GeV Emission in the Region of HESS J1809−193 and HESS J1813−178: Is HESS J1809−193 a Proton Pevatron?

In order to take advantage of the improved resolution at higher energies, we studied the morphology of the sources above 5 GeV. As a first step, we optimized the location and shape of new significant sources seen in the residuals maps. Later, when the analysis was done at lower energies, even more sources (with soft gamma-ray spectra) had to be added to the region of interest to improve the fit and fully consider their effect. Whenever a point source was added to the model, the tool gtfindsrc was used to optimize its location. When instead an extended excess was found, a systematic search for the maximum TS value was done by fitting a uniform disk template along a grid of positions and using a range of radii at each position (changing each in steps of 01). For simplicity, a simple power-law spectrum is used for additional background sources, fitting also the normalization and spectral index each time.

The unidentified 3FGL sources 3FGL J1810.1−1910, 3FGL J1811.3−1927c in the region of HESS J1809−193, as well as 3FGL J1814.0−1757c and 3FGL J1814.1−1734c possibly associated with HESS J1813−178, were removed from the model. They all have a relatively flat spectral energy distribution (SED) and thus are probably part of the extended emission found here with a more detailed analysis. The source 3FGL J1808.5−1952, to the SW of HESS J1809−193, remained in the model, as it has a soft GeV spectrum and is associated with a globular cluster and thus is not part of the TeV source. In the fit, the normalizations of the 3FGL sources located within 7° from the center of the region of interest were set free, while the rest were fixed to the cataloged values. Based on the inspection of the residual analysis obtained after an initial fit with the 3FGL model, it was evident that additional sources from the hard LAT source catalog (3FHL) in the region of interest needed to be added, these included extended sources such as 3FHL J1804.7−2144e (containing the unidentified TeV source HESS J1804−216), 3FHL J1800.5−2343e in the W28 region, and the point source 3FHL J1801.5−2450, classified as a pulsar in the W28 region (Ajello et al. 2017). Both the normalization and spectral index of these 3FHL sources were kept free in the fits.

After improving the model in this way for events above 5 GeV, the new residuals show apparently extended excess emission in the regions of HESS J1809−193 and HESS J1813−178. The GeV excess at the location of HESS J1809−193 seemed to be elongated, approximately following the TeV emission. Figure 1 shows a TS map obtained for HESS J1809−193 using a point source hypothesis, which reveals significant emission.

Zoom In Zoom Out Reset image size

A detailed study of the morphologies was carried out and the results are shown in Table 1. We used different hypotheses for the morphology of the emission, including the point sources reported in the 3FGL catalog for HESS J1809−193. In the case of HESS J1813−178, the point sources from the 3FGL catalog are located relatively far away from the peak of the excess and thus new locations were optimized with the larger data set used here. For HESS J1809−193, the TeV template obtained from the count map shown in the H.E.S.S. discovery paper Aharonian et al. (2007) was also used.

For both H.E.S.S. sources, considerable residuals are seen after subtracting the point source hypothesis model from the data. As can be seen in Table 1, the quantity TS_disk − TS_point = 2 × (log(L_disk) − log(L_point)), a measure of the source extension, is much greater than 16, the typical threshold used to claim that a source is extended (e.g., Acero et al. 2016). Additionally, in both cases, the TS of the disk morphology is greater than the TS of the two point sources, which is another criterion that has been applied to distinguish between an extended source and a combination of point sources (The Fermi LAT Collaboration et al. 2017). The best-fit disk radii, spectral indices, and center locations (J2000) for HESS J1809−193 and HESS J1813−178 are 05 ± 015, 2.22 ± 0.12, R.A. = 18^h10^m22^s, decl. = −19°26' (±02), and 06 ± 006, 2.07 ± 0.09, R.A. = 18^h13^m10^s, decl. = −17°37' (±005), respectively. The radii and location errors are given at the 3σ level.

After subtracting the best-fit model with the uniform disk hypothesis for HESS J1813−178 to the data some photons are observed clustered at the location of the TeV peak. We fit the position of a new point source on top of the disk template for HESS J1813−178, and as can be seen in the table there is some evidence at the 3σ-level for the contribution of a point source besides the extended emission (TS_disk+point − TS_disk = 18, for 4 additional degrees of freedom). The location of this point source was found to be R.A. = 18^h13^m51^s, decl. = −17°51' (and 3' 1σ error circle radius), which is consistent with the peak of the TeV emission. However, the individual TS of the point source obtained in a fit was slightly below the threshold of 25 to claim a detection. This was also true in the other energy intervals studied and we decided to only use a disk in the final model for HESS J1813−178.

For HESS J1809−193 and HESS J1813−178, the extended templates provide a much better fit compared to the point-source hypothesis. In the case of HESS J1809−193, both the disk morphology and TeV counts map template can be considered good representations of the GeV morphology. We used the disk template to estimate the spectral parameters and used the TeV counts map template to estimate a systematic uncertainty that we added to the statistical uncertainty for those parameters. We conclude that extended emission is significantly detected in the region of the TeV sources.

When we fit the model found previously to data in the energy interval 1–500 GeV, several new residuals unrelated to HESS J1809−193 and HESS J1813−178 appeared in different parts of the region of interest, most notably at the location of the unidentified TeV source HESS J1808-204, and close to the point source 3FGL J1809.2−2016c. The emission seems to be extended and we optimized the location and size of a disk template to account for it. The center position and radius of the resulting disk were R.A. = 18^h08^m23^s, decl. = −20°30' and 03, respectively. The source significance and spectral index were 8σ and 3.1 ± 0.7 above 1 GeV. Similarly, we added other sources that are relatively significant and have very soft spectra (with spectral indices in the range 2.6–3.1) at the coordinates (all given in degrees) R.A., decl.: 273.53, −17.8 to the East of HESS J1813−178, 271.65, −21.345 close to W30, 274.492, −20.079 outside the Galactic plane and 275.08, −16.21 about 22 NE of HESS J1813−178. Finally, a 03-radius disk template was found to the south of the region of interest, with the center located at the coordinates R.A. = 272.96, decl. = −24.106 (with spectral index ∼2.8 and overall significance ∼11σ), having the unidentified source 3FGL J1810.8−2412 at the border, both of which might be related to SNR G7.5−1.7. The addition of all these sources was found to have no effect on the morphological studies done at higher energies.

We chose the energy range 0.5–500 GeV to make the final spectral analysis because substantial soft emission not accounted for by our model was seen below 500 MeV near HESS J1809−193, and that complicated the analysis at the lowest energies. The residuals in the final energy range chosen were acceptable and therefore the model was a good representation of the data. As shown in Table 2, the spectra of the GeV sources coincident with HESS J1809−193 and HESS J1813−178 are better described by a simple power law with fewer degrees of freedom than the other models corresponding to "curved" spectra. The spectral indices and integrated fluxes are, respectively, for HESS J1809−193 and HESS J1813−178, 2.22 ± 0.08, (1.68 ± 0.1) × 10⁻⁸ photons cm⁻² s⁻¹ and 2.14 ± 0.04, (2.0 ± 0.11) × 10⁻⁸ photons cm⁻² s⁻¹. Table 2 shows the final TS values, which correspond to detection significances of 19.5σ and 21.6σ for HESS J1809−193 and HESS J1813−178, respectively.

We binned the LAT data in 10 energy bins equally spaced logarithmically in the 0.5–500 GeV range to measure the source flux and plot spectral points next to the global fit. Within each interval, we fixed the spectral indices of the sources associated with HESS J1809−193 and HESS J1813−178 to the value found above and fit the normalization to calculate the photon flux. 95%-confidence level upper limits on the flux are calculated for intervals without a source detection. The individual flux values are consistent with the fits in the entire energy range, and we present the SED models in the following section.

The gamma-ray spectrum of the emission in the region of HESS J1809−193 is extended and shows a spectral index of 2.22 ± 0.05 that is entirely consistent with that measured at TeV energies, 2.23 ± 0.05 (Renaud et al. 2008). Motivated by this fact, we fit the GeV–TeV spectral points to a radiative model from hadronic interactions through Markov Chain Monte Carlo sampling of likelihood distributions of the fit parameters (with a simple power-law particle spectrum) using the numerical code Naima (Zabalza 2015). This code implements a parametrization of the gamma-ray production cross-sections for pp interactions from Kafexhiu et al. (2014). The resulting best-fit particle spectral index and its 1σ error was 2.22 ± 0.02 (χ² = 27 for 14 degrees of freedom). This result is not surprising since the gamma-ray spectrum follows the particle distribution and both of their spectral indices should be similar. Using a particle distribution that is a power law with an exponential cutoff, the resulting cutoff energy in the fit tends toward the upper bound in the allowed range when the range is increased, and thus the data were not able to constrain it. The data and model are shown in Figure 2 for a particle distribution with a cutoff energy of 1 PeV. It should be kept in mind that the simplified model shown here using one single population may be an oversimplification of the real situation. A complete explanation of the morphology of the gamma-ray emission should involve the modeling of cosmic-ray diffusion, which was not taken into account here.

Zoom In Zoom Out Reset image size

The hadronic scenario for HESS J1809−193 is supported by the recent observations of Castelletti et al. (2016) and this work proves that this source could be a cosmic-ray pevatron. From the same fit using Naima (Zabalza 2015), it was found that the total energy needed in the high-energy protons is

$$\begin{eqnarray*}&&6.2\times {10}^{49}\,{\rm{erg}}\,\left(\displaystyle \frac{10\,{{\rm{cm}}}^{-3}}{n}\right){\left(\displaystyle \frac{d}{3{\rm{kpc}}}\right)}^{2}\end{eqnarray*}$$

for a source distance d (in kpc) and a target density n (in cm⁻³).

Typical kinetic energies in SNRs are of the order of 10⁵¹ erg. It is believed that around 10% of this energy can be transferred to cosmic rays (although this value is highly uncertain, see, e.g., Dermer & Powale 2013) and therefore using this conversion efficiency an average density of 6 cm⁻³ is required for a source distance of 3 kpc. As mentioned before, Castelletti et al. (2016) made a thorough analysis of ambient structures in the region of the TeV source. They discovered a system of molecular clouds physically related to the shock front of the SNR G11.0−0.0 at an estimated distance of 3 kpc, coincident with the peak of the TeV emission and with a total proton density in the range 2–3 × 10³ cm⁻³.

The possibility that HESS J1809−193 is accelerating protons to PeV energies is extremely interesting and relevant to the theory of cosmic-ray origin. Although the current paradigm for the origin of cosmic rays requires that SNRs act as pevatrons for at least some time during their evolution, no known SNR shows a gamma-ray spectrum that is compatible with this idea. Only one unidentified pevatron candidate is known, and it is located in the Galactic center region (HESS Collaboration et al. 2016). However, a considerable fraction of the TeV emission from this region can alternatively be explained by leptonic emission from a pulsar population expected to exist near the Galactic center (Hooper et al. 2017). The results presented here do not, of course, prove that HESS J1809−193 is a pevatron but give evidence in favor of a hadronic scenario for the gamma-ray data (specifically at GeV energies). The nature of the GeV to TeV emission has to be confirmed as hadronic in the future (using more detailed multiwavelength observations and modeling) and the spectrum of the source needs to be accurately measured at tens of TeV to confirm the H.E.S.S. results and to explore if the spectrum extends to higher energies than those measured so far. For this purpose, currently operating TeV observatories such as the High-altitude Water Cherenkov Gamma-ray Observatory (Abeysekara et al. 2013) may be able to measure the source spectrum at the highest energies, and future TeV observatories such as the Cherenkov Telescope Array (Hermann & CTA Consortium 2011) will offer a substantial improvement in resolution that is crucial to locate the emission regions. The confirmation of the acceleration of cosmic rays to PeV energies in the HESS J1809−193/G11.0−0.0 system would be a considerable advancement in the understanding of the origin of cosmic rays, as it would be the first case involving a known SNR.

3.1.1. Alternative Scenarios: Contributions from Several SNRs

The SNRs that are seen in the region have small angular sizes and the emission from their shells would be close to point-like for Fermi (see Figure 1). The gamma-ray emission in the region of HESS J1809−193, on the other hand, is better described by an extended hypothesis. One may explain this emission by proposing the existence of a previously unknown extended SNR in the region to account for the gamma-rays. However, deep radio imaging does not show evidence for such an SNR (Castelletti et al. 2016), which would be visible at those wavelengths through synchrotron radiation.

Considering the fact that the known SNRs in the region of HESS J1809−193 are located close to each other in the sky, it might be possible that the LAT observations cannot distinguish the gamma-ray contributions from them. We point out that the shape of the SED at gamma-ray energies can be explained by leptonic or mixed scenarios involving leptonic and hadronic contributions from several or all of the SNRs in the region.

The SNRs are likely located at different distances. For G11.4−0.1, Pavlovic et al. (2014) and Brogan et al. (2004) give values of 8.4 kpc and 9 kpc, respectively, estimated from the radio surface brightness–diameter (Σ − D) relation. Some distances found in the literature for the other SNRs are 8.3 kpc for G11.1+0.1 implied by the same relation (Pavlovic et al. 2014), although if G11.1+0.1 is associated to the pulsar PSR J1809−1917 its distance would be 3.7 kpc (Morris et al. 2002), and 9.8 and 3.0 kpc for G11.0−0.0 (Pavlovic et al. 2014; Castelletti et al. 2016). G11.2−0.3 contains a compact X-ray PWN near its center (Roberts et al. 2003) and it has been studied in detail. Several derived distances or distance constraints for G11.2−0.3 are 10 kpc, 6 kpc, 4.4 kpc, and >4.5 kpc (Pavlovic et al. 2014; Brogan et al. 2004; Kilpatrick et al. 2016; Radhakrishnan et al. 1972, respectively). Green (2004) and Minter et al. (2008) also derived distances to G11.2−0.3 of 4.4 kpc and 5.5–7.7 kpc, respectively.

We may assume that only the sources G11.1+0.1 and G11.0−0.0 are located at comparable distances of 3–4 kpc, and we show next the SED plots with simple leptonic and mixed models that can explain the available data, including the radio fluxes from these two sources. The models shown are not actual fits to the data but simple comparisons of the flux levels assuming the same leptonic spectra are produced in the sources. Although exploring all of the possible parameter space is outside the scope of this paper, we may give estimates of the values for some of the physical parameters required. Since the distances cited above for G11.4−0.1 were obtained from the Σ − D relation whose validity has been questioned strongly in the literature (e.g., Green 2004), we consider that its distance is unknown and include its fluxes in the plots for comparison.

Figure 3 shows a mixed leptonic and hadronic model where the TeV emission comes from high-energy leptons scattering off cosmic microwave background photons (IC-CMB), and the GeV emission, having a spectrum that is softer than expected from IC-CMB interactions, comes from hadronic collisions. The particle distributions in both cases are described by a power law with an exponential cutoff and index 2.2. The cutoff energies are 1 TeV for hadrons and 50 TeV for leptons, while their total energies are $4.8\times {10}^{49}\,{\rm{erg}}\,\left(\tfrac{10\,{{\rm{cm}}}^{-3}}{n}\right)$ and 1.0 × 10⁴⁹ erg, respectively, for a source distance of 3 kpc and target density n. The existence of high-energy leptons gives rise to synchrotron fluxes at a level that depends on the average magnetic field. Although the magnetic field is expected to be amplified in an SNR, with respect to its value in the interstellar medium (e.g., Bell & Lucek 2001), some observations imply relatively low average field values (∼5–10 μG) for some gamma-ray emitting SNRs (e.g., Ajello et al. 2016; Araya2017). Figure 3 shows the resulting synchrotron fluxes and the measured fluxes of three SNRs, chosen based on the previous discussion. Synchrotron fluxes were obtained for a magnetic field of 3 μG. The radio data were taken from the literature (Kassim 1989; Brogan et al. 2006). The particles producing the gamma-rays may thus be supplied by the SNRs in the region and their level of radio emission is compatible with the model. Of course, only observations of X-ray synchrotron emission can provide evidence for leptons having the necessary energies to account for the TeV emission.

Zoom In Zoom Out Reset image size

The gamma-ray SED of HESS J1809−193 is also compatible with a purely leptonic model consisting of bremsstrahlung emission producing most of the GeV fluxes and IC-CMB producing TeV emission. It was found that a single population with an energy distribution in the form of a power law with index 2.35 and an exponential cutoff energy of 70 TeV reproduces the observed high-energy fluxes, as shown in Figure 4. The resulting fluxes from synchrotron emission are compatible with the sum of the individual radio fluxes from G11.1+0.1 and G11.0−0.0. The required total energy in leptons is 5.4 × 10⁴⁹ erg for a distance of 3 kpc and the density of the target material for bremsstrahlung emission is 10 cm⁻³. The level of X-ray synchrotron emission from the same leptons should be detectable even for a relatively low average magnetic field of 3 μG.

Zoom In Zoom Out Reset image size

Finally, as pointed out by Castelletti et al. (2016), star formation activity could also play a role in producing the gamma-rays seen from HESS J1809−193. Several bright H ii regions lie within the boundary of the VHE source (see the next section for details on some of the observed gamma-ray spectra and morphologies of star-forming regions (SFR)).

In the case of HESS J1813−178 similar scenarios to those for HESS J1809−193 could be considered. Interestingly, the spectral indices measured at TeV and GeV energies are compatible (∼2.1). However, the TeV emission seems to be compact while the GeV emission is very extended (as seen in Section 2, the radius found was 06 ± 006 for a disk-like hypothesis). Different populations of cosmic rays might produce independent components of emission at GeV and TeV energies. However, little is known about the effects of, for example, a changing environment in the spectra of gamma-ray emission produced by hadrons or leptons. Remarkably, for example, similar morphologies have been seen at GeV and TeV energies for SNR IC 443 whose gamma-rays are of hadronic origin (Ackermann et al. 2013), but there are different ambient properties throughout the emission regions (Humensky & VERITAS Collaboration 2015).

From the fact that the GeV emission that we found here is very extended while current TeV observations seem to show only a compact source at the highest energies, we only considered here a scenario for the GeV gamma-rays, as the TeV emission might come from a PWN powered by PSR J1813−1749. The spectral shape of the emission can be easily explained by a hadronic accelerator, perhaps the SNR G12.82−0.02 or one of the other SNRs in the region (see a catalog of SNRs by Ferrand & Safi-Harb 2012), and hard to explain in a leptonic scenario where the expected emission is harder (although note that this might not always be true for a PWN, as it is believed to be the case for Vela-X, e.g., Grondin et al. 2013). The role of SNR G12.82−0.02 in the production of the TeV gamma-rays has been explored before by Funk et al. (2007), who found that both leptonic and hadronic scenarios are compatible with the multiwavelength observations of this object. If the GeV emission that we found, on the other hand, is produced by cosmic rays accelerated in this SNR, these particles would have already escaped from the object and would be propagating in the interstellar medium, as the SNR itself has a much smaller angular size (∼3') compared to the ∼12 of the GeV emission found here.

Interestingly, there is a giant SFR known as W33 within the LAT source. It is characterized by having very massive stars and vigorous star formation activity (e.g., Messineo et al. 2015). It subtends 15' and is located at a distance of 2.4 kpc (Immer et al. 2013). It has been proposed that SFRs might be able to accelerate cosmic rays through the combined winds of massive stars and some of these regions have been detected in gamma-rays (e.g., Cesarsky & Montmerle 1983; Bykov & Toptygin 2001; Katsuta et al. 2017), being the one in the Cygnus X region perhaps the most famous example (Ackermann et al. 2011). Known as the Cygnus Cocoon, the GeV emission is very extended (∼4°) and has a relatively hard spectrum (photon index ∼2.2). Recently, Katsuta et al. (2017) found several extended regions of GeV emission within ∼2° of the Galactic coordinates (l, b) = (25°, 0°) having similar photon spectral indices of 2.1 ± 0.2 and no spectral curvature in the LAT range, as well as no clear counterparts at lower energies. They present a plausible scenario where this emission is from a previously unknown SFR supported by the confinement of the GeV emission by bubble-like structures seen at lower wavelengths and related to the putative OB association, as well as similarities with the Cygnus Cocoon. Preliminary results in the Fermi High-latitude Extended Source Catalog show three similar objects tentatively associated to SFRs (Wood et al. 2017).

Although the GeV emission found here is more extended than W33, perhaps this could similarly point to the presence of a previously unknown component of star formation that could be related to W33. The GeV spectral index, the extension and the lack of spectral curvature found here are features that are similar to those seen in the previous examples of SFRs. The diameter of the GeV region is ∼50 pc for a distance of 2.4 kpc, about half the size of the Cygnus Cocoon. Following the usual motivation for hadronic models for the GeV emission from SFRs, due to their spectral shape, Figure 5 shows the gamma-ray SED of the extended emission in the direction of HESS J1813−178 with a hadronic model where the cosmic rays have a distribution in the form of a power law with index 2.15 and a cutoff energy of 80 TeV (the TeV data points from H.E.S.S. are included only for reference, please note that the TeV emission is compact and possibly associated with a PWN). The required total energy in the particles is (normalized to the distance to W33)

$$\begin{eqnarray*}&&4.2\times {10}^{50}\,{\rm{erg}}\,\left(\displaystyle \frac{1\,{{\rm{cm}}}^{-3}}{n}\right){\left(\displaystyle \frac{d}{2.4{\rm{kpc}}}\right)}^{2},\end{eqnarray*}$$

which is similar to the energy found in the Cygnus Cocoon (Ackermann et al. 2011). We note that, as found by Katsuta et al. (2017), a leptonic scenario for typical parameters in SFRs could also be used to explain the GeV gamma-rays (invoking bremsstrahlung emission), and we refer the reader to their work for an example of such a model.

Zoom In Zoom Out Reset image size

Finally, other SNRs are known within the location of the GeV emission: G12.5+0.2, G12.7−0.0, and G13.5+0.2 (Ferrand & Safi-Harb 2012). It is then possible that the extended gamma-rays are contributed by several of these objects that cannot be resolved by the LAT. A combination of accelerators may produce cosmic rays that escape and interact with molecular clouds. Some clouds are known to exist at least near HESS J1813−178 (Funk et al. 2007). As mentioned before, rapid electron diffusion away from a PWN could produce an unusually steep GeV spectrum (Hinton et al. 2011). Even a previously unknown SNR located at a closer distance could be responsible for the emission and explain its extension. More observations are needed toward HESS J1813−178 to understand the environment and the origin of the high-energy particles.