© The Authors 2023

1 Introduction

Supernova remnants (SNRs) are captivating objects that cause a long-lasting impact on the Galactic ecosystem, leaving distinct imprints that can be observed across the entire electromagnetic spectrum. This paper, focused on the source Kes 17 (G304.6+0.1), is part of a series of articles conducted by the author team. The series is dedicated to investigating the association between radio and γ-ray emissions in remnants of stellar explosions. Previous studies in this series were devoted to G338.3–0.0 (Supan et al. 2016), Kes 41 (Supan et al. 2018a,b), and G46.8–0.0 (Supan et al. 2022), all of which are middle-age γ-ray emitting SNRs that interact with their ambient medium.

The initial observations of Kes 17 at radio wavelengths were conducted in the 1970s at 408 and 5000 MHz using the Molonglo and Parkes single-dish telescopes (Goss & Shaver 1970); Shaver & Goss 1970a; Milne & Dickel 1975). The first distance estimate for the remnant was approximately 6 kpc, derived using the uncertain Σ-D relation (Shaver & Goss 1970b). Later, Caswell et al. (1975) established a lower limit of 9.7 kpc based on absorption features in the low-resolution neutral hydrogen (H I) spectrum of gas clouds along the line of sight. This lower limit remained unchanged for nearly five decades, until the recent reanalysis by Ranasinghe & Leahy (2022), which yielded a kinematic distance 7.9 ± 0.6 kpc to the remnant.

Kes 17 was extensively studied in the X-ray domain using data obtained with the X-ray Multi-Mirror Mission (XMM-Newton)¹, Suzaku², and the Advanced Satellite for Cosmology and Astrophysics (ASCA)³ satellites (Combi et al. 2010; Gök & Sezer 2012; Gelfand et al. 2013; Pannuti et al. 2014; Washino et al. 2016). Based on the observed properties of the X-ray emitting gas, Combi et al. (2010) proposed that the source belongs to the mixed-morphology (MM) type of SNRs, which is characterised by a shell-like morphology in radio wavelengths and a filled-centre composition in X rays. Additionally, they suggested the presence of a nonthermal component in the northern, central, and southern regions of the SNR shock front. However, subsequent studies with improved statistics have raised doubts about this possibility and concluded that the X-ray spectrum is dominated by thermal emission (Gök & Sezer 2012); Gelfand et al. 2013; Washino et al. 2016). At the high-energy end of the spectrum, Kes 17 has been linked to a GeV source detected by the Fermi Large Area Telescope (LAT; Wu et al. 2011; Gelfand et al. 2013). No counterpart at TeV energies has been reported so far. The first hint of interaction for the SNR environment came from low-resolution observations of the 1720-MHz maser line of hydroxyl (OH; Frail et al. 1996). A significant breakthrough in linking Kes 17 to the interstellar matter occurred through near-infrared (near-IR) spectroscopic studies, which were crucial in firmly determining the location and characteristics of the shocked H₂ gas (Lee et al. 2011).

This paper is organised as follows: in Sect. 2 we present the first analysis of the radio continuum spectrum for Kes 17. Adopting a standard evolution model, we also derived the SNR age by means of the radio size of the expanding forward shock and 21 cm spectral line observations of the H I gas. Section 3 focuses on investigating the morphology and kinematics of the molecular gas emitting in CO lines. This represents the first study of the CO gas in the region of Kes 17. In Sect. 4 we provide an updated analysis of the Fermi-LAT data covering 14.5 yr. We also investigate the emission mechanism that causes the high-energy flux through a broadband modelling that incorporates the revisited measurements at radio and GeV γ-ray energies. Our findings are summarised in Sect. 5.

2 SNR Kes 17 in radio wavelengths

2.1 Morphology and spectrum

The radio remnant Kes 17 is characterised by non-uniform emission from a complete, albeit irregular, shell structure with an average size of ~7′. This can be observed in the 843 MHz image from the Molonglo Sky Survey (SUMSS, HPBW = 45″ × 50″)⁴ included in the inset of Fig. 2a. The surface brightness of the remnant is ~4.3 × 10⁻²⁰ W m⁻² Hz⁻¹ sr⁻¹ at 843 MHz, while that of the brightest elongated feature (~4′.2 × 1′.2 in size) along the southern periphery is ~0.11 × 10⁻²⁰ W m⁻² Hz⁻¹ sr⁻¹. An arc of enhanced synchrotron emission is detected in the north-west region of the remnant, near 13^h05^m30^s, −62°41′00″. This arc has a size of approximately ~1′.3 × 2′.5 and coincides with a distinctive bend in the shock front of Kes 17. Bright continuum and line emission in the IR wavebands, as reported by Lee et al. (2011), accompanies the radio synchrotron emission along this edge of Kes 17. Another noteworthy feature is an indentation towards the east of the radio shell, at approximately 13^h06^m05^s, −62°41′10″, which might indicate that the SNR shock is wrapped around an external inhomogeneity. The structure of the surrounding matter, revealed for the first time by the analysis of the CO gas, is discussed in Sect. 3.

To construct the global spectrum of the radio continuum emission of Kes 17, we compiled flux density estimates from the literature as well as new fluxes that we measured from publicly available radio surveys. For frequencies below ~160 MHz, the lowest at which Kes 17 has been detected to date, we used the Galactic and Extragalactic All-sky Murchison Widefield Array Survey (GLEAM; Hurley-Walker et al. 2019)⁵. We also used the Southern Galactic Plane Survey (SGPS; McClure-Griffiths et al. 2005)⁶ at 1420 MHz and the S -band Polarisation All Sky Survey (S-PASS; Carretti et al. 2019)⁷ at 2303 MHz. Flux measurements with an error greater than 20% were discarded from the analysis. When information on the primary calibrator was available, the remaining data were brought to the absolute flux scale presented by Perley & Butler (2017). This flux scale is valid for the entire range of compiled frequencies in our analysis (88-8800 MHz) and has an accuracy of ~3%. The set of data points was fitted using the simple power-law model S_ν ∝ ν^α, where S_ν represents the integrated flux at frequency ν, and α is the radio spectral index. During the fitting process, flux measurements were rejected when their dispersion with respect to the model was greater than 2σ of the best-fit values.

The final dataset is reported in Table 1. It constitutes the most complete compilation of radio flux measurements conducted for Kes 17 to date. Figure 1 displays the integrated radio continuum spectrum for this SNR. Our new flux density determinations are represented by filled blue circles. A weighted least-squares fit was applied to the data points, resulting a spectral index α = −0.488 ± 0.023. This value is flatter than the previous measurement (α ≃ −0.54) reported by Shaver & Goss (1970b), which was based on flux estimates at only 408 and 5000 MHz. The synchrotron radiation spectrum we derived for Kes 17 is consistent with electrons that are accelerated via a first-order Fermi mechanism (Onić 2013). Regarding the spectral shape, the straight distribution of flux densities at low radio frequencies (below 100 MHz) indicates that if ionised gas exists, whether it is located either co-spatially or coincidentally intersecting Kes 17 along the line of sight, the free-free absorption it produces does not affect the integrated continuum spectrum of the remnant. Low-frequency turnovers caused by free-free absorption by ionised gas in H II regions (or in their associated lower-density envelopes), as well as at the interface between an ionising shock and its immediate environment, have been measured in the spectra of some SNRs (e.g. Kes 69, 3C 391, 3C 397, Kes 67, Kes 75, W41, Kes 73, 3C 396, and W49B; Castelletti et al. 2021). Whether the forward shock of Kes 17 ionises the western region where SNR interaction with dense gas has been proved in the infrared (Lee et al. 2011) might be determined by radio observations with improved sensitivity and resolution, especially in the low-frequency portion of the spectrum.

Fig. 1 Spectrum of the continuum emission from SNR Kes 17 at radio frequencies, constructed with the fluxes in Table 1. Blue data points denote our new measurements from public survey images, and the green points correspond to literature flux estimates. The straight line represents the best fit with a power-law model in the form Sν ∝ να, which yields a spectral index value α = −0.488 ± 0.023. The shaded darker and lighter pink regions around the straight line denote a variation in the fitted spectral parameters of 1 and 2σ, respectively.

2.2 SNR age and neutral gas properties from radio data

Estimating the dynamical age of supernova remnants involves indirect methods because it cannot be measured directly. So far, all age estimates for Kes 17 have been derived on the basis of spectral fitting parameters to the cold and low- density X-ray emitting gas interior to the radio shell. However, there are large differences in the estimated ages, with values ranging from 2.3 kyr to as high as 64 kyr, depending on factors such as ionisation timescales, τ, and electron densities, n_e (τ ~ 1–3 × 10¹² cm⁻³ s, n_e ~ 0.4–2.3 cm⁻³; Pannuti et al. 2014, and references therein). In addition, according to the suggestion that Kes 17 is a member of the MM SNR group, and considering a thermal conduction model, Gelfand et al. (2013) derived an age range from 2 to 5 kyr. Additionally, they determined an upper age limit of 40 kyr by assuming that clump evaporation into the inter-cloud medium primarily causes the observed X-ray emission.

In this section, we employ a standard evolution model and examine continuum and line emissions at centimetre wavelengths to estimate the age of Kes 17. The relative intermediate extent of the shock front of Kes 17 (as illustrated in the inset of Fig. 2a) compared to other SNRs discovered in the Galaxy, coupled with the absence of optical signatures of radiative shocks in observations from surveys such as the SuperCOSMOS H-alpha Survey (SHS; Parker et al. 2005)⁸ or the Space Telescope Science Institute Digitized Sky Survey (STScI DSS)⁹, lends support to the hypothesis that Kes 17 is in the Sedov expansion stage of its evolution. Based on this picture, the time elapsed since the explosion can be estimated via the relation (Cox 1972) (1)

where R_S is the current shock radius (in pc), n₀ is the ambient interstellar density (in cm⁻³), and ϵ₀ is the initial explosion energy (in units of 0.75 × 10⁵¹ erg). The radius, measured in the radio continuum image of Kes 17 at 843 MHz, is ~3.5′ or ~8 pc according to the revisited kinematic distance 7.9 ± 0.6 kpc to the SNR obtained by Ranasinghe & Leahy (2022) from neutral hydrogen H I absorption features. The ambient interstellar density is well represented by the neutral hydrogen gas density, and we estimated it via n₀ = N_H/L, which is the ratio of the hydrogen column density to its depth in the line of sight in the region of the SNR. To calculate N_H, we considered the 21 cm line emission of H I from the SGPS data. Our focus was to identify any sign of neutral gas that could have been swept up by the SNR shock or by the stellar winds of the progenitor star. If we detected accumulation of H I around the radio continuum boundary of the remnant, it could provide us with a rough estimate of the pre-shock medium density under the assumption that the accumulated atomic gas was uniformly distributed inside the volume of the H I shell before the stellar explosion. We found no neighbouring structure of neutral atomic gas that might be associated with Kes 17, however. Moreover, an inspection of the HI datacube shows the remnant in absorption in the complete velocity range from ~0 km s⁻¹ to the tangent point velocity (υ_TP ≃ −42 km s⁻¹, according to the Galactic rotation curve of Reid et al. 2014). Therefore, we simply hypothesised that n_H could well be represented by the mean density value measured in circular test-areas of radius ≃8′ distributed around the remnant (we tested different values and determined that our result remains consistent within the uncertainties, regardless of the size chosen). Under the common assumption that the H I emission is optically thin, after subtracting an appropriate mean background level to each H I velocity channel, the mean column density around Kes 17, calculated by integrating the H I emission between −31 and −14 km s⁻¹ is N_H ≈ 8 × 10²⁰ cm⁻². This velocity interval agrees with the interstellar molecular matter traced in CO associated in space and velocity with the SNR (further analysis of this topic is provided in Sect. 3). According to all the assumptions we made previously, our analysis produces a number density in the HI ambient environment n₀ ≈ 7 cm⁻³, which is higher than the typical value ~1 cm⁻³ averaged over the cold-, warm-, and hot-gas phases of the interstellar medium (ISM) (McKee & Ostriker 1977). Therefore, using Eq. (1) and adopting the value ~4 × 10⁵⁰ erg for the energy released in the SN event, as derived by Leahy et al. (2020), who incorporated both uniform ISM and stellar wind SNR evolutionary models, we estimate that Kes 17 is approximately 11 kyr old. Notably, when using the kinetic energy 10⁵¹ erg for a canonical SN, the age decreases to roughly 7 kyr. This is not critically different from our calculation within the uncertainties. We are aware that our approach to determining the age of Kes 17 provides a first-order approximation because (i) it assumes that the SNR is in the Sedov stage of its evolution, (ii) the mean number density of atomic hydrogen, as measured from H I data, represents an upper limit because some of the H I may be unrelated gas located behind the SNR, and (iii) our result ignores the possibility of the remnant evolving in an inhomogeneous ambient medium.

Fig. 2 Spatial and spectral distribution of the CO gas towards Kes 17 from ThrUMMS (Barnes et al. 2015). (a) Colour-coded image of the 12CO (in red) and the 13CO (in green) J =1–0 line emissions, integrated from −31 to −14 km s−1. The yellow regions are areas where the two CO isotopologue emissions overlap. The contours (levels 0.034, 0.35, 0.75, and 1.2 mJy beam−1, with a beam smoothed to the 80″ spatial resolution of the CO data) delineate the 843 MHz continuum radiation from Kes 17. For reference, the inset in the upper left corner displays the structure of the SNR shell as observed by the SUMSS (45″ resolution). The overlaid grid consists of the series of boxes (1′.5 in size) used to analyse the kinematics of the CO gas. The H II region G304.465–00.023 in the field is also labelled (Urquhart et al. 2022). (b) Collection of 12CO (in red) and 13CO (in green) J =1–0 spectra extracted from the boxes, numbered from 1 to 42, in panel a. The original datasets were convolved to a resolution of 80″ to reduce the graininess. The yellow contours superimposed on the spectra correspond to the radio continuum emission from Kes 17.

3 Molecular environment of Kes 17

The properties of the molecular gas in the region of Kes 17 as traced by the emission from carbon monoxide (CO), did not receive attention in previous works. Dense interstellar material interacting with the western shock front of the SNR was only revealed in infrared wavebands (Lee et al. 2011). Here, we present the main results of the first study towards Kes 17 that is carried out by using both ¹²CO and ¹³CO emission data in their rotational transition J = 1–0. The datacubes for both species were extracted from the Three-mm Ultimate Mopra Milky Way Survey (ThrUMMS; Barnes et al. 2015)¹⁰. The spatial and spectral resolutions are 72″ and ~0.35 km s⁻¹, respectively, with sensitivities of ~1 K each.

After inspecting the CO data cubes throughout their velocity ranges (−65, +55) km s⁻¹, we only found molecular structures that projected, correspond to the radio continuum emission from the SNR shell in two intervals, with velocity ranges from approximately −45 to −37 km s⁻¹ and approximately −31 to −14 km s⁻¹. The CO structure in the first range peaks at −41 km s⁻¹ and lies in the plane of the sky towards the eastern border of Kes 17, with an angular size of approximately 8′ × 4′ (in the south-north and east-west directions). We assigned a distance ≃5 kpc to this molecular material because its velocity ≃−41 km s⁻¹ is largely consistent with that of the tangent point in the direction of Kes 17. Consequently, we discarded this cloud as possibly associated with Kes 17 (d_SNR ≃ 8 kpc, Ranasinghe & Leahy 2022), and it is not analysed in the following. The emission from the second velocity component originates from a cloud in the eastern part of the SNR shell. The distributions of the integrated ¹²CO and ¹³CO emissions are presented in the colour-coded Fig. 2a, overlaid with the contours of the radio continuum emission from the SNR shock wave. Overall, the ¹²CO compared to the ¹³CO emission appears to be spatially more extended within the region of interest. This difference in distribution can be explained by the fact that¹³CO emission is optically thinner than that from ¹²CO, whereby it gives an account of more internal and denser regions in the cloud (Wilson et al. 2013). Notably, in the brightest part of the uncovered molecular structure, corresponding to the yellowish regions in which the ¹²CO and ¹³CO emissions overlap, Kes 17 exhibits a significant deviation from a spherical symmetry (within the 80″ resolution of the radio image smoothed to match the resolution of the CO data).

In order to gain further insight into the characteristics of the molecular gas, we extracted ¹²CO and ¹³CO spectra across the entire region where the radio continuum emission from and the CO line emission towards Kes 17 show a line-of-sight superposition. To cover this region comprehensively, we employed a grid of 1f5 boxes, as depicted in Fig. 2a. To enhance the signal-to-noise ratio of the spectra, they were smoothed by averaging intensity values within the five nearest-neighbour velocity channels. The resulting spectra, presented in Fig. 2b, show a broad emission region in ¹²CO with a velocity span Δυ ≃ 17 km s⁻¹ and an intensity varying from approximately 2–4 K. Multiple ¹²CO kinematic components contribute to the observed emission, with peak velocities at approximately −30, −25, and −20 km s⁻¹. Examples of profiles showing this behaviour correspond to boxes 16–18, 22–24, and 28–30, all of which lie within the outermost radio contour of the SNR shell. At the easternmost border of Kes 17, these three velocity components appear to be less distinguishable and exhibit blending. We stress that the ¹³CO profiles (Fig. 2b) do not reproduce the triple-peaked structure observed in the ¹²CO gas, but a broad peak at about −20 km s⁻¹, in complete agreement with the velocity of the H I absorption features used to constrain the distance to Kes 17 (Ranasinghe & Leahy 2022). The intensity of the ¹³CO emission peaks is ~1–2 K.

Figure 3 displays position-velocity (p-υ) diagrams of the molecular gas emission. They were constructed by integrating the ¹²CO emission along the RA direction in seven slices, each covering a range of 100″. These slices span the entire cloud of interest and constitute an appropriate tool for effectively capturing the spatial heterogeneity of the individual velocity components observed in the spectral distribution shown in Fig. 2b. By inspecting the p-υ diagrams, it is evident that the molecular emission is mostly concentrated at −20 km s⁻¹, adjacent to the position at which the radio shell of the remnant is highly distorted, and it appears to branch off to the interior of the SNR (panels e–g in Fig. 3). Bright knots are noticeable in the cloud interior.

By combining the CO emission and H I absorption profiles (not shown here) extracted over the brightest part of the molecular concentration emitting at −20 km s⁻¹, we determined that it is located at its far kinematical distance ≃8 kpc. The remaining velocities components of the cloud are at about 7 kpc (−30 km s⁻¹) and 7.5 kpc (−25 km s⁻¹). Taking the associated uncertainties (≃1.0 kpc) in these determinations into account, it can be concluded that these peaks arise from different components of the same structure. The average distance to this structure is estimated to be approximately 7.5 kpc, which is completely compatible with the distance determined for Kes 17 (≃7.9 kpc; Ranasinghe & Leahy 2022). The error in the distance determination for the molecular gas stems from various factors. One of these contributions is the uncertainty in the peak velocity value for each gas component. Additionally, it can be challenging to accurately measure the properties of individual molecular components, especially when they are not completely resolved. Lastly, the use of a Galactic rotation curve, such as the one proposed by Reid et al. (2014), involves assumptions and uncertainties.

We note that although our analysis of the molecular material through ¹²CO and ¹³CO lines supports the coexistence of the discovered eastern cloud and the remnant, we did not observe distinct broadenings in the CO emission that would be attributable to turbulence caused by the impact of the Kes 17 shock front. Therefore, we propose that the spectral behaviour of CO might illustrate a soft contact between the surrounding cloud and the remnant shockwave. Certainly, the process of impacting the cloud might be at an initial stage.

We have also estimated the total mass M and mean density n(H₂) of the molecular gas in the newly detected cloud at υ_LSR ≃ −31 to −14 km s⁻¹ by using both the ¹²CO and ¹³CO (J = 1–0) emissions. The procedure involves calculating the molecular hydrogen column density N(H₂), and deriving both M and n(H₂) from it. N(H₂) is obtained from the integrated emission of the CO by using appropriate conversion factors relating the integrated emission of ¹²CO and the H₂ column density (X₁₂ = 2.0 × 10²⁰ cm⁻² (K km s⁻¹)⁻¹, Bolatto et al. 2013), and also between the column density N(¹³CO) and N(H₂) (X₁₃ = 7.7 × 10⁵, Kohno et al. 2021). We refer to Wilson et al. (2013), who explained the expressions and assumptions (related to local thermodynamic equilibrium) employed for obtaining column densities in detail. The mass was calculated through the relation M = µ m_H Ω D² N(H₂), where µ = 2.8 is the mean molecular mass of the cloud¹¹, m_H is the hydrogen atom mass, and Ω is the solid angle subtended by the cloud located at the distance D. On the other hand, the mean molecular density is n(H₂) = N(H₂)/1, where l denotes the extent of the cloud in the line of sight, which is assumed to be equal to the average of the mean size of the structure in RA and Dec. To integrate the CO emissions, we used a circular region with a radius of 5′ (or ~11 pc at a distance of ~7.5 kpc to the cloud) centred at 13^h06^m30^s, −62°43′20″. From this integration, we derived a molecular column density N(H₂) ≈ 1 × 10²² cm⁻², consistent for the ¹²CO and ¹³CO gases. Therefore, the resulting mean mass and molecular density for the eastern cloud were estimated to be M ≈ 4.2 × 10⁴ M_⊙ and n(H₂) ≈ 300 cm⁻³. The uncertainties in these measurements are about 40% and comprise errors in the distance and the definition of the structure in the plane of the sky, as well as in the velocity space. We also note that the differences in the obtained values using emissions from both ¹²CO and ¹³CO were found to be within 20%. The fact that the values estimates from both ¹²CO and ¹³CO are in agreement indicates that both isotopologues provide consistent measurements and can be used to derive cloud parameters.

Now we focus on the molecular gas distribution towards the western side of the SNR shell. Of particular interest is the absence of CO emission above 3σ (where σ ~ 1 K) that is spatially correlated with the radio continuum bright region, which is roughly 2′ × 4′ in size (centred at 13^h05^m30^s, −62°41′10″). In this region, molecular hydrogen and ionic lines at infrared wavelengths have revealed the expansion of the SN shock in a molecularcloud, as reported by Lee etal. (2019). Basedon these findings, we tentatively propose the existence of a CO-dark gas component west of Kes 17. In this scenario, the gas-phase carbon could be in atomic form, while the hydrogen is in molecular form. A similar phenomenon was observed in CTB 37A (Maxted et al. 2013) and RX J1713.7–3946 (Sano & Fukui 2021). More sensitive CO molecular-line measurements are needed to shed more light on this scenario for Kes 17. It is, however, worth noting that the detection of γ-ray radiation in the direction of the remnant can indirectly trace the dark molecular gas component if it is generated by cosmic-ray collisions with the gas (Wolfire et al. 2010). In the case of Kes 17, a γ-ray excess at GeV energies has indeed been detected in projected coincidence with the remnant. The analysis of this emission is addressed in Sect. 4 of this work. In passing, a peculiar wall-like structure of ¹³CO is observed at a distance of approximately 1′.5 from the outermost radio contour towards the west. However, the straight vertical border of this structure and the absence of a counterpart in the ¹²CO data covering the same region strongly suggest that this is not a real feature.

Fig. 3 Position-velocity (p-υ) diagrams for the 12CO emission towards the field of Kes 17. (a) Regions numbered from 1 to 7 that were used to construct the p-υ diagrams for the eastern cloud associated with Kes 17, shown in panels b–h. The colours indicate the spatial distribution of the 12CO J =1–0 gas integrated in the (−31, −14) km s−1 velocity range as in Fig. 2, and the contours trace the 843 MHz radio continuum emission from Kes 17. (b–h) p-υ diagrams derived from the 12CO line emission depicted in panel a. They were constructed by integrating in the 100″ RA interval indicated by the regions numbered from 1 to 7 in panel a. The horizontal dashed lines mark the extent of Kes 17 in the declination dimension. The colour representation is the same for all p-υ diagrams.

4 Field of Kes 17 at γ-ray energies

The first reports of emission in the γ-ray domain that is spatially projected onto Kes 17 were presented by Wu et al. (2011) and Gelfand et al. (2013), based on statistics of 30- to 39-month data from Fermi-LAT. The latest Fermi-LAT catalogue of GeV sources (4FGL-DR3¹²; Abdollahi et al. 2020, 2022) identifies the observed γ-ray excess directed towards Kes 17 as 4FGL J1305.5–6241. To date, there have been no reports of TeV radiation detected in the field of Kes 17. In this section, we provide an update on the GeV emission in direction to this SNR and investigate the nature of the high-energy photons by modelling the spectral energy distribution (SED) by combining the updated γ-ray data with the firstly obtained radio continuum spectrum of Kes 17 presented in Sect. 2.1.

4.1 Treatment of GeV data from Fermi-LAT

Our analysis comprises the largest statistics of events for Kes 17 to date, consisting of approximately 14.5 yr of continuous data acquisition with the Fermi-LAT, spanning from the beginning of the mission on August 4, 2008, to February 24, 2023¹³. This represents a significant improvement of ~450% in observing time compared to the previous study conducted by Gelfand et al. (2013).

The LAT data were processed using the fermipy module version 1.1.6 (Wood et al. 2021), which uses the Science Tools package version 2.2.0¹⁴. Events were selected using the Pass 8, third release (P8R3) of photon reconstructions and the latest instrument-response functions (P8R3_SOURCE_V6). The region of interest (ROI) used to extract the events was a circle 15° in size, centred at Kes 17 (13^h05^m53^s, −62°42′10″). To extract valid events, we employed the tasks gtselect and gtmktime and applied standard filters for good-time intervals (GTIs)¹⁵ and selecting “source” class events (evtype = 3). An additional cut for the zenith angle at 90° was implemented to minimise potential contamination from cosmic-ray interactions in the upper atmosphere. We considered events with reconstructed energies above 0.3 GeV to mitigate the adverse effects of systematic uncertainties in the effective area and the degradation of the point spread function (PSF) at the lowest energies (Ackermann et al. 2012). Furthermore, we excluded photons above 300 GeV because only a few events occur at the highest energies.

The set of filtered events was used to fit a sky model through a maximum likelihood optimisation procedure (Mattox et al. 1996). For the optimisation, we implemented a binned likelihood analysis over the ROI. The spatial bins were set at 0^°.01 for both morphological and spectral analysis, and the energy range from 0.3 to 300 GeV was divided into 10 logarithmic bins per decade. In the analysis, we included all sources from the LAT 10 year (4FGL-DR3) located within the ROI, except 4FGL J1305.5–6241 in projected coincidence with Kes 17. The models used for the optimisation were gll_iem_v06 for the diffuse Galactic background and iso_P8R2_SOURCE_V6_v06 for the isotropic background. The spectral parameters and normalisations of these models were allowed to vary freely during the optimisation process. Convergence was achieved by means of the Minuit optimiser, fixing the source parameters beyond a radius of 4° from the ROI centre.

4.2 Morphological and spectral characteristics of the GeV emission

To investigate the spatial distribution of the γ-ray emission, we used the fermipy tool tsmap to construct a test-statistics (TS) map. Each pixel in this map represents the likelihood of having a point source at the corresponding coordinates, compared to the null hypothesis of the sky model without the point source. The TS parameter is calculated as TS = 2 ln(L/L₀), where L and L₀ are the likelihoods including and excluding the source in the sky model, respectively. The TS map allows us to assess the significance of the source detection, (Mattox et al. 1996). In Fig. 4 we present the TS map we obtained, overlaid with contours depicting the radio emission from Kes 17. The global TS value after the likelihood optimisation for the GeV source projected in coincidence with Kes 17 was TS ≃ 730, corresponding to a detection significance σ ≃ 27. This represents an improvement of about 2.5 times over the value reported by Gelfand et al. (2013).

We tested the possibility that the GeV emission might be extended by using the extension tool from fermipy, which makes a likelihood analysis by modelling the source as a circular region with variable radius. After convergence, the fitted value of the radius at 95% confidence was 0°09, and the significance of the extension is TS ≃ 16.3 (σ ~ 4). The low significance of the fitted size is indicative that the emission is not significantly extended. We then consider in the following that the γ-ray excess detected by the Fermi-LAT corresponds to a point-like source, and consequently, we modelled it as a point source. Under this assumption, the localise tool yields the following location of this source: RA = 13^h05^m40.71^s ± 01.66^s, Dec = −62°42′01.6″ ± 21.6″, within a 95% confidence limit. From our analysis, the localisation of the source, indicated by a green ellipse in Fig. 4, has been significantly improved by approximately one order of magnitude compared to the previous value reported in the 4FGL catalogue. The molecular gas, traced by the bright IR filaments (Lee et al. 2011, and references therein) could extend to cover the γ-ray region. As we proposed in Sect. 3, this region is suspected of containing CO-dark molecular gas in interaction with the Kes 17 shock front.

To investigate the spectral characteristics of the GeV γ-ray excess detected towards Kes 17, we generated an SED by performing a binned likelihood analysis in the 0.3–300 GeV range, implemented through the fermipy tool sed. To ensure a balance between energy resolution and statistical significance, the data were grouped into five equally spaced bins per decade on a logarithmic energy scale. Additionally, we incorporated energy dispersion corrections to mitigate systematic effects on the fitted spectral parameters.

In our treatment, systematic contributions arise from the effective area, the PSF of the Fermi-LAT, the energy scale, and variations in the spectral parameters due to the normalisation of the diffuse background¹⁶. The contribution related to the effective area is variable and can reach approximately 10% at the extremes of the energy range considered in our analysis. For energies below 100 GeV, the systematic error related to the PSF containment radius is about 5% and increases linearly to about 20% for higher energies. The energy scale uncertainties are within 5% throughout the energy range. On the other hand, systematics associated with the Galactic diffuse background were estimated following the procedure from Abdo et al. (2009), which consists of artificially varying and fixing the normalisation by ±6% with respect to the original fit and examining the resulting variations in the fitted spectral parameters. In the analysis of the broadband SED (Sect. 4.3), both statistical and systematic effects were considered. The contributions from both sources of uncertainties were added in quadrature to obtain a final error band for each energy bin.

The data were fitted using a power-law model dN/dE = ϕ₀(E/E₀)^−Γ, where ϕ₀ is the differential flux (in units of cm⁻² s⁻¹ MeV), E₀ is the pivot energy, and Γ represents the spectral energy index. Through the bin-by-bin likelihood procedure, we obtained a spectral index value (± stat ± syst), that agrees well with the 4FGL value, but it is softer and marginally consistent with that from Gelfand et al. (2013). Our result also agrees with the correlation observed by Acero et al. (2016) between the radio spectral index α and the GeV photon index Γ for SNRs interacting with molecular clouds. The data points are shown in Fig. 5, where the blue and red error bars denote the statistic and systematic uncertainties, respectively. The plot also shows the power-law fit to the data points, as well as the 1σ confidence interval for the fit. The integrated flux is determined to be F(0.3 − 300 GeV) = (2.98 ± 0.14) × 10⁻¹¹ erg cm⁻² s^{−1 17} corresponding to a luminosity L_γ(0.3 − 300 GeV) = (2.22 ± 0.45) × 10³⁵ erg s⁻¹ at the distance of 7.9 ± 0.6 kpc. The phenomenology of how the luminosity in γ-rays for SNRs competes with factors such as distance uncertainties, molecular gas repository densities, or time-evolution effects is complex, and a detailed analysis of this topic is beyond the scope of this work. However, despite this limitation, we can provide a brief comparison of our L_γ estimate for Kes 17 with the estimates derived in a similar energy range (0.1–100 GeV) for emitters identified as advanced (i.e. from middle-aged to old, ≳10 kyr) and young (≲3 kyr) SNRs associated with molecular clouds (for a more detailed discussion, we refer to Acero et al. 2022). For instance, when considering the older sources W44 and IC 443 with high local densities (~10²–10⁴ cm⁻³; Yoshiike et al. 2013; Dell’Ova et al. 2020), they would be 5 or even 60 times more luminous than Kes 17 if we placed them at the distance of ~8 kpc¹⁸. Additionally, the γ-ray luminosity of the mature remnant Cygnus Loop, evolving in a low-density environment (~1–10 cm⁻³, Fesen et al. 2018), would be comparable to our estimate for Kes 17 if they were located at the same distance¹⁹. We also point out that all of these middle-and advanced-age remnants are more luminous by one to two orders of magnitude than the young SNRs Tycho and Kepler, which are expanding in low-density media (~10 cm⁻³, Acero et al. 2022; Zhang et al. 2013)²⁰.

Fig. 4 Test-statistics map with 0°.01 pixel size in the 0.3–300 GeV energy band in the Kes 17 SNR field (see text for details). The blue contours delineating the SNR radio synchrotron emission are from the 843 MHz SUMSS map (levels 0.015, 0.15, and 0.30 Jy beam−1, at the resolution of 45″). The positions and distances of the H II region G304.465–00.023 (Urquhart et al. 2022) and the two known pulsars, PSR J1306–6242 (Kramer et al. 2003) and PSR J1305–6256 (Manchester et al. 2001), lying in the field, are also labelled. The plus symbol and dark-green ellipse within the Kes 17 radio contours mark the best-fit position and 95% confidence region, respectively, for the GeV emission obtained with a point-source spatial template. The dashed brown circle traces the 95% confidence fitted size of the γ-ray source. The inset shows a ~15′ × 15′ close-up view of the Kes 17 region, centred at the position of the γ-ray excess.

Fig. 5 Spectral energy distribution of the γ-ray emission in the region of Kes 17, as detected by Fermi-LAT. The orange line indicates the best fit to the data using the power-law model dN/dE = ϕ0(E/E0)−Γ with (see text for details). The yellow shaded zone corresponds to the 1σ region of the spectral fit. The blue and red error bars represent statistical and systematic uncertainties in the spectral points, respectively.

4.3 Analysis of the spectral energy distribution of Kes 17 from radio to γ rays

In this section, we study the spectral energy distribution of Kes 17 by incorporating the new nonthermal continuum radio spectrum extracted from the SNR shell and the high-energy spectrum obtained from new observations by the Fermi-LAT. For the nonthermal X-ray emission, we used an upper limit derived from Suzaku data by Gelfand et al. (2013). At very high energy, we used an upper limit above 1 TeV from the H.E.S.S. Galactic Plane Survey (Fernández Gangoso 2014; H.E.S.S. Collaboration 2018). To model the multi-wavelength emission, we considered an electron population that produces synchrotron radiation, inverse Compton scattering (IC), and non-thermal Bremsstrahlung. Additionally, we incorporated a proton population that interacts with the surrounding gas, resulting in the subsequent production and decay of neutral pions (π⁰). The parameters characterising these models are presented in Table 2 and were derived using the Naima Python package (Zabalza 2015).

To investigate the plausibility of a scenario in which the leptonic component is dominant, we considered a one-zone model with an electron population distributed in energies according to a power law. The updated radio continuum spectrum, extended to cover frequencies from 88 to 8800 MHz, allowed us to further constrain the spectral index of the synchrotron emission, at α = −0.488 ± 0.023 (Sect. 2.1). We used this measurement to fix the initial power-law index Γ_e = 1−2α of the electron energy spectrum to a value ≈1.9. Then, the X-ray upper limit derived by Gelfand et al. (2013) imposes a tight constrain on the magnetic field (B), the energy cut-off (E_c) and the energy density (W_e) of the electron population, which are degenerate. Following Gelfand et al. (2013), we assumed a magnetic field B = 35 µG, which yields a maximum value of the cut-off energy of E_c ≈ 2 TeV and a total energy density W_e = 4.3 × 10⁴⁸(d_SNR/7.9 kpc)² erg, which appears reasonable for a middle-aged system (see the discussion in Gelfand et al. 2013 about the cut-off in the spectrum due to synchrotron losses of middle-aged to old systems). We first considered the electron population distribution obtained from the radio spectrum synchrotron fit to compute the associated inverse Compton emission in order to reproduce the observed level of measured γ-ray emission. We considered three interstellar radiation fields: the cosmic microwave background (CMB; T_CMB =2.72 K, u_CMB = 0.26 eV cm⁻³), the far-IR radiation (T_FIR = 27 K, u_FIR = 0.415 eV cm⁻³), and the near-IR starlight radiation (T_SL = 2800 K, u_SL = 0.8 eV cm⁻³), computed from the GALPROP²¹ model at the position of the remnant (galactocentric distance of ~6 kpc) (Strong et al. 2004; Porter et al. 2006). However, as shown in Fig. 6, the IC radiation produced by this electron population fails to reproduce the new Fermi-LAT spectrum. Particularly, the shape of the GeV emission cannot be reproduced by a simple electron population with an index of 1.9. Furthermore, the upper limit from H.E.S.S. places strong constraints on the level of IC emission at very high energies, and it constrains the maximum value of the energy break in the electron spectrum to be 1.5 TeV and sets a minimum value for the magnetic field strength at 35 µG (see Table 2). Therefore, we conclude that it is not possible to adequately model the broad-band emission using a purely leptonic scenario, at least within the framework of a one-zone model.

We now investigate a scenario in which the main part of the γ-ray emission is attributed to π⁰ decay. At this stage, it is important to recall that the Fermi source is spatially coincident with a part of the shell that has bright radio emission and exhibits IR filaments. While no molecular structures were detected in the region of bright γ-rays (Sect. 3), the study of molecular hydrogen and ionic lines enabled highlighting the signature of a shock due to an interaction of the SNR shell with a cloud (Lee et al. 2011). Consequently, we considered only the molecular gas in the western region of Kes 17 in our analysis as the dominant contributor to the GeV photon flux through hadronic interactions. In the following, we consider a cloud density of 400 cm⁻³ that agrees with the estimation made by Lee et al. (2011) from the IR emission. This value is an order of magnitude higher than the ~10 cm⁻³ used in Gelfand et al. (2013). To analyse the hadronic origin of the γ-ray radiation, we assumed two models: a power-law proton spectrum with an index ~2.4 and a cut-off above a few TeV (ExpCut-off), and a broken power law (BPW) with indices Γ_p,1 = 2.4 and Γ_p,2 = 3.5 below and above E_b =2 TeV, up to a maximum energy of 10 TeV, which is strongly constrained by the steep spectrum deduced from Fermi data above few tens of GeV and the non-detection of Kes 17 by H.E.S.S. We report all proton dominated model parameters in Table 2. Both models fit the data equally well. To simplify the presentation, we plot only the ExpCut-off model in Fig. 6, which clearly shows that the γ-ray spectrum strongly supports the hadronic origin of the radiation. Because the hadronic γ-ray emission is proportional to the product of the kinetic energy in protons and the density of the medium, these parameters are closely correlated. Assuming that the total mass of the molecular cloud acts as the target material, we derived a total energy of cosmic-ray protons W_p = 2.97 × 10⁴⁹(n_p/400 cm⁻³)⁻¹(d_SNR/7.9 kpc)² erg. As can be appreciated in Fig. 6, the first two data points at the lowest γ-ray energies seem to deviate from a pure power-law shape and appear to be compatible with the so-called “pion-bump” feature observed below a few hundred MeV. As discussed in Tang (2018), the combination of this rising feature in the spectrum with a steep spectrum beyond a few tens of GeV is widely recognised as a characteristic signature of π⁰ decay, illuminating hadronic emission in SNRs. This π⁰ signature is observed in particular in the growing class of advanced-age GeV emitter SNRs that interact with molecular clouds, such as W44 (Giuliani et al. 2011), IC 443 (Ackermann et al. 2013), and W51C (Jogler & Funk 2016). On the other hand, the modelling of the data depicted in Fig. 6 shows that the contribution of Bremsstrahlung radiation from the previously considered electron population, with a similar density, clearly is a minor component in the GeV energy range, and its spectral behaviour does not reproduce the spectrum shape accurately. In the final proton-dominated model, we include this marginal contribution, as well as that from IC emission of an electron population with a total energy density of W_e = 1.5 × 10⁴⁸(d_SNR/7.9 kpc)² erg. The W_e/W_p ratio is 0.05 (see Table 2), which is higher than that in the proton-dominated scenario (0.01) discussed by Gelfand et al. (2013), but significantly lower than their IC-dominated scenario (0.1). This can be attributed to the reduction in W_p as a consequence of the higher density of the target matter considered here. On the basis of the new measurements of the GeV and radio spectra, we conclude that although a single electron population is considered to reproduce the overall synchrotron emission of the remnant, the π⁰ decay process may primarily be responsible for the point-like γ-ray emission detected towards Kes 17 in the GeV energy range.

Fig. 6 Broadband SED from radio to γ rays for Kes 17. The fluxes in the radio band correspond to those in Table 1 (in units of erg cm−2 s−1), while those at γ-ray energies correspond to the new Fermi-LAT data from this work (Sect. 4.2). The upper limits at X-ray and TeV energies where derived from the non-detection of X-ray synchrotron emission and the H.E.S.S. observatory, respectively (see text for details). (a) SED modelling with a unique electron population producing synchrotron and IC radiation over the broadband spectrum. (b) SED modelling with an ExpCut-off proton-dominated model producing synchrotron, IC, and Bremsstrahlung radiation from a parent electron population, and γ-ray emission via the decay of π0 created from a proton population interacting with the surrounding gas (see Table 2 for a complete list of parameters).

5 Summary and conclusions

Based on our comprehensive update on the radio and γ-ray radiations from Kes 17, along with our analysis of the molecular environment brightening in the ¹²CO and ¹³CO (J = 1–0) lines, we propose the following picture: Kes 17 was created in a stellar explosion that occurred approximately 11 kyr ago in an ambient medium with a density of about 7 cm⁻³. The observed spectral shape of the shock front emitting from 88 to 8800 MHz is adequately fitted with a simple power-law model with an index α = −0.488 ± 0.023. The available radio data suggest that no ionised gas is located in, around, or anywhere along the sight-line to Kes 17 that significantly impacts the integrated spectrum. If it were present, this ionised gas might produce an spectral curvature below 100 MHz due to free-free thermal absorption. Low frequency radio data with a higher sensitivity and resolution are key to spatially resolving spectral curvatures due to intrinsic properties related to the shock in Kes 17 and its interaction with the immediate SNR surroundings observed in CO and infrared lines.

The eastern part of Kes 17 is wrapped around a CO cloud. The main evidence for this interaction includes the distortion of the SNR shock and the distance to the CO cloud, which is found to be completely compatible with the distance to the remnant. The average mass and density of this cloud are determined to be 4.2 × 10⁴ M_⊙ and 300 cm⁻³. No appreciable CO emission is detected towards the western region of the radio shell, where molecular hydrogen has been proven to be shocked by the Kes 17 shock front. This suggests the presence of a CO-dark molecular gas. Additionally, no features of atomic hydrogen physically connected to Kes 17 are detected at the sensitivity and resolution of the SGPS data used in this work.

In its evolution, Kes 17 produces γ-ray photons at GeV energies, which have been observed by the Fermi-LAT telescope. The flux and luminosity at 7.9 kpc in the 0.3–300 GeV energy band are estimated to be (2.98 ± 0.14) × 10⁻¹¹ erg cm⁻² s⁻¹ and (2.22 ± 0.45) × 10³⁵ erg s⁻¹, respectively. The spectrum of this high-energy emission has an index . Based on observational evidence and modelling of the broadband SED ranging from radio to γ rays, it has been determined that a purely leptonic (IC) scenario is not favoured as an explanation for the emission from Kes 17 observed at GeV energies. Instead, the evidence suggests that the primary contribution to the γ-ray flux originates from the collision between the western part of the SNR shock front and a dense IR-emitting region. Consequently, our analysis adds Kes 17 to the list of SNRs whose emission at GeV energies is hadron dominated. The γ-ray luminosities measured in this class of remnants exhibit differences that can be interpreted in terms of the amount of molecular gas, which serves as target material for cosmic-ray interactions, as well as time-evolution effects. Future observations conducted using the most advanced instruments operating at the highest energies of the electromagnetic spectrum, coupled with improved resolution and sensitivity, will contribute to refining our understanding of the spectral and morphological behaviour of Kes 17 in the γ-ray regime.

Acknowledgements

We are very grateful for the thorough corrections provided by the anonymous referee. G.C. and L.S. are members of the Carrera del Investigador Científico of CONICET, Argentina. This work was supported by the ANPCyT (Argentina) research project with number BID PICT 2017-3320. This work and collaboration is also supported by the International Emerging Actions program from CNRS (France).

References

All Tables

All Figures

Fig. 1 Spectrum of the continuum emission from SNR Kes 17 at radio frequencies, constructed with the fluxes in Table 1. Blue data points denote our new measurements from public survey images, and the green points correspond to literature flux estimates. The straight line represents the best fit with a power-law model in the form Sν ∝ να, which yields a spectral index value α = −0.488 ± 0.023. The shaded darker and lighter pink regions around the straight line denote a variation in the fitted spectral parameters of 1 and 2σ, respectively.

In the text

Fig. 2 Spatial and spectral distribution of the CO gas towards Kes 17 from ThrUMMS (Barnes et al. 2015). (a) Colour-coded image of the 12CO (in red) and the 13CO (in green) J =1–0 line emissions, integrated from −31 to −14 km s−1. The yellow regions are areas where the two CO isotopologue emissions overlap. The contours (levels 0.034, 0.35, 0.75, and 1.2 mJy beam−1, with a beam smoothed to the 80″ spatial resolution of the CO data) delineate the 843 MHz continuum radiation from Kes 17. For reference, the inset in the upper left corner displays the structure of the SNR shell as observed by the SUMSS (45″ resolution). The overlaid grid consists of the series of boxes (1′.5 in size) used to analyse the kinematics of the CO gas. The H II region G304.465–00.023 in the field is also labelled (Urquhart et al. 2022). (b) Collection of 12CO (in red) and 13CO (in green) J =1–0 spectra extracted from the boxes, numbered from 1 to 42, in panel a. The original datasets were convolved to a resolution of 80″ to reduce the graininess. The yellow contours superimposed on the spectra correspond to the radio continuum emission from Kes 17.

In the text

Fig. 3 Position-velocity (p-υ) diagrams for the 12CO emission towards the field of Kes 17. (a) Regions numbered from 1 to 7 that were used to construct the p-υ diagrams for the eastern cloud associated with Kes 17, shown in panels b–h. The colours indicate the spatial distribution of the 12CO J =1–0 gas integrated in the (−31, −14) km s−1 velocity range as in Fig. 2, and the contours trace the 843 MHz radio continuum emission from Kes 17. (b–h) p-υ diagrams derived from the 12CO line emission depicted in panel a. They were constructed by integrating in the 100″ RA interval indicated by the regions numbered from 1 to 7 in panel a. The horizontal dashed lines mark the extent of Kes 17 in the declination dimension. The colour representation is the same for all p-υ diagrams.

In the text

Fig. 4 Test-statistics map with 0°.01 pixel size in the 0.3–300 GeV energy band in the Kes 17 SNR field (see text for details). The blue contours delineating the SNR radio synchrotron emission are from the 843 MHz SUMSS map (levels 0.015, 0.15, and 0.30 Jy beam−1, at the resolution of 45″). The positions and distances of the H II region G304.465–00.023 (Urquhart et al. 2022) and the two known pulsars, PSR J1306–6242 (Kramer et al. 2003) and PSR J1305–6256 (Manchester et al. 2001), lying in the field, are also labelled. The plus symbol and dark-green ellipse within the Kes 17 radio contours mark the best-fit position and 95% confidence region, respectively, for the GeV emission obtained with a point-source spatial template. The dashed brown circle traces the 95% confidence fitted size of the γ-ray source. The inset shows a ~15′ × 15′ close-up view of the Kes 17 region, centred at the position of the γ-ray excess.

In the text

Fig. 5 Spectral energy distribution of the γ-ray emission in the region of Kes 17, as detected by Fermi-LAT. The orange line indicates the best fit to the data using the power-law model dN/dE = ϕ0(E/E0)−Γ with (see text for details). The yellow shaded zone corresponds to the 1σ region of the spectral fit. The blue and red error bars represent statistical and systematic uncertainties in the spectral points, respectively.

In the text

Fig. 6 Broadband SED from radio to γ rays for Kes 17. The fluxes in the radio band correspond to those in Table 1 (in units of erg cm−2 s−1), while those at γ-ray energies correspond to the new Fermi-LAT data from this work (Sect. 4.2). The upper limits at X-ray and TeV energies where derived from the non-detection of X-ray synchrotron emission and the H.E.S.S. observatory, respectively (see text for details). (a) SED modelling with a unique electron population producing synchrotron and IC radiation over the broadband spectrum. (b) SED modelling with an ExpCut-off proton-dominated model producing synchrotron, IC, and Bremsstrahlung radiation from a parent electron population, and γ-ray emission via the decay of π0 created from a proton population interacting with the surrounding gas (see Table 2 for a complete list of parameters).

In the text