young vs old is not just a matter of age but of the expansion regime).So it is likely that RT-instability is more efficient in stretching the magnetic field in the radial direction in younger systems than in older ones.It is unlikely that X-ray emitting particles are volume filling and hence subject to the effect of RT-instability at the CD.The action of RM-instability at the very shock front -that previous papers (Vink et al. 2022;Ferrazzoli et al. 2023;Zhou et al. 2023) attributed the radial magnetic field to -might also have some dependence of the dynamics of the SNR, making younger (and hence faster) systems more prone to it (but also on ISM conditions, in particular the presence of density gradients).The magnetic field structure in the upstream could also be reshaped by the efficiency of CR acceleration.Unfor-tunately with just four SNRs reported by IXPE so far, it is hard to find any reasonable and robust trend.To make any meaningful attribution we would need a better characterization of these SNRs.For these reasons we defer in-depth discussions on comparisons between the different SNRs to a future publication.

Magnetic field turbulence
While the exact nature of both ordered and turbulent components of the magnetic field still remain uncertain, we can place constraints on their relative magnitudes through the observed PD.Here we use the Bandiera & Petruk (2016) model to describe the connection between observed PD, photon spectral index Γ, and turbulence level of the magnetic field δB/B 0 of a source.This model assumes that a uniform field B 0 is superposed on the random field of the source.The random component has a Gaussian distribution -with a kernel of δ B -isotropic on average, so that the stronger the random magnetic field component, the smaller the PD.We find that for the "All" region of RX J1713, δB/B 0 = 1.6 ± 0.2, while for the P1, P2, and P3 regions it is 1.0 ± 0.2, 0.6 ± 0.1, and 0.6 ± 0.2, respectively.The shock compressed model of Bandiera & Petruk (2016), where δB tan = 4 √ 2δB perp1 , gives an expected downstream PD∼ 69% with tangential direction for a RX J1713 average photon spectral index 2.2, and isotropic upstream magnetic turbulence.There are two options to obtain the IXPE maximum measured value PD≃ 40%.One is to assume that in the upstream there is a net radial magnetic field.In this case, using an extension of the model of Bandiera & Petruk (Bandiera & Petruk 2024), one needs B perp /δB ≃ 1.4 − 1.5 in the upstream, implying δB/B 0 < 1, contrary to spectral model fitting and to expectations from acceleration theory.The other option is to assume that here we see mainly the effect of shock compression, whereas further downstream of the shock, as suggested by the comparison with the Chandra image in Fig. 3 (c), the turbulence has partially re-isotropized (maybe due to other instabilities acting to stretch the radial component).In this case, in order to reduce the PD to the measured values, a lower level of anisotropy with δB tan ∼ 2.5δB perp is required.As an explanation of the different turbulence levels among the SNRs seen by IXPE, Zhou et al. (2023) puts forward properties of the environment, in particular the ambient density.Indeed, the remnant with the smallest detected PD, Cas A, is evolving in a medium with average density of 0.9 ± 0.3 cm −3 (Lee et al. 2014).In the RX J1713 case, the inter-cloud density is instead ∼ 0.01 − 0.1 cm −3 (Cassam-Chenaï et al. 2004;Katsuda et al. 2015;Tsuji & Uchiyama 2016).SN 1006, which has a maximum PD comparable with that of RX J1713, is also in a rarefied environment with density ∼ 0.05−0.085cm −3 (Acero et al. 2007;Giuffrida et al. 2022).Therefore, our findings support the suggestion that high PD peaks are achieved in the presence of low ambient density.The ambient density affects also the maximum size of the Bell instability for the most rapid growth mode.According to Bell (2004), upstream turbulent magnetic fields can be generated by non-resonant streaming instability driven by CR current.We can deduce that this kind of instability is acting in RX J1713 because it can grow if δB >> B 0 , a condition possible only in presence of an isotropic turbulence upstream.Since the shock compression enhances only tangential components, every isotropic turbulence upstream becomes mainly tangential downstream.The length-scale of the Bell instability is given by (Bell 2004): where V s is the shock velocity (value ∼ 3900 km s −1 taken from Tsuji et al. ( 2021)), n 0 the ambient density (value 0.015 cm −3 from Tsuji & Uchiyama (2016)), E max the maximum energy and B 0 the unshocked magnetic field.At a distance of 1 kpc, this λ Bell value corresponds to ∼ 90 arcseconds, equivalent to three times the IXPE angular resolution and similar in size to the regions P1, P2, and P3.Because the smallest regions for which we detect significant polarization are comparable in scale with the Bell (2004) wavelength, we cannot claim that we see the effect of the streaming instability in the turbulence, and a consequent effect in the orientation of the magnetic field.

Bohm factor and age
The IXPE observation of RX J1713 also allows us to test two scenarios linking polarization with other source features: (1) the closer the Bohm factor, η, is to unity, the lower the PD; and, (2) younger SNRs have lower PD (shocks are faster and magnetic fields are more turbulent), whereas older remnants tend to have higher PD.The Bohm factor is defined as the ratio of the mean free path of the particle diffusion to the particle gyroradius, and it is connected to magnetic field turbulence in the quasi-linear region as with B 0 the initial background magnetic field and δB the turbulent field.The observed cutoff energy in the Xray synchrotron emission, ϵ 0 , corresponds to maximum particle energy in the acceleration region, which in turn is connected to the Bohm factor and shock speed v sh (Tsuji et al. 2021): Measurements of the cutoff energy and shock speed can thus provide estimates of the Bohm factor.The closer the Bohm factor to unity, the more turbulent B field on sub-pc scale is, and the lower the polarization would be.RX J1713 is recognized as an accelerator operating at the most efficient rate -i.e. the acceleration proceeds in a regime close to the Bohm limit of η ≃ 1; (Tanaka et al. 2008;Tsuji et al. 2019) -indeed Tsuji et al. (2021) estimate for this source η = 1.4 ± 0.3.Even with a Bohm factor close to unity, RX J1713 can achieve the observed PD only if the magnetic field is tangential to the shock front (e.g., Casse et al. 2001), as CR electrons can spread near the shock in an anisotropic way, with a much slower diffusion across than along the magnetic field lines.
The second scenario is also a possibility, as age estimates of RX J1713 (possibly associated with SNR 393) makes it the oldest SNR observed by IXPE to date.

Comparison with radio data
In Figure 3 (b) we show the RX J1713 ATCA 20 cm image -with a ∼ 70 arcsecond beam size -centered on the IXPE field of view with the magnetic field direction inferred by X-ray polarimetry and radio polarization measured by Lazendic et al. (2004) shown as overlays.Lazendic et al. (2004) detected significant linear polarization of 3 − 7% towards the western shell of RX J1713 in the region observed by IXPE, at 1.4 GHz and 2.4 GHz.The large Faraday rotation measure, RM > 100 rad m −2 (e.g.Lazendic et al. 2004) severely affects the PA at radio frequencies, and limits polarization measurements to patches of the brightest radio filaments.Consequently, they were unable to determine the intrinsic orientation of the magnetic field from available data.In contrast, polarization in the X-ray regime does not suffer from Faraday rotation.For this reason, the IXPE observation represents the first instance in which it has been possible to map the magnetic field in the NW region of RX J1713.As shown in Table 1, the X-ray PD in the considered regions is larger than the ∼ 5% radio value, as expected for the steeper X-ray index (∼ 2.2) with respect to the radio index (∼ 1.5), that allows for higher maximum PD (∼ 69% in the radio, ∼ 77% in the X-rays), and by the smaller volume sampled by the X-ray emission that is less prone to depolarization effects due to different magnetic field orientations along the line of sight (Vink & Zhou 2018).In a recent publication by Cotton et al. (2023), based on MeerKAT data, the radio polarization of the SNRs G4.8+6.2 and G7.7-3.7 raises potential questions pertaining to an age dependence on the magnetic field morphology.Both remnants show evidence for tangential magnetic fields similar to what we infer from IXPE measurements of RX J1713.X-ray observations of G7.7-3.7 (Zhou et al. 2018) indicate a relatively young age for the remnant based on the density and ionization timescale of the plasma.Based on these estimates and the sky position of the remnant, these authors suggest a possible association with SN 386.The remnant G4.8+6.2 is significantly above the Galactic plane and it could be a very-high-energy γ-ray source corresponding to a 4 σ hot spot in the H.E.S.S. significance map (H.E. S. S. Collaboration et al. 2022).Suggestions of a young age and possible association with year 1163 CE Korean "guest star" have been made (Liu et al. 2019).However, the estimated distance -which is highly uncertain -yields either an uncomfortably large remnant or a surprisingly low radio luminosity for such an interpretation.If the remnant is indeed associated with an event from 1163 CE, the tangential fields are in stark contrast to the radial fields observed for SN 1006 based on IXPE measurements.Finally, the 1738-year-old SNR 1E 0102.2-7219 in the Small Magellanic Cloud was found with ATCA to have a tangential magnetic field (Alsaberi et al. 2024).Therefore, there may be emerging evidence that the radial magnetic fields observed in the youngest SNRs give way to tangential magnetic fields in remnants with ages of only 1000-2000 years.

Comparison with gamma-ray data
The IXPE X-ray polarimetry of RX J1713 also provides new input for the debate on the leptonic versus hadronic nature of its γ-ray emission.In Figure 3 (d) we show the magnetic field direction inferred by X-ray polarimetry overimposed on the HESS >2 TeV image (H.E. S. S. Collaboration et al. 2018) of the IXPE field of view.The substantial coincidence of X-ray and γ-ray emission favors a leptonic origin of the latter (Aharonian et al. 2006;H. E. S. S. Collaboration et al. 2018): the same electron population could be the source of both synchrotron and likely inverse Compton emission.In Figure 3 (a) we show the 12 CO distribution compared to the magnetic field direction inferred by IXPE.Indeed, there are some 12 CO molecular clouds in the range of -6 to -16 km s −1 associated with RX J1713 (Sano et al. 2013), with a large uncertainty in their position along the line of sight, but the compressed magnetic field appears to be in correspondence of lower density regions.The hadronic scenario generally invokes dense clumps as the origin of the γ-ray emission (Gabici & Aharonian 2014;Fukui et al. 2021).However, if radial magnetic-field fields are indeed caused by RM instabilities triggered in a clumpy medium (Inoue et al. 2013b), then the tangential magnetic field in RX J1713 suggests that the medium in which RX J1713 evolves is a much less clumpy environment than in those young SNRS for which radial magnetic fields have been reported.This includes the environment of the high-Galactic latitude SN 1006.This argument is not incontrovertible, but the IXPE results on RX J1713 narrow down the possibilities for either explaining radial magnetic fields in young SNRs, or the hadronic γ-ray scenario for RX J1713.Given the interest in the theoretical explanation of the nature of the gamma-ray emission from RX J1713 it would be useful, in the future, to look with IXPE at the other RX J1713 regions to compare magnetic field direction in different regions with different surrounding medium.In particular, the northern spot -where there is less X-ray flux with respect to the γ one -or the south-west that mostly resembles the north-west region observed by IXPE.

CONCLUSIONS
We reported on the IXPE observation of the north western part of the SNR RX J1713 for ∼ 830 ks in the 2−5 keV energy band.We performed a spatiallyresolved search for polarization using both modeldependent and model-independent techniques.Our analysis shows that the polarization direction is normal to the shock and that the average PD in the whole region is 12.5 ± 3.3%.These results are consistent with a model of shock compression of an upstream isotropic turbulence producing a predominantly tangential magnetic field.The other remnants observed by IXPE -Cas A, Tycho, and the NE limb of SN 1006 -all showed radial magnetic fields, making RX J1713 the first example with shock-compressed magnetic fields.A deeper analysis of the nature of this important difference will be carried out in a future work.We estimated the magnetic field turbulence level based on the observed PD and photon spectral index.Our results suggests that either there is a net radial magnetic field in the upstream region, or the turbulence partially re-isotropized downstream.The latter requires a lower level of anisotropy to justify the PD measured in this remnant.We discussed two scenarios: (1) the closer the Bohm factor η to unity -as in RX J1713 -the lower the PD, and (2) younger SNRs have lower PD than older remnants.For the first scenario, the observed PD can be achieved only if the magnetic field is perpendicular to the shock normal, even with η ∼ 1.As for the second scenario, the possible association of RX J1713 with SNR 393 suggests that it is the oldest SNRs observed by IXPE, which may also influence its polarization characteristics.The maximum PD detected, 41.5 ± 9.5%, and the low ambient density, ∼ 0.01 − 0.1 cm −3 , in which RX J1713 is evolving also support the hypothesis that magnetic turbulence, and hence particle acceleration in SNRs, is environment-dependent.We compared the results in the X-ray band and other wavelengths, and found that the average X-ray PD is higher than the 3 − 7% one measured by ATCA in the radio band at 20 cm in the same regions.Because of the large uncertainties in the determination of radio polarization angle, the magnetic field morphology had been unknown.This IXPE measurement thus represents the first map of the magnetic field lines in RX J1713.Future radio polarimetric observation of RX J1713 with, for example, Meerkat would allow to compare the X-ray and radio morphology of the magnetic field.X-ray polarimetry of RX J1713 also provides new input for the debate on the leptonic versus hadronic nature of its γ-ray emission, supporting the former scenario.If radial magnetic-field fields are caused by Richtmeyer-Meshkov instabilities triggered in a clumpy mediumwhich is generally required by the hadronic scenario for the γ-ray emission -then the observed tangential magnetic field suggests that RX J1713 evolves in an environment where the hadronic γ-ray scenario is less favored.The upcoming IXPE observations of relatively co-age supernova remnants such as RCW 86 and Vela Jr. will further shed light on the complex panorama that spatially resolved X-ray polarimetry is unveiling, and allow for a systematic study of their polarimetric properties.Because RX J1713 is located on the Galactic plane, the Galactic Ridge X-ray Background (GRXB, see e.g.Ebisawa et al. 2005) may not be negligible.In order to estimate its value, we extract the spectrum from a Chandra observation (ObsID 10091) of RX J1713 in a region on a chip outside of the remnant.Then, we subtract the spectrum of the associated blank sky, and fit it with an APEC model with electron temperature (kT) fixed at 8 keV, following Katsuda et al. (2015); Tsuji et al. (2019).The estimated 2−5 keV GRXB flux is ∼ 5.9 ± 0.3 × 10 −13 erg cm −2 s −1 , about 15 times smaller than the ∼ 9.161 ± 0.002 × 10 −12 erg cm −2 s −1 source flux from the "All" region.

A.2. IXPE rejected instrumental background
We selected the two local background regions, called East and West (as shown in Figure 1) so to satisfy the following criteria: • encompass regions with surface brightness ≤ 0.45 counts per arcsecond 2 ; • No farther than 6 arcmin from center in order to not be contaminated by spurious edge effects • At least 45 arcseconds (i.e. about 1.5 times the IXPE angular resolution) from the "All" region in order to avoid contamination from the source polarization.
As shown in Table 2, both the individual background regions, and the combined E+W background is found to be unpolarized, with the most stringent limit being PD < 7.3% at 99% confidence level in the 2−5 keV energy band for the E+W background.The intrinsic PD can be obtained by subtracting the PCUBE of the background -after scaling it for the ratio of the regions size -from the PCUBE of the region of interest, thanks to the additive properties of the Stokes parameters, and the uncertainties on the Stokes parameters are linearly propagated to the polarimetric observables according to Kislat et al. (2015).
We also perform a spectropolarimetric study of the background, extracting the I, Q, and U spectra from the background regions East and West.We fit the spectra with a model comprising three components.The first is the GRXB: we model this component as an absorbed (Tbabs with nH = 8.15 × 10 21 cm −2 ) apec with temperature kT = 8 keV (Katsuda et al. 2015;Tsuji et al. 2019).The second component is an absorbed power-law describing the residual SNR emission: we fit this emission using spectra of the same background region on Chandra data (obsid 6370) and find the index (∼ 2.2) and relative flux (1.67 × 10 −12 erg cm −2 s −1 Vs. 1.24 × 10 −12 erg cm −2 s −1 in the 2−5 keV energy band, respectively) with respect to the aforementioned CRXB emission.The third, and more conspicuous, component is the unrejected IXPE instrumental background that we model empirically with an unabsorbed broken power-law.We fix the indexes to 2.3 and −0.378 and the energy break at 3.08 keV by simultaneously fitting the spectra of the background field of ∼ 1.5 Ms worth of publicly available IXPE observations of high Galactic latitude blazars and of other extragalactic point sources: PSR B0540-69 (Obsid 02001299), PG 1553+113 (02004999), IC 4329A (01003601), and 1ES 0229+200 (01006499).
We show the spectropolarimetric fit of the RX J1713 background region in Figure 6.As it was found studying the background region in a model-independent way, no significant polarization is measured for each component.The statistical significance of the fit is given by a χ 2 = 296.66 with 287 degrees of freedom.By comparing the backgroundsubtracted Stokes I spectrum for the "All" region, and the model of the background spectrum, the background flux becomes comparable to, or brighter than the source flux above ∼ 5 keV as shown in Figure 7.For this reason, we select only the 2−5 keV energy band to study the polarization properties.

Figure 1 .
Figure 1.Left: Three-color image of RX J1713: in red the ATCA 20cm image, in green 0.5−7 keV Chandra exposure-corrected mosaic, in blue HESS > 2 TeV excess count map.The magenta contour is the 2−5 keV IXPE emission and the white circle is the IXPE field of view.Right: IXPE count map in the 2−5 keV band smoothed with a 30 arcseconds Gaussian kernel.In purple are marked the large-scale sub-regions of the shell selected for analysis: North (N), South (S), and West (W).The white ellipses are small-scale sub-regions also selected for analysis: P1, P2, and P3.The colorbar represents the intensity per pixel.The background extraction regions are marked by dashed green lines.

Figure 2 .
Figure2.Left: polarization map binned with 2 arcmin-wide pixels.The cyan and green vectors represent the direction of the magnetic field revealed at 2 and 3 σ significance levels, respectively.The length of the vectors is proportional to the PD.The dashed vectors show the 2 σ uncertainty on the magnetic field direction in each pixel to improve the visibility.Right: smoothed polarization map with a 1 arcmin Gaussian kernel.As in the previous panel, the cyan and green vectors represent the direction of the magnetic field revealed at 2 and 3 σ significance levels, respectively.In both panels, we show the IXPE field of view as a white circle, the P1, P2, and P3 regions as white ellipses, while the 2−5 keV IXPE contours are in magenta and the colorbar is the PD in percent.

Figure 3 .Figure 4 .
Figure 3. (a) Mopra 12 CO image of RX J1713 integrated in the -6 to -16 km −1 velocity range.(b) ATCA 20cm image of the NW region of RX J1713: the grey contours are the radio polarization levels, each level corresponding to a 2% PD increase.(c) Chandra exposure-corrected mosaic of 0.5−7 keV images of the NW of RX J1713.(d) HESS >2 TeV excess count map of the NW of RX J1713.In all panels, we show the IXPE field of view as a white circle, the IXPE 2−5 keV contours in magenta, and the P1, P2, and P3 regions as solid white ellipses.In cyan and green we show the magnetic field lines, obtained through Gaussian smoothing of the IXPE data with > 2σ and > 3σ, respectively.

Figure 5 .
Figure 5. Spectropolarimetric fit of the Stokes I, Q, U spectra of the region "All" (data points) and the best-fit model (solid lines).In the bottom panels we show the residuals in terms of (data-model)/error.The Stokes I spectra from the three DUs observations are coded using different colors and shown in the left panel.The Stokes Q and U spectra are grouped for ease of visualization and shown in the right panel.Note that the normalization among the three DUs are known to fluctuate at low energies, but this does not affect the goodness of the fit.
The Imaging X-ray Polarimetry Explorer (IXPE) is a joint US and Italian mission.The US contribution is supported by the National Aeronautics and Space Administration (NASA) and led and managed by its Marshall Space Flight Center (MSFC), with industry partner Ball Aerospace (contract NNM15AA18C).The Italian contribution is supported by the Italian Space Agency (Agenzia Spaziale Italiana, ASI) through contract ASI-OHBI-2022-13-I.0,agreements ASI-INAF-2022-19-HH.0 and ASI-INFN-2017.13-H0, and its Space Science Data Center (SSDC) with agreements ASI-INAF-2022-14-HH.0 and ASI-INFN 2021-43-HH.0, and by the Istituto Nazionale di Astrofisica (INAF) and the Istituto Nazionale di Fisica Nucleare (INFN) in Italy.This research used data products provided by the IXPE Team (MSFC, SSDC, INAF, and INFN) and distributed with additional software tools by the High-Energy Astrophysics Science Archive Research Center (HEASARC), at NASA Goddard Space Flight Center (GSFC).E.Co., A.D.M., R.F., P.So., S.F., F.L.M., F.Mu. are partially supported by MAECI with grant CN24GR08 "GRBAXP: Guangxi-Rome Bilateral Agreement for X-ray Polarimetry in Astrophysics".C.-Y. Ng is supported by a GRF grant of the Hong Kong Government under HKU 17305419.This paper employs a list of Chandra datasets, obtained by the Chandra X-ray Observatory, contained in DOI: 10.25574/cdc.240.We thank Tony Bell, Rino Bandiera, and Damiano Caprioli for the helpful discussions.Facilities: IXPE, Chandra, ATCA, HESS, Mopra Software: CARTA (Comrie et al. 2021), CIAO (Fruscione et al. 2006), ixpeobssim (Baldini et al. 2022), XSPEC (Arnaud 1996) APPENDIX A. BACKGROUND TREATMENT A.1.Galactic X-ray background

Figure 6 .
Figure 6.Spectropolarimetric fit of the Stokes I, Q, U spectra of the background region (data points) and the best-fit model (solid lines).In the bottom panel we show the residuals in terms of (data-model)/error.The Stokes I spectra from the three DUs observations are coded using different colors and shown in the left panel.The Stokes Q and U spectra are instead grouped for ease of visualization and shown in the right panel.

Table 1 .
Lazendic et al. (2004)or the IXPE regions of interest with PCUBE and XSPEC, and ATCA radio observations.In the PD obs and PA obs columns are the observed values, while the ones in the PD and PA for both the IXPE PCUBE and IXPE XSPEC tables are the values after background subtraction.The radio PD fromLazendic et al. (2004)are the average polarization values in a given region plus/minus the standard deviation.

Table 2 .
Stokes parameters and surface area of the background regions in the 2−5 keV energy band.