arXiv:0807.2822v1 [astro-ph] 17 Jul 2008
Very-High-Energy Gamma Rays from a Distant Quasar:
How Transparent Is the Universe?
The MAGIC Collaboration∗
Science 320, 1752 (2008); DOI: 10.1126/science.1157087
Keywords: 3C 279, Quasar, EBL, Cherenkov telescope
MAGIC, designed for a low-energy threshold, has detected
very-high-energy gamma rays from a giant flare of the
distant Quasi-Stellar Radio Source (in short: radio quasar)
3C 279, at a distance of more than 5 billion light-years (a
redshift of 0.536). No quasar has been observed previously
in very-high-energy gamma radiation, and this is also the
most distant object detected emitting gamma rays above 50
gigaelectron volts. Since high-energy gamma rays may be
stopped by interacting with the diffuse background light
in the universe, the observations by MAGIC imply a low
amount for such light, consistent with that known from
Ground-based gamma-ray telescopes are sensitive to the
Cherenkov light emitted by the electromagnetic showers that
are produced by gamma rays interacting in the atmosphere.
These telescopes have discovered, since the first detection (in
1989) of gamma rays in this energy range (from 100 GeV to
several TeV), more than 20 blazars, which are thought to be
powered by accretion of matter onto supermassive black holes
residing in the centers of galaxies, and ejecting relativistic jets
at small angles to the line of sight (1). Most of these ob-
jects are of the BL Lac type, with weak or no optical emission
lines. Quasar 3C 279 shows optical emission lines that allow
a good redshift determination. Satellite observations with the
EnergeticGamma Ray ExperimentTelescope (EGRET)aboard
the Compton Gamma Ray Observatory (CGRO) had measured
gamma rays from 3C 279 (2) and other quasars, but only up to
the energies of a few GeV, the limit of the detector’s sensitivity.
An upper limit for the flux of very-high-energy(VHE) gamma
rays was quoted in (3).
Using MAGIC, the world’s largest single-dish gamma-ray
telescope (4) on the Canary island of La Palma (2200 m above
sea level, 28.4◦N, 17.54◦W), we detected gamma rays at ener-
gies from 80 to >300 GeV, emanating from 3C 279 at a red-
shift of 0.536, which corresponds to a light-travel time of 5.3
billion years. No object has been seen before in this range of
VHE gamma-ray energies at such a distance [the highest red-
shift previously observed was 0.212 (5)], and no quasar has
been previously identified in this range of gamma-rayenergies.
∗The complete list of authors and their affiliations appears at the end of this
The detection of 3C 279 is important, because gamma rays
at very high energies from distant sources are expected to be
strongly attenuated in intergalactic space by the possible inter-
action with low-energy photons (γ + γ → e++ e−). These
photons [extragalactic background light (EBL) (6)] have been
radiated by stars and galaxies in the course of cosmic history.
Their collective spectrum has evolved over time and is a func-
tion of distance. For 3C 279, the range of newly probed EBL
wavelengths lies between 0.2 and 0.8µm (ultraviolet/optical).
Existing instruments that are sensitive only to higher gamma-
ray energies have so far been unable to probe this domain; by
contrast, MAGIC is specifically designed to reach the lowest-
energy threshold among ground-based detectors.
integral flux (100-500 GeV) [cm
22 Feb 2006
23 Feb 2006
Time (Julian Date - 2453000)
optical flux [mJy]
Fig. 1. Light curves. MAGIC (top) and optical R-band data (bottom)
obtained for 3C 279 from February to March 2006. The long-term
baseline for the optical flux is at 3 mJy.
In observations of 3C 279 over ten nights between late Jan-
uary and April 2006 (total of 9.7 hours), the gamma-ray source
was clearly detected (at >6 SDs) on the night of 23 February,
and may also have been detected the night before (Fig. 1). As
determined by the χ2test, the probability that the gamma-ray
flux on all 10 nights was zero is 2.3 × 10−7, corresponding to
The MAGIC Collaboration
Energy, E [GeV]
70 80100200 300400 500
differential flux, dN/dE [TeV
fit to measured spectrum:
dN/dE = N
systematic error band
* = 2.94
* = 0.49
EBL-corrected, Stecker (fast),
Fig. 2. Spectrum of 3C 279 measured by MAGIC. The grey area
includes the combined statistical (1σ) and systematic errors, and
underlines the marginal significance of detections at high energy.
The dotted line shows compatibility of the measured spectrum with
a power law of photon index α = 4.1. The blue and red triangles
are measurements corrected on the basis of the two models for the
EBL density, discussed in the text.
5.04σina Gaussiandistribution[see(7)]. Simultaneousoptical
R-band observations, by the Tuorla Observatory Blazar Moni-
toring Program with the 1.03 m telescope at the Tuorla Obser-
vatory, Finland, and the 35 cm Kungliga Vetenskapsakademien
(Royal Swedish Academy of Sciences) telescope on La Palma,
revealed that during the MAGIC observations, the gamma-ray
source was in a generally high optical state, a factor of 2 above
the long-term baseline flux, but with no indication of short
time-scale variability at visible wavelengths. The observed
VHE spectrum (Fig. 2) can be described by a power law with a
differential photon spectral index of α = 4.1 ± 0.7stat± 0.2syst.
The measured integrated flux above 100 GeV on 23 February
is (5.15 ± 0.82stat± 1.5syst) × 10−10photons cm−2s−1.
The EBL influences the observed spectrum and flux, result-
ing in an exponential decrease with energy and a cutoff in the
gamma-ray spectrum. Several models have been proposed for
the EBL (6). All have limited predictive power for the EBL
density, particularly as a function of time, because many de-
tails of star and galaxy evolution remain uncertain. We illus-
trate the uncertainty in the EBL by using two extreme mod-
els: a model by Primack et al. (8), close to the lowest pos-
sible attenuation level consistent with lower EBL limit from
galaxy counts (9,10); and a “fast-evolution” model by Stecker
et al. (11), corresponding to the highest attenuation of all the
models. We refer to these models as “low” and “high,” respec-
tively. The measured spectra of 3C 279, corrected for absorp-
tion according to these two models, are shown in Fig. 2. They
represent the range for the possible intrinsic gamma-rayflux of
A power-law fit to the EBL-corrected points (12) results in
an intrinsic photon index of α∗= 2.9±0.9stat±0.5syst(low) and
α∗= 0.5 ± 1.2stat± 0.5syst(high). The systematic error is de-
0 0.1 0.20.3 0.4 0.5 0.60.7
Gamma-Ray Energy [TeV]
Energy Threshold of MAGIC
Primack et al. (2005)
Stecker et al. (2006)
Kneiske et al. 2002 (modified)
Opacity For Gamma Rays
τ > 1
τ = 5
τ = 2
τ = 1
Fig. 3. The gamma-ray horizon. The redshift region over which
the gamma-ray horizon can be constrained by observations has
been extended up to z=0.536. The prediction range of EBL models
is illustrated by (8) (thick solid black line) and (11) (dashed-dotted
blue line). The tuned model of (14) (dashed blue line) represents
an upper EBL limit based on our 3C 279 data, obtained on the
assumption that the intrinsic photon index is ≥ 1.5 (red arrow). Lim-
its obtained for other sources are shown by black arrows, most of
which lie very close to the model (14). The narrow blue band is
the region allowed between this model and a maximum possible
transparency (i.e., minimum EBL level) given by (8), which is nearly
coincident with galaxy counts. The gray area indicates an optical
τ > 1, i.e., the flux of gamma rays is strongly suppressed. To il-
lustrate the strength of the attenuation in this area, we also show
energies for τ = 2 and τ = 5 (thin black lines), again with (8) as
termined by shifting the absolute energy scale by the estimated
energy error of 20% and recalculating the intrinsic spectrum.
Furtherdiscussion of the intrinsic spectrumand the spectral en-
ergy density can be found in (7).
The measuredspectrum of 3C 279permits a test of the trans-
parency of the universe to gamma rays. The distance at which
the flux of photons of a given energy is attenuated by a factor e
(i.e., the path correspondingto an optical depth τ = 1) is called
the gamma-ray horizon and is commonly expressed as a func-
tionoftheredshiftparameter(13); weshowthis energy/redshift
relation in Fig. 3. In the context of Fig. 3, we make use of a
by (15) and fine-tuned such that for 3C 279, the intrinsic pho-
ton index is α∗= 1.5. The tuning allows for the statistical and
systematic errors (1 SD, added linearly). Although the intrinsic
spectrum emitted by 3C 279 is unknown,α∗= 1.5 is the lowest
value given for EGRET sources (not affected by the EBL) and
all spectra measured by gamma-ray telescopes so far (16), so
we assume this to be the hardest acceptable spectrum. The re-
gion allowed between the maximumEBL determinedusing the
above procedure and that from galaxy counts (8) is very small.
The results support, at higher redshift, the conclusion drawn
from earlier measurements (15) that the observations of the
VHE Gamma-Rays from a Distant Quasar
Hubble Space Telescope and Spitzer correctly estimate most
of the light sources in the universe. The derived limits are con-
sistent with the EBL evolution corresponding to a maximum
star-formation rate at z ≥ 1, as suggested by (8) and similar
The emission mechanism responsible for the observed VHE
radiation remains uncertain. Leptonic emission models (as-
suming relativistic electrons in the jet as source of the gamma
rays), generallysuccessfulin describingblazardata [e.g.,(17)],
can, with some assumptions, also accommodate the MAGIC
spectrum. Hadronic models [involving relativistic protons, e.g.
(18)] provide a possible alternative. However, a genuine test
of the models can be only obtained with simultaneous obser-
vations at different wavelengths, which are not available for
the observations described here. Future tests of these models
should use observations from sources at all wavelengths from
radio to VHE gamma rays. In the domain of VHE gamma
rays, we can expect important new insights by simultaneous
observations with the Large Area Telescope (LAT), the high-
energy gamma-ray instrument on the Gamma Ray Large Area
Space Telescope [GLAST (19)]. Our observations of this dis-
tant source in VHE gamma rays demonstrate that a large frac-
tion of the universe is accessible to VHE astronomy.
1. R. D. Blandford, M. J. Rees, Astrophys. Lett. 10, 77 (1972).
2. R. C. Hartman et al., Astrophys. J. 385, L1 (1992).
3. F. Aharonian et al., Astron. Astrophys. 478, 387 (2008).
4. E. Lorenz, New Astronomy Review 48,339 (2004).
5. J. Albert et al., Astrophys. J. 667, L21 (2007).
6. M. G. Hauser, E. Dwek, Ann. Rev. Astron. Astrophys. 39, 307
7. Further information on data and methods and additional dis-
cussion are available on Science Online.
8. J. R. Primack, J. S. Bullock, R. S. Somerville, in High Energy
Gamma-Ray Astronomy, American Institute of Physics Confer-
ence Series, F. Aharonian, H. Voelk, D. Horns, Eds. (American
Institute of Physics, Heidelberg, 2005), vol. 745, p. 23.
9. P. Madau, L. Pozzetti, Mon. Not. R. Astron. Soc. 312, L9 (2000).
10. G. Fazio et al., Astrophys. J. Suppl. Ser. 154, 39 (2004).
11. F. W. Stecker, M. A. Malkan, S. T. Scully, Astrophys. J. 648, 774
12. We approximate the intrinsic energy spectrum by dN/dE ∝
E−α∗, where α∗is the intrinsic photon spectral index.
13. G. G. Fazio, F. W. Stecker, Nature 226, 135 (1970).
14. T. M. Kneiske, K. Mannheim, D. H. Hartmann, Astron. Astro-
phys. 386, 1 (2002).
15. F. Aharonian et al., Nature 440, 1018 (2006).
16. According to recent simulations (20), photon indices of <.5 can
not be entirely excluded.
17. L. Maraschi, G. Ghisellini, A. Celotti, Astrophys. J. 397, L5
18. K. Mannheim, P. L. Biermann, Astron. Astrophys. 251, L21
19. S. Funk et al., (GLAST-LAT Collaboration), GLAST and
ground-based gamma ray astronomy, SLAC-PUB-12871; www-
20. F. W. Stecker, M. G. Baring, E. J. Summerlin, Astrophys. J. 667,
21. We thank the Instituto de Astrofisica de Canarias for the excel-
lent working conditions at the Observatorio del Roque de los
Muchachos in La Palma, Canary Islands. The support of the
German Bundesministerium f¨ ur Bildung und Forschung and
Max-Planck Gesellschaft, the Italian Istituto Nazionale di Fisica
Nucleare (INFN) and Spanish Centro de Investigaci` on Cien-
tifica y Technologia is gratefully acknowledged. This work was
also supported by ETH (research grant TH 34/043) and the
Polish MNil (grant 1P03D01028).
The MAGIC Collaboration: J. Albert,1E. Aliu,2H. Anderhub,3
L. A. Antonelli,4
A. Biland,3R. K. Bock,8,9∗G. Bonnoli,12P. Bordas,13V. Bosch-
Ramon,13T. Bretz,1I. Britvitch,3M. Camara,5E. Carmona,8
A. Chilingarian,14S. Commichau,3J. L. Contreras,5J. Cortina,2
M. T. Costado,15,16S. Covino,4V. Curtef,6F. Dazzi,9A. De
Angelis,17E. De Cea del Pozo,18R. de los Reyes,5B. De
Lotto,17M. De Maria,17F. De Sabata,17C. Delgado Mendez,15
A. Dominguez,19D. Dorner,1M. Doro,9M. Errando,2M. Fagiolini,12
D. Ferenc,20E. Fern´ andez,2R. Firpo,2M. V. Fonseca,5L. Font,7
N. Galante,8R. J. Garc´ ıa L´ opez,15,16M. Garczarczyk,8M. Gaug,15
F. Goebel,8M. Hayashida,8A. Herrero,15,16D. H¨ ohne,1J. Hose,8
C. C. Hsu,8S. Huber,1T. Jogler,8T. M. Kneiske,6D. Kranich,3A. La
Barbera,4A. Laille,20E. Leonardo,12E. Lindfors,12S. Lombardi,9
F. Longo,17M. L´ opez,9E. Lorenz,3,8P. Majumdar,8G. Maneva,22
M. Mart´ ınez,2D. Mazin,2M. Meucci,12M. Meyer,1J. M. Miranda,5
R. Mirzoyan,8S. Mizobuchi,8M. Moles,19A. Moralejo,2D. Nieto,5
K. Nilsson,21J. Ninkovic,8N. Otte,8,23† I. Oya,5M. Panniello,15‡
R. Paoletti,12J. M. Paredes,13M. Pasanen,21D. Pascoli,9F. Pauss,3
R. G. Pegna,12M. A. Perez-Torres,19M. Persic,17,24L. Peruzzo,9
A. Piccioli,12F. Prada,19E. Prandini,9N. Puchades,2A. Raymers,14
M. Rib´ o,13
S. R¨ ugamer,1A. Saggion,9T. Y. Saito,8M. Salvati,4M. Sanchez-
Conde,19P. Sartori,9K. Satalecka,11V. Scalzotto,9V. Scapin,17
S. N. Shore,26N. Sidro,2A. Sierpowska-Bartosik,18A. Sillanp¨ a¨ a,21
L. Takalo,21F. Tavecchio,4P. Temnikov,22D. Tescaro,2M. Teshima,8
D. F. Torres,25,18
A. Venturini,9V. Vitale,17R. M. Wagner,8W. Wittek,8V. Zabalza,13
F. Zandanel,19R. Zanin,2J. Zapatero7
L. S. Stark,3
1Universit¨ at W¨ urzburg, D-97074 W¨ urzburg, Germany.
de F´ ısica d’Altes Energies, Edifici Cn., Campus UAB, E-08193
Nazionale di Astrofisica (INAF) National Institute for Astrophysics,
I-00136 Rome, Italy.5Universidad Complutense, E-28040 Madrid,
laterra, Spain.8Max-Planck-Institut f¨ ur Physik, D-80805 M¨ unchen,
3ETH Zurich, CH-8093 Switzerland.
6Technische Universit¨ at Dortmund, D-44221 Dortmund,
7Universitat Aut` onoma de Barcelona, E-08193 Bel-
The MAGIC Collaboration
Deutsches Elektronen-Synchrotron, D-15738 Zeuthen, Germany.
12Universit` a di Siena, and INFN Pisa, I-53100 Siena, Italy.
Universitat de Barcelona Insitut de Ciencias del Cosmos–Institut
d’Estudis Especials de Catalunya (IEEC), E-08028 Barcelona,
Spain.14Yerevan Physics Institute, AM-375036 Yerevan, Armenia.
15Instituto de Astrofisica de Canarias, E-38200, La Laguna,
E-38206 La Laguna, Tenerife, Spain.
INFN Trieste, I-33100 Udine, Italy.18Institut de Cienci` es de l’Espai
[IEEC-Consejo Superior de Investigaciones C´ ıentificas (CSIC)],
E-08193 Bellaterra, Spain.
(CSIC), E-18080 Granada, Spain.20University of California, Davis,
FI-21500 Piikki¨ o, Finland.
Nuclear Energy, BG-1784 Sofia, Bulgaria.23Humboldt-Universit¨ at
zu Berlin, D-12489 Berlin, Germany.
tronomico and INFN, I-34143 Trieste, Italy.
de Recerca i Estudis Avancats, E-08010 Barcelona, Spain.
Universit` a di Pisa, and INFN Pisa, I-56126 Pisa, Italy.
9Universit` a di Padova and INFN, I-35131 Padova,
10University of Ł´ od´ z, PL-90236 Ł´ od´ z, Poland.
16Departamento de Astrofisica, Universidad,
17Universit` a di Udine, and
19Inst. de Astrofisica de Andalucia
21Tuorla Observatory, Turku University,
22Institute for Nuclear Research and
25Instituc´ ıo Catalana
∗To whom correspondence should be addressed.
† Present address: Santa Cruz Institute for Particle Physics, Uni-
versity of California, Santa Cruz, CA 95064, USA.
27 February 2008; accepted 27 May 2008
Supporting Online Material
Figs. S1 to S3
Comments to the Introduction
Ground-based gamma ray astronomy, pioneered by the Whip-
ple Collaboration, has led to a number of important discoveries
in recent years, showing an unexpected wealth of sources in
the very high-energy(VHE) sky above 100 GeV. Among them,
some twenty extragalactic sources were found, all but one are
blazars. Blazars are now known in a broad range of luminosi-
ties, from low-luminosity BL-Lac objects to high-luminosity
quasars. They emit non-thermal radiation from radio waves
to gamma rays, originating presumably in synchrotron emis-
sion and inverse-Compton scattering. Most blazars so far de-
tected with ground-based telescopes belong to the rare sub-
species of BL-Lac objects, that can be interpreted as a kind
of dying quasar running out of fuel, and thinning out at large
cosmological distances (1). 3C 279 is the first blazar of the
quasar type observed to emit gamma rays in the VHE range. It
had been discovered as a source of lower energy gammas by
the EGRET detector (2), with a peak flux higher than any other
source seen by EGRET. Subsequently 3C 279 became a much-
studied source with multi-wavelength observation campaigns
The observations were carried out in the “on/off mode”, which
includes “on” events taken with the telescope pointing at the
source, and a similar amount of “off” events taken off-axis at a
sky location near the source, with very similar operating con-
ditions. The “off” events are necessary for estimating reliably
the background of hadronic events.
The raw data were calibrated (9) and analyzed with the stan-
dard MAGIC analysis and reconstruction software (10). Data
runswith anomaloustriggerrates dueto badobservationcondi-
tions were rejected. The remaining “on” data correspondto 9.7
hours, with 4.9 hours of “off” data. The standard analysis re-
duces the image to a single cluster of pixels by removingimage
noise and tracks far from the main shower, and calculates im-
age parameters from the cleaned image (11). These are used by
a multi-variate method to discriminate gamma rays from back-
ground hadrons; the method is called “Random Forest” (12)
and is a multi-tree classifier algorithm which calculates, from a
combination of image parameters, a separation variable called
All image parameters were checked for consistency between
“on” and “off” data. The excess events were obtained by sub-
tracting suitably normalized “off” data from the “on” data, as
shown in Fig. S1. The variable Alpha is one of the image pa-
rameters. It describes the direction of the main alignment axis
of the pixels in the image, a zero value indicating a shower
seemingly coming (in the camera plane) from the known posi-
tion of the source. Alpha was not included in the multivariate
gamma-hadronseparation. Cuts in hadronness and Alpha were
optimized using data samples from observations of the Crab
Nebula at comparable zenith angles. For the detection (and
the lightcurve) we require hadronness<0.12 and Alpha<12◦,
resulting in an efficiency of 40% for gammas. For these cuts
and with a minimal Size of 100 (Size is the sum of signals, in
photo-electrons), we obtain a peak in the distribution of esti-
mated energy at 170 GeV.
The significance of the signal of 23 February was calculated
according to equation 17 in (13), resulting in 6.15σ, taking the
data and the “on” data. The probability that we deal with a sta-
tistical fluctuation in ten independent observations (trial factor
correspondingto10nights ofobservation)is 3.87×10−9, which
corresponds to a significance of 5.77σ. If “off” and “on” data
are compared bin by bin without smoothing, the significance
is 5.5σ for 23 February, and 5.1σ after correcting for trials.
We also applied a χ2test to the results of ten nights, giving
50.35 for 10 degrees of freedom, or a probability of 2.3x10−7,
which in a Gaussian distribution corresponds to 5.04σ. The
long-term detector stability has been studied by monitoring the
Crab Nebula over a large range of zenith angles, and for sev-
eral years. The observed integrated flux above 100 GeV has
shown no evidence of any variations beyond what is expected
from statistics. We can thus conclude that the observed signal
is indeed an isolated flare from 3C 279.
0 10 2030 4050 60 7080 90
Observation time: 82 mins
Excess events: 166
Excess significance: 6.15
Fig. S1. Histogram of the image parameter Alpha. The image pa-
rameter Alpha obtained for 3C 279 in the night of 23 February 2006,
for “on” (red dots) and “off” (crosses and filled area) pointings, re-
spectively. The dotted line is a simple parabolic fit (without linear
term) to the “off” events between 0◦and 80◦, serving to smoothen
the background for better extrapolation towards Alpha=0; the ex-
cess events are obtained with respect to this line.
The reconstruction of gamma ray energy uses an approxi-
mation derived from image parameters. It is obtained by the
the Random Forest method using Monte Carlo gamma sam-
ples. For the reconstruction of the energy spectrum, we used a
low-energy analysis, making also use of the timing properties
of the air showers with a lower energy threshold (≈110 GeV).
The cuts for obtaining the spectrum were slightly loosened, re-
sulting in a gamma efficiency of 50%. The obtained spectrum
The MAGIC Collaboration
was unfoldedfor detector effects using (14). The spectrum was
cross-checked with two alternative methods (one of them is the
standard analysis and a second one also uses of the timing in-
formation). All three spectra agree well, as shown in the figure
(main paper) as grey area, the envelope shape for 1σ errors of
the spectra. The predicted energy resolution is ≈23%, reason-
ably constant above 150 GeV. Systematic errors of our detector
(from atmospheric density and transmission uncertainties, the
photon detection efficiency, non-linearity in the signal chain,
are discussed in detail in (15). Uncertainties result in a possible
systematic energy shift of up to ±20%, which translates into a
shift of ∓25% in the limit set for the gamma ray horizon (en-
ergy for which the optical depth τ = 1). The uncertainty in the
spectral slope coefficient is ±0.2.
Attenuation of gamma rays from distant sources and EBL
Absorption of high-energy gammas by the extragalactic back-
ground light (EBL) depends on the column density of low-
energy photons traversed by the gamma rays.
rays of observed energy Eγ, the maximal cross section for
pair production at redshift z occurs at wavelength λ
1.24µm(Eγ/1TeV)(1 + z)2. Folded with the shape of the
cross section, MAGIC’s energy range specifically probes (with
3C 279) backgroundwavelengths of 0.2 µm−2 µm. This high-
lights the importance of the ultraviolet-to-near-infrared wave-
length range in modifying the spectra of sources at gigaparsec
distances (16). Knowledgeofthe amountof EBL alongthe line
of sight to gamma ray sources is based on bolometric measure-
ments and galaxy counts. Bolometric measurements are gener-
ally hamperedby foregroundemission in our own Galaxy (17).
Since the light emittedfromgalaxies changeswith time follow-
ing the star formation history (18) and with the ageing stellar
populations, the EBL has changed over time, too. An in-depth
review can be found in (19).
Various models for the EBL have been published. In (20),
several parameterized models using different assumptions for
the evolving EBL are introduced, taking into account magnetic
fields, gas and dust, the relative abundance of elements heavier
than Helium, and the star formation rate. A range of parame-
ters is proposed such that the results agree with what is known
from deep galaxy surveys and infrared background measure-
ments available in 2002. In a recent publication (21), two dif-
ferent backward-evolution EBL models were used to calculate
the intergalactic photon density from z = 0 to z = 6 (called
“Base” and ‘Fast” evolution). In this report we use the “Fast”
evolution model as predicting maximally high absorption. In a
different publication (22), a semi-analytical forward-evolution
approach was used to obtain the evolving EBL density. This
modelpredictsaratherlow EBLlevelatz = 0, justontopofthe
galaxy counts obtained from the Hubble Space Telescope (18).
In fact, the model (22) predicts the EBL level at wavelengths
above 5 µm slightly below the galaxy counts obtained from the
Spitzer (23,24,25) data and the ISOCAM results (26,27). The
model (22) also lies below the FIRAS measurements in the far-
infrared, an undershoot attributed to the abundance of Ultra-
Luminous Infrared Red Galaxies (ULIRG), only recently dis-
covered and not taken into account in the model. However, the
EBL domain above 3 µm does not influence our result, given
that the highest energy measured by MAGIC in the 3C 279
spectrum is below 500 GeV; EBL photons with λ > 2.5 µm do
not reach the energy necessary for pair production. We have,
therefore, chosen the model (22) as the low limit for the EBL
level in the frequency range relevant for the MAGIC 3C 279
In the discussion of the gamma ray horizon in the main pa-
per, we have used a model tuned to give for 3C 279 the intrin-
sic photon index α∗= 1.5. We call this ad-hoc EBL model
maxEBL and give more details in the following. It is based
on (20), with modified parameters. The ”best-fit” model, de-
fined in (28), is changed by reducing the optical and infrared
star formationrate, byaddinga ”warmdust” componentandby
taking into account a high fraction of UV emission, which fits
the data derived from AGN absorption lines, a method called
proximity effect [e.g., (29)]. Additionally, the dust parameter
for the optical galaxy component is set to a very low value of
E(B-V) = 0.01, to account for the maximum possible optical
flux. In detail, the parameters used [definition in (20)] are the
SFROPT: α = 3.5,β = −1.2,zp= 1.2, ˙ ρ∗(zp) = 0.08,
SFRLIG: α = 4.5,β = 0.0,zp= 1.0, ˙ ρ∗(zp) = 0.09,
and further fesc = 4,c2 = 10−23.4,E(B − V)OPT = 0.01. Con-
trary to the ”best-fit” EBL, the maxEBL model is above the
emissivity data at wavelengths between 0.16 – 1.0 µm [ (20)
and references therein].
All EBL models mentioned contain the evolution of the ra-
diation contributing to the EBL at different redshifts, and use
the Lambda-CDM Universe with Ωλ = 0.7 and Ωm = 0.3,
confirmed by the latest results from the Wilkinson Microwave
Anisotropy Probe (WMAP) (30). The models are graphically
summarized in Fig. S2, along with a number of measurements.
Stecker et al., 2006, (fast evolution)
max-EBL (this work)
Primack et al., 2005
Cosmic Microwave Background
Mazin & Raue, upper limit 2007
HESS upper limit, 2006
Fig. S2. EBL Models. Some of the models for EBL, for z=0, and
measurements at various wavelengths. The green line is based
on (20), with parameters adapted to the results from the latest
galaxy counts. The dotted line is from (31). The shaded verti-
cal band indicates the range of frequencies corresponding to the
VHE Gamma-Rays from a Distant Quasar
Comparing the effect of different EBL models on our mea-
sured spectrum we add the following comments:
• The high model (21) leads to an intrinsic spectrum of
3C 279 with a (fitted) index of α∗= 0.5 ± 1.2 (see main
paper), a rise in the VHE gamma ray spectrum very diffi-
cult to reconcile with extrapolationfrom EGRET data and
• The maxEBL model has been derived to be in agreement
with an intrinsic blazar spectral index of α∗= 1.5, and
can be considered as an upper limit on the EBL density
respecting the constraints from the measured VHE blazar
spectra (16,31). Note that the values for the optical and
infrared star formation rate (SFR) and for the optical dust
parameterareclose to their lowerlimits, andnoAGN con-
tribution has been taken into account.
• The low EBL derived from (22) apparently gives an ac-
ceptable result, resulting in a most probable intrinsic
power law with a spectral index α∗= 2.9 ± 0.9 (see main
Some of the models discussed above seem less likely in the
light of the present measurements, requiring the sources to
exhibit a redshift-dependent hardening of their spectra. We
should also note that the EBL density may be higher than de-
rived here, if VHE photons are speculated to undergo interac-
tions beyond standard physics [e.g., (32)].
Spectral Energy Distribution
Blazars represent the most violent variant of the AGN phe-
nomenon,relatedtotheactivityofsupermassive(M = 108−109
solar masses) black holes harbored in the center of galaxies.
Their emission, extending from radio wavelengths to high-
energy gamma rays, is produced within a jet of plasma ejected
at relativistic speeds and pointing close to the Earth.
The spectral energy distribution (SED) of blazars is charac-
terized by two broad bumps, the first one peaking between the
infrared and the X-ray band, the second one in the gamma ray
domain. The first peak traces the synchrotron emission of rela-
tivistic electrons spiraling along the lines of the magnetic field
in the jets; the high-energy peak is presumably due to the in-
verse Compton scattering between these relativistic electrons
and low-energy photons. The latter can be the synchrotron
anism (34,35)]. Ambient radiation can enter the jet either from
an accretion disk or from surrounding gas clouds. These mod-
els ultimately rely on electrons (or pairs) accelerated within the
flow and thus are called leptonic models.
Other possibilities (hadronic models) involve gamma ray
emission due to accelerated protons and ions. Such emission
can result from pion production in hadronic interactions, pair
production,or be due to synchrotronradiation [e.g., (36,37,38,
39)], and includes the associated production of neutrinos and
ultrahigh-energy cosmic rays (40). Although the energy losses
of protons and ions are suppressed by their large mass and the
low matter density in the acceleration zone, they become size-
able at ultrahigh energies where photo-productionof pions can
take place in collisions with low-energy synchrotron photons.
Until now the majority of known extragalactic sources of
VHE photons belongs to the class of BL-Lac objects, whose
spectra show a weak signature of thermal radiation from out-
side the jet and hence favor the SSC scenario.
[mostly Flat Spectrum Radio Quasars (FSRQs)], such as
3C 279the situationwith thermalradiationis different(theblue
bumpinthe spectrum),andtheEC process is likely todominate
if the gamma ray emission zone is close to the origin of the jet.
The validity of the EC mechanism is not undisputed, however,
the central source of thermal radiation [e.g., (41,42)].
Hadronic emission models require much larger magnetic
field strengths than leptonic models, in order to confine the
more massive and energetic particles to the jet. They gener-
ically produce harder VHE spectra than leptonic models. In
low-peaked blazars at high redshifts such as FSRQs, however,
the spectrum is steepened due to attenuation by the EBL, ren-
dering differences in the spectra difficult to detect. 3C 279 is
amongthe nearest FSRQs, and the energythreshold of MAGIC
is low enough to permit observing this source.
The EGRET gamma ray telescope (operating in the past
decade and sensitive between 20 MeV and few GeV) observed
at several occasions bright and hard (photon index around 2)
emission from 3C 279. In particular, during an extensive cam-
paign held in 1996 the source showed a large (a factor 10 in
amplitude) and short (∼ 2-3 days) gamma ray flare (4) with
episodes of relatively fast variability, including an increase by
a factor of 2.6 in ∼8 hours. Naively extrapolating this behavior
to the MAGIC band one could expect bright and short episodes
also in this band. However, in the standard leptonic framework
the detection of photons with energy above 100 GeV emitted
by 3C 279 is somewhat surprising. Indeed, a careful consid-
eration of the conditions expected in the source raises several
problems, related to the high energy required for the electrons
to emit photons with such high energy and the strong opacity
to gamma rays expected inside these sources. The detection of
strong VHE emission from 3C 279 thus is stretching leptonic
models to or beyond their limits.
The simplest EC model (“one-zone”) assumes that the bulk of
the emission is produced within a single region. In order to
properly model the emission, good coverage of the entire SED
is required, since only with known position and luminosity of
both peaks is it possible to fully constrain the physical param-
eters [e.g., (43)]. In the case of the 3C 279 flare, we only have
the simultaneous measurement in the optical R–band with the
VHE spectrum derived by MAGIC. With this input we tried to
reproducethe observed VHE spectrum assuming a simple one-
zone model [see (44) for a full description].
The MAGIC observations contribute data to the important
energy range above 1025Hz, poorly studied thus far because ly-
ing above the range of satellite-borne instruments and below
that of most ground-based Cherenkov telescopes. The SED in
The MAGIC Collaboration
Fig. S3 shows the MAGIC points corrected according to the
two extreme EBL models discussed in the paper, together with
measurements at lower frequencies from different epochs. In
the same figure we show the generic one-zone model described
here. The model uses the following assumptions and param-
eters: The source is spherical with radius R = 5 × 1016cm,
and it is in motion with a bulk Lorentz factor Γ = 20 at an an-
gle θ = Γ−1= 2.9 deg with respect to the line of sight. The
magnetic field has an assumed intensity B = 0.15 G. The par-
ticles, with a total density n = 2 × 104cm−3, follow an energy
distribution represented by a broken power law extending from
γ1= 1 to γ2= 3 × 105, with indices n1= 2 and n2= 3.7 below
and above the break at γp= 2.5 × 103. We model the external
radiation field assuming that a fraction τ = 0.005 of the disk
emission, a black body with luminosity 6 × 1045erg/s, is di-
luted into the Broad Line Region, a sphere with radius 4×1017
Fig. S3. Spectral energy distribution for 3C 279. Observations
at different source intensities, over the years from 1991 to 2003,
from (45) and (46), with MAGIC points (2006) at the high-energy
end (right). The MAGIC points are corrected according to the two
extreme EBL models of (22) and (21) (blue and red triangles, re-
spectively); the former clearly gives a more probable result with a
spectral index of α∗= 2.9 ± 0.9. An emission model and its indi-
vidual components are also shown: the solid line represents the
total model emission, fitted only to the (blue) MAGIC points and the
(black) triangle from the KVA telescope. The individual components
of the emission model are synchrotron radiation (dotted line), disk
emission (long-dashed), synchrotron-self-Compton (short-dashed),
and external Compton (dot-dashed).
Accounting for attenuation of the gamma rays due to pair pro-
duction in collisions with photons from the integrated light of
the dominant radiation process. The emitted spectrum also ap-
pearsto beverysimilar to the spectra of manyhigh-peakedBL-
Lac objects, although their corresponding synchrotron peak is
at much higher energies. In the frameworkof inverse-Compton
scattering models this implies that external photon fields may
act as scattering target, although seemingly not to the extent to
create internal opacity due to pair production in photon-photon
collisions (47). The inverse-Compton models predict a steep
decline for two reasons: for one, the spectrum of the highest
energyelectrons, judgingfrom the synchrotronspectrum, tends
to be steep; and second, VHE electrons scatter with ambient
photons in the Klein-Nishina regime, reducing the scattering
efficiency. An EBL larger than the integrated light of known
galaxies would imply that the generic leptonic emission mod-
els must undergo a major revision, perhaps involving proton
synchrotron and cascade emission. The latter would shed new
light on the correlation between active galactic nuclei and the
directions of the highest energy cosmic rays (48).
1. A. Cavaliere, V. D’Elia, Astrophys. J. 571, 226 (2002).
2. R. C. Hartman et al., Astrophys. J. 385, L1 (1992).
3. L. Maraschi et al., Astrophys. J. 435, L91 (1994).
4. A. E. Wehrle et al., Astrophys. J. 497, 178 (1998).
5. U. Bach et al., Astron. Astrophys. 464, 175 (2007).
6. W. Collmar et al., Multifrequency observations of the blazar
3C 279 in January 2006, to be published in Proc. of the 6th
INTEGRAL workshop “The Obscured Universe” (Moscow, July
2-8, 2006), eds. S. Grebenev, R. Sunyaev, C. Winkler, ESA SP
622 (2006) (preprint arXiv:0710.1096).
7. R. C. Hartman et al., Astrophys. J. 553, 683 (2001).
8. M. Boettcher et al., Astrophys. J. 670, 968 (2007).
9. M. Gaug et al. (MAGIC Collab.), Calibration of the MAGIC Tele-
scope, Proc. of the 29th ICRC, Pune, 5-375, astro-ph/0508274
10. T. Bretz et al. (MAGIC Collab.), Standard Analysis for the
MAGIC Telescope, Proc. of the 29th ICRC, Pune, 5-315,
11. A. M. Hillas, in Proc. 19th International Cosmic Ray Confer-
ence, La Jolla (1985).
12. J. Albert et al., Nucl. Inst. Meth. A 588, 424 (2008).
13. T. Li, Y. Ma, ApJ 272,317 (1983).
14. J. Albert et al., Nucl. Inst. Meth. A 583, 494 (2007).
15. J. Albert et al., Astrophys. J. 674, 1037 (2007).
16. F. Aharonian et al., Nature 440, 1018 (2006).
17. K. Mattila, Mon. Not. R. Astron. Soc. 372, 1253 (2006).
18. P. Madau, L. Pozzetti, Mon. Not. R. Astron. Soc. 312 L9 (2000).
19. M. G. Hauser, E. Dwek, Ann. Rev. Astron. Astrophys. 39, 307
VHE Gamma-Rays from a Distant Quasar Download full-text
20. T. M. Kneiske, K. Mannheim, D. H. Hartmann, Astron. Astro-
phys. 386, 1 (2002).
21. F. W. Stecker, M. A. Malkan, S. T. Scully, Astrophys. J. 648, 774
22. J. R. Primack, J. S. Bullock, R. S. Somerville, Observational
Gamma Ray Cosmology, In: High Energy Gamma Ray Astron-
omy, volume 745 of American Institute of Physics Conference
Series, Eds. Aharonian, Voelk, and Horns, page 23 (2005).
23. G. Fazio et al., Astrophys. J. Suppl. Ser. 154, 39 (2004).
24. C. Papovich et al. Astron. Astrophys. J. Suppl. Ser. 154, 70
25. H. Dole et al., Astron. Astrophys. 451,417 (2006).
26. T. S. Metcalfe et al., Astron. Astrophys. 407, 791 (2003).
27. D. Elbaz et al., Astron. Astrophys. 381, 848 (2002).
28. T. M. Kneiske, T. Bretz, K. Mannheim, D. H. Hartmann, Astron.
Astrophys. 413, 807 (2004).
29. J. Scott et al., Astrophys. J. Suppl. 130, 67 (2000).
30. D. N. Spergel et al., Astrophys. J. Suppl. Ser. 148, 175 (2003).
31. D. Mazin, M. Raue, Astron. Astrophys. 471, 439 (2007).
32. A. De Angelis, M. Roncadelli, O. Mansutti, Phys. Rev. D 76,
33. L. Maraschi, G. Ghisellini, A. Celotti, Astrophys. J. 397, L5
34. C. D. Dermer, R. Schlickeiser, Astrophys. J. 416, 458 (1993).
35. M. Sikora, M. C. Begelman, M. J. Rees, Astrophys. J. 421, 153
36. K. Mannheim, Astron. Astrophys. 269, 67 (1993).
37. K. Mannheim, P. L. Biermann, Astron. Astrophys. 251,L 21
38. F. A. Aharonian, New Astronomy 5, 377 (2000).
39. A. M¨ ucke, R. J. Protheroe, R. Engel, J. P. Rachen, T. Stanev,
Astroparticle Physics 18, 593 (2003).
40. K. Mannheim, T. Stanev, P. L. Biermann, Astron.Astrophys. 260,
41. E. J. Lindfors, E. Valtaoja, M. T¨ urler, Astron. Astrophys. 440,
42. E. J. Lindfors et al., Astron. Astrophys. 456, 895 (2006).
43. F. Tavecchio, L. Maraschi, G. Ghisellini, Astrophys. J. 509, 608
44. L. Maraschi, F. Tavecchio, Astrophys. J. 593, 667 (2003).
45. L. Ballo et al., Spectral Energy Distributions of 3C 279 Revis-
ited, Astrophys. J. 567, 50 and references therein (2002).
46. W. Collmar et al., Proceedings of the 5th INTEGRAL Workshop
on the INTEGRAL Universe (ESA SP-552), 16-20 February
2004, Munich, Germany. p.555, (2004).
47. A. Reimer, Astrophys. J. 665, 1023 (2007).
48. J. Abraham et al. (The Pierre Auger Collaboration), Science
318, 938 (2007).