Gamma-ray emission from the globular clusters Liller 1, M80, NGC 6139, NGC 6541, NGC 6624, and NGC 6752
P. H. T. Tam, A. K. H. Kong, C. Y. Hui, K. S. Cheng, C. Li, T.-N. Lu
aa r X i v : . [ a s t r o - ph . H E ] J a n Gamma-ray emission from the globular clusters Liller 1, M80,NGC 6139, NGC 6541, NGC 6624, and NGC 6752
P. H. T. Tam , A. K. H. Kong , , C. Y. Hui , K. S. Cheng , C. Li , and T.-N. Lu Institute of Astronomy and Department of Physics, National Tsing Hua University,Hsinchu, Taiwan Department of Astronomy and Space Science, Chungnam National University, Daejeon,South Korea Department of Physics, University of Hong Kong, Pokfulam Road, Hong Kong Golden Jade Fellow of Kenda Foundation, Taiwan [email protected]
ABSTRACT
Globular clusters (GCs) are emerging as a new class of γ -ray emitters, thanksto the data obtained from the Fermi Gamma-ray Space Telescope. By now,eight GCs are known to emit γ -rays at energies >
100 MeV. Based on the stellarencounter rate of the GCs, we identify potential γ -ray emitting GCs out of allknown GCs that have not been studied in details before. In this paper, we reportthe discovery of a number of new γ -ray GCs: Liller 1, NGC 6624, and NGC 6752,and evidence for γ -ray emission from M80, NGC 6139, and NGC 6541, in which γ -rays were found within the GC tidal radius. With one of the highest metallicityamong all GCs in the Milky Way, the γ -ray luminosity of Liller 1 is found to bethe highest of all known γ -ray GCs. In addition, we confirm a previous reportof significant γ -ray emitting region next to NGC 6441. We briefly discuss theobserved offset of γ -rays from some GC cores. The increasing number of known γ -ray GCs at distances out to ∼
10 kpc is important for us to understand the γ -ray emitting mechanism and provides an alternative probe to the underlyingmillisecond pulsar populations of the GCs. Subject headings:
Gamma rays: stars — Globular clusters: individual (Liller 1,M80, NGC 6139, NGC 6441, NGC 6541, NGC 6624, and NGC 6752) — Pulsars:general 2 –
1. Introduction
Globular clusters (GCs) are very efficient in forming close binary systems and their dece-dents such as low-mass X-ray binaries (LMXBs), cataclysmic variables, and millisecond pul-sars (MSPs). It is commonly believed that the high stellar encounter rate of GCs facilitatesthe formation of these binary systems through dynamical interactions (e.g., Verbunt & Hut1987). Pooley et al. (2003) and Pooley & Hut (2006) have shown by X-ray observations thatthe number of close binary systems in a GC scales with the stellar encounter rate (Γ c ) ofthe GC. Recently, Hui et al. (2010) found that a cluster with higher Γ c and metallicity hostsmore MSPs.Radio and X-ray observations have revealed about 140 MSPs in 26 globular clus-ters (Manchester et al. 2005). However, the presence of much stronger X-ray emitters cancontaminate the X-ray observations of MSPs. Abdo et al. (2009b) reported a number ofMSPs in the Galactic field using data obtained using the Fermi Gamma-ray telescope. Be-cause MSPs are the only known steady γ -ray sources in GCs, γ -ray observations serve as analternative channel in studying the underlying MSP populations in GCs.The Large Area Telescope (LAT), onboard the Fermi Gamma-ray Space Telescope , isa pair-production telescope designed to detect γ -rays with energies between ∼
20 MeV and >
300 GeV. It operates in a survey mode in which it scans the whole sky every 3 hours. The68% containment radius of individual photons is 0 . ◦ ∼ − ph cm − s − above 100 MeV in one year of survey-mode observations (Atwood et al. 2009). Such sen-sitivity has enabled the discovery of γ -rays from 8 GCs (Abdo et al. 2010b,c), including47 Tucanae (Abdo et al. 2009a) and Terzan 5 (Kong et al. 2010). From the known γ -rayluminosity of individual MSPs in the Galactic field (Abdo et al. 2009b), they would not bedetected at distances of several kpc. It is therefore believed that the γ -rays from GCs donot come from individual MSP, but from a population of MSPs.The radiation mechanism of γ -rays remains unclear. In the pulsar magnetospheremodel (e.g., Venter & de Jager 2008), γ -rays up to a few GeV is radiated from the MSPsthrough curvature radiation. On the other hand, inverse Compton (IC) processes resultedfrom energetic particles up-scattering low-energy photons, such as starlight and infraredlight, may give rise to γ -rays of MeV to TeV energies (Bednarek & Sitarek 2007; Cheng et al.2010). It is worth noting that γ -ray observations at >
100 GeV have not yet resulted in anydetections (Anderhub et al. 2009; Aharonian et al. 2009). γ -rays from those GCs with high encounter rate. We reportthe discovery of γ -rays from the GCs Liller 1, M80, NGC 6139, NGC 6541, and NGC 6752.
2. Cluster sample and Fermi Observations
It has long been suggested that the stellar encounter rate of a GC is a measure of thenumber of LMXBs, the precedents of MSPs (see, e.g. Verbunt & Hut 1987; Gendre et al.2003). Assuming that each MSP emits similar amount of γ -rays, the expected γ -ray lumi-nosity of a GC is proportional to the number of MSPs it hosts, and thus to the encounterrate Γ c , as supported by the findings made by Abdo et al. (2010c). Here we estimate theencounter rate by Γ c ∝ ρ r /σ where ρ is the central luminosity density, r c the core radius,and σ the velocity dispersion at the cluster center. We adopted the values of σ as presentedin Gnedin et al. (2002), and those of ρ and r c in Harris (1996, 2003 version). Then weranked the GCs according to the encounter rate divided by the squared distance compiledin Harris (1996, 2003 version), giving a measure of the relative expected γ -ray flux for eachGC. Most of the eight known γ -ray GCs (47 Tucanae, ω Centauri, M62, NGC 6388, Terzan 5,NGC 6440, M28, and NGC 6652) and three source candidates (NGC 6541, NGC 6752, andM15) are ranked high, i.e. at the top 23 in the ranking. Only NGC 6652 gave a relativelylow expected γ -ray flux. Among the top 20, we identify the following GCs that have notbeen reported as γ -ray source or source candidate before (in descending order of expected γ -ray flux): Liller 1, M22, NGC 2808, NGC 362, NGC 6540, NGC 1851, Terzan 6, M80, andNGC 6397. On the other hand, nearby γ -ray emission has been reported for NGC 6441 andNGC 6624 (Abdo et al. 2010c). In this work, we searched for γ -rays from the GCs in theabove list, as well as looked deeper into the cases of NGC 6441, NGC 6541, NGC 6624, andNGC 6752. NGC 6139 (ranked 30th) was also studied.The γ -ray data used in this work were obtained between 2008 August 4 and 2010August 21. We used the Fermi Science Tools v9r15p2 package to reduce and analyze thedata provided by the Fermi Science Support Center . Only those data that passed themost stringent photon selection criteria (i.e. the “diffuse” class) were used. To reduce thecontamination from Earth albedo γ -rays, we excluded events with zenith angles greater than105 ◦ . The instrument response functions (IRFs) “P6 V3 DIFFUSE” were used. http://fermi.gsfc.nasa.gov/ssc/data/analysis/scitools/
3. data analysis and results
Photons with energies between E min and E max that come from a circular region-of-interest (ROI) around the GCs were included in the analysis. The ROI, as well as E min and E max , for those GC coincident with γ -ray emission are given in Table 1. We sub-tracted the background contribution from all sources in the first Fermi/LAT catalog (1FGL;Abdo et al. 2010b) within the circular region of 15 ◦ radius around the GC position, aswell as diffuse emission by including Galactic diffuse model (gll iem v02.fit) and isotropicbackground (isotropic iem v02.txt). We assumed single power laws for all first Fermi/LATcatalog sources considered, except for γ -ray pulsars of which the spectra follow power lawswith exponential cut-off (Abdo et al. 2010a,c).We created the “Test-statistic (TS; Mattox et al. 1996) maps” in the neighboring regionof the clusters using the tool gttsmap . These TS maps are created by moving a putativepoint source through a grid of locations on the sky and maximizing − log(likelihood) at eachgrid point, while stronger and presumably well-identified sources are included in each fit.Weaker sources would be identified in these maps. Such TS maps around Liller 1, M80,NGC 6139, NGC 6441, NGC 6541, NGC 6624, and NGC 6752 are shown in Figs. 1 to 4centered on the best-fit centroid of the γ -ray emission (to be determined in the following).Photons with energy between E min and E max as stated in Table 1 were used in generatingthe maps, except for Liller 1 whose map was generated using photons with energy between2–30 GeV to avoid contamination from the strong Galactic background due to its Galacticposition of l = 354 . ◦
84 and b = − . ◦ γ -ray emission within the tidal radius of Liller 1, M80, NGC 6139,NGC 6541, NGC 6624, and NGC 6752. In addition, we found significant γ -ray emission nextto NGC 6441, previously reported in Abdo et al. (2010c). In all these cases, a γ -ray sourcecandidate in the neighborhood of the corresponding GCs is detected above the modeledbackground emission. We did not find significant γ -ray emission within the tidal radiusof other candidates: M22, NGC 2808, NGC 362, NGC 6540, NGC 1851, Terzan 6, andNGC 6397.The TS values and the corresponding significances (= √ T S ) of the γ -ray emissioncoincident with the GCs are shown in Table 2. The number of gamma-ray photons associatedwith the corresponding γ -ray source in the best-fit model provided by the likelihood tool gtlike are also listed. Since the significance values for a number of positions were calculated, thissignificance has to be corrected according to a “trial factor” to give a “post-trial” significance.This trial factor accounts for the increased probability of finding a fake signal with a numberof search positions. We estimate the trial factor as N GC × N bin , where N GC is the numberof GCs searched, and N bin the number of bins within the tidal radius of each GC. We have 5 –searched for γ -rays from 30 GCs(including those with the highest encounter rate normalizedby its distance and those first mentioned in Abdo et al. 2010c), therefore N GC = 30. For N bin , the number of bins is obtained by dividing the area of the tidal radius of an GC (wetake the tidal radius of Liller 1, 13 ′ , as the averaged tidal radius) by the area of each bin(0 . ◦ × . ◦ .
01 squared degrees). This gives N bin ∼
15. After accounting for the trial factor,the “post-trial” detection significance values are obtained. Since the TS value for adjacentgrid points are correlated in the likelihood analysis procedure, the post-trial significance canbe considered as conservative estimates.We then used gtfindsrc to determine the positions of the γ -ray source candidates and gtlike (unbinned likelihood analysis) to obtain their spectra and their TS values. The inte-grated photon fluxes and energy fluxes are given for the energy range 100 MeV to 100 GeV,using simple extrapolation from the respective models. For the three cases of most significantdetection (i.e., Liller 1, NGC 6624, and the γ -ray source next to NGC 6441), photons be-tween 300 MeV and 100 GeV are divided into 6 logarithmically-equally spaced energy bins.The flux in each bin was obtained by gtlike using photons in the corresponding bin, whileboth the normalization and photon indices of all γ -ray sources and the diffuse backgroundcomponents, namely Galactic diffuse and isotropic components, were set to be free. Only forcases where the photon statistics is not enough for a satisfactory fit, the photon index wasfixed at the best-fit value shown in Table 2.We also performed a long-term temporal analysis of the 7 γ -ray sources, in which the 2-year data were binned into several periods depending on the photon statistics. No significant γ -ray variability nor flaring period was found, indicating that the sources are stable inradiating γ -rays.In the following section, we discuss the results of each of the GCs in more details.
4. Gamma-ray properties from individual clusters4.1. Gamma-ray emission within cluster’s tidal radius
The encounter rate of Liller 1 is the highest among the target list mentioned in Section 2.As seen in Fig. 1, γ -ray emission from the direction of Liller 1 was found. Fitting thephoton spectrum ( dN/dE ) with the single power law, the TS value of this source is 107,corresponding to a detection significance of 10.3 σ (i.e. 9.4 σ post-trial). The best-fit positionis R.A. = 263 . ◦
20, Dec. = − . ◦ ± . ◦ σ uncertainty). For localization of 6 –a typical γ -ray source, the systematic position error is estimated to be about 40% of thestatistical error (Abdo et al. 2009c) . The best-fit position is 7 . ′ . ′ E dN/dE ) of the γ -ray emission between 300 MeV and 100 GeV. The photon spectrum ofLiller 1 can be well fit by a single power law with an index of 2 . ± .
1, and the integralphoton flux between 100 MeV and 100 GeV is F . −
100 GeV = (6 . ± . × − cm − s − in this model. On the other hand, the parameters provided by the power law with anexponential cut-off (PLE) model with all three parameters being free are not well constrained;we then fixed the cut-off energy in this model at 2 GeV and 10 GeV, and obtained smallervalues of TS (i.e. 90 and 100, respectively). We therefore do not consider this model to besignificant. We found that the rightmost spectral point shown in Fig. 5, that corresponds to40–100 GeV emission, were detected with a TS value of 16, corresponding to a significanceof 4 σ . At a distance of 9.6 kpc (Harris 1996, 2003 version; which is within the range(8.3 ± γ -ray luminosity of Liller 1 is L . −
100 GeV =(5 . ± . × erg s − , which is the largest among all detected γ -ray emitting GCs by now. Among the GCs in the target list, M80 was found to be a possibly γ -ray emitting GC.This GC is located at the Galactic coordinate of ( l, b ) = (352 . ◦ , . ◦ ◦ and included photons withenergies 200 MeV to 100 GeV in the analysis. Inside the tidal radius of M80 we foundevidence for γ -ray emission that peaks at a TS value of 27, corresponding to a significanceof 5.2 σ (i.e. 3.9 σ post-trial), as shown in Fig. 1 (region A ). We therefore regard M80 as apossible γ -ray emitter. In addition, it is possible that the possible γ -ray emission region mayextend well outside the GC, i.e. into region B . Formally we cannot separate the whole γ -ray emission into two emitting regions. Assuming that the γ -ray emitting region coincidentwith the GC (i.e. region A ) is a separate source, its best-fit position is R.A. = 244 . ◦ − . ◦ ± . ◦ . ′ γ -ray source canbe well fit by a single power law with an index of 2 . ± .
2. The integral γ -ray flux is F . −
100 GeV = (5 . ± . × − cm − s − . The low statistics of photons does not allowfor accurate determination of the parameter values in the PLE model; we therefore do notconsider this model. At a distance of 10 . +0 . − . kpc (Brocato et al. 1998), L . −
100 GeV =8 . +7 . − . × erg s − . We assume similar systematic errors for other GCs as well. γ -ray emitting region B (TS=30) is R.A. = 244 . ◦
42, Dec. = − . ◦ ± . ◦ γ -ray source can be fit by a singlepower law with an index of 2 . ± .
2. The integral γ -ray flux is F . −
100 GeV = (1 . ± . × − cm − s − . The encounter rate of NGC 6139 is the 30th highest in our target list. We found evidencefor γ -ray emission at the position of NGC 6139. The TS value of this source candidate is32, corresponding to a detection significance of 5.6 σ (i.e. 4.5 σ post-trial), obtained in thepower law model. We therefore regard this detection as a marginal detection. The TS maparound NGC 6139 is shown in Fig. 2. The best-fit position is R.A. = 246 . ◦
83, Dec. = − . ◦ ± . ◦ . ′ . ′ γ -ray source is confirmed, the γ -ray spectrum of NGC 6139 can be wellfit by a single power law with an index of 2 . ± .
2. The integrated photon flux between100 MeV and 100 GeV is F . −
100 GeV = (9 . ± . × − cm − s − . The low statisticsof photons does not allow for accurate determination of the parameter values in the PLEmodel; we therefore do not consider this model. At a distance of 10.1 kpc (Harris 1996, 2003version), the γ -ray luminosity of NGC 6139 is L . −
100 GeV = (1 . ± . × erg s − . Inside the tidal radius of NGC 6541, we found a possible γ -ray source that peaks atTS=19, corresponding to a significance of 4.4 σ (i.e. 2.8 σ post-trial). Therefore, we considerthis source to be possibly detected, consistent with the finding of Abdo et al. (2010c). TheTS map is shown in Fig. 2. The best-fit position is R.A. = 272 . ◦
04, Dec. = − . ◦ ± . ◦ . ′ . ′ − . ± . F . −
100 GeV = (9 . ± . × − cm − s − .This flux is consistent with the upper limit of 1 . × − cm − s − as given in Abdo et al.(2010c), noting that their flux was calculated using the PLE model. Moreover, our derivedintegrated energy flux of E . −
100 GeV = (6 . ± . × − erg cm − s − are consistent with thatgiven in Abdo et al. (2010b). The PLE model was not used in the likelihood fit because of theputative nature of the γ -ray emission. At a distance of (6 . ± .
7) kpc (Lee & Carney 2006),the γ -ray luminosity of the γ -ray source candidate is L . −
100 GeV = 3 . +2 . − . × erg s − . 8 – The γ -ray source 1FGL J1823.4-3009 lies close to NGC 6624 (Abdo et al. 2010b,c).Using a larger data set, we confirmed this γ -ray source and obtained a TS value of 121,corresponding to a detection significance of 11.0 σ (i.e. 10.1 σ post-trial). The best-fit positionwas found to be R.A. = 275 . ◦
86, Dec. = − . ◦ ± . ◦ . ′ . ′ . ± .
1, in which the integrated γ -ray flux is F . −
100 GeV = (2 . ± . × − cm − s − . The above power-law index and ourderived integrated energy flux of E . −
100 GeV = (2 . ± . × − erg cm − s − are consistentwith that given in Abdo et al. (2010b). On the other hand, the parameters provided bythe PLE model (with all three parameters free) are not well constrained; we then fixed thecut-off energy in this model at 600 MeV, 1 GeV, 1.5 GeV, 2 GeV, 3 GeV and 5 GeV, andobtained TS values of 111, 120, 121, 120, 118, and 116, respectively). Fixing the cutoffenergy at 1.5 GeV, the power law index was found to be 0 . ± .
2, and the integrated γ -rayflux was found to be F . −
100 GeV = 1 . +0 . − . × − cm − s − . Since this model gives the sameTS(= 121) as the power law, it is a model as good as the power-law model. We found thatthe two rightmost spectral points shown in Fig. 6, that together correspond to 14–100 GeVemission, were detected with a TS value of ∼
10 each, corresponding to a significance of 3 σ each. At a distance of 7.9 kpc (Harris 1996, 2003 version), the γ -ray luminosity of NGC 6624(using the power-law model) is L . −
100 GeV = (1 . ± . × erg s − . We found significant γ -ray emission from NGC 6752. This source has a TS value of 49,corresponding to a significance of 7.0 σ (i.e. 6.0 σ post-trial), in the power law model. Asshown in Fig. 3, the source is compact and located close to the core of NGC 6752. Thebest-fit position is R.A. = 287 . ◦
57, Dec. = − . ◦ ± . ◦ . ′ . ± .
2. The integrated photon flux between 100 MeV and 100 GeV is F . −
100 GeV =(6 . ± . × − cm − s − . This flux is consistent with the upper limit of 7 × − cm − s − as given in Abdo et al. (2010c), noting that their flux was calculated using the PLE model.The parameters provided by the PLE model (with all three parameters free) are not wellconstrained; we then fixed the cut-off energy at 600 MeV, 1 GeV, 3 GeV and 10 GeV, and 9 –obtained smaller values of TS (i.e. 23, 27, 34, and 42, respectively). We therefore do notconsider this model to be significant. At a distance of (4 . ± .
1) kpc (Harris 1996, 2003version), the γ -ray luminosity of NGC 6752 is L . −
100 GeV = (1 . ± . × erg s − . A γ -ray source next to NGC 6441 was first reported by Abdo et al. (2010c). Using alarger data set, we also found a source (a TS value of 101 is obtained with the PLE model,corresponding to a detection significance of 10.0 σ , i.e. 9.1 σ post-trial) with the best-fitposition of R.A. = 267 . ◦
63, Dec. = − . ◦ ± . ◦ . ′ . ′
00) but it is still possiblethat some emission is from within the tidal radius given the current statistical and systematicerrors. We obtained a single power-law fit with an index 2 . ± .
1, in which the integrated γ -ray flux is F . −
100 GeV = (4 . ± . × − cm − s − . This fit gives a TS value of 73. The photonspectrum can better be fit by a PLE where the photon index, Γ, is 0 . ± . ± T S = 100. To further constrain the spectral parameters, weproceeded to fix Γ at 0.2 to 2.0 (with steps of 0.2) while letting the normalization and E c free,and found that TS=101 for Γ γ =0.2, 0.4, and 0.6 and the TS value decreases with increasingΓ. If one fixes Γ γ at 0.4, E c = (1 . ± .
1) GeV is obtained. In this case, the integrated γ -rayflux is F . −
100 GeV = (1 . ± . × − cm − s − . The significance of the PLE model overthe single power law model can be estimated by T S
PLE − T S PL = 28, which correspondsto 5.3 σ . Therefore, we consider the PLE model to be a significantly better model than thepower law model. The PLE model spectrum, as well as the data binned in different energies,is depicted in Fig. 7. At a distance of 11.7 kpc (Valenti et al. 2004), the γ -ray luminosityof NGC 6441 (using the power law model) is L . −
100 GeV = (3 . ± . × erg s − , or L . −
100 GeV = (1 . ± . × erg s − (using the PLE model).
5. discussion
In this work, we report the discovery of three new γ -ray emitting GCs: Liller 1,NGC 6624, and NGC 6752, and we found evidence for γ -ray emission from M80, NGC 6139,and NGC 6541. We also confirmed the γ -ray detection next to NGC 6441 first mentionedby Abdo et al. (2010c). We searched through various X-ray and radio catalogs for plausible γ -ray sources in the neighborhood of the GCs. We found two pulsars next to Liller 1, asshown in Fig. 1 (left panel). The spin-down luminosity of PSR J1733-3322 and PSR J1734- 10 –3333, is 8 . × erg s − and 5 . × erg s − , respectively (Morris et al. 2002), i.e. morethan an order of magnitude lower than the observed γ -ray luminosity of 10 erg s − . To-gether with their large offsets from the γ -ray emitting region, the association of the observed γ -rays with either pulsar is unlikely. In all cases, the GCs seem to be the only plausiblecounterparts of the observed γ -rays (we refer to region A in the case of M80). Without clearcounterparts in other wavelengths, the nature of the γ -ray emitting region B shown in Fig. 1(right panel) remains unclear, and its relation with M80 cannot be confirmed or ruled outat this stage.Several MSPs are known in NGC 6441, NGC 6624, and NGC 6752 (all located closeto its core) but none has yet been uncovered in other cases (e.g., Hui et al. 2009). Wespeculate that there is a population of MSPs – the only known steady γ -ray emitters inGCs – in the γ -ray emitting GCs that we have reported here. For example, the brighterX-ray population of M80 is in many ways similar to 47 Tucanae, the first γ -ray GC known,while the fainter X-ray sources may differ (Heinke et al. 2003). The discovery of γ -raysfrom M80, if confirmed, would suggest that it also hosts a population of MSPs. Under theassumption that the observed γ -ray luminosity depends solely on the number of MSPs (e.g.in the magnetospheric model), the predicted number of MSPs in the cluster using Eq.(1) ofAbdo et al. (2010c) is listed in Table 2. Future multiwavelength observations of any MSPsshould help in understanding the γ -ray production mechanisms and the underlying MSPpopulations.The γ -ray spectra are in general consistent with single power law model, and shows noconvincing evidence for cut-off at high energies. It is mainly due to limited photon statistics.However, the cases of the more significant cases of Liller 1 and NGC 6624, in which γ -ray arefound within the tidal radius, suggests that > > γ -ray emission is associated with the GCs, one striking feature is the displacementof the γ -ray emission from the GC cores. We briefly discuss this feature in the context oftwo radiation scenarios commonly discussed in the literature.In the pulsar magnetosphere model, one expects the emission to coincide spatially withthe MSP population. While MSPs should concentrate in the GC core in general, the finitenumber of MSPs in a GC (i.e., ∼
10, see Table 2, except for Liller 1) may invoke some offsetof the MSPs – and thus the γ -rays they emit – from the core. Furthermore, in addition tothe MSP population formed in the core, some MSPs are formed near the tidal radius of theGCs. The superposition of γ -rays from these two populations may not coincide with the 11 –GC core, but is at a certain distance from the core within the tidal radius. However, thisscenario cannot accommodate the case of the γ -ray source next to NGC 6441.In the inverse Compton models, the γ -ray emitting region depends not only on thedistribution of the underlying MSPs, but also on the cooling timescale and the diffusiontimescale of the accelerated particles. When the diffusion timescale is much smaller than thecooling timescale, the γ − rays may come from the outskirts of the GCs.In the cases of Liller 1 and NGC 6624, the gamma-ray spectrum (cf. Fig. 5 and Fig. 6)can extend up to energies >
40 GeV, which cannot be explained in terms of curvature ra-diation inside the light cylinder. Recently, Cheng et al. (2010) have suggested that inverseCompton scattering between background relativistic electrons/positrons in pulsar wind andthe soft photons from the galactic disk can produce the observed gamma-rays from globularclusters. They predict that if the inverse Compton upscattered Galactic infrared photonsare responsible for GeV photons, then the Compton upscattered galactic optical photonscan produce gamma-rays up to 100 GeV. They also predict that the gamma-rays are diffuseand are emitted from region much beyond the core of the globular cluster. If this is true thecenter of gamma-ray emission regions is affected by three factors: the proper motion of theglobular cluster, the asymmetric diffusion coefficient and nearby optical/IR external sourceslike stars and nearby regions containing an enhanced amount of dust.The discovery of several more γ -ray emitting GCs reported in this work has allowedpopulation studies to be carried out in order to better understand the relationship betweenthe γ -ray properties of GCs and other parameters of GCs. Abdo et al. (2010c) found that thestellar encounter rate scales with the γ -ray luminosity, based on a sample of 8 γ -ray emittingGCs. While the γ -ray luminosity should correlate with the number of MSPs in the clusters inboth models, it is expected that the low-energy photon energy density is also a good estimatorof γ -rays in the IC models. Recently, using a total of 15 GCs that are γ -ray sources or sourcecandidates, including those reported in this work and those previously known (Abdo et al.2010c), Hui et al. (2011) have shown that correlations also exist between γ -ray luminosityand metallicity, as well as γ -ray luminosity and optical/infrared photon energy density.To conclude, we report the detection of new γ -ray emitting GCs, based on a rankinglist that replies on the encounter rate of the GCs. The γ -ray luminosity of Liller 1 is thehighest of all known γ -ray GCs. Its non-detection before this work may be due to its positionthat lies on the Galactic plane, where Galactic diffuse emission is very strong. Liller 1 hasvery high encounter rate and metallicity, and may indeed host a large population of MSPsthat have not been uncovered. Some γ -ray emitting GCs presented here are at distances of ∼
10 kpc away from us, whereas Abdo et al. (2010b) present the upper limits of known GCsout to a distance of 6 kpc only. Therefore, we have demonstrated that the study here has 12 –gone a further step towards finding new sources even at larger distances. This then helps toexpand the sample of known γ -ray GCs for further studies.We acknowledge the use of data and software facilities from the FSSC, managed by theHEASARC at the Goddard Space Flight Center. CYH is supported by research fund ofChungnam National University in 2010. KSC is supported by a GRF grant of Hong KongGovernment under HKU700908P, and AKHK is supported partly by the National ScienceCouncil of the Republic of China (Taiwan) through grant NSC96-2112-M-007-037-MY3 andNSC99-2112-M-007-004-MY3. REFERENCES
Abdo, A. A., et al. (Fermi/LAT Collaboration) 2009a, Science, 325, 845Abdo, A. A., et al. (Fermi/LAT Collaboration) 2009b, Science, 325, 848Abdo, A. A., Ackermann, M., Ajello, M., et al. (Fermi/LAT Collaboration) 2009c, ApJS,183, 46Abdo, A. A., et al. (Fermi/LAT Collaboration) 2010a, ApJS, 187, 460Abdo, A. A., et al. (Fermi/LAT Collaboration) 2010b, ApJS, 188, 405Abdo, A. A., et al. (Fermi/LAT Collaboration) 2010c, A&A, 524, 75Aharonian, F., et al. (HESS Collaboration) 2009, A&A, 499, 273Anderhub, H., et al. (MAGIC Collaboration) 2009, ApJ, 702, 266Atwood, W. B., et al. (Fermi/LAT Collaboration) 2009, ApJ, 697, 1071Bednarek, W., & Sitarek, J. 2007, MNRAS, 377, 920Brocato, E., Castellani, V., Scotti, G. A., Saviane, I., Piotto, G., & Ferraro, F. R. 1998,A&A, 335, 929Cheng, K. S., Chernyshov, D. O., Dogiel, V. A., Hui, C. Y., & Kong, A. K. H. 2010, ApJ,723, 1219Gendre, B., Barret, D., & Webb, N. 2003, A&A, 403, L11 13 –Gnedin, O. Y., Zhao, H. S., Pringle, J. E., Fall, S. M., Livio, M., & Meylan, G. 2002, ApJ,568, L23Harris, W. E. 1996, AJ, 112, 1487 (2003 version)Heinke, C. O., Grindlay, J. E., Edmonds, P. D., Lloyd, D. A., Murray, S. S., Cohn, H. N., &Lugger, P. M. 2003, ApJ, 598, 516Hui, C. Y., Huang, H. H., Cheng, K. S., Taam, R. E., & Becker, W. 2009, AstronomicalSociety of the Pacific Conference Series, 404, 149Hui, C. Y., Cheng, K. S., & Taam, R. E. 2010, ApJ, 714, 1149Hui, C. Y., Cheng, K. S., Wang, Y., Tam, P. H. T., Kong, A. K. H., Chernyshov, D. O., &Dogiel, V. A. 2011, ApJ, 726, 100Kong, A. K. H., Hui, C. Y., & Cheng, K. S. 2010, ApJ, 712, L36Lee, J. W. & Carney, B. W. 2006, AJ, 132, 2171Manchester, R. N., Hobbs, G. B., Teoh, A., & Hobbs, M. 2005, AJ, 129, 1993Mattox, J. R., et al. 1996, ApJ, 461, 396Morris, D. J., et al. 2002, MNRAS, 335, 275Ortolani, S., Barbuy, B., Bica, E., Zoccali, M., & Renzini, A. 2007, A&A, 470, 1043Pooley, D., et al. 2003, ApJ, 591, L131Pooley, D., & Hut, P. 2006, ApJ, 646, L143Valenti, E., Ferraro, F. R., & Origlia, L. 2004, MNRAS, 351, 1204Venter, C., & de Jager, O. C. 2008, ApJ, 680, L125Verbunt, F., & Hut, P. 1987, Proceedings of the IAU Symposium No. 125, The Origin andEvolution of Neutron Stars, 187
This preprint was prepared with the AAS L A TEX macros v5.0.
Table 1: Region-of-interest (ROI), minimum photon energy ( E min ), and maximum photon energy ( E max ) in the analysisof the GCs GC name ROI center ROI radius E min E max R.A. (J2000) Dec. (J2000)( ◦ ) ( ◦ ) ( ◦ ) (GeV) (GeV)Liller 1 263 . − .
39 5 0.3 100M80 244 . − .
02 10 0.2 100NGC 6139 246 . − .
90 6 0.3 100NGC 6441 267 . − .
89 5 0.3 100NGC 6541 272 . − .
85 6 0.3 100NGC 6624 275 . − .
16 6 0.3 100NGC 6752 287 . − .
96 10 0.2 100
Table 2: Summary of γ -ray results of the GCsGC name Spectral TSSignificance N pha Photon index b Cutoff energyPhoton flux c Energy flux d Distance e Luminosity f N MSPg model ( σ ) (Γ) ( E c , GeV) (kpc)Liller 1 PL 107 10.3 966 2 . ± . · · · ±
23 53 ±
18 9.6 59 ±
20 410 +480 − M80 PL 27 5.2 125 2 . ± . · · · . ± . . ± . +0 . − . +7 . − . +123 − NGC 6139 PL 32 5.6 154 2 . ± . · · · . ± . . ± . ± +114 − NGC 6441 PL 73 8.5 451 2 . ± . · · · ±
10 23 ± ± · · · h PLE 101 10.0 380 0.4 i . ± . ± . ± . ± · · · NGC 6541 PL 19 4.4 109 2 . ± . · · · . ± . . ± . . ± . . +2 . − . +49 − NGC 6624 PL 121 11.0 362 2 . ± . · · · ± ± ± +104 − PLE 121 11.0 326 0 . ± . ± . ± . . ± . · · · NGC 6752 PL 49 7.0 130 2 . ± . · · · . ± . . ± . . ± . . ± . +15 − a Modeled photon number associated with the corresponding γ -ray source in the best likelihood fit b All the quoted errors are statistical and 1 σ for one parameter of interest. c Integrated 0.1–100 GeV photon flux in unit of 10 − cm − s − d Integrated 0.1–100 GeV energy flux in unit of 10 − erg cm − s − e Distance adopted from Harris (1996, 2003 version), except for M80 (Brocato et al. 1998), NGC 6541 (Lee & Carney 2006), andNGC 6441 (Valenti et al. 2004). f erg s − g Predicted number of MSPs in the cluster using Eq.(1) of Abdo et al. (2010c), under the assumption that the observed γ -ray luminositydepends solely on the number of MSPs h In this case, it is unlikely that the number of MSPs is the sole factor in determining the observed γ -ray luminosity due to the offset i Model parameters without quoted errors are fixed at the value given.
16 –Fig. 1.— The Test-statistic (TS) maps of regions of 2 ◦ × ◦ centered on the best-fit centroids(labeled by crosses) of the γ -ray emission from Liller 1 ( left ) and M80 ( right ). The size ofeach cross indicates the 1- σ statistical error in the determination of the centroid position.The color scale shows the TS value of every bin of an area 0 . ◦ × . ◦
1. The light blue circlesrepresent the tidal radius of the respective GCs compiled in Harris (1996, 2003 version).
Left : PSR J1733-3322 and PSR J1734-3333, are marked by two white diamonds. The 68%error positions of the LAT first catalog sources 1FGL J1732.3-3243c and 1FGL J1730.0-3408c(treated as background) are shown as eclipses.
Right : The position of the γ -ray MSP PSRJ1614-2230 (Abdo et al. 2009b) is shown as a cross. 17 –Fig. 2.— The TS maps of regions of 2 ◦ × ◦ centered on the best-fit centroids (labeled bycrosses) of the γ -ray emission from NGC 6139 ( left ) and NGC 6541 ( right ). The circlesrepresent the tidal radius of NGC 6139 ( left ) and NGC 6541 ( right ) (Harris 1996, 2003version). The meanings of the color scale and the size of the crosses are the same as inFigure 1. 18 –Fig. 3.— The TS maps of regions of 2 ◦ × ◦ centered on the best-fit centroids (labeled bycrosses) of the γ -ray emission from NGC 6624 ( left ) and NGC 6752 ( right ). The TS map ofa region of 2 ◦ × ◦ centered on the best-fit centroid (labeled by a cross) of the γ -ray emissionnext to NGC 6624. Left panel : The circle represents the tidal radius of NGC 6624 (Harris1996, 2003 version). The white diamond shows the position of both PSR B1820-30A andPSR B1820-30B.
Right panel : The outer and inner circle represents the tidal radius andhalf-mass radius of NGC 6752, respectively (Harris 1996, 2003 version). The meanings ofthe color scale and the size of the crosses are the same as in Figure 1. 19 –Fig. 4.— The TS map of a region of 2 ◦ × ◦ centered on the best-fit centroid (labeled bya cross) of the γ -ray emission next to NGC 6441. The circle represents the tidal radius ofNGC 6441 (Harris 1996, 2003 version). The meanings of the color scale and the size of thecross are the same as in Figure 1. 20 –Fig. 5.— The spectrum ( E × dN/dE ) of the γ -ray emission found within the tidal radius ofLiller 1. Photons between 300 MeV and 100 GeV are divided into 6 logarithmically-equallyspaced energy bins and The derived flux for each bin is plotted at the corresponding meanenergy. The solid line indicates the best-fit single power law model. The rightmost pointhas a TS value of 16, corresponding to a significance level of four. 21 –Fig. 6.— The spectrum ( E × dN/dE ) of the γ -ray emission found within the tidal radiusof NGC 6624. Photons between 300 MeV and 100 GeV are divided into 6 logarithmically-equally spaced energy bins and The derived flux for each bin is plotted at the correspondingmean energy. The solid line and the dashed line indicate the best-fit single power law modeland the PLE model with E c = 1 . ∼
10, corresponding to a significance level of about three each. 22 –Fig. 7.— The spectrum ( E × dN/dE ) of the γγ