FERMI OBSERVATIONS OF TeV-SELECTED ACTIVE GALACTIC NUCLEI

At energies above approximately 100 GeV (hereafter the TeV energy regime), ground-based γ-ray observatories have detected 96 sources over the past two decades. The pace of discovery in this energy regime has been particularly high since the inception of the latest generation of instruments: H.E.S.S., CANGAROO, MAGIC, and VERITAS (Weekes 2008, for recent review). Online catalogs, such as TeVCat,⁵⁸ present continuously updated views of the TeV γ-ray sky. The majority of the TeV sources are galactic; however, 30 extragalactic sources have also been detected, of which 28 correspond to active galactic nuclei (AGNs), the other two being recently detected starburst galaxies. Most (25) of these TeV AGNs are blazars, an AGN sub-category in which the jet of relativistic plasma ejected from the core is roughly co-aligned with our line of sight and hence appears Doppler boosted. The majority (24) of the TeV blazars belong to a further sub-category, the BL Lac objects (from BL Lacertae, the prototype for the class), which do not have significant emission or absorption features in their optical spectra, making it difficult to measure their redshift directly.

The first blazar detected at TeV energies was Markarian 421 (Mrk 421; Punch et al. 1992), at a redshift of z = 0.031. It is seen to be highly variable, with flux varying between ∼0.15 and >10 the flux of the Crab Nebula⁵⁹ (ϕ_Crab). Doubling timescales as short as 15 minutes have been observed during flares (Gaidos et al. 1996). Mrk 421 has a hard spectrum, with mean photon index of Γ = 2.5, and has shown clear evidence of spectral variability during flaring episodes (Krennrich et al. 2002a). The detection of such a distant object was interpreted in the context of γ-ray attenuation through pair-production (Gould & Schréder 1967) to produce a limit on the power density of extragalactic background light (EBL; Stecker & de Jager 1993). Mrk 421 was detected by EGRET, the predecessor of the Fermi Large Area Telescope (Thompson et al. 1993; Kanbach et al. 1988), and was reported in the third EGRET catalog (3EG; Hartman et al. 1999) with a detection significance of ∼10σ. However, EGRET did not have sufficient sensitivity to make detailed measurements on the short timescales required to match the TeV observations.

Since this initial discovery, 28 AGN sources have been detected at TeV energies, the most distant reported being 3C 279, at a redshift of z = 0.54. Like Mrk 421, many of these have shown evidence for variability, undergoing episodic flaring activity with short doubling timescales. To date, the most extreme example of variability has been observed from PKS 2155 − 304 (Aharonian et al. 2007a), in which the flux was seen to reach a maximum of ∼15 ϕ_Crab with doubling times as short as 225 s. However, approximately half of the TeV blazars show no evidence for variability, e.g., PKS 2005 − 489 (Aharonian et al. 2005a) and PG 1553+113 (Aharonian et al. 2006c; Albert et al. 2009). The detections of more distant objects lead to tighter constraints on the level of the EBL, suggesting that its density is close to the minimum level required from galaxy counts (Aharonian et al. 2006a). The spectra of the majority of TeV blazars are adequately described by a simple power law,⁶⁰ with Γ ⩾ 2, and with those of the more distant sources being considerably softer, up to Γ ≈ 4 (e.g., Acciari et al. 2009b; Albert et al. 2007b, 2008b; Aharonian et al. 2006c, 2005a). The peak in the measured γ-ray spectrum (in νF_ν representation) of these objects lies outside the energy range of the TeV instruments, making it difficult to fully constrain models of emission using TeV observations alone. The spatial and spectral properties of the TeV-detected AGNs are presented in Tables 1 and 2, together with references to TeV observations of each source.

Table 1. AGN Detected at TeV Energies

Name	αJ2000	δJ2000	Typea	z	Ref.
Blazars
RGB J0152+017	01h 52m 396	+01° 47' 17''	HBL	0.080	1
3C 66A	02h 22m 396	+43° 02' 08''	IBL	0.444b	2,3c
1ES 0229+200	02h 32m 486	+20° 17' 17''	HBL	0.140	4
1ES 0347 − 121	03h 49m 232	−11° 59' 27''	HBL	0.188	5
PKS 0548 − 322	05h 50m 408	−32° 16' 18''	HBL	0.069	6
RGB J0710+591	07h 10m 301	+59° 08' 20''	HBL	0.125	7
S5 0716+714	07h 21m 534	+71° 20' 36''	LBL	0.300	8
1ES 0806+524	08h 09m 492	+52° 18' 58''	HBL	0.138	9
1ES 1011+496	10h 15m 041	+49° 26' 01''	HBL	0.212	10
1ES 1101 − 232	11h 03m 376	−23° 29' 30''	HBL	0.186	11
Markarian 421	11h 04m 273	+38° 12' 32''	HBL	0.031	12
Markarian 180	11h 36m 264	+70° 09' 27''	HBL	0.046	13
1ES 1218+304	12h 21m 219	+30° 10' 37''	HBL	0.182	14
W Comae	12h 21m 317	+28° 13' 59''	IBL	0.102	15
3C 279	12h 56m 112	−05° 47' 22''	FSRQ	0.536	16
PKS 1424+240	14h 27m 004	+23° 48' 00''	IBL	...	17
H 1426+428	14h 28m 327	+42° 40' 21''	HBL	0.129	18
PG 1553+113	15h 55m 430	+11° 11' 24''	HBL	0.09 − 0.78	19
Markarian 501	16h 53m 522	+39° 45' 37''	HBL	0.034	20
1ES 1959+650	19h 59m 599	+65° 08' 55''	HBL	0.048	21
PKS 2005 − 489	20h 09m 254	−48° 49' 54''	HBL	0.071	22
PKS 2155 − 304	21h 58m 521	−30° 13' 32''	HBL	0.117	23
BL Lacertae	22h 02m 433	+42° 16' 40''	LBL	0.069	24,25c
1ES 2344+514	23h 47m 048	+51° 42' 18''	HBL	0.044	26
H 2356 − 309	23h 59m 079	−30° 37' 41''	HBL	0.167	27
Others
3C 66B	02h 23m 114	+42° 59' 31''	FR1	0.02106	28
M 87	12h 30m 494	+12° 23' 28''	FR1	0.004233	29
Centaurus A	13h 25m 276	−43° 01' 09''	FR1	0.00183	30

Notes. ^aSee notes for Table 3 for explanation of object types. ^bThe redshift of 3C 66A is considered to be uncertain. ^cDetection of E > 1 TeV emission from 3C 66A and BL Lacertae was first claimed by Neshpor et al. (1998, 2001). The measured fluxes are not consistent with the later measurements made with more sensitive instruments. References. (1) Aharonian et al. 2008a; (2) Neshpor et al. 1998; (3) Acciari et al. 2009b; (4) Aharonian et al. 2007c; (5) Aharonian et al. 2007b; (6) Superina et al. 2008; (7) Ong et al. 2009b; (8) Teshima et al. 2008; (9) Acciari et al. 2009a; (10) Albert et al. 2007b; (11) Aharonian et al. 2006a; (12) Punch et al. 1992; (13) Albert et al. 2006b; (14) Albert et al. 2006a; (15) Acciari et al. 2008; (16) Albert et al. 2008b; (17) Ong et al. 2009a(18) Horan et al. 2002; (19) Aharonian et al. 2006c; (20) Quinn et al. 1996; (21) Nishiyama 1999; (22) Aharonian et al. 2005a; (23) Chadwick et al. 1999; (24) Neshpor et al. 2001; (25) Albert et al. 2007a; (26) Catanese et al. 1998; (27) Aharonian et al. 2006b; (28) Aliu et al. 2009; (29) Aharonian et al. 2003a; (30) Aharonian et al. 2009a.

Download table as:  ASCIITypeset image

The Large Area Telescope (LAT) is a pair conversion telescope on the Fermi Gamma-Ray Space Telescope (formerly GLAST), launched in 2008 June. The Fermi LAT instrument, described in detail in Atwood et al. (2009), detects γ rays with energies between 20 MeV and >300 GeV (hereafter denoted the GeV energy regime). The bulk of the Fermi observational program is dedicated to a sky survey, in which the full γ-ray sky is observed every 3 hr. This survey is optimized to produce a uniform exposure to the sky on timescales of months and to facilitate the monitoring and detection of variable and flaring γ-ray sources on shorter timescales. In its first three months of operation, the Fermi LAT mapped the γ-ray sky with a sensitivity and precision that exceeds any previous space mission in this energy regime. A list of the 205 brightest sources (10σ or greater) found during that period, and their properties, has been published by the Fermi-LAT collaboration to guide multiwavelength studies with other instruments (Abdo et al. 2009b). This collection of sources is henceforth referred to as the 0FGL list, with individual sources denoted as 0FGL JHHMM.M±DDMM. In addition, an in-depth study of the population of 0FGL sources most likely associated with AGNs has been made (Abdo et al. 2009a); this population is commonly referred to as the LAT bright AGN sample (LBAS).

Only eight of the 28 TeV-detected AGNs were detected by EGRET and included in the 3EG catalog (see Table 4); six are BL Lacs. The majority of these 3EG GeV–TeV sources were discovered as TeV emitters only with the advent of the latest generation of TeV observatories. With the previous generation of instruments only two such extragalactic GeV–TeV sources were established, despite dedicated programs to observe 3EG sources (e.g., Fegan et al. 2005), indicating the degree of mismatch between the sensitivities and effective energy ranges of these instruments. According to the blazar sequence theory, BL Lacs are the least luminous class of blazars in the GeV regime, with their emission peaking at higher energies, which results in their having hard photon indices in the GeV domain. In fact, above a few 10 s of GeV, BL Lacs are often relatively brighter than other AGNs. With a rapidly falling sensitivity above ∼5–10 GeV, EGRET preferentially detected the more luminous, lower-energy-peaked blazars. In contrast, the Fermi LAT has a relatively flat effective area at high energies (8000 cm² for on-axis γ rays at E > 1 GeV) and an energy response that extends beyond 300 GeV, making it much more sensitive to the hard BL Lac sources than EGRET was. Of the 0FGL sources, 14 are AGNs detected at TeV energies.

One of the most powerful tools for probing the physics underlying the emission from AGNs is the dedicated multiwavelength observational campaign, in which simultaneous observations are made across the full spectrum. Generally, such observations of the TeV blazars reveal two non-thermal components: one at lower energies, peaking in the UV to X-ray regime, and showing clear evidence of polarization, and a second peaking in the γ-ray regime. The low-energy component is commonly interpreted as resulting from synchrotron emission from relativistic electrons in the jet, while the high-energy component results from a different process, such as inverse-Compton scattering of lower energy photons in the region of the jet, or the decay of π⁰ particles produced in interactions of relativistic protons. In many such campaigns, significant correlation between the X-ray and TeV γ-ray emission has been detected (e.g. Buckley et al. 1996; Aharonian et al. 2009b), suggesting that a single population of relativistic particles is responsible for the emission in both regimes (see Katarzyński et al. 2005, for further discussion of X-ray/TeV correlations). Leptonic mechanisms, such as the synchrotron self-Compton (SSC) and external-Compton (EC) processes, and hadronic mechanisms have been invoked to explain the broadband emission and correlated time variability seen between the low-energy and high-energy components (see Böttcher 2007, for further discussion). Instances of isolated γ-ray variability have also been detected in some cases (e.g., Krawczynski et al. 2004), and it is probable that no simple mechanism will fully explain the considerable variety of the blazar behavior.

Despite the participation of instruments across a wide range of the spectrum, until recently, multiwavelength campaigns have been largely unable to probe the energy range between ∼150 keV and ∼150 GeV, as no instrument with sensitivity matched to the day-to-week timescales of typical campaigns has existed. As such, although the synchrotron component has been well measured from radio to X-ray, the full extent of the higher energy γ-ray component has not. In the case of the TeV blazars, ground-based γ-ray instruments have generally measured the falling edge of the high-energy component (in νF_ν representation), which has usually been consistent with a featureless power law. The rising edge and peak of the emission have, however, been inaccessible, and hence models of the emission have been unconstrained in this energy range. In many cases, very different emission mechanisms have been invoked, and can explain the data equally well. With the launch of Fermi, which has the sensitivity to measure the emission from the brighter TeV blazars on the day-to-week timescales, a large part of this gap in coverage has been closed. A recent multi-wavelength campaign on PKS 2155 − 304 was the first in which the rising and falling edges of the high-energy spectral energy distribution (SED) were simultaneously measured with precision (Aharonian et al. 2009c). Measurement of the full broadband spectrum and the pattern of correlation between the optical, X-ray, GeV γ-ray, and TeV γ-ray emission removes degeneracies in modeling of this object which were present in the results from previous campaigns that could not measure the high-energy component fully.

In this paper, we present the results of the first 5.5 months of Fermi-LAT observations of the known TeV blazars and of those AGNs for which upper limits exist at TeV energies. The motivation for this study is two-fold: (1) to present as complete a picture of the high-energy emission as possible by combining the GeV and TeV results on these objects, and (2) to help guide future TeV observations. For each object detected by Fermi, a power-law fit to the GeV spectrum is presented, as are light curves on monthly timescales. For the brighter sources, light curves on ten-day timescale are also given. The GeV power-law spectra are extrapolated to TeV energies assuming absorption on the EBL, yielding predictions for TeV emission for the simplest case where there is no curvature in the intrinsic spectra of the objects. For those objects which are not detected by the Fermi LAT, upper limits in the GeV range are presented and extrapolated to TeV energies.

The primary objects selected for this study are the 25 blazars and three radio galaxies detected at TeV energies. These are listed in Table 1 with their coordinates, the AGN subclass of the object, redshift, and references to the initial detection at TeV energies. In summary, 19 high-frequency-peaked BL Lacs (HBLs), three intermediate-frequency-peaked BL Lacs (IBLs), two low-frequency-peaked BL Lacs (LBLs), one flat-spectrum radio quasar (FSRQ), and three Fanaroff–Riley radio galaxies (type FR1) have been detected by TeV instruments, the most distant having a redshift of z = 0.54. Table 2 lists the parameters of a power-law fit to the TeV spectra for these objects, where available: the integral flux (ϕ ± Δϕ) and photon index of the fit (Γ ± ΔΓ) and the threshold energy for the observation (E_thres), such that the differential spectrum is

$$\frac{dN}{dE} = (\Gamma -1)\frac{\phi }{E_\mathrm{thres}} \left(\frac{E}{E_\mathrm{thres}}\right)^{-\Gamma } = F_{200} \left(\frac{E}{200\,\mathrm{GeV}}\right)^{-\Gamma }.$$

The differential flux at 200 GeV (the median threshold of the measurements), F₂₀₀, is calculated from the TeV power-law spectrum and presented in the table to compare the TeV objects at a single energy lying within the domain of the Fermi-LAT observations. For some objects, multiple TeV spectra have been measured, either by different instruments, in different epochs, or when the object is in different flux states. Where possible, the spectrum corresponding to a low-flux state is listed.

In addition, we search for GeV emission from 68 objects for which TeV upper limits were published from observations with the Whipple 10 m telescope (Horan et al. 2004; de la Calle Pérez et al. 2003; Falcone et al. 2004), HEGRA (Aharonian et al. 2004), MAGIC (Albert et al. 2008a) and H.E.S.S. (Aharonian et al. 2005c, 2008b). These targets are listed in Table 3 with the lowest flux upper limit published.

Table 3. AGNs with Published Upper Limits at TeV Energies

Name	αJ2000	δJ2000	Typea	z	TeV limit		Ref.
					Flux [ϕCrab]	Energy (GeV)
III Zw 2	00h10m310	+10°58'30''	Sy1	0.089	<0.027	> 430	1
1ES 0033+595	00h35m526	+59°50'05''	HBL	0.086b	<0.11	> 390	2
NGC 315	00h57m489	+30°21'09''	FR1	0.016	<0.05	> 860	3
4C+31.04	01h19m350	+32°10'50''	RG	0.060	<0.14	> 760	3
1ES 0120+340	01h23m086	+34°20'49''	HBL	0.272	<0.032	> 190	4
1ES 0145+138	01h48m298	+14°02'19''	HBL	0.125	<0.015	> 310	5
UGC 01651	02h09m386	+35°47'50''	RG	0.038	<0.07	> 790	3
BWE 0210+1159	02h13m052	+12°13'11''	LBL	0.250	<0.012	> 530	1
RGB J0214+517	02h14m179	+51°44'52''	HBL	0.049	<0.17	> 430	6
PKS 0219-164	02h22m010	−16°15'17''	LBL	0.698	<0.27	> 1780	3
NGC 1054	02h42m157	+18°13'02''	Sy	0.033	<0.02	> 860	3
NGC 1068	02h42m407	−00°00'48''	Sy2	0.004	<0.013	> 210	5
V Zw 331	03h13m576	+41°15'24''	LBL	0.029	<0.09	> 870	3
NGC 1275	03h19m482	+41°30'42''	FR1	0.018	<0.03	> 850	3
RX J0319.8+1845	03h19m518	+18°45'34''	HBL	0.190	<0.033	> 190	4
B2 0321+33B	03h24m412	+34°10'46''	NLSy1	0.063	<0.10	> 400	7
1ES 0323+022	03h26m139	+02°25'15''	HBL	0.147	<0.015	> 210	5
4C +37.11	04h05m493	+38°03'32''	RG	0.055	<0.05	> 800	3
1ES 0414+009	04h16m523	+01°05'54''	HBL	0.287	<0.057	> 230	4
3C 120	04h33m111	+05°21'16''	RG	0.033	<0.004	> 230	5
MG J0509+0541	05h09m260	+05°41'35''	IBL	...	<0.11	> 960	3
4C+01.13	05h13m525	+01°57'10''	BL Lac	0.084	<0.10	> 1010	3
Pictor A	05h19m497	−45°46'45''	FR2	0.034	<0.014	> 220	5
PKS B0521 − 365	05h22m580	−36°27'31''	LBL	0.055	<0.042	> 310	1
EXO 0556.4-3838	05h58m062	−38°38'27''	HBL	0.302	<0.051	> 220	5
PKS 0558-504	05h59m474	−50°26'52''	NLSy1	0.137	<0.018	> 310	1
1ES 0647+250	06h50m466	+25°03'00''	HBL	0.203	<0.13	> 780	3
UGC 03927	07h37m301	+59°41'03''	UnC	0.041	<0.09	> 1090	3
3C 192.0	08h05m350	+24°09'50''	RG	0.060	<0.20	> 930	3
RGB J0812+026	08h12m019	+02°37'33''	BL Lac	...	<0.031	> 220	5
3C 197.1	08h21m336	+47°02'37''	FR2	0.130	<0.05	> 960	3
PKS 0829+046	08h31m489	+04°29'39''	LBL	0.174	<0.06	> 1000	3
NGC 2622	08h38m110	+24°53'43''	Sy1	0.028	<0.05	> 400	7
1ES 0927+500	09h30m376	+49°50'26''	HBL	0.188	<0.052	> 230	4
S4 0954+65	09h58m472	+65°33'55''	LBL	0.368	<0.096	> 300	6
MS1019.0+5139	10h22m126	+51°24'00''	BL Lac	0.141	<0.07	> 920	3
1ES 1028+511	10h31m185	+50°53'36''	HBL	0.361	<0.29	> 400	6
RGB J1117+202	11h17m063	+20°14'07''	HBL	0.139	<0.030	> 610	5
1ES 1118+424	11h20m480	+42°12'12''	HBL	0.124	<0.12	> 430-500	6
Markarian 40	11h25m362	+54°22'57''	Sy1	0.021	<0.21	> 430	6
NGC 3783	11h39m017	−37°44'19''	Sy1	0.010	<0.025	> 220	5
NGC 4151	12h10m326	+39°24'21''	Sy1.5	0.003	<0.07	> 790	3
1ES 1212+078	12h15m112	+07°32'05''	HBL	0.136	<0.17	> 920	3
ON 325	12h17m521	+30°07'01''	LBL	0.130	<0.22	> 400-430	6
3C 273	12h29m067	+02°03'09''	FSRQ	0.158	<0.014	> 300	1
MS 1229.2+6430	12h31m314	+64°14'18''	HBL	0.164	<0.17	> 300-430	6
1ES 1239+069	12h41m483	+06°36'01''	HBL	0.150	<0.20	> 400-430	6
1ES 1255+244	12h57m319	+24°12'40''	HBL	0.141	<0.11	> 350-500	6
RGB J1413+436	14h13m437	+43°39'45''	RG	0.089	<0.06	> 400	7
RX J1417.9+2543	14h17m567	+25°43'26''	HBL	0.237	<0.023	> 190	4
OQ530	14h19m466	+54°23'15''	LBL	0.151	<0.058	> 300	6
1ES 1440+122	14h42m483	+12°00'40''	HBL	0.162	<0.033	> 290	5
RGB J1629+401	16h29m013	+40°08'00''	NLSy1	0.271	<0.09	> 400	7
RX J1725.0+1152	17h25m044	+11°52'15''	HBL	0.018	<0.046	> 190	4
I Zw 187	17h28m186	+50°13'10''	HBL	0.055	<0.086	> 300-350	6
1ES 1741+196	17h43m578	+19°35'09''	HBL	0.083	<0.053	> 350-500	6
3C 371	18h06m507	+69°49'28''	LBL	0.051	<0.19	> 300	6
Cyg A	19h59m284	+40°44'02''	FR2	0.056	<0.03	> 910	3
PKS 2201+04	22h04m177	+04°40'02''	BL Lac	0.028	<0.08	> 950	3
PG 2209+184	22h11m539	+18°41'50''	RG	0.070	<0.13	> 400	7
RBS 1888	22h43m416	−12°31'38''	HBL	0.226	<0.009	> 170	5
HS 2250+1926	22h53m074	+19°42'35''	FSRQ	0.284	<0.009	> 590	1
2QZ J225453.2-272509	22h54m532	−27°25'09''	BL Lac	0.333	<0.016	> 170	5
PKS 2254+074	22h57m173	+07°43'12''	LBL	0.193	<0.05	> 900	3
NGC 7469	23h03m156	+08°52'26''	Sy1	0.017	<0.006	> 250	5
PKS 2316-423	23h19m058	−42°06'49''	HBL	0.055	<0.014	> 190	5
1ES 2321+419	23h23m525	+42°10'55''	HBL	0.059	<0.03	> 890	3
1ES 2343-151	23h45m384	−14°49'29''	IBL	0.224	<0.012	> 230	1

Note ^aBL Lac, LBL, IBL, HBL–BL Lac object; FSRQ: Flat Spectrum Radio Quasar; FR1 and 2: Fanaroff-Riley 1 and 2 galaxy; Sy 1, 1.5, and 2: Seyfert 1, 1.5, and 2; RG: Radio Galaxy; NLSy1: Narrow-Line Seyfert 1; UnC: Unclassified; see the SIMBAD (http://simbad.u-strasbg.fr/simbad) and NED (http://nedwww.ipac.caltech.edu) databases. ^bTentative measurement, see comment in the text. References. (1) Aharonian et al. 2008b; (2) de la Calle Pérez et al. 2003; (3) Aharonian et al. 2004; (4) Albert et al. 2008a; (5) Aharonian et al. 2005c; (6) Horan et al. 2004; (7) Falcone et al. 2004.

Download table as:  ASCIITypeset images: 1 2

From these 96 target objects, the 18 listed in Table 4 are identified or associated with sources in the 0FGL list and LBAS sample (Abdo et al. 2009b, 2009a). Those lists were limited to sources with TS > 100 in three months of Fermi-LAT data⁶¹; in this study, the criterion to claim a detection and derive a spectrum is lowered to TS > 25, and the period of observation is increased to 5.5 months. A total of 38 sources are detected by Fermi, of which 21 are jointly detected at GeV and TeV energies. We also give an upper limit for TeV sources not detected by Fermi.

Fermi-LAT data from the 5.5 month interval from MJD 54682 to MJD 54842 are processed with the standard analysis chain ScienceTools (ST; version V9R11). The latest instrumental response functions (IRFs; version P6_V3) are used to characterize the point-spread function (PSF) and effective area during the analysis. These IRFs offer a distinct improvement over those used in the 0FGL analysis, which did not properly account for the presence of remnants of non-triggering events in the tracker. This change results in a systematic increase of ∼15% in the derived flux from γ-ray sources, and a possible change in the spectral index, which is most pronounced for softer sources. However, despite these improvements, the P6_V3 IRFs are based on pre-flight calibrations, and to be conservative, only events in the energy range from 200 MeV to 300 GeV are retained for analysis (see for example. Abdo et al. 2009b).

The data for each of the AGN targets are analyzed in an identical manner. Low-level processing of the spacecraft data is applied automatically in a pipeline, reconstructing the energy, arrival direction, and particle type of the primary. Events reconstructed from a region of interest (ROI) of 10° around the target location are extracted from this database and filtered such that only those having the highest probability of being a photon (those in the so-called "diffuse" event class) and having an angle of <105° with respect to the local zenith (to suppress the background from the Earth albedo) are retained.

For each target, a background model is constructed, consisting of a diffuse galactic component, predicted by the GALPROP program (Strong et al. 2004a, 2004b), a diffuse power-law component (for the extragalactic and instrumental background) and any of the point sources from the Fermi three-month catalog (Abdo et al. 2009b)⁶² which overlap the ROI. The spectrum for each of the point sources is modeled as a power law. An unbinned maximum likelihood method (Cash 1979; Mattox et al. 1996), implemented as part of the ST by the gtlike program, is used to optimize the parameters to best match the observations.

The validity of the optimized emission model is verified by producing a TS map for each region. This is done using the ST gttsmap program, which adds a test source to each location over a prescribed region and calculates the improvement in log-likelihood with the inclusion of the test source. Statistically compelling sources, not accounted for by the model, are identified visually in the map and added to the background in another iteration of gtlike. In addition, high-resolution TS maps are produced for each source of interest and the centroid of the emission and contours defining the 68%, 95%, and 99% probability regions calculated. During the construction of the 0FGL list, a systematic error of ≈1' in the reconstruction of the centroids of emission from well-known bright sources was identified, and this error is folded into the emission contours displayed in Figure 1. If the centroid of emission on the map is $\vec{r}_{\rm c}$ and the distance from the centroid to the (statistical) error contours are defined parametrically by the functions rⁱ_stat(θ), where i = 1, 2, 3 for the 68%, 95%, and 99% probabilities (Pⁱ = 0.68, 0.95, 0.99), then the contours which account for systematic errors are defined as

$$r_\mathrm{syst}^{\,i}(\theta) = \sqrt{r_\mathrm{stat}^{\,i}(\theta)^2 + (\sqrt{-2\ln (1-P^{\,i})}\times 1\mbox{$^\prime $})^2}.$$

For well-detected sources, the error contours are roughly circular, with the systematic error dominating. For weakly detected sources, the contours can be irregularly shaped, with the statistical component dominating.

Zoom In Zoom Out Reset image size

Among the outputs from the gtlike program are the optimized values of the model parameters, the covariance matrix describing their variances and correlations, and the TS of each source, which indicates the significance of the source detection. A contour describing the statistical error on the spectrum (called butterfly diagram) is computed from these values and plotted to indicate the 1σ confidence range of the fitted power-law model. If the power law is written as F(E) = dN/dE = F₀(E/E₀)^−Γ, with the normalization parameter F₀ ± ΔF₀, photon index Γ ± ΔΓ, and covariance cov(F₀, Γ), the contour is defined by

$$\begin{eqnarray} &&\frac{\Delta F^2}{F^2} = \frac{\Delta F_0^2}{F_0^2} - \frac{2\,\mathrm{cov}(F_0,\Gamma)}{F_0}\log \left(\frac{E}{E_0}\right) + \Delta \Gamma ^2 \log ^2\left(\frac{E}{E_0}\right).\nonumber\\ \end{eqnarray} \tag{ 1 }$$

The narrowest point in the butterfly occurs at E_d = E₀exp [cov(F₀, Γ)/F₀ ΔΓ²], the so-called decorrelation energy. For each source, the butterfly is drawn between the lowest energy used in the analysis (either 0.2 GeV or 1 GeV) and the energy of the highest photon detected from the source, subject to the constraint of E < 300 GeV.

For the sources with a detection significance of TS < 25 (∼5σ), upper limits on the integral flux above 200 MeV are computed assuming a photon index arbitrary fixed at 1.5 and 2.0. For the very bright LAT sources, the energy range is divided into two bins (200 MeV–1 GeV and 1 GeV–300 GeV), and spectra are fitted to each bin separately. This analysis is limited to sources for which TS > 100 in each of the energy bins, ensuring that a sufficiently accurate spectrum can be derived in each. For all detected sources, systematic errors on the flux and index of the power-law fit to the full Fermi energy range, caused by systematic errors in the IRFs used in the analysis, are evaluated using the "bracketing" method of Abdo et al. (2009d).

In order to make predictions for the TeV energy domain and to make comparisons between the Fermi-LAT and TeV spectra, the best-fit spectrum and butterfly are extrapolated up to 10 TeV. This assumes that the intrinsic spectrum of the emission is described by a single power-law extending over the GeV and TeV energy range, the simplest and least model-dependent assumption that can be made. Above a few hundred GeV, the photons interact with the infrared photons from the EBL as they propagate through the universe, modifying the detected spectrum from a simple power law. Franceschini et al. (2008) provide tabulated values of the optical depth as a function of the redshift, which are used to compute the flux detectable between 200 GeV and 10 TeV from the extrapolated GeV spectrum. Their EBL model is consistent with experimental measurements—the lower limits from galaxy counts and upper limits from observations of TeV blazars—and is widely used. However, it is not necessarily the final word on EBL density (see, e.g., Krennrich et al. 2008), and any errors in the model will propagate into the extrapolation of the GeV spectra to TeV energies. In addition to absorption on the EBL, TeV photons may undergo absorption in the neighborhood of the source (see, e.g., Aharonian et al. 2008c), which must also be modeled and accounted for to unfold the intrinsic accelerated spectrum from the detected spectrum. However, such modeling is beyond the scope of this paper.

Light curves for each source are produced with time bins of 10 and 28 days (following the lunar cycle). The light curves are produced by binning the events by their arrival times and performing an independent likelihood analysis for each temporal bin with the same model (same background sources and number of free parameters) as in the fitting of the time-averaged spectrum. The probability that the light curve is consistent with being flat, from a χ² fit to a constant value, is also computed, and used to evaluate the hypothesis that the fitted spectra, averaged over the full period, are valid.

Of the 28 TeV-selected sources studied here, 21 are detected by Fermi with TS > 25. This degree of connection between the TeV blazars and the GeV regime was not found by EGRET and the previous generation of TeV instruments, and is evident now only as a result of the improved sensitivity and greater overlap between the effective energy ranges of Fermi and the current generation of TeV instruments.

The majority of the TeV blazars detected by Fermi have a photon index Γ < 2 in the GeV regime, the median index is Γ = 1.9. In contrast, the populations of 42 BL Lacs and 57 FSRQs from the LBAS sample have median indices of Γ = 2.0 and 2.4, respectively. The TeV blazars are amongst the hardest extragalactic objects detected by Fermi.

Of the 68 extragalactic objects with TeV limits which were considered here, a total of 17 are detected in the GeV regime. These too have a hard median index of Γ = 1.95, indicating that they, perhaps, are good targets for deeper observation with TeV instruments. Of these 17, only one is not a blazar (NGC 1275, an FR1 radio galaxy), one is an FSRQ (3C 273), and the remainder are BL Lacs. That the majority of these 16 blazars are BL Lacs, rather than following the ratio of FSRQs to BL Lacs found in LBAS, is most likely a result of the way these objects were selected for observation originally by the various TeV groups, rather than anything inherently fundamental. These 15 objects break down as follows: three LBLs, one IBL, and 11 HBLs.

For each of the 38 sources detected by Fermi, power-law fits to the data and extrapolations into the TeV regime are presented in Figure 2 for GeV–TeV sources, and from the authors on request for the GeV sources with TeV upper limits. Additionally, for TeV sources not detected by Fermi, upper limits on the spectra are presented in Figure 3. To justify the validity of these "averaged spectra," light curves on 28-day timescales are presented in Figure 4 for all GeV-detected objects, and on 10-day timescales in Figure 5 for the brighter GeV emitters. In the 5.5 months of data analyzed here, the majority of sources do not show evidence of flux variation on the timescales tested. For a subset of the sources, as discussed below, a TS map is presented in Figure 1. The remainder are available from the authors on request.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In addition to the figures, the parameters of the power-law spectra and variability are given in tables. For each of the sources detected by Fermi, Table 5 lists

• 1.  
  the flux and index of the power law, the statistical and systematic errors on these quantities, the decorrelation energy;
• 2.  
  the energy of the highest and fifth highest photons detected within 025 of the source position, which corresponds to >99.9% containment according to the P6_V3 IRFs; and
• 3.  
  the probabilities that the 28-day and 10-day light curves are consistent with a constant value.

For the brightest GeV sources, Table 6 lists the parameters of the power-law fits to the low-energy (0.2 GeV–1 GeV) and high-energy (1 GeV–300 GeV) bands. For the TeV sources not detected by Fermi, Table 7 gives the GeV upper limit over the Fermi energy range. Finally, Table 8 presents the extrapolations of the GeV spectra to the TeV domain, listing the differential flux extrapolated (or measured) at 100 GeV, the integral flux in the TeV band (0.2 TeV–10 TeV) and the photon index found by fitting the EBL-corrected spectrum between 100 GeV and 1 TeV with a power law.

4.1. TeV Sources Detected by the Fermi LAT

4.2. TeV Sources not Detected by the Fermi LAT

4.3. GeV Sources with Upper Limits in the TeV Regime

In the LBAS study, the dependence of GeV spectral index of the FSRQ and BL Lac populations on redshift was presented (Figure 11 of Abdo et al. 2009a). Although it was found that the two populations have different spectral properties, no significant relation between the gamma-ray photon index and redshift was found within each source class. The GeV–TeV sources provide a population in which the effects of spectral evolution with redshift can be studied across a much wider energy range than LBAS (although, admittedly in a much smaller redshift range). The presence of a redshift-dependent spectral break in these sources could be indicative of the effects of absorption on the EBL, and provide experimental evidence for this absorption in a manner independent of any specific EBL-density model. Appendix A discusses the relationship between EBL absorption and the redshift dependent change in spectral index between the GeV and TeV ranges,

$$\begin{eqnarray} \Delta \Gamma &=& \Gamma _\mathrm{TeV} - \Gamma _\mathrm{GeV}\nonumber\\ &\ge & \Gamma _\mathrm{TeV} - \Gamma _\mathrm{Int}\nonumber\\ &=& \delta (z,E^*) \approx \left.\frac{d\tau (E,z)}{d\log E}\right|_{E=E^*}, \end{eqnarray} \tag{ 2 }$$

i.e., in the presence of EBL absorption, and assuming that the intrinsic spectra do not get harder with energy, measured spectral breaks must lie in a region of the (ΔΓ, z) plane defined by ΔΓ ⩾ δ(z, E*), with δ(z, E*) defined by the EBL optical depth.

Figure 6 depicts the difference in the measured TeV and GeV spectral indices, ΔΓ, as a function of the redshift, z, for 15 of the 21 extragalactic GeV–TeV sources.⁶³ It is evident that the difference between the GeV and TeV spectral indices increases with redshift. At low redshifts, the radio galaxies M 87 and Cen A have ΔΓ ≈ 0, as do the near-by BL Lacs. At redshifts greater than 0.1, all of the BL Lacs are consistent with ΔΓ ⩾ 1.5.

Zoom In Zoom Out Reset image size

Pearson's correlation factor, r, is widely used to quantify the correlation between two variables. However, this correlation factor does not take measurement errors into account. Thus, to evaluate the significance of the correlation, a series of Gaussian random variables centered on the values of the measured spectral changes, ΔΓ_i, and with width equal to their measurement errors are used to generate multiple simulated data sets. The full width at half-maximum of the distribution obtained for r gives an estimate of the error.⁶⁴ The value obtained is r = 0.76 ± 0.14, which indicates a clear correlation. To check the robustness of this result, the Kendall rank, defined by

$$\begin{equation} \tau _K=\frac{2S}{N\cdot (N-1)}, \end{equation}$$

where N is the size of the data set (N = 15), has been calculated. For each pair of points from the data set (ΔΓ, z), those in the same order are assigned a value of +1, and the others assigned −1. The sum, S, of these N · (N − 1)/2 combinations is then constructed (Gleissner et al. 2004, and references therein), giving a value of τ_K, ranging from −1 ⩽ τ_K ⩽ +1, indicating the degree of correlation. The error on τ_K is calculated in the same way as for r. The same conclusion is established with this test, τ_K = 0.68 ± 0.15.

The effects of systematics present in Figure 6 must be considered when determining the validity of concluding that it reflects the effects of EBL absorption. Such systematics are difficult to evaluate in a quantitative manner, Appendix B discusses possible contributions in more detail. In light of these difficulties, it cannot be claimed with 100% certainty that the observed deficit of sources at ΔΓ ≈ 0 for large redshift is a real effect. We anticipate that the TeV AGNs not detected by the LAT in this study will ultimately be detected and that this figure will then be limited only by the selection bias at TeV energies. Finally, EBL absorption is not the only effect that could cause a redshift dependent spectral break, evolution of the intrinsic spectra with redshift also cannot be excluded, although this may be difficult to reconcile with the LBAS study, which did not detect any evolution in the larger LAT BL Lac sample.

In 5.5 months of observation, the Fermi LAT has detected GeV emission from 21 TeV-detected AGNs, and from 17 AGNs previously observed by TeV groups and for which upper limits have been published at TeV energies. Whereas EGRET detected only a small number of such TeV sources, and detected those only with integration times of months and years, Fermi has detected the majority of the TeV blazars in its first few months of operation. Fermi observations will help TeV observatories optimize the limited observation time available each year, and will be useful in evaluating possible targets for observation. Fermi has sufficient sensitivity to participate meaningfully in simultaneous multi-wavelength campaigns and to measure spectra and light curves from the brighter blazars with a resolution of days to months. In future campaigns, it can be expected that the high-energy emission will be as well covered by Fermi and the TeV observatories as the low-energy peak is by instruments in the radio to X-ray bands.

Many of the TeV sources exhibit an increasing spectrum (Γ < 2) in the GeV range confirming the presence of a high-energy peak in νF_ν representation. This is the first large-scale characterization of the full γ-ray emission component for this class of energetic AGNs. More detailed modeling of these γ-ray sources, which is beyond the scope of this paper, will become possible as more data are acquired by Fermi and the flight-calibrated IRFs become available (extending the effective energy below 100 MeV). The MAGIC-II and H.E.S.S. II instruments will increase the range over which the sensitivities of TeV instruments overlap with Fermi at lower energies, producing better coverage at energies between 50 and 200 GeV where a number of TeV sources seem to have turnovers in their measured spectra.

The intrinsic spectrum for some of the TeV sources can be well described by a single power-law across the energy range spanned by the Fermi LAT and the TeV observatories, with any breaks in the measured γ-ray spectra between the two regimes being consistent with the effects of absorption with a model of minimal EBL density. For other objects, however, a softening of the intrinsic spectrum is required to match the TeV measurements. This could be due to softening intrinsic to the IC component–itself reflecting curvature in the relativistic electron or seed photon distributions.

Based on an extrapolation of the GeV spectra measured by Fermi to TeV energies, a number of previously observed AGNs are good candidates for re-observation with TeV instruments: 1ES 0033+595 and 1ES 0647+250 are two such AGNs.

Redshift-dependent evolution is detected in the spectra of objects detected at GeV and TeV energies. The most reasonable explanation for this is absorption on the EBL, and as such, it would represent the first model-independent evidence for absorption of γ rays on the EBL. Future observations with Fermi and TeV instruments have the potential to probe τ(E, z) in a more quantitative manner.

The Fermi-LAT Collaboration acknowledges the generous support of a number of agencies and institutes that have supported the Fermi-LAT Collaboration. These include the National Aeronautics and Space Administration and the Department of Energy in the United States, the Commissariat à l'Energie Atomique and the Centre National de la Recherche Scientifique / Institut National de Physique Nucléaire et de Physique des Particules in France, the Agenzia Spaziale Italiana and the Istituto Nazionale di Fisica Nucleare in Italy, the Ministry of Education, Culture, Sports, Science and Technology (MEXT), High Energy Accelerator Research Organization (KEK) and Japan Aerospace Exploration Agency (JAXA) in Japan, and the K. A. Wallenberg Foundation, the Swedish Research Council and the Swedish National Space Board in Sweden. Additional support for science analysis during the operations phase is gratefully acknowledged from the Istituto Nazionale di Astrofisica in Italy. This research has made use of NASA's Astrophysics Data System Bibliographic Services, the NASA/IPAC Extragalactic Database, operated by JPL, Caltech, under contract from NASA, and the SIMBAD database, operated at CDS, Strasbourg, France.

Facilities: Fermi LAT

Writing the measured TeV spectrum as F_TeV(E), the unabsorbed (intrinsic) spectrum as F_Int(E), and the redshift-dependent optical depth due to EBL absorption as τ(E, z), the effects of the EBL on the measured spectrum are given by

$$\begin{equation} F_\mathrm{TeV}(E) = e^{-\tau (E,z)} F_\mathrm{Int}(E). \end{equation} \tag{ A1 }$$

If both the measured and intrinsic spectra can be approximated as power laws over the range of the TeV observations, $F_{\rm TeV}(E) \approx C_{\rm TeV} (E/E_0)^{-\Gamma _{\rm TeV}}$ and $F_{\rm Int}(E) \approx C_{\rm Int} (E/E_0)^{-\Gamma _{\rm Int}}$, then the EBL effects must also be given by a power law, say $e^{-\tau (E,z)} \approx C_\tau (E/E_0)^{-\delta (z,E^*)}$. The power-law index of the EBL absorption, δ(z, E*), is a function of the redshift and the energy range over which the TeV observations are made (denoted for convenience as a dependence on some energy, E*, at which the measured TeV data most constrain the fitted spectrum). Equating the two expressions for the absorption to first order in x = log(E/E*) gives δ(z, E*) ≈ dτ(x, z)/dx|_{x = 0}(see Vassiliev 2000; Stecker & Scully 2006, for further discussion of linear and polynomial expansions of the EBL optical depth). Equation (A1) relates δ(z, E*) to the intrinsic and measured spectral indices

$$\begin{equation*} C_\mathrm{TeV} (E/E_0)^{-\Gamma _\mathrm{TeV}} = C_\tau C_\mathrm{Int} (E/E_0)^{-\delta (z,E^*) - \Gamma _\mathrm{Int}},\ \mathrm{or} \end{equation*}$$

$$\begin{equation} \delta (z,E^*) = \Gamma _\mathrm{TeV} - \Gamma _\mathrm{Int} \sim \Gamma _\mathrm{TeV} - \Gamma _\mathrm{GeV}, \end{equation} \tag{ A2 }$$

where it is further assumed that the measured GeV spectral index can be used as a proxy for the intrinsic index in the TeV regime. This is equivalent to assuming that there is no curvature in the intrinsic spectrum between the GeV and TeV energy regimes and that absorption on the EBL does not affect the spectrum in the LAT energy range, which is true for all reasonable EBL models for sources with z < 0.5. For GeV–TeV detected sources, this suggests that the variable ΔΓ = Γ_TeV − Γ_GeV probes the effects of EBL absorption in a manner independent of any specific model of EBL density.

For real sources, there is curvature in the intrinsic spectrum between the GeV and TeV regimes, and the measured Γ_GeV is not a perfect estimator for Γ_Int in the TeV regime. In general, it has been found that the curvature in the differential spectra is concave, and the intrinsic spectrum at TeV energies is expected to be softer than the measured GeV spectrum. Therefore, it is expected that for real sources

$$\begin{equation} \Delta \Gamma = \Gamma _\mathrm{TeV} - \Gamma _\mathrm{GeV} \ge \Gamma _\mathrm{TeV} - \Gamma _\mathrm{Int} = \delta (z,E^*), \end{equation} \tag{ A3 }$$

so that sources would occupy a region of the space of (ΔΓ, z) above some curve defined by the EBL.

A number of components contribute to the systematics in Figure 6 and are addressed below: (1) systematic errors on the points themselves and (2) the effects of the criteria used to select targets for the study.

The systematic errors on the measurements of Γ_GeV for the individual sources are presented in Table 5, and are generally smaller than statistical errors, with the largest being ≈0.25. Similarly, the TeV groups estimate and report systematic errors on Γ_TeV for each of the detected TeV sources. See, for example, Aharonian et al. (2006e) for a discussion of systematic error estimation with H.E.S.S. In general, the GeV and TeV systematic errors are too small to explain the trend in Figure 6, in which ΔΓ changes by >2.0 over the range of redshift in question.

The sources in this study are subject to a two-stage selection process, which may lead to regions of the phase-space of (ΔΓ, z) being inaccessible due to limitations imposed by of the sensitivities of Fermi and the TeV instruments. This could, in turn, lead to a false correlation being evident in the data—for example, see Figure 7 from the LBAS study (Abdo et al. 2009a) for a correlation which might incorrectly be claimed based on a sensitivity-limited sample. Targets were originally selected by TeV astronomers for observation with TeV instruments. Given the sensitivities of those instruments and the amount of time dedicated to each target, some fraction were detected, leading to 28 sources used in this study. The Fermi sensitivity further restricts the sample to the 15 GeV–TeV sources displayed in Figure 6. Given the complexity (and randomness) of the selection process, it is almost impossible to quantify where its "sensitivity" limits are in the space of (ΔΓ, z). In the null hypothesis that there is no EBL effect present in the results of Figure 6, we attempt to evaluate in a qualitative manner whether the lack of sources at ΔΓ ≈ 0 for larger redshifts could arise from a selection effect. The primary selection of targets is based on detection at TeV energies. Sources in this region of the plot would have harder TeV spectra than those actually detected at the larger redshifts. Since, it is unlikely that TeV astronomers are deliberately biasing the sample toward softer distant AGNs, the major effect producing this bias would have to result from the sensitivity limit of the instrument. The instrumental sensitivity directly limits the space of (ϕ, Γ), requiring ϕ>ϕ_Lim(Γ) for detection (as in the LBAS figure cited above), and these limits transfer to the space of (ΔΓ, z) only through convolution with the source function f(ϕ, Γ; z) = d²P/dϕ dΓ. Therefore, a sharp cut-off in this space should not be expected, rather a slower decrease in source counts into the "forbidden" region. In the no-EBL hypothesis, and further assuming that evolution in the source function, the primary consideration is whether the decrease in measured flux with distance coupled with the sensitivity limit would lead to an evolution in the population of detectable sources with redshift. There is very little published material addressing the flux sensitivity of current TeV instruments as a function of spectral index, however it is reasonable to presume that TeV instruments are more sensitive to sources with harder spectra, than to those with softer: they have better background rejection and an improved PSF at higher energies. Indeed, TeV instruments often use "hard cuts" to improve the sensitivity for hard sources. In this case, it would be expected that TeV instruments would preferentially detect harder sources (with lower fluxes) at larger redshifts, whereas this is exactly the opposite of what is actually observed, i.e., that the majority of the distant AGNs are softer—no hard, weak TeV AGNs have been detected to date.