Extended Gamma-ray Emission from the G25.0+0.0 Region: A Star Forming Region Powered by the Newly Found OB Association?
We report a study of extended -ray emission with the Large Area Telescope (LAT) onboard the Fermi Gamma-ray Space Telescope, which is likely to be the second case of a -ray detection from a star-forming region (SFR) in our Galaxy. The LAT source is located in the G25 region, around . The -ray emission is found to be composed of two extended sources and one point-like source. The extended sources have a similar sizes of about . An diameter sub-region of one has a photon index of ; and is spatially coincident with HESS J1837069, likely a pulsar wind nebula. The other parts of the extended sources have a photon index of without significant spectral curvature. Given their spatial and spectral properties, they have no clear associations with sources at other wavelengths. Their -ray properties are similar to those of the Cygnus cocoon SFR, the only firmly established -ray detection of an SFR in the Galaxy. Indeed, we find bubble-like structures of atomic and molecular gas in G25, which may be created by a putative OB association/cluster. The -ray emitting regions appear confined in the bubble-like structure; similar properties are also found in the Cygnus cocoon. In addition, using observations with the XMM-Newton we find a candidate young massive OB association/cluster G25.18+0.26 in the G25 region. We propose that the extended -ray emission in G25 is associated with an SFR driven by G25.18+0.26. Based on this scenario, we discuss possible acceleration processes in the SFR and compare them with the Cygnus cocoon.
Subject headings:acceleration of particles – cosmic rays – gamma rays: ISM – ISM: bubbles; open clusters and associations – X-rays: stars
Most massive stars are formed in clusters by the collapse of giant molecular clouds (e.g., Lada & Lada, 2003). They produce strong radiation fields and stellar winds, which disrupt their natal molecular clouds and create bubble structures around them. Such central OB associations/clusters and the accompanying bubbles constitute massive star-forming regions (SFRs). SFRs could be sites of -ray emission, if the wind energy is efficiently transferred to the acceleration of particles. Many models are proposed for acceleration processes in massive SFRs, such as diffusive shock acceleration (DSA) at the wind boundary and stochastic acceleration by magnetic turbulence (see e.g, Cesarsky & Montmerle, 1983; Bykov & Toptygin, 2001). However, the limited number of -ray detected SFRs limits the possibility of studying the acceleration process in detail.
Observations with the Fermi Large Area Telescope (LAT) have recently revealed the extended -ray emission from an SFR in the Cygnus X region (Ackermann et al., 2011, hereafter MA11). The -ray emission of this SFR (a.k.a. the Cygnus cocoon) spatially extends over a large diameter of and appears associated with the cavities probably created by a massive OB association, Cyg OB2. The observed hard spectrum (photon index ) suggests that the -ray emitting particles are accelerated in the region. The accelerated particles are plausibly powered by Cyg OB2. However there is another possibility that the particles escaped from the nearby supernova remnant (SNR) Cygni.
Recent TeV observations have found an emission whose morphology is consistent with that of the LAT observations (Bartoli et al., 2014). The observed energy spectrum is smoothly connected to that of the LAT observations, and extends to a few TeV. Prior TeV observations have also revealed the existence of relativistic particles of energies of up to a few tens of TeV in this region (Abdo et al., 2007, 2009; Bartoli et al., 2012). If all the observed rays are attributed to the same object, its energy spectrum must steepen at TeV energies.
The -ray detection from the Cygnus cocoon is evidence that relativistic particles exist in the SFR. This is the only firm case of a -ray detection from an SFR, which may indicate that the -ray detection of the Cygnus cocoon is a special case, e.g., the relativistic particles are not accelerated at the SFR but are runaway particles from the SNR. More studies are needed to answer the question of whether SFRs are sites of the particle acceleration.
In this paper, we report a study of LAT observations of extended -ray emission, which is likely to be the second example of a -ray detection of an SFR. The LAT source is located in the region around (hereafter the G25 region). The region was studied by Lande et al. (2012) based on two years of LAT data. They found an extended source which is spatially coincident with the TeV source HESS J1837069. In this paper we re-analyze this complex region using 57 months of LAT data. There is no known large () SFR associated with the G25 region. However, in this direction, Rahman & Murray (2010) recently found copious 8 emission within a region of the brightest free-free emission reported by Murray & Rahman (2010). They claim that the emission is due to a star-forming complex with diameter , which is created by hidden massive OB association(s). The putative OB association might be difficult to detect optically due to heavy extinction, if it is distant from us. In this paper, we investigate the G25 region using X-ray and radio data to check whether a massive SFR exists. An optically hidden massive OB association is expected to be detectable in X-rays, because of the reduced extinction compared to that in the optical. Here we analyze in particular an unidentified X-ray source AX J1836.30647 located in this region (Sugizaki et al., 2001) using XMM-Newton data.
This paper is organized as follows. The observation and the analysis of the LAT sources in the G25 region is reported in Section 2. We describe the XMM-Newton observations and analysis of AX J1836.30647 in Section 3. In Section 4, we discuss the origin of the observed -ray and X-ray emissions. Based on the discussion, we investigate possible acceleration mechanisms for the radiating particles. In Section 5, we summarize our results and the outlook for advances with future observations.
2. Fermi-LAT Data
2.1. Observation and Data Reduction
The Fermi Gamma-ray Space Telescope was launched on 2008 June 11. The LAT onboard Fermi is a pair-conversion telescope equipped with solid state silicon trackers and a cesium iodide calorimeter, sensitive to photons in the very broad energy band from MeV to GeV. The LAT has a large effective area ( cm above 1 GeV for on-axis events), instantaneously viewing sr of the sky with a good angular resolution (per-photon 68% containment radius better than above 1 GeV). Details of the LAT instrument and data reduction are described in Atwood et al. (2009).
The LAT data used here were collected during months from 2008 August 4 to 2013 May 1. We selected Pass 7 Reprocessed SOURCE class events and rays with Earth zenith angles greater than were excluded to reduce the -ray background from the Earth limb, an intense source of rays from cosmic-ray collisions with the upper atmosphere. We also applied a cut that excludes time intervals during which any part of the region we analyzed (see Section 2.2) was beyond the 100 zenith angle limit. We used the P7REP_SOURCE_V15 instrument response functions for the analysis (Ackermann et al., 2012). Note that we made a basic confirming check for the sources analyzed in this paper using the Pass 8 data with the surrounding sources in the 3FGL catalog (Acero et al., 2015). The results do not show strong quantitative differences from those in the following analyses. ﬁ
2.2. Analysis and Results
The G25 region contains four sources in the Fermi-LAT Second Source Catalog (2FGL; Nolan et al. (2012)): 2FGL J1835.50649, J1836.80623c, J1837.30700c, and J1834.70705c. Using the first two years of LAT data, Lande et al. (2012) find an extended -ray source in this region. The authors claim that two of the 2FGL sources are not distinct sources but an approximation for the extended source, and the other two 2FGL sources are unrelated background sources. They also find that the extended TeV source HESS J1837069 is coincident with the LAT extended source. However the LAT flux is about two times higher than what is expected from the TeV observation. In addition, the LAT source is about two times larger than the H.E.S.S. source, with a different shape, and the peak emission of the H.E.S.S. source is located on the edge of the LAT source reported by Lande et al. (2012). This may indicate that other -ray sources contaminate the reported extended source in the LAT band. The First Fermi-LAT Catalog above 10 GeV (1FHL; Ackermann et al. (2013)), based on the first three years of data, reveals another new source (1FHL J1839.40708) located south of the extended source. This region also includes two other 1FHL sources, J1834.60703 and J1836.50655e, which correspond to 2FGL J1834.70705c and the extended source mentioned above, respectively. The detection of the new source in 1FHL is consistent with the suggestion of other -ray sources in this region. In this section we re-analyze this complex region using 57 months of LAT data, which is more than double the observation time considered by Lande et al. (2012).
To characterize the -ray sources in the region, we use gtlike, part of the Science Tools analysis package (v9r32p5)111Available at the Fermi Science Support Center, http://fermi.gsfc.nasa.gov/ssc, and pointlike (Kerr, 2011; Lande et al., 2012). With gtlike and pointlike, we perform binned maximum likelihood fits on the observed rays to optimize parameters of an input model taking into account the energy dependence of the point-spread function (PSF). We first evaluate the spatial extent of the G25 region by using pointlike; the pointlike algorithm is optimized for speed to handle a large number of sources. With the obtained spatial extent, we measure spectral energy distributions (SEDs) of the region by using gtlike. In this analysis, the test statistic (TS) is defined as between a model including a test source or parameter and a model without that source or parameter (the null hypothesis). For each model the likelihood is maximized over the free parameters before TS is evaluated. The significance of the improvement of the maximum likelihood with the additional source or parameter can be evaluated from the TS value, because the value is expected to approximately follow the distribution with additional degrees of freedom (Mattox et al., 1996) in the case that the null hypothesis is a complete description of the region under study.
For the spatial analysis, we select a region around G25, and we include only energies above 3 GeV, because the LAT has better angular resolution at higher energies (68% containment radius better than above 3 GeV). In addition the spatial confusion between the G25 source and the Galactic diffuse component is reduced at higher energies, since the flux ratio of G25 to the Galactic diffuse emission increases with energy. We use spatial bins and divide the 3–500 GeV energy range into 17 bins evenly on a log scale. In the spectral analysis, we use a larger region, , around G25 in the energy range 0.2–500 GeV, given the relatively large LAT PSF at lower energies. The size of the spatial bins is set at ; the energy range is divided into 26 equal bins on a log scale.
The input model for the likelihood analysis includes cataloged sources in the G25 region, nearby sources outside the region, a model for the Galactic diffuse emission, and an isotropic component. A square region in the vicinity of G25 (hereafter the G27 region; see Figure 1) appears to contain extended sources rather than a combination of 2FGL point sources. To perform precise measurements of G25, we also evaluate spatial and energy distributions of this nearby background region. For the analysis we set positions and spectral parameters of the other background sources equal to those in the 2FGL catalog except for 2FGL J1834.30848 and LAT pulsars (PSRs) described in the LAT Second Pulsar (2PC) Catalog (Abdo et al., 2013). 2FGL J1834.30848, which is associated with SNR W41, was recently reported to be spatially extended (Castro et al., 2013; H. E. S. S. Collaboration et al., 2015). After H. E. S. S. Collaboration et al. (2015), we model its spectrum as a power law and its spatial distribution as a symmetrical 2D Gaussian centered at with . When LAT PSRs are associated with 2FGL sources, we adopt positions and spectral functions described in the 2PC catalog instead of the 2FGL catalog. Here 2FGL J1813.41246, J1826.11256, J1833.61032, and J1839.00539 are replaced by LAT PSRs J18131246, J18261256, J18331034, and J18380537 respectively. We include LAT PSR J18351106, which has no counterpart in the 2FGL catalog, in the model. Figure 1 shows that a point-like source (source N) is located off the Galactic plane. Since there is no corresponding source in the 2FGL catalog, source N is also added into the model. Note that there is a corresponding source (3FGL J1824.3-0620) in the 3FGL catalog. Source N is modeled as a point source located at . The spectral shape is modeled with a simple power law with an exponentially cut-off function. The Galactic diffuse emission is modeled using gll_iem_v05.fit while an isotropic component (extragalactic diffuse backgrounds plus residual charged particles) is modeled using iso_iem_v05.txt. Both background models are the standard diffuse emission models released by the LAT Collaboration222Available from the Fermi Science Support Center, http://fermi.gsfc.nasa.gov/ssc/data/access/lat/BackgroundModels.html. We note that an analysis using gll_iem_v05_rev.fit instead of gll_iem_v05.fit does not significantly affect spectral parameters obtained in this analysis. The former includes diffuse emission from molecular gas beyond the solar circle that was omitted from the latter. At the low longitudes of the present analysis, the differences between the two models are unimportant.
In all analyses, the normalizations of the diffuse components are set free unless otherwise mentioned. In the GeV analysis, we set the normalizations free for sources within the fitting region. We also set the spectral parameters free for the sources of photon cm s in the energy range of 3–100 GeV in the 2FGL catalog. In the GeV analysis, because of the large fitting region, there are many fitting parameters. To reduce free parameters, we first set the normalizations free for sources of photon cm s in the energy range of 0.3–100 GeV within the fitting region. Spectral parameters of the sources with the fluxes photon cm s are also set free. Note that we use the GeV fluxes from 2FGL because the catalog does not provide the fluxes of GeV. We fit the data using this model and then use the parameters from this first fit for the GeV analysis. In the subsequent analysis, we determine the freedom of the spectral parameters based on the fitting fluxes in the energy range of 0.2–500 GeV (flux) and the angular distance from the center of the fitting region (): we set the normalizations free for the sources with flux photon cm s in the region; in the region, we set the normalization free for the sources with flux photon cm s and set the normalizations and spectral shapes free for the sources with flux photon cm s. We also include sources outside the fitting regions but within and for the analyses above 0.2 GeV and 3 GeV respectively, with their parameters fixed at those given in the 2FGL and 2PC catalogs.
2.2.1 Analysis Procedure
Here we describe how we evaluate the spatial and energy distributions of G25 as well as of nearby background region G27. Modeling G27 is necessary for precise measurements of G25. The region contains three 2FGL sources (2FGL J1839.30558c, J1840.30413c, and J1841.20459c) and a bright identified source, .
In the first step, we construct an input model using these sources (hereafter “”). We fit data above 3 GeV with the by using gtlike. Figure 1 shows a smoothed count map of a region around G25 above 3 GeV, a corresponding background model map, and the background-subtracted counts map. The background map includes contributions from the modeled point sources, Galactic diffuse emission and isotropic diffuse background whose normalizations are set at the best-fit values obtained by the fit. The -ray excesses associated with the G25 and regions are clearly visible in the background-subtracted map.
Figure 1 might suggest that the -ray emissions of G25 and G27 are extended rather than a combination of 2FGL sources. If the sources are extended, additional point sources (not in 2FGL) are expected to be needed to approximately model the observed -ray emission with a combination of point sources, given that the accumulated time of this dataset (about 57 months) is more than twice as long as that for the 2FGL catalog. To evaluate in detail the spatial distributions of these regions with this dataset, we added multiple point sources to explain the observed -ray emission instead of the 2FGL sources within the regions.
We focus first on G25, because the sum of TS values of the 2FGL sources within the G25 region is larger than that of the region. To determine positions of the multiple point sources, we generate a TS map333A TS map is constructed by stepping a trial point source through a grid of positions, maximizing the likelihood and evaluating the likelihood test statistic of the trial source at each position. of G25 using data above 3 GeV. The input model for the first TS map is constructed by removing the four 2FGL sources within the G25 region from . In the TS map, TS at each grid position is calculated by placing a point source with a power-law energy distribution with photon index set free. Based on the TS map obtained we modify the input model by adding a point source at a grid position whose TS is the largest in the TS map and then re-evaluate the TS map. We iteratively add point sources into the modified model until the peak value of the TS in the map is less than 25 (). After the procedure, we obtain a model with the newly-detected point sources in G25. Next, based on the obtained model, we iteratively add point sources to the region by the same procedure as for G25. We exclude from the model the three 2FGL sources within the region. This modified model is used as an input model for the TS map of G27. Note that in the region is not removed. Since the G25 and regions are spatially close, positions of the multiple sources for each region might affect each other. To minimize the effect, we relocalize the position of each source of G25 in descending order of TS by using pointlike. Then we repeat the same procedure for the sources in G27. Note that the re-localizations do not significantly change the results. Finally we obtain a model with the newly detected point sources in the G25 and G27 regions (“”; see also Table 1).
Note. – Positions of the sources are illustrated in Figure 2.
Note. – Positions of the sources are illustrated in Figure 23.
Tables 1 and 2 list the eight and six point sources that are needed to adequately model the -ray emission from the square regions of G25 and G27, respectively (see also the left panels of Figures 2 and 23). Note that the number of point sources has doubled from 2FGL. The high density of point sources suggests possible extension of the observed -ray emission. To check whether these sources are real point sources or an approximation of extended sources, we evaluate models with extended -ray emission associated with the two regions. The method is described in detail in Section 2.2.2 and Appendix A.
Finally, using the obtained spatial distributions, we measure spectral energy distributions (SEDs) for each source within the regions in the 0.2–500 GeV energy range. We also check the significance of a possible steepening of the SEDs. To do this, we perform likelihood-ratio tests between a power-law function (the null hypothesis), and either an exponentially cut-off power-law or a smoothly broken power-law function (the alternative hypotheses). The exponentially cut-off power-law function is described as
where is 1 GeV. The photon index , cutoff energy , and normalization factor are free parameters. The smoothly broken power-law function is described as
where is 1 GeV. The photon indices below the break, above the break, break energy , and normalization factor are free parameters. The resulting test statistic and provide information on the curvature of the SEDs. The results are described in detail in Section 2.2.3 and Appendix A.
2.2.2 Spatial Distribution
As described in Section 2.2.1, eight point sources are needed if the observed -ray emission of G25 should be explained only by point sources. We modify (see Section 2.2.1) and construct a new model (“”) where we replace the sources in G27 by a spatial distribution estimated in Appendix A. The left panel in Figure 2 shows a background-subtracted map of G25 using the best-fit values of . Here the subtracted background model contains background sources outside the G25 region in addition to the contributions from the Galactic diffuse emission and isotropic diffuse background. In the inset of the right panel, we show a point source simulated using a power-law of photon index 2.1, a typical spectral shape for sources in this region (see Table 5). The high concentration of the point sources in the region rather might be explained by one or more extended sources.
To check for possible extension of the observed -ray emission, we perform the following steps: (i) we adopt as an input model and select the point source of the highest TS among the eight (i.e., in Table 1). (ii) we change the spatial distribution of the source from a point to an elliptical shape and fit the spatial parameters with pointlike. The spectral shape is not changed: a power-law function with its parameters free. In the case that the TS of any of the other point sources decreases to less than 25 as a result, we remove them and refit the spatial shape of the target source. (iii) we evaluate the significance of the spatial extension based on TS (Lande et al., 2012). If the value is larger than 14 ( for three degrees of freedom; Mattox et al. (1996)), we adopt the resulting elliptical shape. In this case, we re-optimize the positions of the remaining point sources in descending order of TS by using pointlike, since a change of the spatial shape might affect the other sources. If , we keep the source as a point source in the model. (iv) we continue to the source of the next highest TS among the remaining sources, and then repeat (ii)–(iv) until the source of the lowest TS is evaluated.
Here we describe details of the application of each step in the procedure. We first fit using an elliptical shape. The resulting elliptical size is so large () that the ellipse includes sources and . Because the TS of the two sources decreases to less than 25, we remove both and refit the elliptical shape. This source is named “G25A” as shown in Table 3. In Table 3, G25A has of 158, which means that the source is significantly () extended. This is not surprising given that the G25A region is explained by three point sources (, , and ) in . We compare the likelihood of the model of the elliptical shape (G25A) and that of the three point sources. The resulting value is . We note that this does not necessarily mean that the elliptical shape explains the observed -ray emission better than the three point sources because the two models are not nested. Here we adopt the elliptical shape, given that it has higher likelihood even though it has 5 fewer degrees of freedom. Next, we select source and fit it using an elliptical shape; was already removed in the previous procedure. Table 3 shows that the resulting elliptical shape “G25B” includes four point sources (, , , and ) in the ( of 126). Since the resulting TS of each of those point sources is less than 25, we remove them and refit the elliptical shape as in the procedure of . We also compare likelihood between models with the G25B region and the four point sources. The resulting value is . Here we adopt the model with the ellipse, given that it has higher likelihood even though it has 9 fewer degrees of freedom. The remaining point source is only . We fit using an elliptical shape and find a of 7, which means that this source is not significantly extended. We keep this source a point source and rename it “G25C”. Finally we obtain “ where we replace the point sources of G25 in for G25A, B, and C.
|Name||Spatial||Center position||Positional||Semi major||Semi minor||AnglebbMeasured counter-clockwise from the Galactic longitude axis to the major axis.||TS||Sources|
|model||(, )||erroraaThe positional error of the center position at 68% confidence level||axis (deg)||axis (deg)||(deg)||in Table 1|
Figure 2 (middle) and Table 3 show the resulting spatial distributions and the corresponding parameters respectively. Since the extended sources G25A and G25B are spatially close, we check whether one elliptical template is enough to explain rays from the two sources. We fit its spatial extent and spectral index. The resulting spatial shape is shown in Figure 2 (right). Here we perform a likelihood ratio test by setting this model (one ellipse) as the null hypothesis and the model of two ellipses (G25A and G25B) as the alternative hypothesis. The improvement of TS is 31, which corresponds to significance of 4.0 given that the alternative hypothesis has seven more free parameters (Mattox et al., 1996; Lande et al., 2012). Note that the model with the two ellipses can take an independent spectral shape for each region, while the model with the one ellipse can take only one spectral shape. The better likelihood might result from such a difference of spectral shapes rather than because of the difference of spatial shapes. Actually, as we show in Section 2.2.3, a part of the G25B region (G25B1) has a much harder spectrum than the other regions. To check this possibility, we add the G25B1 template into the two models and re-fit them. If the above likelihood difference is mainly caused by the spectral discrepancy, the likelihood values of the two models will be similar. The spectral shape of G25B1 is also assumed to be a simple power law. We find that the improvement of TS between the two models is almost the same () as the models without G25B1. In addition, we test a model with a template of the HESS J1837069 region (see Figure 2) instead of G25B1. The resulting likelihood does not change, which indicates that the resulting TS is not sensitive to the spatial shape of the added template.
2.2.3 Spectrum and Temporal Variability
Figures 3, 4, and 5 show spectra of the sources in G25 (G25A, B, and C). The SEDs are obtained by dividing the 0.2–500 GeV energy band into eleven, eleven, and eight logarithmically spaced energy bins respectively. All the sources are fitted with a simple power-law function in each energy bin with the photon index fixed at the value obtained from the broad-band fitting over the 0.2–500 GeV energy range (see below). Since the LAT has better angular resolution at higher energy (Section 2.1), we divide the G25A and G25B regions, which are relatively large () into three sections and evaluate the SED above 3 GeV for each section (see Figure 2). The regions are divided to meet the condition that each section contains a bright peak and has TS . The spectral shapes are fitted with a power-law function. The resulting TS values for G25A1, A2, and A3 are 71, 130, and 52 respectively, while those of G25B1, B2, B3, and C are 66, 61, 54 and 27 respectively. The resulting spectral parameters are summarized in Table 5.
|Name||Parameter||Value||Stat. error||Sys. error aaPropagated uncertainties of the Galactic diffuse model. See text for details.||Sys. error bbPropagated uncertainties of the LAT effective area. See text for details.|
|method 1||method 2|
|G25A||Flux ccThe flux is integrated over 0.2–500 GeV in units of ph cm s.||1.13|
|G25B||Flux ccThe flux is integrated over 0.2–500 GeV in units of ph cm s.||0.60||-0.24/+0.22|
|Name||Parameter||Value||Stat. error||Sys. error aaPropagated uncertainties of the Galactic diffuse model. See text for details.||Sys. error bbPropagated uncertainties of the LAT effective area. See text for details.|
|method 1||method 2|
|G25A1||Flux ccThe flux is integrated over 3–500 GeV in units of ph cm s.||1.5||-0.02/+0.01||-0.07/+0.09|
|G25A2||Flux ccThe flux is integrated over 3–500 GeV in units of ph cm s.||2.0|
|G25A3||Flux ccThe flux is integrated over 3–500 GeV in units of ph cm s.||1.6||-0.08/+0.09|
|G25B1||Flux ccThe flux is integrated over 3–500 GeV in units of ph cm s.||1.0|
|G25B2||Flux ccThe flux is integrated over 3–500 GeV in units of ph cm s.||1.5|
|G25B3||Flux ccThe flux is integrated over 3–500 GeV in units of ph cm s.||1.7|
|G25C||Flux ccThe flux is integrated over 3–500 GeV in units of ph cm s.||0.45||-0.03/+0.05||-0.02/+0.04|
Figure 3 suggests that the spectral shape of G25A is a simple power law in the 0.2–500 GeV band. Quantitatively, the test for spectral curvature shows that and , which means that no significant curvature is present. For the power-law model the photon index is and the integrated 0.2–500 GeV flux is photon cm s. Figure 4 shows clearly that the region G25B has a hard energy spectrum above . This hard spectrum mainly comes from the region G25B1 (see Table 5). Therefore we separately measure SEDs of G25B1 and the other regions: G25B2 plus G25B3 (hereafter G25B). We fit an SED of G25B1 using data; in this energy range the LAT has the best angular resolution and so contamination from the Galactic diffuse emission component and the other sources is expected to be reduced relative to lower energies. The restricted energy range is sufficient for measuring the SED of G25B1 because of insignificant fluxes at lower energy due to the hard spectrum (see Figure 4). The spectral fitting results are summarized in Table 5. We fit the spectrum of G25B in the 0.2–500 GeV band with the SED of G25B1 fixed at the above values. We find no significant curvature in the spectrum ( and ). The photon index is and the integrated 0.2–500 GeV flux is photon cm s. We fit the source G25C using a simple power law. We do not find any significant spectral curvature ( and ) due to the low statistics. The photon index is and the integrated 0.2–500 GeV flux is photon cm s.
We also evaluate the systematic errors for the obtained SEDs. The systematic errors in the spectral analysis are mainly due to uncertainties associated with the model of the underlying Galactic diffuse emission and uncertainties of the effective area of the LAT. The uncertainties of the Galactic diffuse emission are primarily due to imperfections in the Galactic diffuse model and/or the contributions from discrete sources not resolved from the diffuse background. We evaluate the uncertainties of the Galactic diffuse emission by measuring the dispersion of the fractional residuals in ten regions near G25 as shown in Figure 1 (middle), where the Galactic diffuse component dominates. The size of each box is . The fractional residuals, namely (observedmodel)/model, are calculated for each region using the energy range above 1 GeV data. We adopt as the uncertainty of the Galactic diffuse model the second largest value among the ten residuals (90% containment). From the results, the uncertainty of the model is evaluated as 4%. We apply this uncertainty over the entire energy range of 0.2–500 GeV. Systematic uncertainties of the effective area are 10% at 0.1 MeV, decreasing to 3% at 0.3 GeV, constant over 0.3–10 GeV, and increasing to 13% at 500 GeV (Sgro, C. on behalf of the Fermi-LAT collaboration, 2013). Total systematic errors are set by adding in quadrature the uncertainties due to the Galactic diffuse model and the effective area; the systematic errors are dominated by the uncertainties of the Galactic diffuse model below 10 GeV and by the uncertainty of the effective area above 100 GeV. We evaluate the systematic error due to the Galactic diffuse emission for the fitting parameters in the energy range 0.2–500 GeV and 3–500 GeV (see Table 4 and 5) using the above method (method 1). We also evaluate the systematic error using alternative Galactic diffuse models as in de Palma et al. (2013) (method 2). Eight models were tested: the parameters allowed to vary among the models are the radial distribution of the cosmic-ray sources (SNR-like or pulsar-like), the size of the cosmic-ray halo (4 kpc or 10 kpc) and the spin temperature of atomic hydrogen (150 K or optically thin). In this analysis, a single scaling factor is allowed to vary for each alternative diffuse emission model. The obtained systematic errors are marginally consistent with the ones estimated by method 1 (see Figures 3, 4, and 5). We also estimate the systematic error due to the effective area for the fitting parameters by calculating bracketing instrument response functions as in Ackermann et al. (2012). In most cases, the systematic errors due to uncertainties of the effective area are not important compared to those due to the Galactic diffuse model.
We also measure the SED of the region of HESS J1837-069 above 3 GeV. Assuming that G25B is composed of the H.E.S.S. source and the other source, we divide G25B into an elliptical shape of the H.E.S.S. source ( in size; see Figure 2) and the other region. Figure 6 shows the resulting SEDs combined with the SED of the H.E.S.S. source (Aharonian et al., 2006). The figure shows that the LAT SED of HESS J1837-069 smoothly connects to the H.E.S.S. data. The parameters obtained with the power-law model are photon index and integrated 3–500 GeV flux photon cm s. The result is consistent with what we obtained for G25B1 given that G25B1 contains the region of HESS J1837-069 and is slightly larger than HESS J1837-069. On the other hand, Figure 6 also shows that the G25B region except for HESS J1837-069 has a hard spectral tail above GeV, while no clear hardening appears in the SEDs of G25B2 and G25B3 (see Figure 4). This suggests that the spatial region of the spectral hardness is larger than HESS J1837-069, but is mostly concentrated in the G25B1 region with its diameter of . In addition, given that the hard tail remains up to a few hundred GeV, may be larger than the previously reported size, even in the TeV energy range.
To check for time variability in the 0.2–500 GeV range, we divide the whole LAT observation period (about 57 months) into 50, 30, and 4 bins for G25A, B, and C respectively. The number of time bins is determined to meet the condition that the sources in each time bin have . In each time bin, the fluxes of the sources are evaluated by performing a gtlike fit with the Galactic diffuse and isotropic components fixed to the best-fit values obtained for the whole time range in the energy range 0.2–500 GeV. In addition, the photon indices of all sources are fixed at the values obtained from the broad-band fitting. We do not find any indication of variability for any sources in the period spanned by the observations. We also check the time variability of G25B1 in the 3–500 GeV range by performing the same procedure as above and do not find any indication for variability.
3. XMM-Newton Data
3.1. Observations and Data Reduction
The AX J1836.30647 field, centered at (RA, Dec) = (279.078, 6.787) (J2000), was observed with XMM-Newton (Jansen et al., 2001) on 2007 September 18 for about 17 ks. The field is located between the -ray sources G25A and G25B (see the right panel of Figure 2). Data were acquired with the European Photon Imaging Camera (EPIC), which consists of three cameras: MOS1, MOS2, and pn (Strüder et al., 2001; Turner et al., 2001). The full-frame mode and the thick blocking filter were used for the observations. The raw data are processed following standard procedures with the Science Analysis System (SAS) software444http://xmm.esac.esa.int/sas/ version 13.5.0: event files of pn and MOS are extracted and cleaned with epproc and emproc respectively555http://xmm.esac.esa.int/sas/current/doc/epicproc/node3.html. In addition, a count map of the X-ray data is created using emosaic, which combines MOS1, MOS2, and pn images. The spectral analyses are performed with XSPEC (v12.8.1)666https://heasarc.gsfc.nasa.gov/xanadu/xspec.
3.2. Analysis and Results
Figure 7 (left) clearly shows that multiple () point-like sources are concentrated near the center of the field. Such a high concentration strongly suggests that these sources are collectively a stellar association/cluster. To study energy spectra of these sources, we divide the sources into three groups: (i) sources in the magenta circle and outside the cyan circle; (ii) those in the cyan circle; (iii) those outside of both circles (see Figure 7). Information about sources in groups (i) and (ii) is summarized in Table 6. The source positions and the values of HR3 we use are from the 3XMM-DR4 catalog (XMM-Newton Survey Science Centre, 2013). HR3 is a hardness ratio, which is defined as (H-M)/(H+M), where M and H are EPIC count rates in the ranges 1–2 keV and 2–4.5 keV respectively (see the 3XMM-DR4 catalog for more details). We do not analyze sources in group (iii), because most of them must be foreground/background sources unrelated to the putative association/cluster. In group (ii), three bright sources are analyzed; the others are too dim. We extract the events for each source in a circular region with radius displayed in Figure 7 (right). In the spectral analyses, the background region is selected from blank sky within the field of view as shown in Figure 7 (left).
The brightest X-ray source (0.5–12 keV) in this region is P1, and, as shown in Figure 8, clearly displays thermal features. We fit the spectrum with an absorbed thermal model of wabs*apec ( model) using XSPEC (the abundances of Anders & Grevesse, 1989, are used). The resulting parameters are tabulated in Table 7. In the fit, we fix the abundances at 0.2 and 1, both of which can explain the observed data. Since the X-ray image suggests the central sources are members of a stellar association, we also fit the observed spectrum with a typical spectral model for such sources. Here we adopt an absorbed two-temperature thermal model of wabs*(apec + apec) ( model). Typical temperatures are and (e.g., Nazé, 2009; Townsley et al., 2011). In the fit, the normalization of the second component is fixed at 20% of the first component because of low statistics. The ratio of the normalizations is chosen to be a typical value for O stars (Nazé, 2009). Note that the ratio does not strongly influence the resulting parameters (the column density and temperatures change by % when changing the ratio from 10 to 30%). Table 8 shows that both the and models can explain the data. Both models are dominated by the low-temperature ( keV) component and the obtained is almost the same for both models.
Sources Q1–Q15 are relatively dim in the 0.5–12 keV range. Most have similar HR3 values around 0 as shown in Table 6, suggesting they have similar spectral shapes. Given this and their proximities (see Figure 9), we combine the 15 sources to study their general energy spectrum (hereafter Q). First we fit the spectrum with an absorbed power-law function of wabs*pow (PL model). The resulting parameters are tabulated in Table 9. We also fit the spectrum with and models (see Table 7 and 8). In the fit, we fix the abundance and the ratio of the normalization in the same manner as Source P1. All models can explain the observed events. The obtained is different between the and models, because the temperature of the model is very high ( keV) while the low-temperature component is dominant in the model.
Here we evaluate a range of values for the model with a solar abundance. We fix the value of and then fit the model; we re-fit for different values of to find the value where the resulting null hypothesis probability is 5%. This enables us to evaluate a range of where the null hypothesis probability is %. The resulting range of is and for P1 and Q respectively. Note that of Q can be less if is allowed to be very low ( keV). When is set at , the resulting and are keV and 2–4 keV respectively, which are similar to those in Table 8.
|( cm)||(solar)||(keV)||( cm)||( erg cm s)||( erg cm s)|
Note. – The parameters bracketed with [ ] are fixed in the fit. and represent the observed and the absorption-corrected fluxes in the energy bands 0.5–8 keV and 2–8 keV respectively.
|( cm)||(solar)||(keV)||( cm)||(keV)||( erg cm s)||( erg cm s)|
Note. – norm is set at in the fit. The other notes are the same as in Table 7.
|( cm)||( erg cm s)||( erg cm s)|
Note. – The notes are the same as in Table 7.
Sources R1–R4 are also relatively dim but their spectral shapes must be different from Sources Q1–Q15 because of their high values of HR3 (; see Table 6). Therefore we separately combine their events to study their general spectral shape (hereafter R). As expected, the resulting spectrum is heavily obscured in the low-energy band ( keV; see Figure 10). The spectrum is fitted with the model with the temperature fixed at 10 keV and the PL model with the photon index fixed at 1.5, a typical value for non-thermal extragalactic sources. We fix these parameters at reasonable values due to low statistics of the data. In addition we do not apply the model because low-energy photons are too obscured to determine the low-temperature component. The resulting parameters are tabulated in Table 7 and 9. Both models can explain the data with high column densities of cm. Note that we do not analyze sources S1–S4, which have soft spectral shapes (HR3 ), because of the limited statistics.
We also analyze sources in group (ii). Here we combine events of M1 and M2 because of their proximity and low statistics (hereafter M). Since we cannot analyze their detailed spectra because of low statistics (see Figures 11 and 12), we apply the model fixed at the best-fitting spectral shape of . Only and normalizations are set free in the fit. The resulting parameters are displayed in Table 8. Note that the results are only representative since higher statistics are needed to determine the X-ray properties in detail.
4.1. Counterparts to the Gamma-ray Emissions
In our -ray analysis of Section 2.2, the G25 region is divided into two extended regions (regions G25A and G25B) and one point-like source (G25C). For each region/source, we discuss what kind of astrophysical objects are responsible for the observed rays based on the results of our analysis.
4.1.1 Region G25A
G25A is an elongated -ray emitting region (; see Table 3) on the Galactic plane. In Section 2.2.3, we divided the region into three sections (G25A1, A2, and A3) and analyzed them. Given their similar spectral shapes and proximity, they probably originate from the same celestial object (see Table 3 and Figure 2). The observed hard spectral shape () is different from the Galactic diffuse emission (see Figure 3). We infer that the rays of G25A originate not from the Galactic diffuse emission but from a discrete source.
Since PWNe and SNRs are the most prominent candidates for extended -ray sources, we first investigate the possibility that G25A is composed of such sources. To date, no PSR with spin-down luminosity erg s or PWN is known in the G25A region. Although there is no established PWN, the G25A region contains a candidate SNR G24.7+0.6 (right panel in Figure 2). This source was found in the radio band by Reich et al. (1984), who reported extended emission with a size of and a polarized filled central core with a hard spectrum (spectral index of ). They concluded that the radio source is a candidate composite SNR whose PWN is powered by an undetected PSR. Recent observations at 20 cm with better angular resolution (Helfand et al., 2006) confirm that there is a central source of radius. On the other hand, the brightness of the surrounding emission is no brighter than the background emission (see Figure 13). This suggests that the candidate PWN (G24.7+0.6) is the central source. The previously-reported surrounding emission is probably irrelevant to the candidate PWN; it may be diffuse emission from the Galactic plane. No X-ray or TeV detection has been reported from this region. If G24.7+0.6 is responsible for the rays, the -ray spatial size should be much larger than that in the radio band and its spatial shape should be very asymmetric. This suggests that the -ray emission of G25 is unlikely to come from G24.7+0.6.
Next we consider the possibility that G25 is an unknown PWN. In this case, the observed rays are expected to come from the inverse Compton (IC) emission by relativistic electrons. As calculated in Section 2.2.3, the -ray spectrum is represented by a power-law function with a photon index of 2.14 over three decades (0.2–500 GeV). To reproduce such a spectral shape, a distribution of relativistic electrons is required to be where is momentum of the electrons and is about 3.0. The index is much softer than a typical value of 2.0 indicated by multi-wavelength observations of other PWNe. Given the absence of PWNe or energetic PSRs in the regions, the rays of G25A are unlikely to come from a PWN.
We do not find any evidence from multi-wavelength data that SNR shells exist in this region. Molecular clouds are also candidates for discrete extended -ray sources. If relativistic particles escape from nearby acceleration sites such as SNRs, enhanced rays are expected from the molecular clouds illuminated by the accelerated particles (e.g., Aharonian & Atoyan, 1996; Rodriguez Marrero et al., 2008). To check this possibility, we examine whether molecular clouds exist toward the G25A region. We use observations of the CO line, which traces molecular clouds, from the Galactic Ring Survey (GRS; Jackson et al. (2006)). Figures 14 and 15 show the maps for G25, where we integrate them in 5 km s steps from 5 to 125 km s. The figure indicates no molecular cloud covering the entire region of G25A at any distance. It is unlikely that the rays of G25A come from molecular clouds.
There is another candidate for Galactic extended -ray sources: SFRs. LAT observations revealed the extended -ray emissions from the massive SFR Cygnus cocoon (MA11). In Section 4.3, we will discuss in detail the possibility of an SFR origin for the observed rays.
4.1.2 Region G25B
G25B is also an elongated -ray region (; see Table 3) on the Galactic plane. We divided the region into three sections (G25B1, B2, and B3) and analyzed them. As for G25A, the rays of G25B1, B2, and B3 can be ascribed to discrete sources. In addition, we found that G25B1 clearly displays a hard spectral shape (), while the regions G25B2 and B3 have softer spectra (). The consistency of spectral shapes and the proximity of G25B2 and B3 indicate that they probably originate from the same celestial object (see Table 3 and Figure 2). In this section, we first consider sources associated with G25B1 and then G25B2 and B3 (G25B; see Section 2.2.3).
G25B1 is spatially coincident with (see Figure 2). The X-ray observation found PSR J1838-0655 embedded in a PWN with an extent of at the edge of (Gotthelf & Halpern, 2008). The H.E.S.S. source is a TeV PWN powered by this PSR with a spin-down luminosity of erg s. The SED measured by the LAT smoothly connects to that measured by H.E.S.S., which suggests that photons of the LAT and H.E.S.S. data have the same origin (see Figures 4 and 6). G25B1 has a photon index of about 1.5 (see Section 2.2.3), which is consistent with the expected relativistic electron distribution of PWNe: . G25B1 is most likely a PWN powered by PSR J18380655. The -ray luminosity of G25B1 is calculated to be erg s in 3–500 GeV. Here we adopt 6.6 kpc for the distance to following Gotthelf & Halpern (2008) who assume that the source is associated with an adjacent massive star cluster. The ratio of the -ray luminosity to a spin-down luminosity of the PSR is about 3%. Combined with the TeV luminosity, the ratio of the total -ray luminosity to the spin-down energy is about 6%.
As stated in Section 2.2.3, the -ray size of the PWN probably ranges from to , although it is difficult to determine a precise size in the GeV band because of contamination from the G25B2 and B3 regions. In addition, our analysis indicates that the spatial size in the TeV range might be larger than the previously-reported value ( in size), as suggested by a preliminary report based on more accumulated H.E.S.S. data (Marandon et al., 2008). To estimate relativistic particles responsible for the rays, we consider a leptonic model where the rays are from IC scattering. We adopt the simple assumption that the spectral distribution of the electrons has a cut-off power-law function: . The target photon fields for IC scattering are the CMB, infrared, and starlight photons adopted from the GALPROP code (Porter et al., 2008). The temperature and energy density of the CMB, infrared, and starlight photons are eV and 0.26 eV cm, eV and 1.2 eV cm, and 0.30 eV and 3.0 eV cm respectively. We vary the normalization of the electron distribution, index , and a cut-off momentum to reproduce the observed -ray data. As shown in Figure 6, this simple model can explain the observed rays. Given the above discussion, here we use the -ray morphology measured for in the LAT band. Note that the choice of the LAT morphology does not significantly affect the resulting parameters. The obtained total energy of the electrons ( GeV) is erg with the index and TeV .
Next we consider the source of the rays from the G25B region. The spectral shape of the region is similar to that of G25A. In addition, there is no PWN, SNR, or PSR with spin-down luminosity erg s in the region. As for G25A, those celestial objects are unlikely to explain the rays from the region. On the other hand, Figure 14 shows that molecular clouds at km s overlap the regions. However the molecular clouds are spatially much larger than the G25B region. Although we cannot exclude the possibility that a part of the molecular clouds is illuminated by relativistic protons, there is no strong support that the rays come from the molecular clouds.
In summary, G25B1 associated with is most likely a PWN. On the other hand, G25B has no clear association like G25A. Interestingly the region has similar -ray features as G25A (see Section 4.3 for details). In that section, we will discuss the possibility that the rays of the regions originate in an SFR.
4.1.3 Source G25C
The spectrum of G25C () is distinctly harder than the Galactic diffuse emission. This source is detected at high energies GeV and in this energy range its SED is not sensitive to uncertainties of the Galactic diffuse model (see Figure 5). Therefore we conclude that G25C is a discrete -ray object despite its relatively low statistics ( detection) on the Galactic plane. Although we classify it as a point source, G25C has of 7, which might indicate spatial extension. Actually in the residual map (Figure 2), G25C appears as a spatially-extended structure rather than a clear point-like shape. However much greater statistics would be needed to resolve a detailed spatial distribution for this source.
No identified source is associated with G25C. The hard photon index of 2.1 disfavors a PSR origin, since -ray PSRs usually have an energy cutoff of 1–10 GeV (see e.g., Abdo et al., 2013). Interestingly the hard spectral shape of G25C is almost the same as for G25B. Given its spatial proximity and possible spatial extension, G25C may originate from the same celestial object that powers extended source G25B. In addition, possible TeV emission may be seen to the south of the G25B region in the H.E.S.S. image (Figure 2).
4.2. Counterparts to the X-Ray Emissions
In Section 3.2, we reported the X-ray emissions from multiple point sources in the G25.18+0.26 region. Here we discuss counterparts to these X-ray emissions.
Before identifying the sources, we compare our results with past X-ray studies in this direction. ASCA GIS found an unidentified X-ray source AX J1836.30647 at this direction (Sugizaki et al., 2001). The observed flux of the source is erg cm s in a range of 0.7–10 keV, which is 76% of the total flux of sources in group (i) and (ii) obtained in our analysis. Both fluxes are similar and the difference probably comes from the fact that the ASCA source consists of multiple X-ray sources that extend over in diameter; it is difficult to determine precise spectral parameters of such a faint extended source with ASCA. Swift XRT also observed this source for 8 ks on 2007 March, but did not detect it; the upper limit of the flux is erg cm s in 0.3–10 keV (Degenaar et al., 2012). The authors conclude that AX J1836.30647 is a strongly variable or transient source because the upper limit is much lower than the flux reported by ASCA. However they derive the upper limit on the assumption that the ASCA source is a point source. We consider that this incorrect assumption is the reason for the very low XRT upper limit on the source. As stated above, the ASCA flux is almost the same as the total flux in our results, based on the XMM-Newton data that was taken only six months after the Swift observation. This suggests that the total flux of the sources is almost stable.
4.2.1 Sources in Group (ii)
We separately treat the sources in group (ii) from those in group (i), because the group (ii) sources M1, M2, and N1 are spatially coincident with known H ii regions. Figure 16 (left) shows that source N1 is coincident with H ii region G25.294+0.307, which has of 39.6 km s (Lockman, 1989). The velocity corresponds to a kinematic distance of 2.8/12.6 kpc. The region appears surrounded by an IR bubble called N37 (Churchwell et al. (2006); see also the right panel in Figure 16). This is generally understood to be a phenomenon due to winds of putative massive star(s) sweeping up the surrounding dust to create a bubble, ionize the region, and illuminate the bubble by heating the dust. The H ii region and the bubble are probably associated with the young cluster Alicante 6 reported by Marco & Negueruela (2011). They find a significant population of massive stars located at a distance of kpc by using optical and IR data. The estimated distance is consistent with the closer kinematic distance of the H ii region. They find that most massive stars of the cluster are of later types than O stars and only two stars are O-type (O7 II and O7 V) stars (see Figure 16). They also find that a number of B0-1 V stars concentrate in the cavity of the bubble. Such spatial coincidence strongly suggests that the stars are physically associated with N37. The authors also state that the O7 II star located near N37 is a possible ionizing source to create the bubble. Given the facts listed here, the H ii region G25.294+0.307 and the probable associated bubble N37 are most likely to be associated with the young cluster Alicante 6.
Figure 16 shows that source N1 appears to be spatially confined within the bubble N37. Given the spatial association, source N1 is likely to be associated with the bubble. If this is the case, the observed X-ray luminosity is calculated to be erg s in 0.5–8 keV. The X-ray image may suggest that the source N1 is extended and consists of unresolved multiple sources, given that the PSF of the EPIC has a full width at half maximum of .777http://xmm.esac.esa.int/external/xmm_user_support/documentation/ uhb_2.1/node14.html
Sources M1 and M2 are coincident with the H ii region G25.220+0.289 recently found by Anderson et al. (2011) as shown in Figure 16 (left). The H ii region has of 42.4 km s, which is similar to that of G25.294+0.307. This indicates that the sources M1 and M2 are also associated with Alicante 6. Interestingly the O7 II star of Alicante 6 is located between the H ii regions G25.294+0.307 and G25.220+0.289. If the star ionizes G25.294+0.307 as suggested by Marco & Negueruela (2011), G25.220+0.289 may be also ionized by this star. Actually the 8 m IR map shows a filament-like structure coincident with G25.220+0.289, which may be swept up and excited by the O star (Figure 16). Given the observational evidence, the H ii region G25.220+0.289 may be associated with Alicante 6. The observed total X-ray luminosity is calculated to be erg s in the 0.5–8 keV band. We note that the sources M1 and M2 have no counterpart massive stars in Alicante 6. This seems inconsistent with our expectation that the sources M1 and M2 are massive stars based on the derived high X-ray luminosity. This might be because molecular clouds obscure the sources in the optical wavelength. Actually Table 8 shows that obtained with the model is , which corresponds to the optical extinction mag, when we apply the relation (Ryter, 1996). More X-ray statistics would be needed to determine reliable values of the spectral parameters for the interpretation.
4.2.2 Sources in Group (i)
The sources in group (i) clearly concentrate around the center of the magenta circle in Figure 7. As displayed in Table 8, the of sources P1 and Q have similar values of cm for a typical spectral shape of OB-association members, when we adopt the model (see Section 3.2). This suggests that the sources are located at the same distance: they are associated in physical space. In the model, is almost the same as that of the model for source P1 and the low-temperature component () is dominant in both models. The choice of the models does not affect the interpretation of P1. On the other hand, toward Q in the model is cm, which is significantly different from cm for the model. In the model, the obtained temperature of 6–7 keV is too hot for usual stellar association members. In addition, such a low means that sources Q1–Q15 are most likely to be nearby (within a few kpc). We can roughly estimate to be mag based on the using the relation (Ryter, 1996). Given such a low extinction and the close distance, most members of the association G25.18+0.26 should be detectable in the optical band. This is inconsistent with the fact that no clear optical counterpart has been found for G25.18+0.26. Therefore the model is unlikely to represent physical features of Q. The spectrum of Q also can be explained by the PL model with of and . Typical non-thermal extragalactic sources are represented by a PL model with of 1.5–2.5. However most of the sources Q1–Q15 must not be extragalactic sources because the derived is lower than the total column density of H i in this direction ().888http://heasarc.nasa.gov/cgi-bin/Tools/w3nh/w3nh.pl Such a concentration of non-thermal sources is unlikely in our Galaxy. Therefore we conclude that the model best represents the physical parameters of sources P1 and Q1–Q15 and that these sources consist of a stellar association/cluster in our Galaxy. Hereafter we call this object G25.18+0.26.
The young cluster Alicante 6 is also located in this direction (see Section 4.2.1). However the sources in group (i) are unlikely to be associated with this object for the following reasons. First if they are really associated, the absorption-corrected X-ray luminosities of the individual sources are in the range erg s in the 0.5–8 keV band. To calculate the X-ray fluxes, we adopt the model with abundance 1.0 in Table 8. The fluxes of sources Q1–Q15 are calculated by fitting the normalizations with the best-fitting model for source Q with the spectral shape fixed. The obtained high luminosities mean that most sources in group (i) should be massive stars. As shown in Figure 16 (left), however, the massive stars of Alicante 6 do not coincide with the X-ray sources, indicating that they are not associated. In addition, the morphology of the IR image suggests that the most luminous star is likely to be located near the center of the cyan circle (the group (ii) region; see Section 4.2.1). However the X-ray image strongly indicates that the most massive stars concentrate around the center of the magenta circle, if we assume that sources in the groups (i) and (ii) are located at the same distance. The inconsistency also supports our conclusion that the OB association/cluster G25.18+0.26 is not associated with Alicante 6.
Given that there is no optical counterpart to G25.18+26, the OB association is likely to be farther away than Alicante 6 (). of sources P1 and Q are , which is relatively high given that the total Galactic H i column density at this direction is about . This suggests that G25.18+0.26 is at a distant location, which is consistent with our expectation. However we should note that the values of corresponds to mag when we adopt a relation of (Ryter, 1996). In this case the counterparts to the X-ray sources are expected to be visible in the optical band, which is inconsistent with the observational evidence. This may be explained by an uncertainty of the relation between and . Actually Vuong et al. (2003) indicates . When we adopt this relation, the converted is mag. When we adopt the highest estimated value of of mag, the OB association is expected to be invisible in the optical band. Another possible explanation is that the actual values are higher than the obtained ones. Our analysis shows that could go up to (Section 3.2). In addition, the value of (Q) is calculated for combined spectra of Q1–Q15 because of the limited statistics. Some of the integrated sources might be foreground ones, which would make the value of smaller than the real values of the OB association G25.18+0.26. More observations in the X-ray and at other wavelengths will determine the precise distance of G25.18+0.26.
Rahman et al. (2013) claim another candidate OB association, SFC 10, in this direction using near-IR observations, which are less affected by obscuration than the optical band. They select the candidate association based on an observed excess number density of IR sources compared with that of the surrounding region. The reddening of the selected stars indicates that the association is distant ( kpc). The claimed association has a larger size than that of this work but the most dense part is almost spatially coincident with the X-ray association (see Figure 18). This may indicate that the IR and X-ray association have the same origin. More studies are needed to confirm this indication.
The spectrum of R indicates of cm independent of the choice of model (see Section 3.2). The column density is much higher than those of sources P1 and Q () or the total Galactic H i column density in this direction ( cm). This means that sources R1–R4 are embedded in dense gas and obscured with mag using the relation . One potential explanation for this obscuration is that these sources are members of G25.18+0.26; young associations/clusters usually have such sources obscured by their parental molecular clouds. Another interpretation is that the sources are extragalactic. To determine their associations, we would need more data to precisely study individual sources.
4.3. SFR Scenario
As discussed in Section 4.1, there are no clear associations with G25A, B, or C. Interestingly the G25A and G25B regions show similar characteristics: elongated morphologies with similar surface brightnesses and hard energy spectra. Given their proximity, one celestial object plausibly could be responsible for the -ray emission of both regions. Although G25C may have similar characteristics, we need more statistics to study it in detail. Here we discuss -ray emission of G25A and G25B (hereafter G25). Note that the exclusion of G25C does not change our conclusion in the discussion.
We propose a scenario that the observed -ray emission is due to a massive SFR. The first evidence to support the scenario is that the -ray properties of G25 are similar to those of the Cygnus cocoon, the only firm case of -ray detection from an SFR in our Galaxy (MA11): both -ray sources are spatially extended; their energy spectra are described as a power law with hard photon indices of 2.1–2.2 without any significant spectral curvature at least up to GeV; because of their large sizes (–), the LAT is able to obtain SEDs of five sub-regions for each source and finds a uniform spectral shape () in the regions. These similarities suggest that the -ray production process in G25 might be similar to that in the Cygnus cocoon, i.e., a massive SFR.
In this section, we will provide three more lines of observational evidence to support our proposition. First, if a massive SFR exists in this region, an accompanying bubble structure is expected. Although there is no clear evidence that a massive SFR is associated with this region, Rahman & Murray (2010) recently claim a candidate massive SFR in this direction based on radio and IR observational data. In Section 4.3.1, we investigate the surrounding gas distributions to confirm the bubble structure that they report. Second, if the bubble mentioned in Section 4.3.1 constitutes an SFR, a massive OB association/cluster is expected in this region. Actually the X-ray observation in the present work reveals that the OB association/cluster G25.18+0.26 resides in the region (see Section 4.2). In Section 4.3.2, we provide observational evidence in support of the possibility that the bubble is an SFR created by the OB association G25.18+0.26. Finally, if the observed rays come from the SFR proposed in Section 4.3.2, the rays are expected to be largely confined in the bubble (i.e., a cavity delineated by the claimed bubble). This is a morphological property revealed by the LAT observation of the SFR Cygnus cocoon. In Section 4.3.3 we examine relations between the spatial structures of the rays and the claimed bubble, to check whether the morphology follows our expectation. Note that the mechanism of the confinement will be discussed in Section 4.5.2.
4.3.1 Bubble Structure
Toward the direction , Rahman & Murray (2010) find copious 8 emission, which has a bubble structure associated with H ii regions with line-of-sight velocities of 100 km s. Their interpretation is that the emission is associated with a star-forming complex created by hidden massive OB association(s). They estimate the distance to the bubble as 6.1 kpc based on the median velocity ( km s) of the H ii regions.
Figure 17 shows that G25 is spatially coincident with the reported bubble, although a factor of larger. To study the bubble structure in detail, we investigate the CO map for the velocity range around km s in this direction (Figure 18). Figure 18 clearly shows a bright arch-like structure around for in the range 110–120 km s. The structure can also be seen in the IR map and composes the western part of the claimed bubble (hereafter cardinal directions are in Galactic coordinates unless otherwise mentioned). We also overlay H ii regions with km s around this region. The positions and velocities of the H ii regions are taken from the Green Bank Telescope H ii Region Discovery Survey (Anderson et al., 2011) and the Boston University Catalog999Available at http://www.bu.edu/iar/files/script-files/research/hii_regions/index.html. The latter compiles previous studies of ultra-compact H ii regions (Araya et al., 2002; Watson et al., 2003; Sewilo et al., 2004), classical H ii regions (Lockman, 1989), and H ii regions that are truly diffuse (Lockman et al., 1996). Some of the H ii regions are located on the arch (Figure 18). These can be naturally interpreted as swept-up gas illuminated by putative massive stars. These results support an interpretation that the arch is a part of the bubble created by a massive OB association.
Based on the CO data, the arch structure is expected to exist at least from 110 to 120 km s and probably beyond 124 km s. Such large dispersion of the velocity is difficult to be explained only by the motion of the Galactic arm, suggesting that the dispersion is probably explained by local motion of the clouds. The mean velocity of the gas distribution is km s, which is higher than the velocity at the tangent point (113 km s) for the direction . Figure 18 shows the existence of clouds above the velocity at the tangent point, which means that these clouds locally move away from us. This motion may be caused by an expansion of the bubble. Since the arch-like molecular clouds are located near the tangent point and a part of them exists above the velocity at the tangent point, here we adopt the distance to the tangent point ( kpc) as the distance to the clouds. The estimated distance of 7.7 kpc is greater than the distance estimated by Rahman & Murray (2010) (6.1 kpc). The authors derive the distance from the mean velocities of eight H ii regions which they assume are associated with the bubble. However one of the H ii regions has of 84.8 km s. As discussed above, a possible bubble is expected to exist around of 110 km s. Even considering the dispersion of for the bubble, the associated objects are most likely to exist in the velocity range of 95–125 km s. Therefore the H ii region with of 84.8 km s is unlikely to be associated with the bubble. When we exclude this object, the mean velocity of the other seven H ii regions is 111 km s, which is very close to the velocity adopted in this paper (113 km s). Therefore we adopt 7.7 kpc for the distance.
Using the CO map, we confirmed the western shell structure of the claimed bubble. On the other hand, we cannot find any clear shell structures for the other parts of boundaries of the bubble (see Figure 18). Rahman & Murray (2010) delineate the boundary based on the morphology of the IR image and positions of the H ii regions, which they consider associate with the bubble. Except for the clear western shell structure, however, no bright shell structure is found in the IR map either. Therefore the other boundaries may be located at different positions from the assumed ones. Actually we find that an arch-like structure extends from the western part to the northern one for in the range of 114–119 km s (see Figure 18). The arch exists outside the boundary of the candidate bubble. At the southern boundary, we find the H ii regions with km s around , which is outside the claimed bubble. This may suggest that the southern boundary of the bubble is located at or beyond these H ii regions. These findings indicate the possibility that the bubble is more extended than that defined in Rahman & Murray (2010): we delineate a possible boundary of the bubble that is in size as shown in Figure 18. Hereafter we call this structure “the G25 bubble”. Note that the eastern boundary passes thorough molecular clouds at beyond the velocity at the tangent point (113 km s). These high-velocity clouds may be pushed away and compressed by a putative massive OB association/cluster. Figure 15 shows that these high-velocity clouds concentrate around , which supports this possibility.
The bubble with a size of has a physical size of pc at 7.7 kpc. This size is comparable to the thickness of a dense part of the Galactic plane, which means that ambient gas and clouds in the north and south are generally expected to be more sparse than those in west and east on the Galactic plane. Figures 18 and 19 show that the intensities of the CO and the H i emissions at the northern and the southern boundaries are indeed less than those in the western and the eastern boundaries. Given the non-uniform gas distributions, the bubble is expected to be elongated in the Galactic-latitude direction if we assume that the bubble is created by a powerful stellar object on the Galactic plane. The morphology of the possible boundary is consistent with this expectation.
We also investigate the spatial distribution of H i gas using the VLA Galactic Plane Survey (VGPS) (Stil et al., 2006). Figure 19 shows that a spatial distribution of H i is very similar to that of CO above 110 km s in Figure 18. At lower velocities the spatial distribution of H i gas is not similar to that of the molecular clouds: we cannot find any clear spatial correlation between them for the other velocities at this direction. If we assume a powering source such as a massive OB association or/and an SNR around , then the wind of the source would sweep the surrounding molecular clouds and H i gas away; such pressure would make a bubble whose spatial structures are similar regardless of their initial distributions. Therefore the clear spatial correlation between the molecular clouds and the H i gas can be naturally understood if they are shells of the bubble created by an OB association or/and an SNR.
4.3.2 Does OB Association G25.18+0.26 Create a Bubble?
Intense free-free emission is detected from and around this region (Murray & Rahman, 2010). Since free-free emission is expected to be mainly due to reprocessed ionizing photons emitted by young massive stars, the intense emission suggests that massive OB association(s) reside in the region. Therefore an OB association/cluster is the most prominent candidate for powering the G25 bubble (Section 4.3.1), although no such celestial object has previously been found within the bubble at the distance around 7.7 kpc. A candidate for such massive OB associations/clusters is the newly-found OB association/cluster G25.18+0.26 (see Section 4.2). In the following, we provide observational evidence to support this possibility. In the case that G25.18+0.26 creates the bubble, the radius of G25.18+0.26 () will be physically large, 13 (/7.7 kpc) pc. The relatively large size indicates that the object is an OB association rather than an OB cluster (see e.g., Portegies Zwart et al., 2010). Hereafter we call the system of the G25 bubble created by the OB association G25.18+0.26 “SFR G25”.
Since an OB association pushes the surrounding gas away to create a bubble, it should be located inside the bubble. G25.18+0.26 meets this criterion as shown in Figure 20. of sources P1 and Q are (see Table 8), which is relatively high given the fact that the total Galactic H i column density at this direction is about . This suggests that G25.18+0.26 has a distant location. Our assumed distance of 7.7 kpc is consistent with this expectation. However we note that the value of seems to be small for 7.7 kpc, given that the association is located in the inner Galaxy (). If no foreground molecular clouds obscure the association, the obtained may be consistent with . The issue of the value and the distance is discussed in Section 4.2.2. Here we assume that G25.18+0.26 is located at 7.7 kpc.
The physical size of the G25 bubble is large, pc. To create such a large bubble, the responsible OB association should be massive. To check whether the OB association G25.18+0.26 is massive, here we roughly estimate a total mass from its X-ray luminosity. The X-ray luminosity functions (XLFs) of the OB associations/clusters are known to have similar shapes (e.g., Wright et al., 2010); the similarities could be understood in terms of a universal shape of the initial mass function for OB associations and an empirical relation between the mass and the X-ray luminosity of a star. Therefore we expect that the X-ray luminosity of the OB association is roughly proportional to its mass (see also Appendix B).
We estimate the X-ray luminosity of G25, assuming that sources P1 and Q1–Q15 are members of G25.18+0.26. Note that R1–R4 are not included since their association with G25 is uncertain (see Section 4.2.2). The total luminosity of the sources is estimated to be of erg s in the 2–8 keV band using the total absorption-corrected flux of sources P1 and Q based on the model with the solar abundance (see Section 3.2). The luminosity is integrated above 2 keV to reduce the dependency on . The luminosities of the individual sources are in the range erg s (see Section 4.2): the estimated luminosity is integrated for X-ray sources with luminosities above erg s in the 2–8 keV band. Next we estimate the X-ray luminosity of a known OB association for a comparison with that of G25.18+0.26. Here we adopt Cyg OB2 because it is one of the most massive OB associations (; Wright et al. (2010)) and located at a distance of only kpc (Hanson, 2003). In Appendix C, we estimate the corresponding luminosity as erg s.
Using the derived luminosity, we roughly estimate a total mass of G25.18+0.26 to be , where and are the 2–8 keV luminosities of G25.18+0.26 and Cyg OB2 respectively, and is the mass of Cyg OB2 of . Note that we may miss some X-ray sources of G25.18+0.26 above erg s due to the sensitivity limit of the XMM-Newton observation. In addition at least some of sources R1–R4 may be members of the association. Given these considerations, we probably underestimate the total luminosity of G25.18+0.26. However such effect would be a factor of and not crucial for this rough estimation. In any case, the derived mass is comparable to that of Cyg OB2, which means that G25.18+0.26 is one of the most massive OB associations in our Galaxy if it is located at 7.7 kpc.
4.3.3 Gamma-ray Association with the G25 Bubble
As shown in Figure 20, the -ray emission of G25 is spatially well matched with the G25 bubble suggested based on the molecular-cloud map in Section 4.3.1. In the following, we compare spatial structures between the -ray emissions and molecular clouds. In Figure 20 (left), the G25 emission appears to extend over the boundary of the G25 bubble at the northwest. Interestingly the molecular clouds with the highest velocities (the right panel) have a spatial break at the northwestern boundary and a bright part of the -ray emission appears to extend through the break. This can be interpreted as a “breakout” of the bubble: the rays may come from the outflow of high-energy particles. In the western and southwestern parts the rays and molecular clouds are anti-correlated: the rays appear to extend in cavities between dense parts of the molecular clouds. In the southeastern and northeastern parts, bright -ray emission comes from just inside the molecular-cloud boundaries but not within the clouds.
We can summarize the relations between the spatial structures of -ray emission and molecular clouds as below: (i) the rays are anti-correlated with the molecular shells and clumps so that the rays appear to be confined in cavities delineated by dense shells of the G25 bubble (except for the northwestern boundary where a breakout may occur); (ii) the bright -ray emission comes from regions near (but not on) the dense molecular-cloud shell or clump (i.e., the western and eastern boundaries); (iii) no strong -ray emission is detected near the sparse molecular-cloud boundaries (i.e., the northeastern and southern boundaries). The spatial properties show that the -ray emission is confined in the G25 bubble, which is consistent with our proposition that the rays come from SFR G25. These spatial properties will be compared with those of the Cygnus cocoon SFR in Section 4.5.1.
4.4. High-energy Particles in SFR G25
Here we discuss spectral distributions of the high-energy particles radiating the observed rays. We adopt the SFR scenario of Section 4.3 and focus on the rays of the G25A and G25B regions. Model spectral distributions of the relativistic particles are adjusted to reproduce the observed rays. We consider three kinds of emission mechanisms: the -decay emission due to high-energy protons, and the Bremsstrahlung and the Inverse Compton (IC) scattering processes by high-energy electrons. To calculate the -ray emission via these mechanisms, we adopt the target gas density () and the target photons estimated in Appendix D.
The spectral distributions of the relativistic particles are assumed to be cut-off power-law functions of the form . The spectral index of is assumed to be the same between the relativistic protons and electrons, while the cutoff momentum of