A MULTI-WAVELENGTH INVESTIGATION OF THE UNIDENTIFIED GAMMA-RAY SOURCE HESS J1708−410

Within the XMM-Newton field of view a number of point sources are detected. No obvious extended emission is seen with the exception of contaminating reflections from a nearby bright X-ray source at energies above 2 keV (see Figure 1). Two ROSAT hard-band X-ray sources exist to the northeast: 03 away lies J171011.5−405356, and slightly farther away is low-mass X-ray binary (LMXB) 4U 1708−40, either of which could be responsible for the reflection patterns.

We employ the SAS function edetect_chain for source detection and search in two energy bands of 0.3–2 keV and 2–10 keV for sources with a probability P < 10⁻¹³ of arising from random Poissonian fluctuations. This rigorous cutoff threshold is necessary due to the aforementioned contaminating streaks. Ten sources are detected, although we remove one source which lies directly along the arcs of reflection and has zero counts in the soft band. The source detection algorithm also attempts to determine source extension via a Gaussian model, though all detections are consistent with a point source. Table 1 lists the X-ray data. We define the hardness ratio as HR = (C_hi − C_lo)/(C_hi + C_lo) where C_lo and C_hi are MOS counts in the 0.3–2 keV and 2–10 keV bands, respectively. The only source with a cataloged SIMBAD counterpart is source 2, which coincides with the field star HD 326944 of magnitude ∼10 and type G0. Sources 1 and 3 lie within 1'' of USNO-B1.0 12th and 18th magnitude sources, respectively. Only sources 1 and 2 have sufficient counts to extract meaningful point-source spectra, though fits to these sources are still accompanied by large errors. Source 1 is best matched with an absorbed power law, with hydrogen column density hardly constrained at n_H = 2.0^+2.7_−2.0 × 10²¹ cm⁻² (90% single parameter errors). As such, we fix n_H at 2.0 × 10²¹ cm⁻² and measure an index of 2.9^+0.5_−0.4 and unabsorbed 0.3–10 keV flux of 1.3 ± 0.3 × 10⁻¹³ erg cm^-2 s⁻¹. Source 2 (HD 326944) is best modeled by an absorbed MEKAL plasma, with n_H again essentially unconstrained at $1.3_{-0.6}^{+5.4\times 10^5} \times 10^{16} \, \rm cm^{-2}$. With n_H fixed at 1.0 × 10¹⁶ cm⁻² we find a temperature of $kT = 0.49_{-0.10}^{+0.09} \, \rm keV$ and unabsorbed 0.3–10 keV flux $4.2_{-0.7}^{+0.5} \times 10^{-14} \, \rm erg \, cm^{-2} \, s^{-1}$.

The 73 ms time resolution of the PN camera allows searches for periodicities from a possible pulsar candidate to be conducted; the MOS data sets have a time resolution of 2.6 s, insufficient to search for a signal from a typical rotation powered pulsar. To this end, for each of the three brightest point sources, we barycenter the events, extract a light curve, and search for periodicities within the PN data set with the Xronos function powspec; no significant periodicities are detected.

We analyzed ¹²CO (J 1 → 0) data from the survey of Dame et al. (2001) to search for molecular cloud associations with J1708, constrain the target material density, and estimate the distance to J1708. Figure 2 shows the CO velocity profile along the line of sight to J1708 in the single 015 × 015 pixel which covers the TeV source. A single major molecular complex is visible in this profile with a probable distance of ≈3 ± 1 kpc, adopting the Galactic rotation curve of (Brand & Blitz 1993) and assuming a near solution. Effectively all massive star formation occurs in molecular clouds, and all known Galactic TeV sources are related to massive stars or their end products (PWN, SNR, etc.). As such, this 3 ± 1 kpc distance (corresponding to the near side of the molecular ring of the Milky Way) seems the most likely location for HESS J1708. We therefore assume a distance of 3 kpc in all subsequent discussions and scale distances by d₃ ≡ d/(3 kpc). A more speculative alternative possibility is that J1708 is associated with the ∼01 cavity suggested by the Spitzer image of Figure 3 and comparable in size to the HESS source. The region of low infrared emission toward the center of J1708 also seems to correspond to a hole in the nearby (∼1 kpc) CO emission. We could perhaps be looking at a bubble blown by an SNR or stellar wind. There is, however, no evidence of a stellar cluster at the center of this bubble, which undermines this hypothesis.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

An upper limit on the density inside the source can be derived from the column density toward the source and the observed source size (θ = 10' FWHM) n ⩽ n_H/θ D. For the measured molecular column around 3 kpc of ${\approx }5 \times 10^{21} \rm \, cm^{-2}$ (24 K km with a conversion factor $X \approx 2 \times 10^{20} \, \rm cm^{-2}$ (K km)⁻¹) this corresponds to $n \le 200 \, d_3^{-1} \, \rm cm^{-3}$. Note that the dominance of bremsstrahlung over IC on the cosmic microwave background radiation (CMBR) at photon energies of 1 TeV requires $n > 240 \, \rm cm^{-3}$ for an E⁻² electron spectrum (Hinton & Hofmann 2009). We therefore neglect bremsstrahlung as an emission mechanism in the discussions below.

In the absence of any obvious extended X-ray emission from J1708, we fit the flux and use this as an upper limit. To calculate this upper limit, we define a region free of X-ray point sources and scattered light contamination within the innermost (8σ) HESS contour. This region encompasses 18% of the flux from the HESS spectral integration area. We extract the spectrum from this region with the SAS task evselect, and generate response files with the tasks arfgen and rmfgen. The background is taken from a region near the outer edge of the chip, exterior to the HESS contours. Spectra from both the MOS1 and MOS2 chips are fitted to a simple absorbed power law. This fit yields a hydrogen column density of $(2.1\pm 1.2)\times 10^{21} \rm \, cm^{-2}$, which can be compared to calculations from H i maps of the total galactic column density. Kalberla et al. (2005) estimate a column of $1.34\times 10^{22}\ \rm cm^{-2}$, and Dickey & Lockman (1990) tabulate $1.65\times 10^{22}\ \rm cm^{-2}$. The fit hydrogen column density is therefore consistent with a Galactic origin.

Scaling up the unabsorbed power-law flux by the flux ratio of 1/0.18 (i.e., assuming matching X-ray and TeV morphology) provides a 90% flux limit of $1.8\times 10^{-12} \rm \, erg \, cm^{-2} \, s^{-1}$ in the 0.3–10 keV range and a limit of $3.2\times 10^{-13} \rm \, erg \, cm^{-2} \, s^{-1}$ between 2 and 4 keV. The 2–4 keV limit is far less sensitive to the estimated column density to the source, so we quote this value in the following discussion.

While no deep radio images are available in this locale, we can extract upper limits from two major surveys of the southern Galactic plane. The first epoch Molonglo Galactic Plane Survey is a radio continuum survey at 843 MHz with a beam of 43''× 65'' at the source position, and rms sensitivity of 1–2 mJy beam⁻¹ (Green et al. 1999). There is no sign of diffuse emission close to J1708 in the relevant Molonglo image; one point-like source exists within the HESS contours, but does not coincide with any of the XMM-Newton sources. We calculate a conservative mean surface brightness of $3 \rm \, mJy \, beam^{-1}$ within the source region and integrate over the 006 × 008 extent to derive a flux density upper limit of 270 mJy. Higher frequency radio upper limits are available from the Parkes 2.4 GHz survey of the southern Galactic plane; images have a resolution of 104 (≈ the size of J1708) with rms noise of approximately $12 \rm \, mJy \, beam^{-1}$ (Duncan et al. 1995). We see no indication of a source at the position of J1708, and derive an upper limit of $400 \rm \, mJy \, beam^{-1}$. As the source size is approximately the same as the beam size, we place a flux density upper limit of 400 mJy at 2.4 GHz.

The simplest reasonable injection model is an electron power law with a high-energy cutoff. We explore three source ages with this model: 1, 10, and 100 kyr. All ages yield comparable fits to the VHE data, though source parameters must be varied. Enhanced cooling for older sources requires a greater value for the total energy and high-energy cutoff, as well as a lower magnetic field to allow high-energy electrons to survive and IC upscatter to TeV energies. For a 10 kyr source, one can match the data with a total injection energy of 5 × 10⁴⁷ d²₃ erg, an electron power law of slope −1.8, and a high-energy exponential cutoff at 14 TeV. The high-energy cutoff is required to push the X-ray flux below the new XMM-Newton limit; 14 TeV is the minimum allowed high-energy cutoff which still matches the HESS data. One could fit the HESS data with a cutoff energy slightly greater than 14 TeV (this would better match the highest energy HESS points, while worsening the match to the lower energy HESS points), though this would require a lower magnetic field due to the X-ray constraint on synchrotron radiation. For a 14 TeV cutoff, the radio and X-ray limits impose an already rather stringent upper limit on the magnetic field of 4 μG for this spectral shape. A spectral index softer than −1.8 is precluded by the radio upper limits. Such a hard injection spectrum would be remarkable, as a slope of −1.8 is not required for any other galactic VHE source. For this electron spectrum, the synchrotron component peaks in the UV, and IC emission peaks around 1 TeV. Klein–Nishina effects suppress near-IR scattering at the highest energies, so only far-IR and cosmic microwave background (CMB) photons contribute to the VHE regime, with far-IR the dominant component. Figure 4 shows the SED in this scenario, as well as the SED for ages of 1 and 100 kyr. For a source age of 1 kyr and the same power-law index, we require a cutoff at 8 TeV, an energy of 5 × 10⁴⁷ d²₃ erg, and a maximum magnetic field of 4 μG. At this young age, the radio upper limits prove far more constraining than the X-ray limits. A more aged source, at 100 kyr, has a total energy of 6 × 10⁴⁷ d²₃ erg, and requires a cutoff at 45 TeV and a very low magnetic field of 2 μG in order for energetic electrons to survive the synchrotron and IC cooling process.

We also explore the consequences of implementing a low-energy cutoff in the injection spectrum. As discussed in Section 1, a low-energy cutoff is expected for a PWN paradigm; in the case of the Crab Nebula, Kennel & Coroniti (1984) suggest a minimum injection energy of ∼ 1 TeV at the wind termination shock, while Wang et al. (2006) infer a minimum low-energy cutoff of 5 GeV for G 359.95–0.04. We inject an electron power law of slope −2, and a low-energy cutoff of 10 GeV. This low-energy cutoff allows the softer electron spectrum to remain consistent with the radio limits. For an age of 10 kyr, we require an exponential cutoff at 16 TeV, and a maximum magnetic field of 5 μG. Matching the VHE data requires a total of 6 × 10⁴⁷ d²₃ erg in electrons. Such a value is reasonable for a PWN origin; one of the best-studied galactic broadband sources, the Vela X nebula surrounding the Vela pulsar, is estimated to possess ∼10⁴⁸ erg in the form of leptons (de Jager 2007). A 1 kyr aged source requires an exponential cutoff at 12 TeV, magnetic field of 5 μG, and energy of 6 × 10⁴⁷ d²₃ erg. Finally, a 100 kyr aged source requires an exponential cutoff at 60 TeV, magnetic field of 2 μG, and energy of 8 × 10⁴⁷ d²₃ erg.

A hadronic scenario relaxes the strictures imposed by the XMM-Newton upper limit, and instead one finds the radio upper limits more constraining. A proton power law with index −2 and cutoff at 50 TeV provides an adequate fit to the HESS data, requiring energy E = 8 × 10⁵⁰(n/1 cm^-3) d²₃ erg. The timescale for pion production via p–p interactions of $\tau _{pp} \approx 1.5 \times 10^{8} \, (n/1 \, \rm cm^{-3})^{-1}$ yr (Blumenthal 1970) is significantly greater than the expected age of the system, so the proton spectrum is treated as static. For this proton spectrum, we calculate the photons from proton–proton interactions and subsequent π⁰ and η-meson decay following (Kelner et al. 2006). Proton–proton interactions also yield π^± mesons which decay into secondary electrons, which we evolve in a 5 μG field. The secondary electron spectrum is a function of the age of the system, so we evolve over 1, 10, and 100 kyr, as shown in Figure 5. IC and synchrotron fluxes from the resultant secondary electron spectrum are subsequently calculated, indicating that secondary IC is unimportant, while for high source ages one must take care to ensure that the synchrotron component does not exceed the radio upper limits. We observe the π⁰ gamma rays as a broad peak in the GeV to TeV range, indicating that a Fermi detection would not be surprising.

The steep (Γ = −2.5) HESS TeV spectral index could imply that electrons responsible for this emission are cooled. As already discussed, in a leptonic scenario a high magnetic field is precluded (given the low TeV to X-ray flux ratio), which leaves a large age as the remaining cooling possibility. Indeed, for a source age of 1 kyr cooling is negligible, while at 10 kyr cooling is just beginning to affect the electrons responsible for upscattering photons to the VHE regime (which necessitates the slightly greater value for the high-energy cutoff). The mean electron energy required to IC scatter the CMB seed photons to TeV energies (E_TeV) is $E_e = \rm (18\ TeV)\,E^{1/2}_{ \rm TeV}$, so the presence of gamma rays up to 50 TeV indicates electrons must be accelerated to ∼100 TeV levels. For more modest 10 TeV electrons, the cooling timescale for electrons IC scattering off the CMBR is $\tau _{{\rm IC}} = 100 \, (E/10 \, \rm TeV)^{-1}$ kyr. In regions of very low magnetic field the IC timescale is comparable to the synchrotron cooling lifetime $\tau _{{\rm sync }} = 90 \, (E/10 \, \rm TeV)^{-1} (B/3 \, \mu \rm G)^{-2}$ kyr. Regardless of whether electrons lose their energy primarily through IC or synchrotron, based solely on cooling timescales one might estimate J1708 to be as old as 100 kyr for very low ambient magnetic fields. One can compare this age with estimates of the diffusion timescale. A source subtending 007 has radius R = 3.7d₃ pc. Then for a cooling limited electron source with Bohm diffusion $t_{{\rm diff}} = 19 \, d_{3}^2 \, (B/3 \, \mu \rm G) (E/10 \, \rm TeV)^{-1}$ kyr. This indicates a problem; it becomes difficult to confine the VHE electrons required to power the HESS source long enough for a cooling break to develop. This argues against a source age as high as 100 kyr, since for the canonical electron power-law index of −2 we found above that significant numbers of electrons must persist up to 60 TeV to combat the high degree of cooling over such a lengthy period.

Another possibility of explaining the steep HESS spectrum is that there is a cutoff in the electron spectrum not far above the HESS range (at some tens of TeV) corresponding to either a maximum in the acceleration energy or the escape of the highest energy electrons. A truncated accelerator provides a simple explanation, but even if the accelerator persists up to the PeV range, 100 TeV electrons tend to escape in 2 kyr or faster for Bohm diffusion. The SED fits indicate that a source aged 10 kyr will suffer some cooling due to the low magnetic field required by the X-ray limit, but the location of the IC peak still depends critically on the value of the high-energy cutoff. A cutoff, rather than fully cooled, distribution of electrons therefore seems the most natural explanation for the HESS data in the leptonic scenario, justifying our assumed injection spectrum of a power law with an exponential cutoff. Such a cutoff also argues against an age of 100 kyr, noted above.

For an age of 10 kyr one class of potential counterparts, SNRs, could easily occupy the volume assumed for a 3 kpc distance, since this implies a very reasonable mean expansion velocity of $360 \, d_3 \, (\theta /0\mbox{$.\!\!^\circ $}07) \, (T/10 \, {\rm kyr})^{-1} \rm \, km \, s^{-1}$. Middle-aged pulsars and their corresponding wind nebulae offer another possibility. The reverse shock from the parent SNR is expected to reach the PWN in ∼7 kyr (Reynolds & Chevalier 1984), which (as in the case of the Vela PWN; Blondin et al. 2001) can displace the PWN far from the parent pulsar where the magnetic field can be quite low and the scenarios described above remain plausible. Another possibility stems from the high transverse velocity of pulsars, which typically cross their parent SNR shells after ∼40 kyr (Gaensler & Slane 2006). At this late stage, the pulsar has often escaped its original wind bubble, leaving behind a relic PWN (van der Swaluw et al. 2004). In this environment, a lack of X-ray emission is not surprising, as the ambient magnetic field can be very low, while VHE photons arise from IC scattering of the relic electrons.

In the hadronic scenario, the high-energy cutoff at 50 TeV most likely stems from either the maximum accelerator energy or the diffusive escape of the highest energy particles. The Bohm diffusion timescale predicts a loss of 50 TeV particles in ≈6 kyr in a 5 μG field. The assumed source age of 10 kyr would therefore be reasonable. The new XMM-Newton limit places few constraints on the model. In the radio regime, however, the flux scales as the number of electrons n_e × B^3/2, or since we assume constant injection of particles, flux scales as the age of the system T × B^3/2; for J1708, we find $(B/10 \, \mu {\rm G})^{3/2} \,(T/50 \rm \, kyr) < 1$. Though this limit is not very constraining, it does indicate that our assumptions for age and magnetic field are reasonable. The energy requirement of E = 8 × 10⁵⁰(n/1 cm^-3) d²₃ erg is consistent with a supernova origin for ambient densities of ∼10 cm⁻³.