X-ray and Radio Campaign of the Z-source GX 340+0 II: the X-ray polarization in the normal branch

IXPE conducted a $\approx$200 ks observation of GX 340$+$0 from 2024 August 12–16. The details of the observations are listed in Table 1. We used ixpeobssim software (v31.0.1) to bin the event data (Kislat et al., 2015; Baldini et al., 2022) into the high-level products (e.g., energy-dependent light curves, Stokes I, Q, and U spectra, and model-independent polarization properties with the PCUBE algorithm). The event data were calibrated and weighted according to the calibration files corresponding to recent calibration epoch of 2024 July 1 (obssim20240701_alpha075; Di Marco et al., 2022). For the extraction of the source photons, we adopted a 1.6$\arcmin$ circular region. Since the count rate of the source was higher than 2 c/s/arcmin² (see Figure 1), we do not perform any background rejection or subtraction (as suggested in Di Marco et al., 2023). For the model-independent analysis, the polarization information from individual DUs were combined and reported in Section 3.3.1. For the spectral analysis, we re-binned the I spectrum with the optimal binning scheme (Kaastra & Bleeker, 2016) and included a systematic error of 1% in the spectrum to account for differences in the residuals on application of identical model to each IXPE-DU spectra.

The Neutron star Interior Composition Explorer (NICER; Gendreau et al., 2016) is a non-imaging instrument onboard the International Space Station, comprising 56 silicon drift detector and X-ray concentrator optic pairs (52 operational) in focal plane modules (FPM). NICER observed GX 340$+$0 between 2024 August 12 and 2024 August 16 (see Table 1), within the IXPE observation interval. The raw data were reduced and processed using HEASOFT version 6.33 and the NICER Data Analysis Software (NICERDAS) version 12 (2024-02-09_V012) using calibration version xti20240206. We employed the following filtering criteria in generating the good time intervals: NICER being outside of the South Atlantic Anomaly; angular offset for the source of $<54\arcsec$; Earth limb elevation angle ELV $>20\arcdeg$; undershoot rate of 0–500 c/s/FPM; overshoot rate of 0-30 c/s/FPM, and a minimum cutoff rigidity such that COR_SAX $>$ 1.5. The processing was done using the threshfilter=DAY as the observations were taken during the orbital day and thus the lowest energy data may be affected. To investigate the branch-resolved behavior of the source, we first combined the observations using niobsmerge. These filtering criteria resulted in a total exposure of 15.3 ks for scientific analysis. We generated spectral products (including response matrices) using nicerl3-spect, where we grouped the spectra with the optimal binning scheme (Kaastra & Bleeker, 2016) and rebinned the spectra such that each bin had a minimum of 25 counts. We also generated light curves in various energy bands (the 0.5–10 keV light curve binned at 50 s is shown in Figure 1 in orange circles) using nicerl3-lc. Using the time intervals for which we identified the source to be in the different branches (see Section 3.1 for details on the state identification and Table 2 for the NB intervals), we constructed the spectrum from the merged event file for the NB intervals and extracted the appropriate response and auxiliary response files. For estimating the background, we made use of the SCORPEON model (Markwardt et al., 2024).

To supplement the IXPE observation with broad X-ray energy coverage, we conducted a ToO observation with AstroSat (Singh et al., 2014), running 2024 August 13–14. For the observation we used the Large Area X-ray Proportional Counter (LAXPC; Yadav et al., 2016, 2017) as the primary instrument. We procured the level 1 data of the LAXPC observation from AstroBrowse¹¹1https://astrobrowse.issdc.gov.in/astro_archive/archive/Home.jsp and reduced the data using relevant tools in LAXPCsoftware22Aug15²²2http://astrosat-ssc.iucaa.in/uploads/laxpc/LAXPCsoftware22Aug15.zip (Antia et al., 2021; Misra et al., 2021). The pipeline also includes the tools to filter the South Atlantic Anomaly passage and Earth occultation intervals and extract the energy-dependent light curves, spectra, background spectra, and background light curves. For the analysis, we considered only LAXPC 20 as LAXPC 30 was turned off early in the mission and LAXPC 10 indicated abnormal gain variation. To minimize the background contribution over 3-20 keV, we used the layer 1 data for the extraction of higher-level products (e.g., light curves and spectra). We show the scaled light curve (binned at 50 s) from the LAXPC observation in Figure 1 as green crosses. The spectrum from LAXPC was rebinned according to the detector response Due to response uncertainties, we included a 2% systematic error in the spectrum across the complete energy range (3–20 keV).

Insight-HXMT (Zhang et al., 2020) observed GX 340$+$0 from 2024 August 15, 00:44:17 to 2024 August 16, 08:28:41. The data were extracted from all three instruments using the Insight-HXMT Data Analysis software (HXMTDAS) v2.06³³3The data analysis software is available from http://hxmten.ihep.ac.cn/software.jhtml., and filtered with the following standard criteria: (1) pointing offset angle less than $0.04^{\circ}$; (2) Earth elevation angle larger than $10^{\circ}$; (3) the value of the geomagnetic cutoff rigidity larger than 8 GV; (4) at least 300 s before and after the South Atlantic Anomaly passage. To avoid possible contamination from the bright Earth and nearby sources, we only use data from the small field of view (FoV) detectors (Chen et al., 2018). The software also provides the relevant response and background files for the spectral analysis. The HE data are not used in our spectral analysis as the HE spectra were dominated by the background. The energy bands adopted for spectral analysis are 2–10 keV for LE and 10–20 keV for ME. The scaled light curves from Insight-HXMT-LE in 3–10 keV and binned at 50 s are shown in figure 1 as purple circles.

We observed GX 340$+$0 with the Giant Metrewave Radio Telescope (GMRT). Utilizing the wideband receiver backend of GMRT, data were recorded in two frequency bands: band 4 (central frequency 750 MHz, bandwidth 400 MHz) and band 5 (central frequency 1260 MHz, bandwidth 400 MHz) on 2024 August 12 and 2024 August 14, respectively (Proposal ddtC367, for simultaneous coverage with the IXPE campaign). The raw data were downloaded in the FITS format and converted to the Common Astronomy Software Applications (CASA; CASA Team et al., 2022) measurement set format. The data were calibrated (3C286 was used for flux and bandpass calibration, and J1717-398 was used for phase calibration) and imaged using the automated continuum imaging pipeline CASA-CAPTURE (Kale & Ishwara-Chandra, 2021). Manual flagging and calibration were performed as required. Eight rounds of self-calibration (both phase and amplitudes) were completed within each pipeline run to obtain the final cleaned images. No sources were detected at the position of GX 340$+$0 in the band 4 observations, providing a 3$\sigma$ upper limit of 1.1 mJy at 750 MHz, where we use the root mean square (RMS) of the noise from a large area in a source-less region of the image (close to the source position) as $\sigma$. The band 5 data were heavily affected by radio frequency interference (RFI), but the image shows a tentative point source close to the known position of GX 340$+$0 with a flux density, $S_{\nu}$, of 4.5$\pm$0.7 mJy at 1.26 GHz (10% flux density error added in quadrature to the RMS in a large source-less region to obtain the error in the maximum).

The Australia Telescope Compact Array (ATCA) observed GX 340$+$0 on 2024 August 15 between 13:43:40 and 15:49:40 UT. During the observation ATCA was in a relatively extended 1.5 km configuration⁴⁴4https://www.narrabri.atnf.csiro.au/operations/array_configurations/configurations.html. Data were recorded simultaneously at central frequencies of 5.5 and 9 GHz, with 2 GHz of bandwidth at each frequency band. We used PKS B1934$-$638 for flux and bandpass calibration, and the nearby calibrator J1631$-$4345 for phase calibration. Data were flagged, calibrated, and imaged using standard procedures within the (CASA version 5.3.1; CASA Team et al. 2022). Imaging used a Briggs robust parameter of 0, balancing sensitivity and resolution. Fitting for a point source in the image plane, GX 340$+$0 was detected at both central frequencies, where $S_{\nu}=0.70\pm 0.05$ mJy at 5.5 GHz and $0.59\pm 0.05$ mJy at 9 GHz. This corresponds to a radio spectral index, $\alpha$, of $-0.35\pm 0.25$ (defined as $S_{\nu}\propto\nu^{\alpha}$, where $\nu$ is the frequency). Such a radio spectral index could be consistent with a steep spectrum from transient ejecta, or a flat radio spectrum from a steady jet.

To explore intra-observational variability, we also imaged the source on 20-minute timescales. We found that at both central frequencies the flux density decreased for the first $\sim$half of the observation, before remaining stable during the second half. At 5.5 GHz the flux density decreased from $1.07\pm 0.15$ mJy at the start of the observation to $0.44\pm 0.10$ mJy by the middle of the observation. During the second half of the observation, the 5.5 GHz flux density then remained stable (within errors), measuring $0.50\pm 0.10$ mJy at the end of the observation. At 9 GHz, the flux density was measured to be $0.82\pm 0.08$ mJy at the start, dropping to $0.32\pm 0.06$ mJy by the middle of the observation, remaining stable (within errors) to the end of the observation $0.34\pm 0.10$ mJy.

Previous IXPE observations of Z-sources have indicated that the X-ray polarization from these systems is dependent on the branch position. Thus, we need to identify the source state throughout the observation to interpret the nature of X-ray polarization and therefore the emission components. We constructed the HID using all the X-ray instruments using different definitions of hardness (typically limited by the energy range from the instruments) and have shown them in Figure 2.

From the HID, GX 340$+$0 traced the complete Z-track during the observation. The HID from NICER (count rate in 3–10 keV and HR as the ratio between 5–10 and 3–5 keV bands) can identify the epochs when the source was in the HB and the NB while the HID from AstroSat-LAXPC (count rate in 3–20 keV and HR as the ratio between 10–20 and 6–10 keV bands) highlights the track from the hard apex to the EFB. The Insight-HXMT observation (HID constructed using the LE and ME light curves in the 6–10 and 10–20 keV bands, respectively) were mainly in the NB but for some duration the source seemed to indicate an excursion in the FB (as seen below the HR of 0.8 in bottom right panel of figure 2).

We note that the FB and the EFB are not identifiable in the HID created using the soft X-ray bands (i.e., using the hardness ratio from energy bands in 3–10 keV). We construct the HID from the AstroSat-LAXPC observation in the energy bands used for NICER, i.e., 3–10 keV for the count rate and HR between the 5–10 and 3–5 keV bands (shown in figure 3). We observe that a single track similar to the NB is visible in the HID. The points corresponding to the FB and the EFB are identified from Figure 2 and marked as black circles in Figure 3. Additionally, the AstroSat observation shows a significant overlap with NICER observation during one of the FB excursions. The light curves from both observations seem to indicate the presence of a flare (during the FB) but the HR between the 5–10 and 3–5 keV bands was similar for both the FB and NB. Therefore it is challenging to accurately identify any FB intervals solely based on data from the soft X-ray instruments used here (i.e., NICER or IXPE). For this study, we identified FB intervals using the AstroSat-LAXPC and Insight-HXMT HID and extrapolate the intervals to the data gaps (if present) using the fact that Z-sources continuously move along the Z-track without jumping between branches (e.g. Church et al., 2006; Homan et al., 2007; Sriram et al., 2011; Lin et al., 2012, B23). A possible caveat in the state identification is that during the intervals where the source was covered by only NICER and IXPE, there may be some FB/EFB intervals which have been misclassified as NB. The similarity of the FB/EFB and NB in the HID in the soft X-ray emission (e.g., Figure 3) would suggest a similarity in the spectrum of these two branches (at least below $\sim$10 keV). Since the states seem to be indistinguishable in the HID, we assume that the inclusion of these segments during the NB doesn’t affect the spectral or polarimetric results significantly.

To understand various spectral components from the GX 340$+$0 during the NB, we extracted the spectra for all instruments using the methods described in their respective sections (Sections 2.1-2.4) for the intervals corresponding to the NB (Table 2). During the epochs of the AstroSat and Insight-HXMT observations, we excluded any FB/EFB intervals as determined in section 3.1. We conducted the spectral and spectro-polarimetric modeling in xspec v12.14.1. For fitting of the spectra, we used PGstat for NICER spectra (as the background estimation for NICER observations is done by SCORPEON modeling and results in non-Poissonian background) and $\chi^{2}$ for rest of the spectra. We used $\chi^{2}$ as the test statistic and report $1\sigma$ confidence intervals for the estimated parameters.

Since the focus of the analysis is to identify the spectral components responsible for the X-ray polarization, we utilized the joint coverage of IXPE and NICER to identify the spectral components. We also included the spectrum from AstroSat-LAXPC as it overlapped with NICER and IXPE and allows us to extend the spectral range to 20 keV, placing a strong constraint on the hard X-ray emission, primarily the non-thermal component. At the softest energies ($<$0.6 keV), the NICER spectrum was dominated by the noise peak due to additional optical loading from the orbital day. Thus we ignored the NICER spectrum below 0.6 keV. In case of spectra from other instruments, we excluded the energy intervals where the background counts were higher than the source counts. We note that inclusion of Insight-HXMT spectra in the joint fit resulted in bad fit for all instruments. A better fit was obtained if we allowed one set of parameters for IXPE, NICER, and AstroSat-LAXPC and one set of parameters for Insight-HXMT, which could be due to calibration mismatch between these instruments. Since the inclusion of Insight-HXMT spectra did not provide any additional information over joint IXPE, NICER, and AstroSat-LAXPC spectra, we report the analysis of the latter set of spectra.

Noting a high absorption column modeled in the literature of the source (e.g. Church et al., 2006; Iaria et al., 2006), we modeled the absorption column with tbabs and used the abundances from Wilms et al. (2000) and cross-sections from Verner et al. (1996). Initially, we modeled spectra from individual observatories to identify the minimum number of additive components required to sufficiently model them. We fit the spectra with tbabs*nthcomp which assumes that the non-thermal component arises from a Comptonizing medium (Zdziarski et al., 1996; Życki et al., 1999), which is up-scattering photons from an accretion disk. We observed that this combination resulted in an unacceptable fit statistic ($\chi^{2}$/degrees of freedom or d.o.f.=2092/202). The residuals clearly indicated the presence of a softer component, which we modeled with a blackbody (bbodyrad). The inclusion of a blackbody component resulted in a lower test statistic ($\chi^{2}$/d.o.f.=1032/200) but the fit was still not acceptable. The NICER and IXPE spectra indicated the presence of iron line emission that was modeled with a gaussian. For the AstroSat-LAXPC data, the normalization of the gaussian component was observed to be lower than other instruments (which could be due to calibration mismatch). The resulting $\chi^{2}$ of 435.20 for 196 d.o.f. was still not acceptable, where the residuals indicated the presence of a softer component and possibly an edge at 1.8 keV. We described this component as an accretion disk modeled with a soft multicolor blackbody (i.e., diskbb) and tied the inner disk temperature to the seed photon temperature for the nthcomp component ($\chi^{2}$/d.o.f.=168/195). After the inclusion of diskbb, the residuals at the highest energies were accounted for as now the Comptonized emission could describe the emission at higher energies without the need to compensate at softer energies⁵⁵5We note that the current spectral model is similar to the Western model with an addition of a soft and truncated accretion disk to explain the excess in the softer energies ($\lesssim 2$ keV). This particular combination of models has also been tested in previous spectral investigations of GX 340$+$0 by B23, where a comparison to the Eastern model counterpart (i.e., softest thermal component is a blackbody while component $>1$ keV is the accretion disk) is discussed extensively.. We note that although the reduced $\chi^{2}$ is already acceptable at this stage, the fitting statistic of the NICER spectra could be further reduced by modeling the edge-like residuals with a multiplicative component edge. After the inclusion of the edge component, the test statistic of $\chi^{2}$/d.o.f.=151.8/193 is acceptable (as compared to the previous case of model without the edge using either Akaike Information Criterion (AIC) or Bayesian Information Criterion (BIC)). The spectral parameters are noted in Table 3 To compute the confidence intervals for various spectral parameters, we sampled the parameter space using Markov Chain Monte Carlo sampler in xspec. We used the Goodman-Weare algorithm and initialized 30 walkers close to the best fit parameters and ran for a total 9000 steps after a 1980 burn-in steps. The confidence intervals were determined from the marginalized distribution of the sampled parameter space. We also show the count spectra from various instruments, the unfolded model and residuals from all the spectra at various stages of spectral modeling in Figure 4.

Using the NB intervals as defined in Section 3.1, we investigated the polarization properties in a model independent manner (see Section 3.3.1), and using the spectral decomposition from Section 3.2 we investigate the spectro-polarimetric properties of GX 340$+$0 in Section 3.3.2.

GX 340$+$0 has been observed over two epochs with IXPE: once in 2024 March where it was observed to be in the HB (B24, La Monaca et al., 2024a) and again in 2024 August where it showed a complete tracing of the Z-track, but spending the majority of the time in the NB. In this work we have investigated the polarimetric properties of the source in the NB using the 2024 August observation (see B24 for a detailed discussion of the 2024 March observations). The observed 2–8 keV PD during the NB ($1.22\pm 0.25$%, see Table 4) is lower than the PD measured during the HB (4%; B24, ), but the PAs are consistent across both branches. The similarity in the PA indicates that the emission component in the hard X-rays is likely similar across both branches. Assuming the polarization is from a single component, i.e., a Comptonizing medium in both cases, would suggest that the emission has depolarized as the source moves from HB ($9$%; B24, ) to NB (2.4%; Table 4). A similar decrease in net polarization was observed in simulations by Gnarini et al. (2022), where the hard state polarization (similar to the HB in Z-sources) was observed to be higher than the soft state polarization (similar to the NB/FB/EFB in Z-sources). We also estimated the PD and PA for the HB and the FB/EFB segments in the present observation but the duration was not sufficient to detect significant polarization, with an MDP99 of 1.54% and 3.28% respectively. The HB upper limit was lower than the reported value in B24 or La Monaca et al. (2024a), which could be due to inclusion of the hard apex in the current analysis while the source stayed completely in the HB during the IXPE observation in 2024 March.

In the softer energies, the X-ray polarization was not significantly detected and the discrepancy in the angle reported by B24 and La Monaca et al. (2024a) during the HB was not constrained in the NB. According to B24, the accretion disk emission in 2–2.5 keV was the primary contributor towards the PA difference. In the spectral modeling of the NB data (Section 3.2), we note that the disk appeared more truncated than during the HB (e.g. B23, ) with a lower contribution in the 2–8 keV band (the flux from diskbb component in computed using the parameters in B23 is 5 times higher than disk flux using parameters from Table 3) Therefore, a non-detection of the PA discrepancy is consistent with the spectro-polarimetric decomposition suggested by B24.

The X-ray polarization during the NB in GX 340$+$0 is similar to the other Z-sources, e.g., Cyg X-2: $1.85\pm 0.87\%$ (Farinelli et al., 2023), XTE J1701$-$462: $<1.5\%$ (at 99% confidence; Cocchi et al., 2023), GX 5$-$1: $2.0\pm 0.3\%$ (Fabiani et al., 2024), Sco X-1: $1.0\pm 0.2\%$ (La Monaca et al., 2024b) and Cir X-1: $1.3\pm 0.2\%$ (Rankin et al., 2024). Out of these sources, only XTE J1701$-$462 and GX 5$-$1 have been observed in the HB as well (which also match the GX 340$+$0 HB polarization Cocchi et al., 2023; Fabiani et al., 2024, B24). The similarity of polarization levels in these sources over different states, along with a similarity in timing signatures (e.g. Hasinger & van der Klis, 1989), hint at a similarity in the geometry of emission components. In most of these cases, an increasing trend of PD is observed with energy, reinforcing the slab-like geometry for the corona (if the polarized emission originates from the Comptonized medium Gnarini et al., 2022).

During the first GMRT observation on 2024 August 12, GX 340$+$0 was in the NB. No radio emission was detected at 700 MHz, with a 3-sigma upper-limit of 1.1 mJy. During the second GMRT observation 2 days later (on 2024 August 14), GX 340$+$0 had transitioned to the HB branch and the 1.26 GHz observations tentatively detected GX 340$+$0 with a flux density of 4.5$\pm$0.7 mJy. At the start of the ATCA observations the next day (2024 August 15), the target was initially within the FB, transitioning back to the NB during the observation (Figure 1). The ATCA observations detected radio emission from GX 340$+$0 with a flux density of $0.70\pm 0.05$ mJy at 5.5 GHz and $0.59\pm 0.05$ mJy at 9 GHz. The measured flux densities could be consistent with emission from both a compact or transient jet, where the errors in the radio spectrum do not allow us to discriminate between the two. However, imaging the source on shorter timescales suggests that the radio emission was found to be fading at both frequencies during the first half of the ATCA observation (by a factor of 2–3) before remaining steady for the second half. Z-sources typically display fading radio emission as the source moves from the HB to the FB, with radio flares detected around the transition between the NB and HB, indicating the launching of ejecta (e.g., Fender et al., 2004a; Motta & Fender, 2019). Our results fit into this picture, where ejecta are expected to be launched over the HB$\rightarrow$NB transition, which then fade as the source moves towards the FB. We propose that the ATCA observations detected the fading of the flaring associated with this transition, before stabilizing as the source started to move back from the FB to the NB (when we expect it to begin to brighten; e.g., Bradshaw et al. 2003; Migliari et al. 2007; Bałucińska-Church et al. 2010; Church et al. 2012).

The results from the GMRT observations also fit the general picture of radio behavior of Z-sources; the non-detection during NB occurs due to a spectral turnover of the synchrotron emission at $>$700 MHz, likely arising from free-free or synchrotron self-absorption (see discussions in B24, ). Z-sources are at their radio brightest during the HB, where the emission likely arises from a steady jet (e.g., Migliari et al., 2007). As such, the tentative detection at 1.26 GHz on 2024 August 15 as the source moves along the HB could be radio emission from a bright, steady jet.

The spectral analysis of GX 340$+$0 suggests that the continuum emission is comprised of three components: an accretion disk, a blackbody, and Comptonized emission ( B23, and table 3). With recent IXPE observations, we can shed some light on the contribution to the polarized emission from these components. The energy dependence of the polarization in the HB suggests that at least two components are contributing to the polarization and that they may have a polarization angle discrepancy between them (B24, La Monaca et al. 2024a). In the NB, however, the X-ray polarization can be explained by a single component whose PA aligns with the PA of the component in the 2.5–8 keV band in the HB (see table 4 and B24, ). This suggests that in the HB and NB, the same component is providing the polarized emission while there might be another contributing component in the HB which is not significant in the NB. One of the components that could produce this is a low temperature accretion disk. In the HB, the disk emission has a significant contribution in the 2–2.5 keV band (see Figure 2 of B23) while in the NB, the disk emission was significantly lower (see Figure 4).

The simulations of weakly magnetized NS LMXBs (Gnarini et al., 2022) suggest that typical polarization values in HB and NB (which would correspond to hard and soft states in Gnarini et al. 2022) match with a slab like-corona. But the simulations also suggest that the PA variation is drastically different for both states. The difference in the PA is due to the assumption that the disk is polarized, and has a significant contribution in the 2–8 keV. If the thermal component is unpolarized (e.g. a blackbody component from the surface), then the PA dependence on energy will be caused by a single component (i.e. Comptonized emission). In case of GX 340$+$0, we find a similar PA in HB and NB which would support this assumption and motivate simulations with different geometrical parameters to explain the observations. Alternatively, the polarized emission can also arise from a broad SL. But the emission from SL is expected to have a strong variation of PA as a function of energy and the expected PD of the SL would be typically 1.5% or lower (Bobrikova et al., 2024) and since our observations have shown constant PA as a function of energy (at least greater than 3 keV for both the HB and NB), and a higher PD in HB we can rule out SL as a possible origin for polarized emission. Another possibility is that the polarized emission is arising from the reflection component. La Monaca et al. (2024a) have explored this for the HB observation of GX 340$+$0 and suggested that it is difficult to disentangle if the observed PD is arising from solely reflection component or solely Comptonized emission. Similar test for the NB observation (Table 4) suggest a similar degeneracy is present in the NB as well. If the polarization is arising from solely reflection then it has a value of 29% in HB (La Monaca et al., 2024a) and 12% in NB.