The GeV-TeV Connection in Galactic gamma-ray sources
aa r X i v : . [ a s t r o - ph ] O c t Submitted to the Astrophysical Journal
The GeV-TeV Connection in Galactic γ -ray sources S. Funk , O. Reimer , D. F. Torres , J. A. Hinton ABSTRACT
Recent observations with atmospheric Cherenkov telescope systems such asH.E.S.S. and MAGIC have revealed a large number of new sources of very-high-energy (VHE) γ -rays from 100 GeV – 100 TeV, mostly concentrated along theGalactic plane. At lower energies (100 MeV – 10 GeV) the satellite-based in-strument EGRET revealed a population of sources clustering along the GalacticPlane. Given their adjacent energy bands a systematic correlation study betweenthe two source catalogues seems appropriate. Here, the populations of Galacticsources in both energy domains are characterised on observational as well as onphenomenological grounds. Surprisingly few common sources are found in termsof positional coincidence and spectral consistency. These common sources andtheir potential counterparts and emission mechanisms will be discussed in de-tail. In cases of detection only in one energy band, for the first time consistentupper limits in the other energy band have been derived. The EGRET upperlimits are rather unconstraining due to the sensitivity mismatch to current VHEinstruments. The VHE upper limits put strong constraints on simple power-lawextrapolation of several of the EGRET spectra and thus strongly suggest cutoffsin the unexplored energy range from 10 GeV – 100 GeV. Physical reasons forthe existence of cutoffs and for differences in the source population at GeV andTeV energies will be discussed. Finally, predictions will be derived for commonGeV–TeV sources for the upcoming GLAST mission bridging for the first timethe energy gap between current GeV and TeV instruments. Kavli Institute for Particle Astrophysics and Cosmology (KIPAC), SLAC, CA 94025, [email protected] Stanford University, W. W. Hansen Experimental Physics Lab (HEPL) and KIPAC, Stanford, CA94305-4085, USA. [email protected] ICREA & Institut de Ciencies de l’Espai (IEEC-CSIC) Campus UAB, Fac. de Ciencies, Torre C5, parell,2a planta, 08193 Barcelona, Spain. [email protected] School of Physics and Astronomy, University of Leeds, Leeds LS2 9JT, UK. [email protected]
Subject headings: gamma rays: observations; Galaxy: general; (ISM:) supernovaremnants
1. Introduction
In recent years the knowledge of the Galactic VHE γ -ray sky above 100 GeV hasbeen greatly improved through the detection and subsequent study of many sources, mostlyby means of ground-based Imaging Atmospheric Cherenkov telescope systems such as theHigh Energy Stereoscopic System (H.E.S.S.) or the Major Atmospheric Gamma-ray Imag-ing Cherenkov Observatory (MAGIC). Currently known Galactic VHE γ -ray emitters in-clude shell-type Supernova remnants (SNRs) (Aharonian et al. 2006a, 2007a,b), Pulsar WindNebulae (PWNe) (Aharonian et al. 2005a, 2006b,c), γ -ray binaries (Aharonian et al. 2006d;Albert et al. 2006), Molecular clouds (Aharonian et al. 2006e) and possibly also clusters ofmassive stars (Aharonian et al. 2007c). These various source classes were discovered bothin pointed observations using H.E.S.S. and MAGIC as well as in a systematic survey ofthe inner Galaxy performed with the H.E.S.S. instrument. The highest energy photonsdetected from these source classes reach ∼
100 TeV (Aharonian et al. 2007a), currently rep-resenting the end of the observable electromagnetic spectrum for astrophysical objects. It isnatural to investigate the relationship of these TeV sources to sources at lower energies aswill be done in this work focusing on Galactic sources. The closest energy band for whichdata exist is that studied by the Energetic Gamma Ray Experiment Telescope (EGRET)aboard the Compton Gamma-Ray Observatory with an energetic coverage from 100 MeV –10 GeV (Hartman et al. 1999). The GeV sky has a distinctively different overall appearancecompared to TeV energies. In particular focusing on our Galaxy, the most prominent featureof the GeV sky is the dominant diffuse emission from cosmic ray (CR) interactions in theGalaxy, while the TeV sky due to the steeply falling energy spectrum of the diffuse com-ponent is dominated by individual sources. However, several prominent γ -ray sources areknown to emit at both GeV and at TeV energies, the Crab Nebula being the most prominentexample (Weekes et al. 1989; Nolan et al. 1993; Aharonian et al. 2004, 2006f; Albert et al.2007a).In this paper the relationship between Galactic EGRET and VHE γ -ray sources will beassessed in a systematic way. For cases with a positional coincidence between a VHE and anEGRET source (in the following called “coincident sources”) all currently known Galacticobjects will be considered. For cases in which a source is detected only in one band – the“non-coincident sources” – we focus on the region covered by the H.E.S.S. Galactic planesurvey (GPS) during 2004 and 2005 (Aharonian et al. 2005b, 2006g) (Galactic longitude 3 – ± ◦ , Galactic latitude ± ◦ ) so that a statistical assessment of the “non-connection” canbe made. EGRET was unable to perform detailed studies of the γ -ray sky above 10 GeV,partly due to back-splash of secondary particles produced by high-energy γ -rays causing aself-veto in the monolithic anti-coincidence detector used to reject charged particles. Theupcoming Gamma Ray Large Area Space Telescope (GLAST) Large Area Telescope (LAT)will not be strongly affected by this effect since the anti-coincidence shield was designed ina segmented fashion (Moiseev et al. 2007). Moreover, the effective area of the GLAST-LATwill be roughly an order of magnitude larger then that of EGRET. The GLAST-LAT missionwill therefore for the first time fully bridge the gap between the energy range of EGRETand current VHE instruments. Part of the study presented here can be seen as preparatorywork for GLAST-LAT studies of sources in the largely unexplored energy band between 10and 100 GeV.From the 2004 and 2005 H.E.S.S. GPS 22 VHE γ -ray sources have been reported in theInner Galaxy. The third EGRET catalogue (Hartman et al. 1999) represents the companionto the VHE source catalogue above an energy threshold of 100 MeV (with peak sensitivitybetween 150 and 400 MeV, depending on the γ -ray source spectrum). It lists 271 sources,17 of which are located within the H.E.S.S. GPS region. Whilst the EGRET range currentlyrepresents the nearest energy band to VHE γ -rays, for the very few EGRET sources detectedall the way up to ∼
10 GeV, there is still an unexplored energy band of roughly one decadebefore the VHE γ -ray energy range begins at ∼
100 GeV (it should be noted that EGRETdoes have some sensitivity beyond 10 GeV: Thompson, Bertsch & O’Neal (2005) reportedthe detection of ∼ ◦ of a source listed in the third EGRET catalogue). Comparing the instrumentalparameters of VHE instruments and EGRET there is a clear mismatch both in angularresolution and in sensitivity as can be seen in Figure 1. In a ∼ ∼ −
80 more sensitive (in termsof energy flux E dN/dE ) than EGRET above 1 GeV in the Galactic Plane for the exposureaccumulated between 1991 and 1995 (corresponding to the third EGRET catalogue). As-suming a similar energy flux output in the two different bands this mismatch implies at firstsight that H.E.S.S. sources are not likely to be visible in the EGRET data set. Conversely(again under the assumption of equal energy flux output), VHE γ -ray instruments shouldbe able to detect the majority of the EGRET sources, as has been suggested in the past.Figure 2 compares the energy fluxes νF ν for EGRET sources and H.E.S.S. sources in theinner Galaxy. Clearly, the EGRET sources do not reach down as low in energy flux as theH.E.S.S. sources, a picture that will change once the GLAST-LAT is in orbit as depicted bythe GLAST-LAT sensitivity (dashed line). In reality the na¨ıve expectation of equal energyflux output in the GeV and TeV band can easily be wrong in Galactic γ -ray sources for 4 – Energy (eV) ) - s - ( e r g c m n F n -13 -12 -11 -10 -9 Crab NebulaINTEGRAL EGRETGLAST H.E.S.S.MAGIC
H.E.S.S. SurveyGLAST Inner Galaxy
Energy (eV) A ng l e f o r % c on t a i n m e n t ( d e g ) -1 H.E.S.S.EGRETGLAST (all)GLAST (thin convertor)
Fig. 1.—
Left:
Integral sensitivities for current, past and future γ -ray instruments (5- σ sensitivity for E > E multiplied with E assuming a spectrum of E − ). The solidlines show the nominal instrument sensitivities (for a typical observation time as spec-ified below), the dashed curves show the actual sensitivities for the Inner Galaxy asappropriate for this work. INTEGRAL’s (IBIS/ISGRI) sensitivity curve (solid green)shows the sensitivity for an observation time of 10 s, a typical value in the InnerGalaxy. The EGRET curves (brown) are shown for the whole lifetime of the mis-sion (periods 1–9) for the Galactic anti-centre (solid) which received the largest expo-sure time and has a lower level of diffuse γ -ray emission than the Inner Galaxy andfor the position of RX J1713.7–3946 (dashed), a typical position in the Inner Galaxydominated by diffuse γ √ . Right:
Energy-dependence of the angular resolution for selected γ -rayinstruments expressed by the 68%-containment radius of the point-spread function (PSF).As can be seen, the angular resolution of GLAST becomes comparable with current VHEinstruments at high energies, whilst at the lower energy end GLAST and EGRET havecomparable resolutions. 5 –various reasons: EGRET sources may not emit comparable energy fluxes in the VHE γ -rayband but rather exhibit cut-offs or spectral breaks in the energy band between EGRET andH.E.S.S. (this is certainly the case for pulsed magnetospheric emission from pulsars, see forexample Aharonian et al. 2007d). Furthermore, H.E.S.S.-like instruments are typically onlysensitive to emission on scales smaller than ∼ ◦ . If any of the EGRET sources are extendedbeyond 1 ◦ without significant sub-structure on smaller scales (not precluded given the poorangular resolution of EGRET), current Imaging Cherenkov instruments may not be able todetect them since these sources would completely fill the field of view (FoV) and be removedby typical background subtraction methods (see for example Berge, Funk & Hinton 2007).Given the upcoming launch of GLAST and the recent H.E.S.S. survey it seems timely tostudy the relationship between GeV and TeV emitting sources in more detail. )) -1 s -2 log10(Energy Flux (ergs cm -13 -12.5 -12 -11.5 -11 -10.5 -10 -9.5 -9 -8.5 -8 E n t r i e s H.E.S.S. EGRETGLAST
Fig. 2.— Distribution of integrated energy flux νF ν for sources in the Inner Galaxy discussedhere. For EGRET the energy flux between 1 GeV and 10 GeV, for the H.E.S.S. sources, theenergy flux between 1 TeV and 10 TeV is shown. Also shown is the sensitivity predictionfor the GLAST-LAT for a typical location in the Inner Galaxy (l=10, b=0).Section 2 describes the data and analysis methods used in this study, section 3 describesthe sources detected in both energy bands, and section 4 focuses on sources detected in onlyone of the two energy regimes. In section 5 astrophysical implications of the study arediscussed. 6 –
2. Analysis methods
For the sources discussed in this study locations and source spectra in the EGRETband (Hartman et al. 1999) and in the VHE γ -ray band are required. For the inner Galaxy,dedicated upper limits at the specific position of the γ -ray sources in the respective otherband were determined. For the EGRET data these upper limits (at 1 GeV) were derived atthe nominal positions of the H.E.S.S. sources based on a reanalysis of the data used for theproduction of the third EGRET catalogue, applying the standard EGRET likelihood fittingtechnique (Mattox et al. 1996). For the H.E.S.S. data, 2 σ upper limits at the nominal posi-tion of each EGRET source were estimated. This was done by scaling the flux correspondingto the H.E.S.S.-point-source sensitivity in 25 hours (1% of the Crab) by the square-root ofthe ratio of 25 hours to the published exposure time at the position of the EGRET source(taken from Aharonian et al. 2006g). Figure 3 shows all H.E.S.S. and all EGRET sources within the HESS GPS region. Oneproperty of EGRET and VHE γ -ray sources becomes immediately apparent: only a minorfraction of the H.E.S.S. sources coincide within the considerably larger location uncertaintycontours of EGRET GeV sources. Given the rather poor angular resolution of EGRET (68%containment radius of the PSF: 1 . ◦ at 1 GeV) coupled with typically rather limited photonstatistic any systematic assessment of positional matches between EGRET and H.E.S.S.sources is dominated by the localisation error on the EGRET source position. The likelihoodsource position uncertainty contour (PUC) as given in Hartman et al. (1999) have been usedto check for VHE γ -ray sources within these regions on the sky. While most of the VHEsources are extended, their extension is rather small on the scale of the EGRET positionaluncertainty and therefore a source is classified as “coincident” if the centre of gravity of theVHE emission is within the EGRET likelihood PUC. For large sources such as e.g. the SNRRX J1713.7–3946 (HESS J1713–395) this approach is clearly an oversimplification, albeit itis the one used at this stage of this study.The number of spatially coincident sources depends on the EGRET PUC chosen in theinvestigation. For the H.E.S.S. GPS-region, no VHE γ -ray source is located within the 68%positional confidence contour of any EGRET source. Relaxing the coincidence criterion, twoVHE γ -ray sources are found within the 95%-confidence contour of EGRET source positions(shown in red in Figure 3) and an additional three VHE γ -ray sources are located withinthe 99%-confidence contours (shown in orange in Figure 3). Outside the H.E.S.S. GPS-region, no systematic statistical assessment of the non-coincident sources is possible due to 7 –Fig. 3.— Map of the H.E.S.S. GPS region taken from Aharonian et al. (2006g). PublishedH.E.S.S. sources are marked as squares. EGRET sources are shown with their 95% positionalconfidence contours from the 3EG-catalogue. The red (orange) contours and labels denotethose 3EG sources for which a H.E.S.S. source centroid is located within the 95% (99%)confidence contour, the blue contours denote the 95% PUCs for the EGRET sources forwhich no VHE emission is detected.the highly non-uniform exposure of the observations with the limited-FoV VHE instruments.Nevertheless, is it relevant to note that four additional coincident sources are found outsidethe H.E.S.S. GPS-region within the Galactic plane (defined here as as latitude range of 8 –EGRET VHE γ -raySource SourceWithin 68% Within 95% Within 99%PUC PUC PUCWithin the H.E.S.S. GPS3EG J1639–4702 HESS J1640–4653EG J1744–3011 HESS J1745–3033EG J1800–2338 None HESS J1800–2333EG J1824–1514 HESS J1826–1483EG J1826–1302 HESS J1825–137Outside the H.E.S.S. GPS3EG J0241+6103 MAGIC J0240+6133EG J0617+2238 None MAGIC J0616+2253EG J0634+0521 HESS J0632+0583EG J1420–6038 HESS J1420–607Table 1: Positionally coincident EGRET and H.E.S.S. sources within our Galaxy for threeconfidence levels (68%, 95%, and 99%) of the positional uncertainty of the EGRET source. ± ◦ ): HESS J1420–608 (Aharonian et al. 2006c) in the Kookaburra region is located withinthe 68% confidence contour of 3EG J1420–6038. The other three coincident Galactic sourcesare located within the 99% confidence contours of EGRET sources. The Crab Nebula is notlisted in Table 2 although it has been detected by EGRET (Nolan et al. 1993) as well as byall major VHE γ -ray instruments (Weekes et al. 1989; Atkins et al. 2003; Aharonian et al.2004, 2006f; Albert et al. 2007a). The reason for this is that in the 3EG catalogue only theposition of the Crab pulsar is given, whereas the PUC of the off-pulse emission (i.e. theNebula emission) has not been published thus far.These coincident cases are discussed further in section 3. Table 1 summarises the VHE γ -ray sources located within EGRET confidence contours inside and outside the H.E.S.S. GPSregion. Within the Galactic Plane survey region 17.1 square degrees (corresponding to 3%of the total GPS region) are covered by the EGRET 95%-confidence contours. Randomisingthe distribution of H.E.S.S. sources in the region (flat in longitude; Gaussian shape inlatitude with a mean of –0.2 ◦ and a width of 0.34 ◦ as shown in Aharonian et al. 2006g) theprobability for spatial coincidence between these two populations can be established. For the95% confidence contours ∼ . ∼ . ∼ . ∼ Besides the test for positional coincidence a test of spectral compatibility, based on thesimple assumption of a spectral extrapolation by a single power-law between the EGRETand the H.E.S.S. ranges has been performed. To assess the spectral match the quantity σ comb has been defined in the following way: σ comb = q σ + σ . E . S . S . (1)To determine σ , the spectral index of the EGRET source has been varied (around thepivot point of the EGRET best fit) until the extrapolation to 1 TeV matches the H.E.S.S.flux at that energy. This pivot point of the EGRET best fit is the energy at which the erroron the index becomes independent from the error on the normalisation. This resulting indexis called Γ match and σ = (Γ match − Γ ) / (∆Γ ) (2)(where Γ and ∆Γ is the EGRET index and its error taken from Hartman et al. 1999).Consequently, σ is a quantity that describes by how much the EGRET index has to bealtered (with respect to the error on this index) to match the H.E.S.S. spectrum at 1 TeV.In the same way σ H . E . S . S . , is determined by changing the H.E.S.S. spectral index until theflux matches the EGRET flux at 1 GeV (to avoid biases due to spectral cut-offs at the highend of the H.E.S.S. energy range the spectra were fitted only below 1 TeV in cases with clearspectral curvature). The two quantities σ and σ H . E . S . S . are finally added in quadrature toyield σ comb , describing how well the two spectra can be extended into each other by a linearextrapolation. It should be noted that for the procedure described here, only the statistical(not the systematic) errors on the spectral indices are taken into account. For cases with asource detection only in one band, the same procedure can be applied using the upper limitin the other band (with the obvious difference that only the extrapolation from the source 10 –spectrum onto the upper limit can be performed, not the other way around). For cases inwhich the power-law extrapolation with the nominal source photon index turns out to belower – and therefore non-constraining – to the upper limit the corresponding measure σ or σ H . E . S . S . is set to zero (i.e. the spectra are compatible). In several (but not the majorityof) cases the EGRET spectrum can be preferentially fit by a higher order spectral shape(e.g. an exponential cutoff or a broken power-law) as will be discussed in section 4.
3. VHE γ -ray sources with EGRET counterparts Only a few coincident sources between the GeV and the TeV band have been reported sofar. The VHE γ -ray sources that positionally coincide with EGRET sources are summarisedin Table 1. Whilst the positional coincidences between EGRET and VHE γ -ray sources mightall be chance coincidences as shown in the previous section, in the following all positionalcoincidences within the 99% EGRET PUCs will be considered. Some of the properties ofthe sources and their respective source classes will be discussed along with an investigationon their spectral compatibility as introduced in the previous section. For EGRET sources in the Galactic plane, the only firm identifications with sources atother wavelengths are for pulsars, based on matching radio or X-ray periodicity (Thompson et al.1994). For many of the remaining Galactic EGRET sources, counterparts have been sug-gested, but the angular resolution of the instrument and the strong diffuse γ -ray back-ground in the Galactic plane prevented unambiguous identifications. In VHE γ -rays, severalsource classes have been firmly identified as has been discussed e.g. in Funk (2006), basedon matching morphology, positional coincidence or periodicity. However, the majority ofGalactic VHE γ -ray sources also remain unidentified. Table 2 summarises potential coun-terparts of VHE sources in the coincident cases. While some of these identifications arerather solid (as e.g. in the case of the γ -ray binaries LS 5039 (Aharonian et al. 2006d) andLSI +61 303 (Albert et al. 2006)), in most of the other cases the identification of (even) theVHE γ -ray sources (with relatively small PUC of ∼ ′ ) lack any evidence of associationbeyond positional coincidence. In the cases where a firm identification exists, the VHE γ -raysource can be used to shed light on the nature of GeV source, assuming a physical relation-ship as shown for the Kookaburra region (Reimer & Funk 2007). Such studies demonstratethat observations with VHE γ -ray instruments can provide templates necessary to pin downthe nature of unidentified EGRET γ -ray sources with suggestive but unproven counterparts. 11 –With the upcoming advent of the GLAST-LAT instrument this approach will become veryuseful for associating the GeV emission as measured by a large-aperture space-based γ -rayinstrument with narrow FoV but superior spatial resolution observations of ground-basedVHE γ -ray instruments. Provided that the physical associations discussed in this sectionand shown in Table 2 are confirmed (as e.g., through more sensitive measurements withGLAST-LAT), three long-suspected classes of Galactic GeV sources (SNRs, PWNe and Bi-nary systems) could finally be conclusively established. In the following these different sourceclasses will be briefly discussed in the context of this study.EGRET source VHE γ -ray source Potential CounterpartWithin the H.E.S.S. GPS3EG J1639–4702 HESS J1640–465 G338.3–0.0 (SNR/PWN)3EG J1744–3011 HESS J1745–3033EG J1800–2338 HESS J1801–233 W28 (SNR)3EG J1826–1302 HESS J1825–137 G18.0–0.7 (PWN)3EG J1824–1514 HESS J1826–148 LS 5039 (Binary)Outside the H.E.S.S. GPS3EG J0241+6103 MAGIC J0240+613 LSI +61 303 (Binary)3EG J0617+2238 MAGIC J0616+225 IC443 (SNR/PWN)3EG J0634+0521 HESS J0632+058 Monoceros3EG J1420–6038 HESS J1420–607 Kookaburra (PWN)Table 2: Coincident sources and potential counterparts to the VHE γ -ray sources (and hencealso to the associated EGRET sources). The counterparts are classified into the sourceclasses shell-type SNRs, PWNe and γ -ray binaries. PWNe are currently the most abundant class amongst the identified Galactic VHE γ -ray sources, it is not therefore surprising that PWN are found as potential counterpartsto the coincident sources. The first example for a coincident PWN is HESS J1825–137 –located within the 99% confidence region of 3EG 1826–1302. This source (Aharonian et al.2006b) is currently the best-known example for an offset γ -ray PWN and as such representsa prototype for a new class of γ -ray sources. HESS J1825–137 shows a steepening of theenergy spectrum with increasing distance from the central pulsar. This property, as well asthe observed difference in size between the VHE γ -ray emitting region and the X-ray PWNassociated with the pulsar PSR B1823–13 (Gaensler et al. 2003) can be naturally explained 12 –by different cooling timescales for the radiating electrons of different energies. In this regardit will be important to study this region with the GLAST-LAT in the GeV band to confirm(or refute) this picture. Another example of VHE γ -ray PWN is HESS J1420–607, one of twolikely PWN in the previously discussed Kookaburra region, which is located within the 68%confidence region of 3EG J1420–6038. The Crab Nebula is not listed in Table 2 althoughbeing a prominent GeV and TeV source because no position for the Crab off-pulse (nebular)emission has been published for the GeV data. For previously unidentified sources suchas HESS J1640–465 (G338.3–0.0) an association with the X-ray PWN (Funk et al. 2007) issuggestive but not firmly established at this point. Shell-type SNRs constitute another prominent class of VHE γ -ray sources. However,the two most prominent VHE γ -ray shell-type SNRs RX J1713.7–3946 and RX J0852.0–4622(Vela Jr.) are not prominent GeV emitters even though they are (up to now) the brighteststeady VHE γ -ray sources in the sky after the Crab Nebula. Also Cas A and RCW 86 havebeen reported as VHE γ -ray sources (Aharonian et al. 2001; Albert et al. 2007c; Hoppe et al.2007) but have not been detected by EGRET. Sturner & Dermer (1995); Esposito et al.(1996); Romero, Benaglia, & Torres (1999); Torres et al. (2003) assessed the relationship be-tween unidentified EGRET sources at low Galactic latitude and SNRs and found a statisti-cally significant correlation between the two populations at the 4–5 σ level, were however notable to firmly and uniquely identify individual SNRs as EGRET sources. The GLAST-LATwill shed more light on the GeV emission in this source as well as in the whole population ofGalactic SNRs. By measuring shape and level of the high-energy γ -ray emission the GLAST-LAT might allow for a distinction between hadronic and leptonic emission models as discussedin section 5. Other potential shell-type SNR counterparts related to this analysis are W 28(HESS J1801–233 and 3EG J1800–2338), IC443/MAGIC J0616+225 (Albert et al. 2007b),and the Monoceros Loop SNR (HESS J0632+058 and 3EG J0634+0521) (Aharonian et al.2007e), although in particular in the latter case, the morphology of the VHE γ -ray sourcedoes not lend support to an association with the SNR shell. γ -ray binaries Three binary systems: PSR B1259–63/SS 2883, LS 5039 and LSI +61 303, have nowbeen established as VHE γ -ray sources (Aharonian et al. 2005c, 2006d; Albert et al. 2006;Smith et al. 2007). The latter two of these objects have long been considered as likely coun- 13 –terparts to EGRET sources (Kniffen et al. 1997; Tavani et al. 1998; Hartman et al. 1999;Paredes et al. 2000), however, a definitive identification could not be achieved in the GeVwaveband so far. The VHE γ -ray emission is undoubtedly related to the binary system(as e.g. in LS 5039 established through the detection of characteristic periodicity, matchingthe orbital period of the binary system), strengthening the case that the GeV emission isalso associated to these binaries. Recently, the MAGIC collaboration presented evidencefor VHE γ -ray emission from the black-hole X-ray binary Cyg X-1, during a flaring state inX-rays (Albert et al. 2007d). There is no evidence so far for GeV emission from this object. As described in section 2.2, a test for compatibility between EGRET and H.E.S.S. energyspectra based on a single power-law extrapolation has been performed, calculating for eachof the coincident cases in the H.E.S.S. GPS region a measure of spectral mismatch: σ comb .Figure 4 shows the result of these extrapolations. The values for the spectral compatibilityparameter σ comb are rather small. The largest value, potentially indicative of a spectralmismatch, is found for the case of the γ -ray binary association LS 5039 (3EG J1824–1514 andHESS J1826–148). However, this value is completely dominated by the small statistical erroron the H.E.S.S. power-law fit below 1 TeV (error on the photon index: ∆Γ stat = ± . sys = ± . σ comb can be interpreted as the probability density function for σ comb forrandomly selected H.E.S.S. and EGRET sources and is shown in Figure 5 as a red histogram.This distribution should be compared to the measured distribution of σ comb for positionallycoincident pairs (black histogram). Even though the distribution for the scrambled sourcesshows a tail to large values of σ comb , most random pairings result in values of σ comb <
5. AKolmogorov test yields a probability of 89% that the two distributions are based on a commonunderlying distribution. Thus a spectral compatibility based on a power-law extrapolationof a typical (randomly picked) EGRET and a typical H.E.S.S. source is expected to occur by 14 –
Energy (eV) ) - s - d N / d E ( e r g c m E -13 -12 -11 -10 -9 = 1.4 comb s = 0.9 s Energy (eV) ) - s - d N / d E ( e r g c m E -13 -12 -11 -10 -9 = 3.1 comb s = 2.3 s Energy (eV) ) - s - d N / d E ( e r g c m E -13 -12 -11 -10 -9 = 3.5 comb s = 3.5 s Energy (eV) ) - s - d N / d E ( e r g c m E -13 -12 -11 -10 -9 = 4.3 comb s = 0.6 s Energy (eV) ) - s - d N / d E ( e r g c m E -13 -12 -11 -10 -9 = 0.7 comb s = 0.7 s Fig. 4.— Spectra for the positionally coincident EGRET and H.E.S.S. sources within theH.E.S.S. GPS region. Sources for which the H.E.S.S. source is located within the 95%confidence level are shown in red, whereas those within the 99% confidence contour (as givein Table 1) are shown in orange. The EGRET “butterfly” is determined from the 3EGcatalogue (Hartman et al. 1999), the H.E.S.S. spectral points are taken from the respectivepublication. For HESS J1826–148 and HESS J1825–137, which have significantly curved TeVspectra, only the spectral points below 1 TeV have been fitted and used for the extrapolation.Large values of σ comb indicate mismatches between the spectra at GeV and at TeV energies.chance even in the absence of a physical association. This is perhaps not surprising, giventhat both EGRET as and H.E.S.S. spectra have typical photon indices of ∼ . ∼ Comb s Significances N u m b e r o f e n t r i es Randomly pickedPositionally coincident
Kolmogorov Prob = 0.89
Fig. 5.— Distribution of σ comb . The red histogram shows the distribution for the spectralconsistency parameter σ comb of all possible combinations of EGRET sources with H.E.S.S.sources within the GPS region. The black histogram shows the same distribution for the 5cases of positional coincidences.
4. Inner Galaxy γ -ray sources detected in only one band In this section the remainder (and majority) of sources in the H.E.S.S. GPS region willbe discussed. These are the sources which do not have a counterpart in the neighbouringenergy band. In Section 4.1 EGRET sources without a VHE γ -ray counterpart will bediscussed, section 4.2 investigates VHE γ -ray sources without an EGRET counterpart. γ -ray counterpart Here those EGRET sources are addressed with a 99%-confidence centroid position regionwhich does not contain a reported VHE γ -ray source centroid. This sample consist of 12EGRET detections, with E >
100 MeV fluxes ranging between 0.4 and 3.1 × − cm − s − and photon indices of the power-law fits between ∼ σ upper limits on the VHE flux at 1 TeV for the nominal EGRET positionwere determined. This was done by scaling the H.E.S.S. sensitivity for a 5 σ point sourcedetection (1% of the Crab in 25 h under the assumption of a photon index of 2.6) to theactual exposures as published for the H.E.S.S. GPS region (Aharonian et al. 2006g). As 16 –EGRET H.E.S.S. σ comb Source Upper Limit(10 − ergs cm − s − )3EG J1655-4554 0.4 1.33EG J1710-4439 1.5 16.33EG J1714-3857 0.2 1.53EG J1718-3313 1.0 03EG J1734-3232 0.6 1.43EG J1736-2908 0.3 3.53EG J1746-2851 0.2 15.73EG J1809-2328 0.5 6.43EG J1812-1316 2.1 1.03EG J1823-1314 0.4 03EG J1837-0423 0.6 03EG J1837-0606 0.4 5.5Table 3: EGRET sources without a VHE γ -ray counterpart in the H.E.S.S. GPS region. TheH.E.S.S. differential upper limits (2 σ ) at 1 TeV for a point-source analysis, are derived fromthe H.E.S.S. 2004–2005 exposure at the nominal EGRET position as described in the text(under the assumption of a photon index of 2.6).described previously, the spectral compatibility parameter σ was determined accordingto Equation 2. For cases in which the EGRET extrapolation with the nominal 3EG photonindex undershoots the H.E.S.S. upper limit, σ is set to zero. The resulting plots areshown in Figure 6. For Gaussian errors σ represents the probability that the true GeVspectrum would pass through the HESS upper limit.In seven of the twelve cases the H.E.S.S. upper limit does not impose a strong con-straint on an extrapolation of the EGRET spectrum ( σ < . σ >
5. For these casesthe VHE γ -ray data strongly suggest a spectral turnover (cutoff or break) well below theH.E.S.S. range. Such behaviour is not surprising for some Galactic source classes. For theEGRET-detected pulsars a cutoff in the energy spectrum is seen in many sources within the EGRET energy regime (and therefore clearly well below the VHE range). Indeed, forthree out of the four EGRET sources for which a spectral change is implied by the H.E.S.S. 17 – Energy (eV) ) - s - d N / d E ( e r g c m E -14 -13 -12 -11 -10 -9 = 1.3 s Energy (eV) ) - s - d N / d E ( e r g c m E -14 -13 -12 -11 -10 -9 = 16.3 s Energy (eV) ) - s - d N / d E ( e r g c m E -14 -13 -12 -11 -10 -9 = 1.5 s Energy (eV) ) - s - d N / d E ( e r g c m E -14 -13 -12 -11 -10 -9 = 0 s Energy (eV) ) - s - d N / d E ( e r g c m E -14 -13 -12 -11 -10 -9 = 1.4 s Energy (eV) ) - s - d N / d E ( e r g c m E -14 -13 -12 -11 -10 -9 = 3.5 s Energy (eV) ) - s - d N / d E ( e r g c m E -14 -13 -12 -11 -10 -9 = 15.7 s Energy (eV) ) - s - d N / d E ( e r g c m E -14 -13 -12 -11 -10 -9 = 6.4 s Energy (eV) ) - s - d N / d E ( e r g c m E -14 -13 -12 -11 -10 -9 = 1.0 s Energy (eV) ) - s - d N / d E ( e r g c m E -14 -13 -12 -11 -10 -9 = 0 s Energy (eV) ) - s - d N / d E ( e r g c m E -14 -13 -12 -11 -10 -9 = 0 s Energy (eV) ) - s - d N / d E ( e r g c m E -14 -13 -12 -11 -10 -9 = 5.5 s Fig. 6.— SED of the EGRET source for which no VHE γ -ray source was found within the99% confidence contour. Sources marked with a square show γ -ray emission above 10 GeV inthe EGRET data as reported by Thompson, Bertsch & O’Neal (2005), for sources markedwith a triangle the EGRET data are better described by either a broken power-law or apower-law with an exponential cutoff as shown in Figure 7. 18 –non-detection, a pulsar association has been proposed: 3EG J1710–4439 was unambiguouslyidentified with PSR 1706–44 (Thompson et al. 1994), 3EG J1809–2328 was proposed to beof PWN nature (Braje et al. 2000), and 3EG J1837–0606 was suggested as the counterpartof PSR J1837–0604 (D’Amico et al. 2001). The remaining source in the sample for whichthe spectral extrapolation of the EGRET source is constrained by the H.E.S.S. upper limit,is the Galactic Centre source 3EG J1746–2851. This object is extremely interesting and im-portant, and may be related to the TeV emission detected in this region, however, a properdiscussion falls beyond the scope of this paper.It is interesting to note that an analysis of the EGRET data above 10 GeV (Thompson, Bertsch & O’Neal2005) found eleven EGRET sources with evidence for emission above 10 GeV (at a level ofless than 10% probability that the number of photons seen is a fluctuation of the diffusebackground emission). Five of these sources are located in the H.E.S.S. GPS region. Thesesources are 3EG J1655–4554, 3EG 1710–4439 (PSR B1706–44, with a 6.1 σ detection signifi-cance above 10 GeV) 3EG J1714–3857, 3EG J1746–2851, and 3EG J1837–0606 (all markedwith a white-and-blue square in Figure 6). Interestingly, all of these sources belong to theclass of non-coincident sources, i.e. have no counterpart at VHE γ -ray energies. The char-acteristic cut-off energies of these sources are therefore likely confined to the region below ∼
100 GeV. This emphatically emphasises the existence of cutoffs within the energetic gapleft between the end of the EGRET measurements and the onset of the H.E.S.S. and MAGICobservations.To further investigate the cutoff hypothesis a spectral analysis of the EGRET energyspectra has been performed by means of higher order representations, as has been reportedby Bertsch et al. (2000); Reimer & Bertsch (2001). The EGRET spectra were fitted witha broken power-law and with a power-law with an exponential cutoff: ∂J∂E ( E, K, λ , λ ) = ( K (cid:0) E (cid:1) − λ ( E ≤ K (cid:0) E (cid:1) − λ ( E ≥ ∂J∂E ( E, K, λ, E c ) = K (cid:18) E (cid:19) − λ exp (cid:18) − EE c (cid:19) (4)The χ of the resulting fits were compared to that for a single power-law and an F-testemployed to test if the more complex form was justified. For many γ -ray sources thereis insufficient high-energy data to justify higher order functional fits. However, for fourof the 17 EGRET sources considered in this study the F-test strongly suggests a differentspectral form (with a chance probability < .
05 as discussed in detail in Reimer & Bertsch(2001)): 3EG J1655–4554 is better fit by a power-law with exponential cutoff, 3EG J1710- 19 –4439, 3EG J1736-2908, and 3EG J1746-2851 are best fit with a broken power-law. All ofthese sources have no positional counterpart at TeV energies (and are marked with trianglesin Figure 6). The different spectral representations are shown in red in Figure 7. It isinteresting to note, that out of the four sources mentioned above for which the H.E.S.S. non-detection strongly suggests a cutoff in the energy spectrum, the two sources with the largestincompatibility measure σ are also characterised by a statistically significant cutoff in theEGRET spectrum. In particular, the previously mentioned source 3EG J1746–2851 (GalacticCentre) shows strong indications for an energy break below 10 GeV. The indicated cutoff insome of the EGRET spectra is entirely consistent with the constraining VHE limit based onpower-law extrapolation. The prediction that the other two EGRET sources (3EG J1809–2328, and 3EG J1837–0606) constrained by the H.E.S.S. upper limits show a cutoff in theenergy range between 10 GeV and 100 GeV is therefore well justified and will be tested byupcoming GLAST-LAT observations. Energy (eV) ) - s - d N / d E ( e r g c m E -14 -13 -12 -11 -10 -9 = 1.3 s Energy (eV) ) - s - d N / d E ( e r g c m E -14 -13 -12 -11 -10 -9 = 16.3 s Energy (eV) ) - s - d N / d E ( e r g c m E -14 -13 -12 -11 -10 -9 = 3.5 s Energy (eV) ) - s - d N / d E ( e r g c m E -14 -13 -12 -11 -10 -9 = 15.7 s Fig. 7.— SED at E >
30 MeV for the non-coincident cases in which the EGRET spectrumshows significant deviation from a simple power-law form. The previously reported higherorder spectral representations are shown in red (exponential cutoff for 3EG J1655–4554 andbroken power-law for 3EG J1710–4439, 3EG J1736–2908 and 3EG J1746–2851). 20 – γ -ray sources without an EGRET counterpart In this section the H.E.S.S. sources without a catalogued EGRET counterpart are ad-dressed. At all nominal H.E.S.S. source locations, flux upper limits have been determinedfrom the EGRET data at energies above 1 GeV by means of the EGRET likelihood tech-nique (Mattox et al. 1996). In the determination of the EGRET upper limit, both theGalactic diffuse emission and point-sources exceeding a 5 σ -detection significance thresholdwere modelled and subsequently subtracted. The underlying EGRET exposure correspondsto the first four years of the EGRET mission. As previously discussed, the sensitivity ofEGRET (in terms of energy flux E dN/dE ) is considerably worse than the H.E.S.S. sensi-tivity so that no EGRET detection of a H.E.S.S. source is expected under the assumptionof equal energy flux – which might obviously not necessarily be fulfilled in an astrophysicalsource.Methodologically similar to the previous section, the determination of spectral com-patibility was performed by extrapolating H.E.S.S.-measured VHE spectra to 1 GeV andcomparing the resulting flux to the EGRET upper limit at that energy. The spectral com-patibility parameter σ H . E . S . S . is determined in a similar way to σ . The spectra of H.E.S.S.sources with significant curvature were only fitted from the threshold energy at ∼
100 GeVto 1 TeV. As in previous sections, σ HESS − EGRET describes how well the extrapolated H.E.S.S.spectrum can be accommodated by the EGRET upper limit. The resulting spectral energydistributions (SEDs) of the non-coincident H.E.S.S. sources are shown in Figures 8 and 9.In all cases, the values of σ HESS − EGRET are less than or equal to 1, implying that noEGRET upper limit is violated by the H.E.S.S. extrapolation to 1 GeV, in stark contrast tothe results discussed in the previous section. The most interesting case is that of HESS J1713–395 (RX J1713.7–3946). In this case the power-law extrapolation is at the level of the EGRETupper limit and σ HESS − EGRET = 1. The unconstraining nature of the EGRET upper limitsis simply a consequence of a lack of instrumental sensitivity at GeV energies, worsened inregions of pronounced diffuse γ -ray emission such as the H.E.S.S. GPS region. However,this situation will change significantly in the near future, given the expected sensitivity ofthe GLAST-LAT as also shown in Figures 8 and 9 in which σ HESS − GLAST is calculated for atypical one-year GLAST sensitivity limit in the Inner Galaxy. These numbers suggest thatthe increased sensitivity of the LAT might render common GeV-TeV studies possible. Whilethe EGRET upper limits are currently insensitive to linear extrapolations of the H.E.S.S.spectra, the GLAST-LAT will clearly allow for more sensitive studies. It should, however, benoted, that a linear extrapolation between H.E.S.S. and GLAST-LAT energies most probablyrepresents the “best-case” for any such study: physical models typically show spectra thatharden towards GeV energies, unless a different emission component/process takes over. It 21 –
Energy (eV) ) - s - d N / d E ( e r g c m E -14 -13 -12 -11 -10 -9 HESS J1614-518 = 3.0
HESS-GLAST s = 0.4 HESS-EGRET s Energy (eV) ) - s - d N / d E ( e r g c m E -14 -13 -12 -11 -10 -9 HESS J1616-508 = 10.3
HESS-GLAST s = 0 HESS-EGRET s Energy (eV) ) - s - d N / d E ( e r g c m E -14 -13 -12 -11 -10 -9 HESS J1632-478 = 1.0
HESS-GLAST s = 0 HESS-EGRET s Energy (eV) ) - s - d N / d E ( e r g c m E -14 -13 -12 -11 -10 -9 HESS J1634-472 = 1.5
HESS-GLAST s = 0 HESS-EGRET s Energy (eV) ) - s - d N / d E ( e r g c m E -14 -13 -12 -11 -10 -9 HESS J1702-420 = 2.7
HESS-GLAST s = 0 HESS-EGRET s Energy (eV) ) - s - d N / d E ( e r g c m E -14 -13 -12 -11 -10 -9 HESS J1708-410 = 2.9
HESS-GLAST s = 0 HESS-EGRET s Energy (eV) ) - s - d N / d E ( e r g c m E -14 -13 -12 -11 -10 -9 HESS J1713-381 = 0.3
HESS-GLAST s = 0 HESS-EGRET s Energy (eV) ) - s - d N / d E ( e r g c m E -14 -13 -12 -11 -10 -9 HESS J1713-395 = 17.3
HESS-GLAST s = 1.0 HESS-EGRET s Energy (eV) ) - s - d N / d E ( e r g c m E -14 -13 -12 -11 -10 -9 HESS J1718-385 = 0
HESS-GLAST s = 0 HESS-EGRET s Fig. 8.— (Part 1) SED at E >
30 MeV for the cases in which no EGRET cataloguedcounterpart source was found for the H.E.S.S. source. The dashed arrow shows the predictedupper limit from a one-year GLAST scanning observation, taking into account the galacticdiffuse background. Derived from this is the spectral compatibility parameter σ HESS − GLAST between GLAST and H.E.S.S. assuming a non-detection with GLAST to illustrate thatGLAST will be able to probe the power-law extrapolation from VHE γ -ray energies whereasthe existing EGRET upper limits are unconstraining in this regard. 22 – Energy (eV) ) - s - d N / d E ( e r g c m E -14 -13 -12 -11 -10 -9 HESS J1745-290 = 8.0
HESS-GLAST s = 0 HESS-EGRET s Energy (eV) ) - s - d N / d E ( e r g c m E -14 -13 -12 -11 -10 -9 HESS J1747-281 = 2.5
HESS-GLAST s = 0 HESS-EGRET s Energy (eV) ) - s - d N / d E ( e r g c m E -14 -13 -12 -11 -10 -9 HESS J1800-240 = 3.0
HESS-GLAST s = 0.3 HESS-EGRET s Energy (eV) ) - s - d N / d E ( e r g c m E -14 -13 -12 -11 -10 -9 HESS J1804-216 = 5.0
HESS-GLAST s = 0.9 HESS-EGRET s Energy (eV) ) - s - d N / d E ( e r g c m E -14 -13 -12 -11 -10 -9 HESS J1809-193 = 4.6
HESS-GLAST s = 0 HESS-EGRET s Energy (eV) ) - s - d N / d E ( e r g c m E -14 -13 -12 -11 -10 -9 HESS J1813-178 = 2.0
HESS-GLAST s = 0 HESS-EGRET s Energy (eV) ) - s - d N / d E ( e r g c m E -14 -13 -12 -11 -10 -9 HESS J1834-087 = 3.0
HESS-GLAST s = 0 HESS-EGRET s Energy (eV) ) - s - d N / d E ( e r g c m E -14 -13 -12 -11 -10 -9 HESS J1837-069 = 7.8
HESS-GLAST s = 0 HESS-EGRET s Fig. 9.— (Part 2) SED at E >
30 MeV for the cases in which no EGRET cataloguedcounterpart source was found for the H.E.S.S. sources. The dashed arrow shows the predictedupper limit from a one-year GLAST scanning observation, taking into account the diffuseemission. Derived from this is the spectral mismatch between GLAST and H.E.S.S. assuminga non-detection with GLAST to illustrate the GLAST will be able to probe the power-lawextrapolation from VHE γ -ray energies whereas the EGRET upper limits are unconstrainingin this regard. 23 –remains to be seen if GLAST will detect emission at comparable energy flux and potentiallydetermine the position of the peak in the SEDs. As discussed previously, the tremendousadvantage of the GLAST-LAT over any previous mission is the continuous energy coveragefrom 30 MeV all the way up into the VHE γ -ray range at ∼
300 GeV with significantlyimproved sensitivity and angular resolution, bridging the current energy gap in which someof the physically interesting suggested energy cutoffs occur.
5. Interpretation5.1. Sources detected both at GeV and TeV energies
As previously stated and shown in Table 2, only 9 sources exist which can be charac-terised as coincident Galactic EGRET and VHE γ -ray sources at this moment (5 withinthe inner Galaxy, 4 outside of the H.E.S.S. GPS region). Given the large number of Galac-tic sources in both GeV and TeV γ -rays this number is rather small – and is indicative ofdifferent dominant source classes in these two energy domains. However, for the few caseswhere a positional coincidence may exist some important astrophysical implications as wellas predictions for the upcoming GLAST mission can be drawn.Whilst EGRET and in particular GLAST have sufficiently large FoVs to be able toefficiently observe the whole sky, the limited FoV of imaging VHE γ -ray instruments (typ-ically 4 ◦ diameter) allow for only limited sky coverage. However, for known GeV sourceshigh-angular resolution VHE instruments such as MAGIC and H.E.S.S. with tremendouslyhigher photon statistics at high energies can help in the identification and interpretation ofthe GeV emission. This approach has been followed by Reimer & Funk (2007) for the Kook-aburra complex. In this region of TeV and GeV γ -ray emission, a re-analysis of the EGRETdata taking advantage of the higher spatial resolution images from H.E.S.S. observations,demonstrated that the dominant GeV emission (3EG J1420–6038) is positionally coincidentwith HESS J1420–607 (Aharonian et al. 2006c). This EGRET source has been flagged asconfused in the 3EG catalogue (Hartman et al. 1999) and in the re-analysis 3EG J1420–6038 was found to be partially overlapping with a less intense second GeV γ -ray source.This second GeV source – detected below the nominal detection threshold for EGRET –is apparent in a dedicated analysis at approximately 1/3 of the GeV flux of the dominantsource (Reimer & Funk 2007) and is positionally coincident with the second VHE γ -raysource in the Kookaburra region, HESS J1418–609 (Aharonian et al. 2006c) (associated withthe “Rabbit” PWN). This suggestive morphology match between the GeV data and theH.E.S.S. data thus helped in the interpretation and identification of the confused EGRETsources and made a separation into two individual sources possible. Studies such as this one 24 –show how confused GeV emission regions (in particular in the Galactic plane where the dif-fuse γ -ray background is dominant) may be unravelled using the GeV emission as measuredfrom a large-aperture space-based γ -ray instrument together with narrow FoV but superiorspatial resolution observations provided by ground-based atmospheric Cherenkov telescopes.This approach seems promising for achieving convincing individual source identifications inthe era of the GLAST-LAT. Energy (eV) ) - s - d N / d E ( e r g c m E -13 -12 -11 -10 -9 HESS J1640-465 / 3EG J1639-4702 yr G, Age: 10 m B = 10 yr G, Age: 10 m B = 100 yr G, Age: 10 m B = 10
Fig. 10.— SED for the coincident source HESS J1640-465 along with leptonic IC-modelsfor different magnetic fields and different ages of the system. The purpose of this figureis to demonstrate that rather extreme values for the magnetic-field ( ∼ µ G) or the ageof the system ( ∼ years) have to be invoked to fit such a spectral energy distributionin a leptonic model. These models numerically take into account the time-evolution of theelectron spectrum considering energy losses and injection of electrons in time-steps muchshorter than the age of the system. Synchrotron and IC losses are calculated following theformalism in Blumenthal & Gould (1970). The injection spectrum for the electrons waschosen to have a photon index of 2.5, the Inverse Compton scattering was performed on theCMB only. It should be noted, that the X-ray flux between 2 and 10 keV detected fromthis source is at the level of 10 − erg cm − s − as determined by Funk et al. (2007). Thisrather low X-ray flux renders a connection between the H.E.S.S. and the EGRET source inany leptonic model extremely difficult as demonstrated by this figure.On the other hand, the detection of VHE γ -ray sources with EGRET (or the GLAST-LAT) may help in the interpretation of the TeV data and the modelling of the γ -ray emission 25 –mechanism. Measuring the energy spectrum of a high-energy γ -ray source over 5–6 decadesin energy should provide rather stringent constraints on the γ -ray emission mechanism. TheCrab is in this respect the only good example of a Galactic source for which both an excellentGeV and TeV coverage exists which in turn helped to understand the emission mechanismand the magnetic field strength rather well in comparison to most other γ -ray sources. Withthe advent of the GLAST-LAT many more such sources with a good GeV and TeV coveragecan be expected.Figure 10 shows the SED for the positionally coincident sources 3EG J1639–4702 andHESS J1640–465 (Aharonian et al. 2006g; Funk et al. 2007). The figure shows a rather typi-cal γ -ray SED for a positionally coincident sources (see Figure 4) with a power-law spectrumat TeV energies with photon index 2 . ± .
15 and a similar power-law at GeV energies withphoton index 2 . ± .
18, at an energy flux level an order of magnitude higher than thatat 1 TeV. The EGRET source 3EG J1639–4702 is rather close to the detection significancethreshold (with a TS / -value of 6.4). Taking this SED as representative, several scenar-ios for a common origin of the γ -ray emission are considered. For a hadronic model theshape of this SED can be rather easily fitted, requiring a power-law distribution of primaryhadrons with dN/dE ∝ E − α , with α ≈ . γ -ray emission interpreted as inverse-Compton up-scattering of soft photon fields, matching the shape of the SED requires ratherextreme values for the magnetic field (given the low level of X-ray synchrotron emission fromthis system as reported by Funk et al. 2007) or for the age of the system (given the needto confine the accelerated electrons within the system). This is demonstrated in Figure 10which shows 3 leptonic model curves. In the generation of these models, the time-evolutionof the electron spectrum due to energy losses was taken into account. These energy losseswere calculated according to the formalism described in Blumenthal & Gould (1970). Forhigh energy electrons the energy-loss (cooling) timescale E/ ( dE/dt ) is proportional to 1 /E for losses predominantly via synchrotron radiation or IC in the Thomson regime. In thiscase, for continuous injection of electrons with a power law spectrum dN/dE ∝ E − α , aspectral break to E − ( α +1) will occur. The slope of the IC spectrum (again in the Thomsonregime) is given by Γ = ( α + 1) /
2. In the idealised case of the Thomson cross-section and asingle (thermal) target radiation field the break energy is given approximately by: E break ≈ . t source / yr) − (( U rad + B / π ) / − ) − (T / . µ G andage of 10 years. This curve provides an adequate description of the H.E.S.S. data, but not 26 –the EGRET data due to the characteristic turnover of the γ -ray spectrum at lower energies.The other two curves (solid green and dash-dotted red) are shown to illustrate how the SEDcould be accommodated in a leptonic model and thus how the peak of the IC emission canbe pushed into the EGRET range. Taking a typical Galactic radiation field (which mightnot be realistic as e.g. in binary system with a massive stellar component) either rather highmagnetic fields (green solid) or rather old sources have to be invoked (dash-dotted red). Thehigh-magnetic field scenario would, however, lead to the prediction of a very high X-ray flux.This prediction contradicts the faint X-ray emission detected from this object (at the levelof 10 − erg cm − s − ) as well as in most other Galactic VHE γ -ray sources (where the X-rayemission is typically at the same level or below the VHE γ -ray energy flux). To explain the γ -ray emission of coincident sources through leptonic IC emission, the sources should thusbe rather old to be able to accumulate enough low energy electrons to explain the high GeVflux in a typical Galactic radiation field. They should then, however, either be rather brightX-ray emitters or be very old.VHE γ -ray sources may be detectable using GLAST even if the γ -ray emission is gen-erated by IC scattering on a typical Galactic radiation field, as demonstrated for the SNRRX J1713.7–3946 where a GLAST detection should shed light on the heavily debated originof the TeV emission (Funk et al. 2007b). γ -rays of leptonic origin (produced by IC) might bedistinguishable from those of hadronic origin (produced by π -decay) through their charac-teristic spectral shape, although recent claims have been made that under certain conditionsthe leptonic γ -ray spectra might resemble those of pionic decays (Ellison et al. 2007). Fig-ure 11 shows that the GLAST-LAT will have the sensitivity to measure energy spectra (in5 years of scanning observations) for both hadronic and leptonic emission scenarios, illus-trating that the LAT energy range is particularly well suited to distinguish these models.Measuring the spectral shape of the γ -ray emission through deep GeV observations withthe GLAST-LAT will play an important role in interpreting the currently known TeV γ -raysources. For sources where no positional coincidence has been found for the GeV and TeV do-mains both instrumental and astrophysical explanations can be invoked. 27 –
Energy (eV) ) - s - d N / d E ( e V c m E -1 Inverse Compton - decay p GLAST - hadronic 5 yearsGLAST - leptonic 5 yearsH.E.S.S.
EGRET
Fig. 11.— High-energy SED for the SNR RX J1713.7–3946. The black data points showmeasurements with H.E.S.S., whereas the blue circles and red triangles show simulatedGLAST data, assuming two different models (leptonic and hadronic) for the γ -ray emission(shown as dashed red and solid blue lines). This simulation uses the current best estimateof the LAT performance and illustrate that in principle the GLAST-LAT should be able todetect this prominent shell-type SNR in a 5-years observation or faster, depending on theemission mechanism. This figure has been reproduced from (Funk et al. 2007b). The most obvious reason for a non-detection of a TeV source with EGRET is the sen-sitivity mismatch. In a typical ∼ ∼ −
80 lower than that of EGRET for its entire lifetime (above1 GeV in the Galactic Plane). Additionally, with decreasing detection significance an in-creasing number of EGRET sources are expected to be artificial due to source confusion inthe Galactic plane and in particular due to uncertainties from the model chosen to describethe dominant diffuse γ -ray emission. The GLAST-LAT will inevitably shed more light onall persistent EGRET sources, since these will be rather bright γ -ray sources for the LAT 28 –instrument. However, it should be noted that the brightest Galactic H.E.S.S. sources (suchas RX J1713.7–3946) are not going to be very bright GLAST sources as discussed in theprevious section. Certainly, similar to EGRET, the LAT will (at the lower end of the energyrange) suffer from uncertainties and systematic effects due to intrinsic properties of the ex-perimental approach and in particular due to the modelling of the diffuse γ -ray background,however, at a lower flux level.Another instrumental effect that could render a correlation between GeV and TeVsources unlikely, is the insensitivity of imaging VHE γ -ray instruments to very extendedsources (radius > ◦ ) without significant sub-structure. The EGRET data do not put strongconstraints on the source extension of a typical source in the Galactic plane. Source exten-sions that can be derived from the data are on the scale of the EGRET PSF, i.e. degreescales. The angular resolution (and thus the maximum sensitivity) of VHE γ -ray instrumentson the other hand is of the order of a few arc minutes. The upper limits for H.E.S.S. atthe positions of EGRET sources quoted in this study are derived under the assumption of apoint-like source (with a typical size of the source region of less than ∼ . ◦ rms width). Thesensitivity and thus the upper limit scales roughly linearly with the source size (Funk 2005)and for source sizes in excess of ∼ ◦ , the H.E.S.S. data become completely unconstrain-ing due to the fact that the source size becomes comparable with the size of the FoV andno reliable background estimation can be performed (see Berge, Funk & Hinton 2007, for adescription of the background estimation techniques used). Large-FoV instruments (withpoorer angular resolution) such as Milagro (Atkins et al. 2002), are better suited to detectsources with intrinsically large sizes in VHE γ -rays (with sufficiently high fluxes). However,due to their modest ( ∼ ◦ ) angular resolution, such instruments suffer from problems ofsource confusion similar to those of current GeV measurements. Indeed, several of the re-cently reported Milagro source candidates are coincident with EGRET sources (Abdo et al.2007). Hypothesising that EGRET sources exhibit angular sizes larger than ∼ ◦ , Milagro-type instruments might be better suited to detect large scale emission at VHE γ -ray energies.Again, the GLAST-LAT, with its superior angular resolution to EGRET, will shed more lighton the issue of the intrinsic sizes of GeV sources in the Galactic plane. The constraints onthe power-law extrapolation of EGRET sources by sensitive H.E.S.S. upper limits as de-rived in the previous sections are naturally only valid under the assumption that the VHEcounterpart to the EGRET emission does not exhibit a size much larger than ∼ ◦ . 29 – The non-detection of most TeV sources in the GeV range by EGRET may be due simplyto a lack of instrumental sensitivity. On the other hand, the lack of TeV counterparts tomost bright GeV sources requires the presence of steepening (or cut-offs) between 10 and100 GeV in the spectra of these sources (see section 4 and Figure 7). Steepening in γ -rayenergy spectra between 10 and 100 GeV can occur for many reasons, the most prominent ofwhich are discussed briefly below. Acceleration limits . The maximum energy to which particles are accelerated in a sourcemay be determined by a balance between the acceleration and energy loss timescales, orbetween acceleration and escape timescales, or simply by the lifetime of the source. In thelimit of Bohm diffusion, the escape time of accelerated particles from the source can bewritten as t escape ∼ ( r source / pc) D ( E/ TeV) − ∆ (6)The associated cut-off in the resulting γ -ray emission may occur at much lower energies, asin the case of proton-proton interactions (a factor ∼
20 as shown in Kappes et al. 2007),or close to the primary particle energy, as in the case of inverse Compton scattering in theKlein-Nishina limit (Blumenthal & Gould 1970).
Particle transport may impact on the spectral shape in several ways. For protons de-scribed by a power-law J p ( E p ) = KE − Γ p the γ -rays produced in hadronic interactions areexpected to follow a similar power-law spectrum F γ ( E γ ) ∝ E − Γ γ . Generally, high energyparticles escape more easily leading to a cut-off in the particle and hence γ -ray spectruminside the source. Therefore, due to particle transport, the spectrum of the protons gener-ating the γ -rays through hadronic interactions is not necessarily the same as the one at theacceleration site. In the case of diffusion the proton spectrum at the γ -ray production sitecan instead be written as J p ( E p , r, t ) = c π f, where f ( E p , r, t ) is the distribution functionof protons at an instant t and distance r from the source. The distribution function satisfiesthe diffusion equation (Ginzburg & Syrovatskii 1964). ∂f∂t = D ( E p ) r ∂∂r r ∂f∂r + ∂∂E p ( P f ) + Q, (7)where P = − dE p /dt is the continuous energy loss rate of the particles, Q = Q ( E p , r, t ) is thesource function, and D ( E p ) is the diffusion coefficient. Atoyan, Aharonian & V¨olk (1995) de-rived a general solution for Equation (7). Hence, as has been emphasised by Aharonian & Atoyan(1996), the observed γ -ray flux can have a significantly different spectrum from that expectedfrom the particle population at the source. In the (expected) case of energy-dependent diffu-sion ( D ∝ E − ∆ , with ∆ typically assumed to lie in the range ∼ . − .
0) the γ -ray spectrum 30 –will follow F γ ( E γ ) ∝ E − (Γ+∆) γ . The exact shape of the spectrum will depend on the age ofthe accelerator, duration of injection, the diffusion coefficient, and the location of the targetmaterial.The influence of convection (lower energy cutoff in primary particle spectrum) is typ-ically stronger for low energy (GeV) γ -rays potentially resulting in a VHE γ -ray sourcethat has no EGRET counterpart in cases in which an external accelerator produces primaryhadrons near an active target.Torres, Domingo-Santamaria & Romero (2004) and Domingo-Santamaria & Torres(2006) have recently studied collective wind configurations produced by a number of massivestars, and obtained densities and expansion velocities of the stellar wind gas that is thetarget for hadronic interactions in several examples, showing that these may be sources forGLAST and the TeV instruments in non-uniform ways, i.e., with or without the correspond-ing counterparts in the other energy band. Particle energy losses away from the acceleration site may also produce spectral steep-ening in a very natural way as discussed earlier (see section 5.1). In the case where particleinjection is effectively finished (i.e. the injection rate is much lower than in the past), radia-tive energy losses may produce a rather sharp cut-off in the γ -ray spectrum as e.g. shown in(Funk et al. 2007). For high energy electrons the energy-loss (cooling) timescale E/ ( dE/dt )is proportional to 1 /E for losses dominantly via synchrotron radiation or IC in the Thom-son regime. In this case, for continuous injection of electrons with a power law spectrum dN/dE ∝ E − α , a spectral break to E − ( α +1) will occur. The slope of the IC spectrum (againin the Thomson regime) is given by Γ = ( α +1) /
2. In the idealised case of the Thomson cross-section and a single (thermal) target radiation field the break energy is given approximatelyby: E break ≈ . t source / yr) − (( U rad + B / π ) / − ) − (T / . γ - γ pair-production occurs above a threshold ǫ γ ǫ target > m e c . For stellar systems with ǫ target ∼ ∼
500 GeV. Pairs produced in γ - γ interactionsmay inverse Compton scatter on the same radiation field – leading to the developmentof a cascade (Protheroe, Mastichiadis & Dermer 1992). Attenuation on the interstellar IRand CMB can be neglected below 10 TeV so γ - γ ’cut-offs’ are only expected in compactregions of very high radiation density, for example within binary stellar systems. Theseabsorption/cascade ’features’ may not represent the end of the γ -ray spectrum as emissionmay recover at energies above the resonance. 31 – As demonstrated by Figures 8 and 9 the GLAST-LAT should be able to detect severalof the VHE γ -ray sources in the inner Galaxy, assuming a simple power-law extrapolationof the spectrum from TeV to GeV energies. However, this power-law assumption may notbe valid in several cases, as discussed in the following for the known TeV source classes. Pulsar Wind Nebulae are currently the most abundant VHE γ -ray sources in the Galac-tic plane. The most prominent example is the Crab Nebula (Weekes et al. 1989; Aharonian et al.2004; Atkins et al. 2003; Aharonian et al. 2006f; Albert et al. 2007a). The SED expected ofPWNe does not typically result in significant GeV fluxes: the Crab Nebula, detected through-out both energy bands, seems to be an exceptional case due to its very strong magnetic fieldand relative proximity (2 kpc). Most VHE γ -ray PWNe are expected to be dominated byIC emission for which the energy flux generally turns down at lower energies. The positionof this inverse Compton peak determines detectability for both GeV and TeV instruments.Also the size and flux of the source also obviously affect the detectability with GLAST-LAT.In general the higher the energy of the inverse-Compton peak in these sources, the lower thechance will be to detect them with GLAST. If a large fraction of the GeV emission attributedto EGRET Galactic unidentified sources is related to pulsed magnetospheric emission frompulsars as opposed to emission from the extended wind nebula then a correlation betweenthe H.E.S.S. and EGRET sources in the Inner Galaxy could be expected, given that themajority of the H.E.S.S. sources in this region seem to be PWNe associated with energeticpulsars (Carrigan et al. 2007). However, this expectation may not hold in general due todiversity of parameters like the beaming geometry or different conversion efficiency of thepulsar’s spin-down power into the Nebula and into γ -rays. Shell-type Supernova remnants
The two prominent and bright VHE γ -ray SNRs (RX J1713.7–3946 and RX J0852.0–4622) are not expected to be very bright GLAST-LAT sources. Nev-ertheless, they are probably amongst the more easily detectable TeV sources in the GLAST-LAT band. A detailed simulation of the expected signal from RX J1713.7–3946 shows thatit might be detectable in one year of GLAST-LAT observations depending on the assumedTeV γ -ray emission mechanism as shown in the previous section. Morphological studies inGeV γ -rays will either have to struggle with moderate angular resolution at low energies orwith low photon statistic at high energies. However, spectral studies will be immediatelypossible following a potential detection as shown in Figure 11 for RX J1713.7–3946. ForRX J0852.0–4622 (Vela Junior) the situation is even further complicated by the close-bybright Vela Pulsar. While both of these prominent TeV-emitting objects are rather young( ∼ γ -ray instruments, namely older SNRs which have accumulateda large number of lower energy CRs, but for which the higher energy CRs (those that maygive rise to the TeV emission) have already left the acceleration site. A common detectionboth with GLAST and VHE γ -ray instruments might require a hadronic origin of the γ -rayemission rather than an Inverse compton (IC) origin due to the characteristic turn-over ofthe IC spectrum at lower energies. Gamma-ray Binary systems host a variety of non-thermal phenomena. The TeV de-tected binaries: LS 5039 (Aharonian et al. 2006d), PSR B1259-63 (Aharonian et al. 2005c),LSI +61 303 (Albert et al. 2006) and Cyg X–1 (Albert et al. 2007d) are currently seen ascandidates for detection at GeV energies. γ - γ absorption in binary systems may pro-ducing anti-correlation of the TeV to GeV radiation during the orbit of these systems.These orbital modulations are predicted in basically all models for these systems, irre-spective of the assumptions of a pulsar or a black hole compact object or the processby which high-energy radiation is emitted (see e.g. Dermer & B¨ottcher 2007; Dubus 2006;Paredes, Bosch-Ramon & Romero 2006). Details in predicted light-curves and spectral evo-lution in time are however rather distinctive (Khangulyan, Aharonian & Bosch-Ramon 2007;Sierpowska-Bartosik & Torres 2007).
6. Summary
The main results of the study of the relationship between GeV and TeV sources are:1. There are rather few spatially coincident GeV-TeV sources for the considered Galacticregion.2. Those few positional coincident GeV-TeV sources could occur by chance, the chanceprobability of detecting two coincident sources within the H.E.S.S. GPS region is ∼ γ -ray emission mechanism in these Galactic particle accelerators.Summarising, the study presented here shows that the GLAST-LAT will tremendouslyadvance the study of the relationship between GeV and TeV sources by improving thesensitivity over EGRET by an order of magnitude and in particular by bridging the currentlyuncovered energy range between 10 GeV and 100 GeV.The authors would like to acknowledge the support of their host institutions. In particu-lar, S.F. acknowledges support of the Department of Energy (DOE) and Stanford Universityand would like to thank the whole H.E.S.S. and the GLAST-LAT collaborations for theirsupport and helpful discussions on the topic, in particular Werner Hofmann, Felix Aha-ronian, Benoit Lott, and Seth Digel. DFT is supported by the Spanish MEC grant AYA2006-00530, he acknowledges Juan Cortina and other members of the MAGIC collaborationfor advice and encouragement. OR is supported by the National Aeronautics and SpaceAdministration under contract NAS5-00147 with Stanford University. JAH is supported byan STFC Advanced Fellowship. REFERENCES
Abdo, A., et al., 2007, ApJ 664, 91Aharonian F.A., & Atoyan, A.M. 1996, A&A 309, 91Aharonian, F. A., et al., 2001, A&A 370, 112Aharonian, F. A., et al. (
HEGRA Collaboration ), 2001, ApJ 614, 897Aharonian, F. A., et al. (
H.E.S.S. Collaboration ), 2005a, A&A 435, L17 34 –Aharonian, F. A., et al. (
H.E.S.S. Collaboration ), 2005b, Science 307, 1938Aharonian, F. A., et al. (
H.E.S.S. Collaboration ), 2005c, A&A 442, 1AAharonian, F. A., et al. (
H.E.S.S. Collaboration ), 2006a, A&A 449, 223Aharonian, F. A., et al. (
H.E.S.S. Collaboration ), 2006b, A&A 460, 365Aharonian, F. A., et al. (
H.E.S.S. Collaboration ), 2006c, A&A 456, 245Aharonian, F. A., et al. (
H.E.S.S. Collaboration ), 2006d, A&A 460, 743Aharonian, F. A., et al. (
H.E.S.S. Collaboration ), 2006e, Science 312, 1771Aharonian, F. A., et al. (
H.E.S.S. Collaboration ), 2006f, A&A 457, 899Aharonian, F. A., et al. (
H.E.S.S. Collaboration ), 2006g, ApJ 636, 777Aharonian, F. A., et al. (
H.E.S.S. Collaboration ), 2007a, A&A 464, 235Aharonian, F. A., et al. (
H.E.S.S. Collaboration ), 2007b, ApJ 661, 236Aharonian, F. A., et al. (
H.E.S.S. Collaboration ), 2007c, A&A 467, 1075Aharonian, F. A., et al., (
H.E.S.S. Collaboration ), 2007d, A&A 466, 543Aharonian, F. A., et al., (
H.E.S.S. Collaboration ), 2007e, A&A 469, L1Albert, J., et al. (
MAGIC Collaboration ), 2006, Science 312, 1771Albert, J., et al. (
MAGIC Collaboration ), 2007a, submitted to ApJ (astro-ph/0705.3244)Albert, J., et al. (
MAGIC Collaboration ), 2007b, ApJ 664, 87Albert, J., et al. (
MAGIC Collaboration ), 2007c, A&A in press (astro-ph/0706.4065)Albert, J., et al. (
MAGIC Collaboration