XMM-Newton Measurement of the Galactic Halo X-ray Emission using a Compact Shadowing Cloud
David B. Henley, Robin L. Shelton, Renata S. Cumbee, Phillip C. Stancil
aa r X i v : . [ a s t r o - ph . GA ] N ov Draft version November 8, 2017
Preprint typeset using L A TEX style emulateapj v. 05/12/14
XMM-NEWTON
MEASUREMENT OF THE GALACTIC HALO X-RAY EMISSION USING A COMPACTSHADOWING CLOUD
David B. Henley, Robin L. Shelton, Renata S. Cumbee, and Phillip C. Stancil
Department of Physics and Astronomy, University of Georgia, Athens, GA 30602; [email protected]
Draft version November 8, 2017
ABSTRACTObservations of interstellar clouds that cast shadows in the soft X-ray background can be used toseparate the background Galactic halo emission from the local emission due to solar wind chargeexchange (SWCX) and/or the Local Bubble (LB). We present an
XMM-Newton observation of ashadowing cloud, G225.60 − T h ≈ × K,emission measure E h ≈ × − cm − pc) were not sensitive to the foreground model used. This islikely due to the relative faintness of the foreground emission in this observation. However, the datado favor the existence of a foreground. The halo parameters derived from this observation are in goodagreement with those from previous shadowing observations, and from an XMM-Newton survey ofthe Galactic halo emission. This supports the conclusion that the latter results are not subject tosystematic errors, and can confidently be used to test models of the halo emission.
Keywords:
Galaxy: halo — ISM: clouds — ISM: individual objects (G225.60 − INTRODUCTION
An important result from
ROSAT was the discov-ery of shadows in the soft X-ray background (SXRB),caused by interstellar clouds partially blocking thedistant X-ray emission (Burrows & Mendenhall 1991;Snowden et al. 1991). Analysis of such shadows showedthat hot, X-ray-emitting plasma exists in the halo ofour Galaxy (e.g., Wang & Yu 1995; Kuntz et al. 1997;Snowden et al. 2000). By comparing the X-ray emissionobserved toward and to the side of a shadowing cloud,one can separate the hot halo emission from the fore-ground emission, attributable to hot gas in the LocalBubble (LB; Sanders et al. 1977; Snowden et al. 1990),charge exchange (CX) reactions between solar wind ionsand neutral H and He in the heliosphere and the Earth’sexosphere (Cravens 2000; Robertson & Cravens 2003a,b;Koutroumpa et al. 2006), or a combination of the two(Smith et al. 2014; Galeazzi et al. 2014). Separating theforeground and halo emission is necessary to test modelsfor the foreground emission, and for the origin of the hothalo plasma.More recently,
XMM-Newton and
Suzaku observa-tions of shadowing clouds have been used to constrainthe hot halo emission. These satellites’ CCD cam-eras have higher spectral resolution than
ROSAT ’s pro-portional counter. Such studies obtained halo tem-peratures and emission measures of ∼ × K and ∼ (3–12) × − cm − pc, respectively (Galeazzi et al.2007; Smith et al. 2007; Gupta et al. 2009; Lei et al.2009). However, whereas ROSAT ’s large field of view( ∼ ◦ ) meant that a shadowing cloud and the adjacent off-cloud sky could be observed in a single pointing, XMM-Newton and
Suzaku ’s smaller fields of view ( ∼ . ◦ ∼ . ◦
3, respectively) required that the above-cited shad- owing observations consist of two separate pointings—one toward and one to the side of the cloud under study.While this strategy would be fine if the foreground emis-sion were dominated by a constant source, a time-varyingsource, solar wind charge exchange (SWCX) emission,is now known to be a major, possibly dominant, con-tributor to the foreground emission in the
XMM-Newton and
Suzaku band (Koutroumpa et al. 2007, 2009, 2011).This SWCX emission is variable on timescales of < COBE /DIRBE-corrected
IRAS dust maps(Schlegel et al. 1998) for compact interstellar clouds thatwould potentially cast an X-ray shadow that would fitwithin a single
XMM-Newton field of view. We identifiedthe cloud G225.60 − −
66 in Odenwald 1988;G225 hereafter) as a viable target (see Figure 1(a)). Theoptical depth of this cloud is such that the observed 0.4–1.0 keV surface brightness of the background emissiontoward the cloud is ∼ ∼
60 ks
XMM-Newton ex-posure. Unfortunately, the distance to this cloud is not Another potentially viable target was [RHK93] 9364(Reach et al. 1993), at l = 317 . ◦ b = +83 . ◦
8. However, the con-trast between the on- and off-cloud regions within a single
XMM-
HENLEY ET AL. known. Odenwald (1988) assumed a distance of 200 pc;the clouds in his sample for which he was able to esti-mate distances are at similar distances. If G225 is at adistance of ∼
200 pc, it would be beyond the LB.Here, we present the
XMM-Newton observation of thiscloud, which we used to constrain the Galactic haloX-ray emission. This is the first measurement of thisemission using a single-pointing shadowing observationwith a CCD-resolution spectrometer (Anderson et al.(2010) carried out similar observations with
XMM-Newton , but their target clouds were at low Galacticlatitudes ( b ∼ . ◦ ∼
75% and ∼
40% of the low-Galactic-latitude 1/4 keVforeground to SWCX, respectively. At higher energies,Koutroumpa et al. (2011) attributed approximately halfof the foreground O
VII emission in an
XMM-Newton observation of MBM 12 to SWCX. However, the relativecontributions of LB and SWCX emission to an arbitrary
XMM-Newton observation are not known. Therefore, weconsidered two limiting cases for our foreground model—one in which LB emission dominates, and one in whichSWCX emission dominates (for the latter case, we exam-ined two different SWCX models).The remainder of this paper is organized as follows.The observation and data reduction are described in Sec-tion 2. The spectral model and the results from the spec-tral analysis are presented in Sections 3 and 4, respec-tively. We discuss our results in Section 5. OBSERVATION AND DATA REDUCTION
G225 was observed by
XMM-Newton (Jansen et al.2001) for 90 ks on 2013 Feb 04–06 (observation ID0690500101). The pointing direction was ( α, δ ) =(02 h m . s , − ◦ ′ . ′′ l, b ) = (225 . ◦ , − . ◦ ∼ XMM-Newton
Extended Source Analy-sis Software ( XMM -ESAS; Snowden & Kuntz 2013), asincluded in the Science Analysis System (SAS) version13.5.0. We initially processed the data using the stan-dard SAS epchain and emchain scripts, and then usedthe XMM -ESAS pn-filter and mos-filter scripts toremove periods of soft proton flaring, during which thecount-rate was elevated. After this filtering, 46.6 and64.1 ks of good time remained from the pn and MOS2cameras, respectively.
Newton field was not expected to be as large as for G225. Also,it was not possible to obtain the required exposure from a singlepointing. http://heasarc.gsfc.nasa.gov/docs/xmm/xmmhp xmmesas.html http://xmm.esac.esa.int/sas/ We used the SAS edetect chain script to de-tect sources with 0.5–2.0 keV fluxes exceeding 2 × − erg cm − s − . Such sources were excludedfrom the data using circular exclusion regions. Fora given source, the source exclusion radius was equalto the semimajor axis of the ellipse on which thesource count rate per pixel is 0.2 times the local back-ground count rate. This radius depends on the sourcebrightness relative to the local background. We es-timate that the 0.5–2.0 keV surface brightness of theremaining, unremoved background sources is (3 . ± . × − erg cm − s − deg − (90% confidence in-terval for the whole XMM-Newton field). FollowingHenley & Shelton (2013) and Henley et al. (2014a), webased this estimate on the number density of sourceswith fluxes of 2 . × − to 2 × − erg cm − s − (Moretti et al. 2003) and the measurement of the resid-ual surface brightness after removing sources brighterthan 2 . × − erg cm − s − (Hickox & Markevitch2006). The uncertainty estimate takes into account thevariance in the number of sources due to source cluster-ing (Peebles 1980; Vikhlinin & Forman 1995) in additionto the Poissonian variance—see Henley & Shelton (2013)for details. The above surface brightness is about twicethe typical halo surface brightness (Henley & Shelton2013). The uncertainty on the surface brightness of theunremoved sources does not have a statistically signifi-cant effect on our measurements (Section 4).For each camera, we created an image of the 0.4–1.2 keV quiescent particle background (QPB), using the XMM -ESAS pn back and mos back programs. Theseimages were constructed using a database of filter-wheel-closed data, scaled to our observation using data fromthe unexposed corner pixels that lie outside the field ofview (Kuntz & Snowden 2008). We also used the
XMM -ESAS proton program to create images of the residualsoft proton contamination that remains despite the fil-tering described above. The parameters for the soft pro-ton models were determined from the spectral fitting(see Section 3, below). We subtracted the QPB andsoft-proton images from the corresponding 0.4–1.2 keVimages extracted from our
XMM-Newton data, dividedthese background-subtracted images by the correspond-ing exposure maps, and adaptively smoothed the re-sulting flat-fielded images (using the
XMM -ESAS adapt program). We filled in the chip gaps and the holes inthe data resulting from the source removal using datafrom neighboring pixels. The resulting X-ray images ofG225 from the pn and MOS2 cameras are shown in Fig-ures 1(b) and (c), respectively.In the pn image one can clearly see the shadow castby the cloud: there is a deficit of counts where the 100- µ m intensity, I , is greatest. However, the shadow isnot apparent in the MOS2 image. This difference be-tween the two cameras’ images is not an artifact of theparticle background subtraction—the shadow is appar-ent in the pn image and not the MOS2 image even if wedo not subtract the QPB and the soft proton contamina-tion. Instead, the difference is due to the MOS2 camera’slower sensitivity—for a ∼ × K plasma, say, the 0.4–1.2 keV MOS2 count rate is ∼ MM-NEWTON
SHADOWING MEASUREMENT OF HALO X-RAY EMISSION - : : . - : : . : . - : : . : . Right ascension D ec li n a t i on (a) : . : . - : : . : . : . : . Right ascension D ec li n a t i on (b)
5’ 0 0.1 0.2 0.3 0.4 0.5 0.6 : . : . - : : . : . : . : . : . Right ascension D ec li n a t i on (c) Figure 1. (a)
IRAS µ m image of G225 (Schlegel et al. 1998). The gray scale is in MJy sr − . The black square indicates the XMM-Newton pn field of view. (b) QPB- and soft-proton-subtracted, flat-fielded, adaptively smoothed 0.4–1.2 keV
XMM-Newton pn image ofG225. The chip gaps and the holes in the data resulting from the source removal have been filled using data from neighboring pixels. Thecolor scale is in counts ks − arcmin − . The white contours show the IRAS µ m intensity (1–5 MJy sr − in one-unit steps). The coloredpolygons indicate the spectral extraction regions (see text for details). Note that the polygons used for the spectral extraction follow thepixels in the 100- µ m map, whereas the contours have been smoothed. (c) As in (b), for MOS2. count rates expected over the pn and MOS2 fields, tak-ing into account the variation in the absorbing columndensity of the cloud and the telescope vignetting. Whilethe pn data are indeed expected to exhibit a shadow, theresulting MOS2 count rates are too low to produce a no-ticeable contrast between the on- and off-shadow regions,given the XMM-Newton exposure time.We extracted X-ray spectra from different regionsof the
XMM-Newton field of view, corresponding todifferent absorbing column densities, N H . These col-umn densities were derived from the IRAS I map(Schlegel et al. 1998), using the Snowden et al. (2000) I -to- N H conversion relation. The spectral extractionregions are shown by the colored polygons in Figures 1(b)and (c). These regions outline the I pixels that cor-respond to the following N H ranges: < > × cm − (cyan).Note that, because of the different fields of view, theextraction regions for the MOS2 spectra are slightly dif-ferent from those for the pn spectra.From each region we extracted a pn and a MOS2SXRB spectrum, using the XMM -ESAS pn-spectra and mos-spectra scripts, respectively, and grouped theresulting spectra such that there were at least 50 countsper bin. The spectral extraction scripts also calcu-lated the redistribution matrix file (RMF) and the an-cillary response file (ARF) needed for each spectrum,using the SAS rmfgen and arfgen programs, respec-tively. For each spectrum, we calculated a correspond-ing QPB spectrum using the
XMM -ESAS pn back and mos back programs. As noted above, the QPB spectrawere constructed from a database of filter-wheel-closed
HENLEY ET AL. data, scaled using data from the camera pixels outsidethe field of view. We subtracted from each SXRB spec-trum the corresponding QPB spectrum before carryingout our spectral analysis. SPECTRAL MODEL DESCRIPTION
In order to separate the foreground and halo emission,we used XSPEC version 12.8.1l (Arnaud 1996) to fit anSXRB spectral model simultaneously to the 0.4–5.0 keVspectra extracted from the different regions of the XMM-Newton detectors (we used the spectra from all four re-gions indicated in Figures 1(b) and (c)). Because thepn image exhibits an X-ray shadow whereas the MOS2image does not (Figure 1), we investigated fitting to thecomplete set of pn and MOS2 spectra and fitting just tothe pn spectra. We assumed Anders & Grevesse (1989)abundances.Our SXRB spectral model consisted of componentsrepresenting emission from the foreground, the Galac-tic halo, and the extragalactic background. We also in-cluded components representing parts of the instrumen-tal background that were not removed by the QPB sub-traction (see below). As noted in the Introduction, weexperimented with different models for the foreground,described in the subsections below. In particular, we con-sidered limiting cases in which LB emission (Section 3.1)or SWCX emission (Sections 3.2 and 3.3) dominate theforeground. The details of the other model componentsare as follows.We modeled the Galactic halo emission with asingle-temperature (1 T ) APEC thermal plasma model(Smith et al. 2001; Foster et al. 2012), whose tem-perature and emission measure were free parame-ters. We modeled the extragalactic background us-ing the double broken power-law model described inSmith et al. (2007), but with the overall normaliza-tion rescaled so that the 0.5–2.0 keV surface bright-ness matched that expected from sources below thesource removal flux threshold of 2 × − erg cm − s − (Henley & Shelton 2013; Henley et al. 2014a); as notedin Section 2, this surface brightness is 3 . × − erg cm − s − deg − . These components weresubject to absorption, modeled using the XSPEC phabs model (Ba luci´nska-Church & McCammon 1992;Yan et al. 1998). The absorbing column density, N H ,was different for each spectral extraction region, andwas calculated from the average value of I in eachregion (Schlegel et al. 1998), using the conversion rela-tion from Snowden et al. (2000). These column den-sities were 1.47 (1.50), 2.78 (2.82), 4.94 (4.96), and7 .
00 (7 . × cm − for the yellow, green, magenta,and cyan regions in Figure 1(b) (Figure 1(c)), respec-tively. At the energy of the O VII line, the optical depthin the highest- N H region is 0.66, meaning that the haloO VII emission is attenuated by 48%. In the lowest- N H region, the halo O VII emission is attenuated by 13%.In addition to the above SXRB components, we addedGaussians at ∼ ∼ XMM -ESAS, and hence were not removed http://heasarc.gsfc.nasa.gov/xanadu/xspec/ by the QPB subtraction. The parameters of these lineswere independent for each individual spectrum. In orderto model any residual soft proton contamination that re-mained in the spectra despite the filtering described inSection 2, we added a power-law that was not foldedthrough the instrumental response (Kuntz & Snowden2008; Snowden & Kuntz 2013). For each detector used(pn or MOS2), the spectral index of this component wasthe same for all four spectra, and the normalizations weretied together according to the relative scaling given bythe XMM -ESAS proton scale program. The best-fitparameters of this soft proton component were used tocreate the soft proton images mentioned in Section 2,which were used in the creation of Figures 1(b) and (c).
Foreground Model 1: Local Bubble (LB)
We initially modeled the foreground emission with a1 T APEC thermal plasma model that was not sub-ject to any absorption. The temperature and emis-sion measure of this component were free parameters.Physically, this model represents emission from a hotplasma, like that thought to be in the LB. AlthoughSWCX is now known to be a major, possibly dominant,source of the foreground emission in the
XMM-Newton band (Koutroumpa et al. 2007, 2009, 2011), such a ther-mal plasma model has been found to adequately modelthe foreground emission in CCD-resolution SXRB spec-tra (e.g., Galeazzi et al. 2007; Henley & Shelton 2008;Gupta et al. 2009). Note that we assumed that the LBemission originates entirely in front of the cloud.
Foreground Model 2: C14-SWCX
While using a thermal plasma model for the fore-ground emission appears to provide adequate fits toCCD-resolution SXRB spectra, it is possible that thetrue shape of the foreground spectrum, likely dominatedby SWCX emission, is different from that expected froma hot plasma. If this is the case, then a thermal plasmamodel for the foreground could lead to biases in the best-fit halo parameters. Therefore, in an attempt to avoidsuch biases, we modified our original SXRB model so thatthe foreground component was composed of CX emis-sion lines. For this model, we use CX line ratio datafrom Cumbee et al. (2014, hereafter C14; in that paper,we applied our CX data to a
Suzaku observation of theCygnus Loop, the spectrum of which is different fromthat of the SWCX emission). We refer to this new fore-ground model, which is more physically justified than athermal plasma model, as the C14-SWCX model.This foreground SWCX model consisted of C VI Ly α – ǫ , O VII K α – ǫ , and O VIII Ly α – ǫ emission lines. Forthe O VII K α feature, we modeled the forbidden, in-tercombination, and resonance lines individually. Theoverall normalization of the emission from each ion wasindependent (i.e., we did not constrain the ion ratios apriori ). For each ion, we tied together the lines’ normal-izations using the relative intensities from the CX modeldescribed in C14. These CX line ratios were calculatedfor a collision energy of 1 keV u − (438 km s − ; cf. atypical speed for the slow solar wind is 400 km s − ; e.g., Note that the model used here includes C VI Ly ǫ , which wasnot included in C14. The C VI Ly ǫ /Ly α ratio that we used is 0.0012(R. S. Cumbee & P. C. Stancil, 2014, private communication). MM-NEWTON
SHADOWING MEASUREMENT OF HALO X-RAY EMISSION
XMM-Newton detectors at low energies, we did not include lines fromN VI or N VII in the C14-SWCX model (these ions’ K α lines lie between those of C VI and O VII ).Carter et al. (2010) and Ezoe et al. (2011) used a sim-ilar CX model (based on data from Bodewits 2007) intheir analyses of SWCX enhancements observed duringan
XMM-Newton and a
Suzaku observation, respectively.However, we are unaware of such a model having previ-ously been applied to a shadowing observation.
Foreground Model 3: ACX-SWCX
Our third and final foreground model used theAtomDB Charge Exchange code (ACX; Smith et al.2014), and is referred to here as ACX-SWCX. For eachion receiving an electron via CX, the ACX model uses an-alytic expressions to calculate the most-probable n shelland the distribution of orbital angular momenta, l , forthe captured electron (see Smith et al. 2014 for details).This model then calculates the spectrum produced as theelectron radiatively cascades to the ground state (mainlyusing data from AtomDB 2.0.2; Foster et al. 2012). Therelative strengths of the lines from different ions of thesame element are determined from the ionization bal-ance of the input ion population, which is controlled bythe model’s temperature parameter, assuming that therelative ion populations are in collisional ionization equi-librium (CIE). The relative strengths of lines from differ-ent elements, meanwhile, are governed by the assumedabundances (Anders & Grevesse 1989).For our purposes, we set the ACX model’s swcx and model flags to 1 and 8, respectively (Smith & Foster2014). The former setting means that each ion under-goes a single CX reaction on the line of sight, and is theappropriate setting for studying CX in the context of thediffuse SXRB. The latter setting means that, if the most-probable n shell for electron capture is not an integer, thecaptured electrons are distributed between the two near-est n shells. This setting also means that the “Separable”distribution (Smith et al. 2014, Equation (4)) is used forthe l distribution. SPECTRAL ANALYSIS RESULTS
The spectral fit results are shown in Table 1, for theLB (Section 3.1), C14-SWCX (Section 3.2), and ACX-SWCX (Section 3.3) foreground models. In addition,we show results obtained with no foreground compo-nent in the spectral model (“None”). The upper halfof the table shows the results obtained by fitting simul-taneously to the pn and MOS2 spectra, while the lowerhalf shows the results obtained by fitting just to the pnspectra. The best-fit foreground model parameters arein columns 2 and 3 for the LB and ACX-SWCX fore-ground models, and in columns 4–6 for the C14-SWCXforeground model. For all models, the best-fit halo tem-perature, T h , and emission measure, E h , are in columns 7 h (10 K) H a l o e m i ss i on m ea s u r e , E h ( - c m - p c ) LB foregroundC14−SWCX foregroundACX−SWCX foregroundNo foregroundSolid symbols:Open symbols: pn onlypn + MOS2
Figure 2.
Halo temperatures and emission measures obtainedusing the various foreground models, indicated by color (see key).Solid symbols and error bars (open symbols and dashed error bars)indicate the results obtained from just the pn data (from the pnand MOS2 data jointly). and 8, respectively. Figure 2 compares the halo temper-atures and emission measures obtained using the variousforeground models. Figure 3 shows the pn spectra fromthe regions with the lowest and highest values of N H (yel-low and cyan regions in Figure 1(b), respectively), alongwith the best-fit models obtained using each of the threeforeground models, and using no foreground model. Ingeneral, the fits shown are reasonably good, and the fitsto the spectra that aren’t shown are of similar quality.Overall, the pn data result in tighter constraints onthe halo parameters when used on their own than whencombined with the MOS2 data. The average widthsof the 90% confidence intervals on the halo tempera-ture and emission measure are 0 . × K and 1 . × − cm − pc, respectively, from the pn-only fits, com-pared with 0 . × K and 1 . × − cm − pc, respec-tively, from the joint pn + MOS2 fits. This differencemay be due to the fact that the soft proton contami-nation in the MOS2 spectra is more severe than in thepn spectra (Figure 4). Because the pn spectra result intighter constraints overall on the halo parameters, in thefollowing we shall concentrate on the results obtainedfrom the pn-only fits.The results in Table 1 were obtained assumingthat the 0.5–2.0 keV surface brightness of the extra-galactic background is equal to that expected fromsources below the source removal flux threshold, 3 . × − erg cm − s − deg − (Section 3). The uncer-tainty on this expected surface brightness is ± . × − erg cm − s − deg − (Section 2). We found thatvarying the surface brightness of the extragalactic back-ground model within this uncertainty did not have a sta-tistically significant effect on our best-fit model param-eters. This was mainly because, if we adjusted the nor-malization of the extragalactic model, the normalizationof the soft proton contamination model adjusted itselfto compensate, leaving the other model components notsignificantly affected.For the LB foreground model, while the best-fitforeground temperature, T fg , is rather low, within theuncertainty it is consistent with the range of values foundfrom previous shadowing studies ( T fg ∼ (0.8–1.2) × K;Snowden et al. 2000; Smith et al. 2007; Galeazzi et al.2007; Henley et al. 2007; Henley & Shelton 2008;
HENLEY ET AL.
Table 1
Spectral Fit ResultsForeground HaloForeground T fg Normalization a I (C VI ) b I (O VII ) c I (O VIII ) d T h E h χ /dofmodel (10 K) (L.U.) (L.U.) (L.U.) (10 K) (10 − cm − pc)(1) (2) (3) (4) (5) (6) (7) (8) (9)Joint fits to pn and MOS2 data:LB 1.02 (0.64,1.22) 7.81 (4.16,217.37) · · · · · · · · · · · · · · · < > · · · · · · · · · · · · · · · · · · · · · · · · < > · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · Note . — Values in parentheses are the 90% confidence intervals. a For the LB foreground model, this is the foreground emission measure, E fg , in units of 10 − cm − pc. For the ACX-SWCX foreground model, thisis the normalization of the foreground component, in units of 10 − arcmin − . b Foreground C VI Ly α intensity. As this line ( E = 0 . XMM-Newton band used here, this intensity is not constrained directly,but is instead constrained by the higher-energy Lyman lines via the C14 CX line ratios. c Foreground O
VII K α intensity. We have summed the intensities of the resonance, intercombination, and forbidden lines. d Foreground O
VIII Ly α intensity. Lei et al. 2009; Gupta et al. 2009). Because thisforeground model is relatively faint within the
XMM-Newton band (most of the emission would be emittedbelow 0.4 keV), its emission measure, E fg , is poorlyconstrained. However, it too is consistent (within itsuncertainty) with the results from previous shadowingstudies.Although the physical nature of the C14-SWCX fore-ground model is quite different from that of the LB fore-ground model, for this particular shadowing observationthese two models yield best-fit foreground spectra thatare similar in shape in the XMM-Newton bandpass (com-pare Figures 3(a) and (b)). As a result, the best-fit haloparameters from these two models are very similar. How-ever, the halo parameters are less well constrained whenwe use the C14-SWCX foreground model. This is be-cause, in this model, the foreground C VI , O VII , andO
VIII intensities are completely independent, whereas inthe LB model they are controlled by the foreground tem-perature. This means that there is more freedom in theshape of the foreground spectrum, and as a result morefreedom in the shape of the halo spectrum, and hence inthe halo temperature. Note that the C14-SWCX fore-ground model yields a higher χ than the LB foregroundmodel, despite having one more free parameter.The ACX-SWCX foreground model yields a muchsofter best-fit foreground spectrum than the other fore-ground models. Since this foreground model producesvery little O VII emission, the halo component must pro-duce relatively more O
VII , and as a result this fore-ground model yields a slightly lower halo temperature.However, the difference is only a few × K, and is notsignificant given the error bars.Figure 5 shows χ as a function of halo temperature foreach of the foreground models that we studied. In addi-tion to the best-fit χ minimum at T h ≈ × K, eachcurve also exhibits a local minimum at T h ≈ . × K.At these local minima, the foreground models are harderthan in the best fits, to compensate for the softness ofthe cooler halo models. This means that there is somedegeneracy between the hardnesses of the foreground and halo components. However, the differences in χ between the minima at the lower and higher halo tem-peratures are 17.8, 7.2, and 19.8 for the LB, C14-SWCX,and ACX-SWCX foreground models, respectively, mean-ing that the lower halo temperature is excluded at the >
99% level (∆ χ = 6 .
63 for a single interesting param-eter; Lampton et al. 1976). To put this another way,the observed
XMM-Newton spectra require a soft line-emission component and a hard line-emission component(with temperatures of . . × and ∼ (2.0–2.5) × K,respectively, for models with a temperature parame-ter). Figure 5 shows that models in which the softercomponent is in the foreground and the harder compo-nent is in the halo (i.e., our best-fitting models, with T h ≈ × K) are strongly preferred over models inwhich these two components are switched.Because the ACX-SWCX foreground model yields sim-ilar halo parameters to the other foreground models, de-spite the foreground spectrum being much softer, and be-cause omitting the foreground component altogether stillyields an acceptable fit (reduced χ = 1 . − L max + 2 k, (1)where L max is the maximum likelihood and k is the num-ber of free parameters. The lower the value of AIC, thebetter the model. As we used χ minimization in our fit-ting, we make use of the fact that − L max = χ + C ,where χ is the best-fit value of χ , and C is a constantindependent of the particular model being considered (asonly differences in AIC are meaningful, we can ignore C ).For each foreground model, we calculated the AIC rela-tive to that obtained with no foreground model,∆AIC(Model X) = AIC(Model X) − AIC(No f/g) . (2)For the pn-only fits, the LB, C14-SWCX, and ACX-SWCX foreground models yield ∆AIC = − . − . MM-NEWTON
SHADOWING MEASUREMENT OF HALO X-RAY EMISSION c oun t s s - k e V - (a) LB foreground model pn, lowest N H pn, highest N H HaloForegroundExtragalacticSoft protons −404 ( da t a - m ode l ) s lowest N H highest N H c oun t s s - k e V - (b) C14−SWCX foreground model pn, lowest N H pn, highest N H HaloForegroundExtragalacticSoft protons −404 ( da t a - m ode l ) s lowest N H highest N H c oun t s s - k e V - (c) ACX−SWCX foreground model pn, lowest N H pn, highest N H HaloForegroundExtragalacticSoft protons −404 ( da t a - m ode l ) s lowest N H highest N H c oun t s s - k e V - (d) No foreground model pn, lowest N H pn, highest N H HaloExtragalacticSoft protons −404 ( da t a - m ode l ) s lowest N H highest N H Figure 3.
XMM-Newton pn spectra from the regions of the G225 field with the lowest and highest values of N H (gray and black datapoints in the above plots, corresponding to the yellow and cyan regions in Figure 1(b), respectively), with the best-fit spectral models fromthe fits just to the pn data. For plotting purposes only, the data have been regrouped such that each bin has a signal-to-noise ratio of atleast 3. Plots (a), (b), and (c) show the best-fit models obtained with the LB (Section 3.1), C14-SWCX (Section 3.2), and ACX-SWCX(Section 3.3) foreground models, respectively. Plot (d) shows the fit with no foreground component in the spectral model. For the spectrumfrom the highest- N H region, we also plot individual model components (see key; note that we do not plot the component representing theinstrumental Al line). For the SWCX foreground model, the dotted lines show the contributions to the foreground from C VI and O VII (from left to right; the best-fit foreground O
VIII intensity is zero). and − .
6, respectively. These differences in AICamount to strong (∆AIC < −
5) or decisive (∆AIC < −
10) evidence in favor of including a foreground compo-nent in the model (Liddle 2007). DISCUSSION AND CONCLUSIONS
Foreground Emission
The foreground emission toward G225 appears to berelatively faint in the
XMM-Newton band. Our spectralanalysis implies foreground 0.4–1.0 keV surface bright-nesses of 5.0 (2.3–8.8), 4.1 (2.4–8.2), and 2.6 (0.6–4.4) × − erg cm − s − deg − for the LB, C14-SWCX,and ACX-SWCX foreground models respectively (thevalues in parentheses are the 90% confidence intervals).In contrast, the results of previous XMM-Newton and
Suzaku shadowing studies imply foreground 0.4–1.0 keVsurface brightnesses of (7–18) × − erg cm − s − deg − (Lei et al. 2009; Gupta et al. 2009; Smith et al. 2007;Galeazzi et al. 2007; Henley et al. 2007). The highestof these is from a pair of XMM-Newton pointings onand off an unnamed dusty filament (Henley et al. 2007),which are now known to be contaminated by stronger-than-typical SWCX emission (Koutroumpa et al. 2007;Henley & Shelton 2008).The faintness of the foreground emission limits theamount of physical information about the foregroundthat we can extract from our observation of G225. Forexample, from the C14-SWCX model we obtain only up-per limits on the foreground O
VII K α and O VIII Ly α intensities, and so we cannot constrain the solar wind HENLEY ET AL. −4 −3 −2 −1 channel energy (keV) c oun t s s - k e V - pnMOS2 / 10ExtragalacticSoft protons Figure 4.
XMM-Newton pn (gray) and MOS2 (black) spectrafrom the region of the G225 field with the lowest value of N H , withthe best-fit spectral model obtained with the LB foreground model(thin solid lines). The MOS2 data have been shifted down by afactor of 10. For each spectrum, we also plot the extragalactic andsoft proton components of the model (thick solid and dashed lines,respectively; the other model components are not plotted). Notethat, in the MOS2 spectrum, the soft proton component is brighterrelative to the extragalactic component than in the pn spectrum,indicating more severe soft proton contamination. h (10 K) c LBC14−SWCXACX−SWCX
Figure 5. χ as a function of halo temperature for each of thethree foreground models that we studied (solid line: LB model;dashed line: C14-SWCX model; dotted line: ACX-SWCX model). O /O ion ratio using this model. The temperature ofthe ACX-SWCX model can provide information on thision ratio, albeit under the assumption of a CIE ion distri-bution. At the best-fit temperature of the ACX-SWCXcomponent, 7 . × K, 99% of the oxygen is in the O charge state (from ATOMDB v2.0.2; Foster et al. 2012).Assuming CIE therefore results in a best-fit model fromwhich there is virtually no oxygen SWCX emission inthe XMM-Newton band (the SWCX emission from thismodel in the
XMM-Newton band is mainly from N VI K α and C VI Ly β and Ly γ ).We use the upper limit of the temperature of the ACX-SWCX component, 1 . × K, to place an upper limitof 0.006 on the solar wind O /O ratio (ATOMDB).This is significantly less than the ratio expected for theslow solar wind (0.35; Schwadron & Cravens 2000), sug-gesting that, during the XMM-Newton observation, theportion of the G225 sight line in the heliosphere passedmainly through fast solar wind (for which this ratiois nearly zero; Schwadron & Cravens 2000). This is asomewhat surprising result, as the observation was takenonly ∼ /O ratio) is determined not just by the oxy-gen K lines, but also by lower-energy lines from carbonand nitrogen, and so the low solar wind O /O ratiocould in principle be an artifact of our assuming the de-fault Anders & Grevesse (1989) abundances for the ACXmodel. In practice, this appears not to be the case: ifwe adjust the abundances of carbon, nitrogen, and neonrelative to oxygen so they match those expected for theslow solar wind (von Steiger et al. 2000, specifically, theaverage of the “Max” and “Min” values from their Ta-ble 1) and refit, we find that the halo results are unaf-fected, and the resulting upper limit on the solar windO /O ratio is 0.010, still much lower than the valueexpected for the slow solar wind. However, we note thatthe results for the ACX model could be affected by theassumption of an ion distribution described by a singletemperature.It should also be noted that the sun was less ac-tive during the most recent maximum than during pre-vious maxima (e.g., the sunspot number and the so-lar 1–8 ˚A flux at the most recent maximum were ap-proximately half the values at the 1990 maximum;Winter & Balasubramaniam 2014), which may have af-fected the solar wind structure. Unfortunately, so-lar wind charge distribution data from the SWICS in-strument on board the Advanced Composition Explorer ( ACE ) are unavailable for times after August 2011, whereas our observation was taken in February 2013.Therefore, we are unable to check if the solar wind hadan unusual ion composition prior to and during our ob-servation. Halo Emission
The halo parameters derived from the G225 pn spectraare not sensitive to the particular foreground model usedin the analysis, although omitting the foreground com-ponent altogether does result in a halo temperature thatis ∼
10% lower. This insensitivity to the details of theforeground model is likely due to the relative faintnessof the foreground emission, noted above. If the spectralanalysis carried out here were repeated on a shadowingobservation with bright foreground emission, we wouldexpect to see some sensitivity of the halo parameters tothe assumed form of the foreground emission. We plan totest this in a future study. (Note that this will necessar-ily involve using shadowing observations that consist ofseparate on- and off-shadow pointings, unlike the single-pointing observation studied here).G225 is included in the Snowden et al. (2000) cata-log of X-ray shadows in the
ROSAT
All-Sky Survey,as shadow S2267M661. The intrinsic 1/4 keV halocount rate in the direction of G225 is (947 ± × The absolute abundances of these elements relative to hydrogenare not important here, as hydrogen does not emit in the
XMM-Newton band. The absolute abundances affect only the overallnormalization of the ACX model. This name is derived from the coordinates of the center of the
MM-NEWTON
SHADOWING MEASUREMENT OF HALO X-RAY EMISSION h (10 K) H a l o e m i ss i on m ea s u r e , E h ( - c m - p c ) G225 LB f/gG225 C14−SWCX f/gG225 ACX−SWCX f/gG225 No f/gOther shadowsHenley & Shelton (2013)
Figure 6.
Comparison of our halo measurements with those fromprevious studies. The solid symbols show our pn-only results fromFigure 2. The magenta squares show the results from previous
XMM-Newton or Suzaku shadowing studies: from top to bottom,a
Suzaku study of an unnamed dusty filament (Lei et al. 2009; notethat this result has been rescaled—see text for details), a
Suzaku study of MBM 12 (Smith et al. 2007), an
XMM-Newton study ofMBM 20 (Galeazzi et al. 2007), and a
Suzaku study of MBM 20(Gupta et al. 2009). The black diamonds show results from theHenley & Shelton (2013)
XMM-Newton survey of the halo, forsight lines within 15 ◦ of G225. − counts s − arcmin − . In contrast, our best fithalo models imply 1/4 keV count rates of (240–270) × − counts s − arcmin − . This discrepancy impliesthat a 1 T model cannot adequately model the halo X-ray emission down to photon energies of ∼ ROSAT
All-Sky Sur-vey data (Kuntz & Snowden 2000). In order to obtain areasonable model of the 1/4 keV emission, a ∼ × Kcomponent must be added to the halo model—since sucha component would contribute to the halo’s O
VII emis-sion, its inclusion would affect the best-fit temperatureof the ∼ × K component of our current spectralmodel. Even such a two-temperature model is likelyan approximation of the halo’s true temperature struc-ture, as there may be a continuum of temperatures inthe halo (Shelton et al. 2007; Lei et al. 2009). However,the 1 T halo model used here is still useful for character-izing the emission within the 0.4–5.0 keV XMM-Newton band, and the results obtained from such halo modelscan still be used to test models of the hot halo gas, pro-vided such models’ emission predictions are characterizedin the same way as the observed emission (Henley et al.2010).Figure 6 compares our measurements with those fromprevious
XMM-Newton and
Suzaku shadowing stud-ies. In these studies, the halo emission was character-ized with a single X-ray temperature. The Lei et al.(2009) result was obtained using a different abundancetable from the other studies (Wilms et al. 2000 versusAnders & Grevesse 1989 ). The halo emission is domi-nated in the XMM-Newton / Suzaku band by oxygen K α emission; for a given temperature, the intensity of this region of the sky analyzed by Snowden et al. (2000), rather thanfrom the coordinates of the cloud. These were calculated using v2.0.2 of APEC (Foster et al.2012). If we instead use the Raymond & Smith (1977 and up-dates) code, we obtain count rates ∼ × − counts s − arcmin − higher. Note that Gupta et al. (2009) do not explicitly state whichabundance table they used for their plasma emission components. emission is proportional to R n e n O dl = E h A O / .
2, where n e and n O are the halo electron and oxygen number den-sities, respectively, E h ≡ R n dl is the halo emission mea-sure, and A O is the halo oxygen abundance. Hence, thebest-fit halo emission measure is approximately inverselyproportional to the assumed value of A O . Therefore, inorder to allow a fair comparison with the other results,we have multiplied the Lei et al. (2009) emission measureby A O (Wilms et al.) /A O (Anders & Grevesse) = 0 . XMM-Newton and
Suzaku shadow-ing studies. This agreement implies that the fact thatthese other studies consisted of two separate pointings,which could potentially have had different foregroundbrightnesses (see Introduction), did not adversely affectthe halo results. However, as noted above, the halo re-sults derived from these other studies may be sensitiveto the assumed foreground model.Figure 6 also compares our measurements with re-sults for nearby sight lines in the Henley & Shelton(2013)
XMM-Newton survey of the halo (within 15 ◦ ofG225). In this survey, the foreground model was basedon results from the previously mentioned Snowden et al.(2000) shadow catalog, extrapolated from the 1/4 keV ROSAT band to the 0.4–5.0 keV
XMM-Newton band.The Henley & Shelton (2013) emission measures shownin Figure 6 are typically smaller than that obtained fromG225. One might therefore conclude that there is asystematic error in the Henley & Shelton emission mea-sures, possibly due to the assumed foreground model.However, the Henley & Shelton result that is closestto the G225 results in Figure 6 (obs. 0302500101, at( T h , E h ) = (2 . × K , . × − cm − pc)) is alsothe closest sight line to G225 on the sky (angular sep-aration = 4 . ◦ ∼ ◦ from G225.) Furthermore, theagreement between the G225 measurements and the mea-surement from the nearest Henley & Shelton (2013) sightline supports the conclusion that the Henley & Sheltonresults are well calibrated and not subject to systematicerrors. Such a conclusion is important for when we usethe Henley & Shelton measurements to test models ofthe halo X-ray emission (Henley et al. 2014b).We thank the anonymous referee, whose commentshave helped improve this paper. This research isbased on observations obtained with XMM-Newton ,an ESA science mission with instruments and contri-butions directly funded by ESA Member States andNASA. We acknowledge use of the R software package(R Development Core Team 2008). This research wasfunded by NASA grant NNX13AF69G, awarded throughthe Astrophysics Data Analysis Program, and partiallysupported by NASA grant NNX09AC46G.REFERENCES
Anders, E., & Grevesse, N. 1989, Geochim. Cosmochim. Acta, 53,197 HENLEY ET AL.