Suzaku X-ray Observations of the Fermi Bubbles:
Northernmost Cap and Southeast Claw Discovered with MAXI-SSC

The MAXI mission is the first astronomical payload installed on the Japanese Experiment Module - Exposed Facility (JEM-EF or Kibo-EF) on the International Space Station (ISS). MAXI scans almost the entire (90%) sky twice in every ISS orbit ($\sim$ 90 minutes). MAXI carries two detectors – the Gas Slit Camera (GSC; Mihara et al. 2011) and the aforementioned SSC. Given the spectral resolution of its X-ray CCD, all-sky X-ray images in the energy range of 0.5$-$12 keV can be generated with the MAXI-SSC with a field of view of 1.5${}^{\circ }$ $\times$ 90${}^{\circ }$. The SSC is one of the best instruments to study large diffuse X-ray structures as recently demonstrated in a study of the Cygnus superbubble (Kimura et al. 2013).

While a detailed X-ray spectral study of the Fermi bubbles using the MAXI-SSC data is ongoing, all-sky maps in various X-ray energy bands have been prepared for various uses. Figure 1 shows the X-ray sky maps obtained with the MAXI-SSC taken from 2009 August 18 to 2012 February 01 in the Low-band (L-band = 0.7$-$1.7 keV: $top$-$left$) and Mid-band (M-band = 1.7$-$4.0 keV: $top$-$right$) (Kimura et al. 2013). These maps are exposure-time corrected, while no background is subtracted. Whereas the global structure seen in the L-band image is essentially consistent with the ROSAT all-sky maps taken at 0.75 and 1.5 keV (Snowden et al. 1995), the M-band image shows prominent structures which are unseen in the L-band image. Specifically, we found a “cap” or clump-like structure appearing at the northernmost part of the Fermi bubble at high Galactic latitudes 45${}^{\circ }$ $<$ $b$ $<$ 55${}^{\circ }$ (hereafter, N-cap). Also there appears to be a NPS-like feature visible to the southeast in the L-band image. A “claw” is seen at low Galactic latitudes $-$20${}^{\circ }$ $<$ $b$ $<$$-$10${}^{\circ }$ (hereafter SE-claw) in the L-band image, which is also seen in the $ROSAT$ 0.75 keV map. These new all-sky maps from MAXI-SSC motivated us to obtain follow-up observations using Suzaku as detailed below. Suzaku is suitable for investigating such faint diffuse X-ray emission because of its low instrumental background and superior energy resolution especially below 1 keV.

As a part of an AO8 program in 2013, we carried out four Suzaku observations of $\simeq$ 20 ksec each, pointed “on” and “off” the (i) N-cap and (ii) SE-claw regions (Fig. 1) with a total (requested) duration of 80 ksec. The Suzaku observation logs and pointing positions are summarized in Table 1. The Suzaku satellite (Mitsuda et al. 2007) is equipped with four X-ray telescopes (XRT; Serlemitsos et al. 2007), each carrying a focal-plane X-ray CCD camera (X-ray Imaging Spectrometer, XIS; Koyama et al. 2007). One of the XIS sensors is a back-illuminated (BI) CCD (XIS1), and the other three XIS sensors are front-illuminated (FI) ones (XIS0, XIS2, and XIS3). Because the operation of XIS2 ceased in 2006 November due to contamination by a leaked charge, we use only the other three CCDs. The XIS was operated in the normal full-frame clocking mode with the 3 $\times$ 3 or 5 $\times$ 5 editing mode. Although Suzaku also carries a hard X-ray detector (Takahashi et al. 2007), we do not use the data collected by its PIN and GSO instruments because the thermal emission we describe below is too faint to be detected at 10 keV and no statistically significant excesses over the CXB were found with these PIN/GSO detectors.

The lower panels of Figure 1 show our Suzaku XIS fields of view (FOVs; $\sim$ 18’ $\times$ 18’) onto the MAXI-SSC image for the series of northernmost and southeast observations of the bubbles. Green and white circles denote our previous AO7 pointings as published in Paper-I and our newly obtained AO8 pointings, respectively, while yellow dashed curves indicate the boundary of the Fermi Bubbles as drawn by Su et al. (2010). Note that in the N-cap region, there are some regions with previous Suzaku pointings and the corresponding focal centers of these observations are denoted in the figure by cyan circles; the analysis of these archival data are discussed below.

We conducted all data reduction and analysis with HEADAS software version 6.14 using the calibration database (CALDB) released on 2013 August 13. First, we combined the cleaned event data of the two editing modes using XSELECT. The data corresponding to epochs of low-Earth elevation angles (less than 20${}^{\circ }$ during both night and day) were then excised, as well periods when the Suzaku satellite was passing through (and 500 sec after) the South Atlantic Anomaly. Moreover, we also excluded the data obtained when the Suzaku satellite was passing through the low Cut-Off Rigidity (COR) of below 6 GV. Finally, hot and flickering pixels were removed using SISCLEAN (Day et al. 1998). With these data selection criteria applied, the resulting total effective exposures for all the observed pointings are summarized in Table 1.

Because the BI CCD (XIS1) has higher instrumental background than those of FI CCDs, we extracted the XIS images from only the two FI CCDs (XIS0, XIS3). We made the XIS images in three photon energy ranges of 0.4$-$10 keV, 0.4$-$2 keV, and 2$-$10 keV. In the image analysis, calibration sources located at the corners of the CCD chips were excluded. The images of Non X-ray Background (NXB) were obtained from the night Earth data using XISNXBGEN (Tawa et al. 2008). Since the exposure times for the data toward the N-cap and SE-claw (original data) were different from that of the NXB, we corrected for the differences in the exposure times using XISEXPMAPGEN (Ishisaki et al. 2007) before subtracting them from the original images. In addition, we implemented a vignetting correction by simulating flat sky images with XISSIM (Ishisaki et al. 2007). The XIS0 and XIS3 images were then combined, re-binned by a factor of 4 (i.e., 4 pixels were combined to 1), then smoothed by a Gaussian function with $\sigma$ = 0.07’ (Gaussian kernel radius set as 8 in DS9), and the resultant images are presented in section 3.

With the combined XIS0+3 images, we first ran the source detection algorithm in XIMAGE (Giommi et al. 1992) to differentiate compact X-ray features from diffuse X-ray emission. Through this approach, bright compact emitters which are not associated with catalogued sources were detected (all at approximately $\ge 3\sigma$).

We used all the (FI and BI) CCDs, namely, XIS0, 1, 3 for the spectral analysis to increase the photon statistics. Regarding the bright compact X-ray features described above, we defined the source regions as 2’ radius circles centered on the emission peaks, and the corresponding background regions were set to annuli on the same CCD chips with inside and outside radii of 2’ and 4’, respectively. We made redistribution matrix files (RMFs) and auxiliary response files (ARFs) using XISRMFGEN and XISSIMARFGEN (Ishisaki et al. 2007), respectively. For the diffuse emission analysis, we set the source region to the whole CCD chip which remained after excluding all the compact features detected at significance levels above 3$\sigma$, with typical 2’ radius circles which is sufficient to avoid contamination from the compact sources. Using XISSIMARFGEN and new contamination files (released on 2013 August 13), the ARF files were created assuming the uniform extension of the diffuse emission within 20’ radii orbicular regions (giving the ARF area of 0.35 deg${}^2$). As opposed to the point source analysis, we subtracted as background, the NXB data obtained from the region in the same CCD chip.

We note that even though the most recent version of XISSIMARFGEN replicates well the deterioration of the CCD quantum efficiency below 2 keV, significant differences in the spectral shape of the measured emission continua between XIS 0, 1, and 3 were found below $\simeq$ 0.5 keV due to contamination¹¹1https://heasarc.gsfc.nasa.gov/docs/astroe/prop_tools/suzaku_td/node10.html. Hence we ignored the spectral data below 0.6 keV for XIS 0 and 3, and below 0.5 keV for XIS 1. We also did not use the high energy data above 8 keV for all CCDs due to low photon statistics. Although we constructed light curves for bright X-ray compact features, none of them suggested any significant variability within the short exposures obtained.

Figure 2 shows the obtained 0.4$-$10 keV Suzaku-XIS images (combined XIS0 and 3) for the N-cap field (N-cap_on/off) and the SE-claw field (SE_on/off). Note the bright features seen in the images near the corners of the CCD chips are spurious sources resulting from instrumental artifacts, which were excluded in the spectral analysis. Although the pointing positions are only $\simeq$2${}^{\circ }$ apart in the SE-claw, we can see that the surface brightness of the inner bubble (SE_on) is much higher than that of the SE_off, with ON-OFF flux differences being as large as $\sim$ 10$\sigma$. In contrast, such large variations in surface brightness are not clearly seen in the N-cap regions.

In the figures, magenta crosses show the positions of compact X-ray features detected at $\ge 3\sigma$ level which are probably associated with catalogued background sources from the NED and SIMBAD databases. The other compact features detected here with Suzaku at high significances that miss any identifications in these databases are named “src1$-$5”and indicated as red circles in Figure 2. Given the limited photon statistics and the relatively large PSF of the XIS, it is difficult to determine whether the detected compact features are point-like or extended. In this context, it is important to estimate the expected number of background active galactic nuclei (AGN) likely to be found within each XIS FOV. Following Stawarz et al. (2013), we estimate that two or three uncatalogued AGN are present within each XIS FOV, corresponding to the number of detected compact X-ray features without any optical identification in the present data. In order to investigate the possible associations with background AGN, we looked into the available mid-infrared (MIR) maps of the areas covered by our Suzaku exposures using the NASA Widefield Infrared Survey Explorer (WISE; Wright et al. 2010) satellite data. As a result, we found just one WISE counterpart (indicated as a black cross in Figure 2) which was thus excluded in the spectral analysis described below.