University of Wisconsin - Madison MAD/PH/847

August 1994

Particle Astrophysics
with High Energy Neutrinos
Thomas K. Gaissera , Francis Halzenb and Todor Staneva
arXiv:hep-ph/9410384v1 26 Oct 1994

a
Bartol Research Institute, University of Delaware, Newark, DE 19716, USA
b
Department of Physics, University of Wisconsin, Madison, WI 53706, USA

Abstract
The topic of this review is the particle astrophysics of high energy neutrinos. High
energy is defined as Eν > 100 MeV. Main topics include:
– atmospheric neutrinos and muons from π, K and charm decay. They probe uncharted
territory in neutrino oscillations and constitute both the background and calibration
of high energy neutrino telescopes,
– sources of high energy neutrino beams: the galactic plane, the sun, X-ray binaries,
supernova remnants and interactions of extra-galactic cosmic rays with background
photons,
– an extensive review of the mechanisms by which active galaxies may produce high
energy particle beams,
– high energy neutrino signatures of cold dark matter and,
– a brief review of detection techniques (water and ice Cherenkov detectors, surface de-
tectors, radio- and acoustic detectors, horizontal airshower arrays) and the instruments
under construction.
1 Introduction
The scope of this paper is neutrino astronomy for Eν > 100 MeV. Our main interest is
neutrinos from energetic astrophysical sources such as binary stars and accreting black holes
in Active Galactic Nuclei (AGN). We will also discuss atmospheric neutrinos at some length
because they are the only neutrinos with E > 100 MeV that have yet been detected. At-
mospheric neutrinos are both background and calibration beam for high energy neutrino
astronomy. They are of interest in their own right because they probe an uncharted range of
neutrino oscillation parameter space. For stellar collapse neutrinos and solar neutrinos we
refer the reader to the recent review of Totsuka [1] and, for solar neutrinos to the reviews of
Bahcall et al. [2,3] and Turck-Chieze et al. [4].
In his classic review of cosmic ray showers in 1960 [5], Greisen ends with a discussion of
the prospects for gamma ray and neutrino astronomy at very high energy. He notes that
“Since photons and neutrinos propagate in straight lines, success in their detection will open
up broad new areas of astronomy.” He discussed the relation between photons and neutrinos
from decay of pions produced in the interstellar medium or near a source. He described a
detector very much like the present water Cherenkov detectors (complete with veto shield
against entering muons), and he estimated the rate of interactions of atmospheric neutrinos
in 3 kilotons of sensitive volume to be 500 events per year.
The idea of detecting neutrinos by looking for neutrino-induced upward or horizontal
muons was suggested by Markov & Zheleznykh [6] at about the same time. The process is

νµ + N → µ + anything, (1)

where N is a nucleon in the material surrounding the detector. The muon range increases
with energy. This extends the effective target volume and makes it possible to see neutrino-
induced muons with detectors of moderate size. Two groups (Kolar Gold Fields [7,8] and
Case-Wittwatersrand [9,10]) reported the first observations of atmospheric neutrinos with
the detection of horizontal muons in detectors so deep that the muons could not have been
produced in the atmosphere.
Atmospheric neutrinos are of current interest, despite their long history and apparently
mundane origin, because of the anomalous flavor ratio observed for neutrino interactions
in the large proton decay detectors, Kamiokande [11] and IMB [12]. The essential point is
that, because of their large volume, these detectors can measure interactions of neutrinos
inside the detector. They need not depend on the large external target mass provided by
the long range of energetic muons produced in charged current interactions of muon-type
neutrinos. They can therefore study both νe and νµ interactions. In all, more than a
thousand atmospheric neutrino events have now been measured by the various underground
experiments[11,12,13,14,15,16] The anomaly is that the observed ratio of events produced
by electron neutrinos to those from muon neutrinos is significantly larger than expected.
Several new detectors designed specifically for high energy neutrino astronomy are about
to come into operation. The Baikal experiment [17] has already reported some muon mea-
surements [18]. The DUMAND [19] and AMANDA [20,21] are being partially deployed at
present, and other detectors, such as NESTOR [22] are in advanced prototype stages. A
major stimulus for this activity is the prospect that Active Galactic Nuclei (AGN) may be

2
prolific particle accelerators and beam dumps, and therefore intense sources of high energy
neutrinos.
We have divided our review into three major sections: (a) atmospheric neutrinos, (b)
possible sources of high energy neutrinos of extraterrestrial origin and (c) neutrino detection.
We include some comments about high energy gamma ray astronomy relevant to possible
neutrino sources at the beginning of part (b). We begin with a brief treatment of neutrino
production in cosmic ray cascades, which is relevant both for atmospheric and astrophysical
neutrinos.

2 Neutrino production
Unlike the typical monoenergetic beam produced by a machine, cosmic accelerators produce
power law spectra of ions at high energy,

φp ∝ E −(γ+1) . (2)

The observed high energy cosmic ray spectrum at Earth is characterized by γ ∼ 1.7. In
general, a cosmic accelerator in which the dominant mechanism is first order diffusive shock
acceleration (first order Fermi mechanism), will produce a spectrum with γ ∼ 1 + ǫ, where
ǫ is a small number. The observed spectrum is thought to be steeper than the accelerated
spectrum because of the energy dependence of the cosmic ray diffusion in the galaxy. The
simplest way to understand this is to think of the observer as inside a volume of “contain-
ment” from which the characteristic escape time decreases with energy

τ (E) ∝ E −δ . (3)

If Q(E) is the rate of production of cosmic rays per unit volume, then the observed cosmic
ray density will be
ρCR (E) ∼ Q(E) × τ (E) ∝ E −(2+ǫ+δ) . (4)
For 1 ≤ E ≤ 100 GeV, a value of δ ∼ 0.6 can be inferred from observed ratios of secondary
cosmic ray nuclei (e.g. Li, Be, B) to their progenitors (e.g. carbon and oxygen) [23].
Production of secondary particles (S) is related to the spectrum of accelerated primaries
(P ) by
dPS ∆ ∞ dnP S (ES , EP )
Z
= φP (EP )dEP , (5)
dES λP E S dES
where ∆/λP is the probability of interaction in traversing a small amount (∆) of target. If
the distribution of secondaries depends only on the ratio of energies, x = ES /EP , then the
integral in Eq. (5) becomes
Z 1
φP (ES ) xγ−1 FP S (x) dx ≡ φP (ES ) ZP S , (6)
0

where
1 dσP S dnP S
Z
FP S = d2 pT ES 3
≈ ES . (7)
σ dp dES

3
For γ > 1, F (0) does not contribute to the integral, and the scaling approximation made
here is an adequate approximation for rough estimates. This treatment generalizes readily
to decay chains, such as p → π ± → µ± , etc., and it can be applied to cascades in galactic
and stellar environments as well as in the Earth’s atmosphere. In case of a thick target in
which the primary beam is fully attenuated, the production spectrum of secondaries is given
by Eq. (5) with the replacement
∆ ΛP
→ , (8)
λP λP
where ΛP is the attenuation length of the primary.
In general, the flux of neutrinos from decay of pions is given by [24]
( )
dNν N0 (Eν ) Aπν
= × + (· · ·) , (9)
dEν 1 − ZN N 1 + Bπν cosϑ Eν /ǫπ
where Aπν = ZN π (1 − rπ )γ /(γ + 1), rπ = (mµ /mπ )2 and Bπν is a constant that depends on
nucleon and pion attenuation lengths. The first term inside the curly brackets represents
neutrinos from decay of pions. The energy ǫπ is a characteristic energy that reflects the
competition between decay and interaction in the medium. For cascade development in
the Earth’s atmosphere ǫπ ∼ 115 GeV; it is larger for more tenuous media, such as the
atmosphere of the Sun.
The (· · ·) represents the contributions of other mesons, with ǫK ± ∼ 850 GeV and ǫD± ∼
4 × 107 GeV in the Earth’s atmosphere. Each term also contains the appropriate branching
ratio, e.g. 0.635 for K ± → µ + νµ . For Eν ≪ ǫi , all parent mesons decay, and the neutrino
spectrum is parallel to the primary nucleon flux. For Eν ≫ ǫi the neutrino spectrum steepens
by one power of Eν . In the atmosphere, muons with Eµ ≫ µ c2 × 15 km/cτµ ∼ 2 GeV reach
the surface and stop before they decay.
In the energy range important for contained events (0.1 < Eν < 2 GeV) decay in flight
of atmospheric muons is the dominant source of νe , and an important source of νµ . At much
higher energy, the dominant source of atmospheric νe is

KL0 → π e νe , (10)
The relative contributions of the various decay modes to lepton spectra in the Earth’s at-
mosphere are discussed in detail by Lipari [25]. If we consider neutrino production in as-
trophysical settings, with typical matter densities of 1010 atoms/cm3 , pions and kaons will
always decay and the neutrino spectrum will follow the nucleon spectrum also at very high
energy.
In addition, muon decay will continue to be an important source of neutrinos at high
energy as well as low. Thus we can define three types of neutrino spectra that can be
produced by cosmic rays:
1. Atmospheric neutrinos, which follow the incident cosmic ray spectrum with γ ∼ 2.7
up to ∼100 GeV and steepen toward γ ∼ 3.7 at higher energy. The position of the
bend increases with increasing zenith angle, which generates a characteristic angular
dependence of the atmospheric neutrino spectrum described by the angular factor in
Eq. (9). Electron neutrinos, that come mostly from muon decays have a spectrum with
one power of E steeper.

4
 2. Neutrinos produced by galactic cosmic rays in interactions with interstellar gas. These
extraterrestrial neutrinos should follow the cosmic ray spectrum up to the highest
energies, since all interaction products, including muons, decay.
3. Neutrinos produced by cosmic rays at their acceleration sites and following the hard
(γ ∼ 2.0–2.2) cosmic ray source spectra, which are not yet affected by the energy-
dependent escape from the Galaxy.
These three types of neutrino fluxes are illustrated schematically in Fig. 1.
Atmospheric neutrinos have been detected and studied extensively. Diffuse galactic neu-
trinos should exist with intensities comparable to the diffuse galactic gamma ray back-
ground [26]. They should be detected by a future generation of detectors of sufficient size to
see neutrinos at a rate of several per 105 m2 per year [27]. These two fluxes can be used for
calibration of high energy neutrino telescopes, which have as their principal goal the search
for high energy neutrinos from energetic astrophysical systems.
The existence of neutrinos associated with cosmic ray sources is more problematic. It
requires substantial acceleration in compact sources with sufficient local gas to act as a beam
dump. The possibility (not certainty) of such point sources is suggested by the fact that the
standard model of cosmic ray acceleration by supernova blast waves in the diffuse interstellar
medium cannot accelerate particles to the highest observed energies. An alternative for the
higher energy cosmic radiation is acceleration in compact sources. The argument goes as
follows.
First order Fermi acceleration at supernova blast shocks offers a very attractive model
for a galactic acceleration mechanism, providing about the right power and spectral shape.
Shock acceleration takes time, however, because the energy gain occurs gradually as a particle
diffuses back and forth across the shock front. The finite lifetime of the shock thus limits
the maximum energy per particle that can be achieved at a particular supernova. The
acceleration rate is
dE u2 u
≃K Z e B, so Emax < Z e B L , (11)
dt c c
where u is the shock velocity, Z e the charge of the particle being accelerated and B the
ambient magnetic field. The numerical constant K ∼ 0.1 depends on the details of diffusion
in the vicinity of the shock. The crucial length scale in Eq. (11) is given by L ∼ u T , where
T ∼ 1000 yrs for the free expansion phase of a supernova. Using this kind of argument,
Lagage & Cesarsky [28] show that, in its simplest version (shock velocity parallel to magnetic
field direction, B = BISM ∼ 3µGauss) Emax can only reach energies ∼ < 1014 eV × Z for an
accelerated nucleus. That leaves a large gap of some three orders of magnitude that cannot
be explained by the “standard model” of cosmic ray origin. To reach a higher energy one
has to increase significantly B and/or L.
One possibility is to use the much higher magnetic fields associated with some energetic
astrophysical systems. There is no shortage of such objects in the Galaxy, the most obvious
being neutron stars and black holes, as well as young supernova remnants. Some examples of
possible acceleration sites will be discussed in Section 6 below. Any such compact region with
active particle acceleration would be a likely site for production of high energy neutrinos and
photons through interactions of the accelerated particles with the ambient gas and radiation
fields.

5
 We emphasize that this is not the only possibility. Some argue [29,30] that explaining
the higher energy cosmic rays by a new source is unnatural because it requires fine tuning to
produce a smooth spectrum where cosmic rays from the second source join onto those from
the first.
There are several ways to extend the basic supernova mechanism to higher energies. One
possibility [31] takes advantage of the fact that some supernovas explode into the stellar
wind of a progenitor star rather than the interstellar medium. If the progenitor wind carries
a high enough magnetic field, then higher top energies can be achieved (see Eq. 11). Other
possibilities involve a configuration in which the magnetic field is quasi-perpendicular to the
shock normal [32] or the interaction of high energy cosmic rays with expanding shocks of
several supernovas in an active region [30]. Mechanisms such as these would not be correlated
with point sources of high energy gamma rays and neutrinos.
Perhaps the most exciting possibility at present is the suggestion [33,34] that particle
acceleration plays a central role inside AGN and that interactions of these high energy
particles with dense photon fields and gas in the central regions of AGN [35,36,37,38] will
lead to production of neutrinos of very high energies. This possibility is the subject of §6.
Figure 1 illustrates the window of opportunity for high energy neutrino astronomy. The
steepening at high energy of the spectrum of atmospheric neutrinos, which dominate the total
neutrino flux at low energy, allows the possibility of reasonable signal/background ratios,
shown schematically by the shaded area on Fig. 1. The small angle between the parent
neutrino and the secondary lepton in charged current interactions at high energy allows for
accurate source location and enhanced signal/background ratio in the case of point sources.
The exact position of the crossover from atmospheric to astrophysical neutrinos, depends on
luminosity, distribution and distances of potential sources and on energy response and, in
the case of point sources, angular resolution of the detectors. These will be discussed for a
number of sources in Sections 6–8 below.

3 Atmospheric neutrinos
The atmospheric cascade that produces the cosmic ray neutrino beam is depicted in Eq. (12):

p −→ π + ( + K + . . .) −→ µ+ + νµ
|→ e+ + ν̄ + ν ,
µ e
(12)
− − −
n −→ π ( + K . . .) −→ µ + ν̄µ
|→ e− + ν + ν̄ .
µ e

Protons also produce negative mesons and neutrons produce positive mesons, but the same
charge processes are slightly favored by the steep spectra and the excess of same-charge
mesons in the forward fragmentation region.
Although analytic expressions like Eq. (9) are qualitatively correct, more detailed calcu-
lations are needed for a precise evaluation of the atmospheric neutrino flux. Several compli-
cations must be accounted for:

6
 < 10 GeV
• The primary cosmic ray spectrum is not a simple power law, especially ∼
>
and ∼ 100 TeV. Moreover, it depends on location and direction (because of the
geomagnetic cutoff) and on the epoch of the solar cycle.
• Muon energy loss, decay and polarization must be accounted for.
• The inclusive cross sections do not have exactly scale-invariant forms. Furthermore,
nuclei, as well as nucleons, are involved in the collisions.

3.1 Contained Events

Contained events are those neutrino interactions that originate within the detector’s fiducial
volume and whose interaction products are all contained within that volume. Most such
events have lepton energies in the GeV range or less. In this case, most muons decay, and
all the complications listed above come into play.
Qualitative expectations for the neutrino ratios follow from simple kinematics of the
π → µ → e decay chain. Because of the asymmetry of the π → µ + νµ decay mode, each of
the two neutrinos from muon decay has about the same energy as the neutrino from pion
decay. Thus in a given energy range
νe 1 ν̄µ ν¯e µ−
≈ , ≈ 1, and ≈ + < 1. (13)
νµ 2 νµ νe µ
The excess of νe to ν̄e is a consequence of the excess of protons to neutrons in the incident
cosmic ray beam.
Several detailed calculations [39,40,41,42] of the ∼GeV neutrino fluxes agree with each
other in finding a value for the neutrino flavor ratio within 5% of each other. Attention is
focussed on the flavor ratio because most of the sources of uncertainty in the calculation of
neutrino fluxes cancel in calculating the ratio. (These uncertainties include normalization of
primary spectrum, treatment of geomagnetic effects and parameterization of pion production
in collisions of cosmic ray protons and helium with nuclei of the atmosphere.) Specifically,
the neutrino flux calculations give

νe + 13 ν̄e
Re/µ = = 0.49 ± 0.01 (14)
νµ + 31 ν̄µ
for 0.1 < Eν < 1 GeV [43]. The fact that this expectation is apparently significantly
violated [11,12] is largely responsible for the great interest in atmospheric neutrinos.
The water Cherenkov detectors [11,12] in fact see as many [11] or more [12] electrons
than muons. Part of the difference between expectation and observation is a consequence
of the fact that the acceptance of the detectors spans a larger energy range for electrons
than for muons. Even after accounting for flavor-dependence of the acceptance, however, a
significant discrepancy remains.
Detailed simulations of the detector response, including acceptance effects, lead to the
result that the ratio of ratios for charged leptons from interactions of ν and ν̄ is
(µ/e)data
= 0.60 ± 0.06 ± 0.05 (15)
(µ/e)sim

7
for Kamiokande with 6.2 kT-yrs of data (389 contained, single-ring events) [44]. The corre-
sponding IMB result is

(µ/e)data
= 0.54 ± 0.03 ± 0.05 (16)
(µ/e)sim
for 7.7 kT-yrs of data (507 contained, single-ring events [12]). The water detectors thus show
at least a 4σ discrepancy between observation and expectation for the νµ /νe ratio.
Measurements with tracking calorimeters give mixed results, and the statistical uncer-
tainties are significantly larger than for the water detectors. NUSEX (0.74 kT-yrs, [15]) and
Frejus (1.56 kT-yrs, [14]) are both consistent with expectation. In contrast, Soudan 2 (1.0
kT-yrs, [16]) finds a ratio of ratios similar to Kamiokande. The number of events is relatively
low in all three experiments, and the systematic effects are also different.
In view of the complexity of the analysis involved, it is of interest to ask to what extent
the data from the two water detectors are consistent with each other. Beier et al. [45]
compared the IMB and Kamiokande data at stage when Kamiokande had 4.92 kT years of
data [11] and IMB had 3.4 kT years [46]. They found the data to be fully consistent between
the two experiments.
Table 1 shows a comparison based on the more complete data sets, 7.7 kT years for
IMB [12] and 6.2 kT years for Kamiokande [44]. The comparison is made by converting
Kamiokande data to IMB. Three factors are involved:
1) The larger exposure of IMB (7.7/6.2 = 1.24).
2) Higher muon threshold at IMB–the muon threshold is 300 MeV/c for IMB as compared
to 200 MeV/c for Kamiokande. Using the momentum spectra of Ref. [47], we estimate that
this reduces the Kamiokande muon rate by 0.79.
3) The lower geomagnetic cutoffs at IMB. Using the cutoff effects from Ref. [39], this increases
the muon rate by a factor 1.23 and the electron rate by 1.34. The larger correction factor
for electrons is a consequence of the fact that the electron threshold (100 MeV/c for both
experiments) is lower than the muon threshold, and the geomagnetic filter affects low energies
more than high.
These three correction factors are applied successively to the Kamiokande column to obtain
the converted Kamiokande numbers that may be compared directly with the IMB column.

Table 1: Comparison of contained event rates.

Kamiokande exposure µ-threshold geomagnetic IMB
muons 191 →237 →187 →230 182
electrons 198 →246 →246 →330 325
total 389 →483 →433 →560 507

After the conversion, the Kamiokande muon/electron ratio (230/330) appears somewhat
higher than for IMB (182/325). Another consideration that must be borne in mind is that
the second half of the IMB data was taken during a period of maximum solar modulation.

8
The event rate at IMB is expected to be somewhat lower during maximum solar modulation,
whereas Kamiokande, with its higher geomagnetic cutoff is much less sensitive to modulation
[39]. This effect should be the same for νµ and νe , however. Recently, Beier & Frank [48]
have pointed out that the momentum spectra of the electrons in the two experiments are also
somewhat different. Beam tests of the response of water Cherenkov detectors to electrons
and muons will soon be carried out at KEK. This should help to resolve questions about
the efficiency for discrimination between neutrino flavors as a systematic error in this type
of detector.
Another possible source of systematic error that has been pointed to is the cross section
for charged current interactions of neutrinos in nuclei. The Fermi gas model has been used for
calculation of the spectra of the produced charged leptons by both Kamiokande [49] and IMB
[50]. Recent calculations by Engel et al. [51] include several effects that go beyond the Fermi
gas model. They find no significant shift in the spectra of electrons relative to muons which
would distort the inferred νe /νµ ratio. In addition, there is some direct confirmation of the
Fermi gas model for Eν > 400 MeV in data discussed in Ref. [52]. Another experiment [53],
which appears to show an anomalous result for the muon spectrum in νµ + carbon → µ + . . .,
is in any case below the energy range of interest here (pµ > 200 MeV/c for Kamiokande and
pµ > 300 MeV/c for IMB).
Assuming that the problem with the contained events reflects an intrinsic property of
particle physics (rather than a lack of understanding of detector response or neutrino cross
sections), one needs to know whether there are too few νµ or too many νe . For example,
[54] with the calculation of Refs. [39,41], an interpretation in terms of neutrino oscillations
can be explained by νµ disappearance (e.g. νµ ↔ ντ oscillations), but not by νµ ↔ νe , which
would increase the predicted flux of νe . On the other hand, comparison of the data to the
calculations of Refs. [40,42] (which are some 30% lower than those of Refs. [39,41]) prefers
νµ ↔ νe in order to boost up the predicted νe flux as well as lower the νµ flux. Fogli et
al. [55] have recently reviewed the limits of various neutrino oscillation scenarios that could
explain the atmospheric neutrino anomaly.
A more exotic explanation is the suggestion that there is an excess of electrons due to pro-
ton decay in the mode p → e ν ν̄ [56]. This interpretation requires a calculated atmospheric
flux with a low normalization (so the muon rate is correctly predicted). The atmospheric
flux should also have a shape such that the deficit of electrons from atmospheric νe ’s occurs
preferentially at low energy. The deficit can then be filled in by the characteristic energy
spectrum of a three-body proton decay. The calculation of Ref. [42] has just these features.
A subset of the authors of Refs. [39,40,41,42] is investigating the source(s) of the dif-
ference among the normalization and shapes of the calculations. It appears that the main
cause of the characteristic shape and low normalization of the calculation of Ref. [42] is the
parameterization of pion production in collisions of protons with light nuclei [57].
The production of π ± with Eπ < 2 GeV is significantly lower in the parameterization of
Ref. [42] than in Refs. [39,41].
This uncertainty could be reduced by comparison to measurement of muon fluxes at high
altitude. Existing high altitude data [58,59,60] have large uncertainties, but they somewhat
favor the higher flux calculations. A new set of experiments [61] should be able to fix
the normalization to perhaps 10%. If the contained neutrino anomaly is due to νµ ↔
ντ oscillations, then the derived parameters suggest the effect should also show up as an

9
apparent deficit in the neutrino-induced upward muon sample. We return to this question
in the next section.

3.2 Upward Muons

Neutrino-induced muons are of interest for two reasons: first, they extend the study of the
atmospheric neutrino spectrum to higher energy, and second, they are the expected signal
of high energy astrophysical neutrinos. In this section we first discuss the relation between a
spectrum of neutrinos from any source and the muon flux that it produces. We then review
the current status of muons produced by atmospheric neutrinos including the extent to which
these measurements restrict the neutrino oscillation interpretations of the contained event
anomaly. In later sections of the paper we use the formulas of this section to discuss possible
signals of high energy astrophysical neutrinos.

3.2.1 Neutrino-induced muons

The detection of neutrino interactions inside the detector volume becomes more difficult for
higher energy neutrinos because of the steeply falling neutrino spectrum. It is possible to
enhance the effective volume of the detector by looking for muons generated in charged-
current interactions of νµ (ν¯µ ) in the rock below the detector. The effective detector volume
is then the product of the detector area and the muon range in rock Rµ . TeV muons have
a typical range of one kilometer in rock, which leads to a significant increase in effective
detector volume. The technique works only for muons entering the detector from below
or near the horizontal (upward going muons), because the the background of downward
atmospheric muons dominates any neutrino-induced signal from above.
The average muon energy loss rate is
* +
dE
= −α(E) − β(E) × E, (17)
dX

where X is the thickness of material in g/cm2 . The first term represents ionization losses,
which are approximately continuous, with α ∼ 2 MeV g−1 cm2 . The second term includes
the catastrophic processes of bremsstrahlung, pair production and nuclear interactions, for
which fluctuations play an essential role. Here β ∼ 4 × 10−6 g−1 cm2 .
The energy-dependent energy loss rates for each process are tabulated in Ref. [62]. The
critical energy above which the radiative processes dominate is

Ecr = α/β ≈ 500 GeV. (18)

To treat muon propagation properly when Eµ > Ecr requires a Monte Carlo calculation of
the probability Psurv that a muon of energy Eµ survives with energy > Eµmin after propagating
a distance X [63]. The probability that a neutrino of energy Eν on a trajectory through a
detector produces a muon above threshold at the detector is [64,63]
Eν dσν
Z
Pν (Eν , Eµmin ) = NA dEµ (Eµ , Eν ) Reff (Eµ , Eµmin ), (19)
0 dEµ

10
where Z ∞
Reff = dX Psurv (Eµ , Eµmin , X). (20)
0
The flux of νµ -induced muons at the detector is given by a convolution of the neutrino
spectrum φν with the muon production probability (19) as
Z
φµ (Eµmin, θ) = Pν (Eν , Eµmin ) exp[−σtot (Eν ) NA X(θ)] φν (Eν , θ). (21)
min
Eµ

The exponential factor here accounts for absorption of neutrinos along the chord of the
Earth, X(θ). Absorption becomes important for σ(Eν ) ∼ > 10−33 cm2 or Eν > 107 GeV.
∼
The event rate is now simply calculated by multiplying Eq. (21) with the effective area
of the detector. One can now tabulate Pν→µ for a given muon energy threshold and fold
it with fluxes of neutrinos of different origin to calculate the expected event rate. Figure 2
shows Pν for two values of muon threshold energy, 1 GeV and 1 TeV. The solid lines are for
ν and the dashed lines for ν̄.
There is some uncertainty in the calculation of the neutrino cross sections for Eν ≫
10 TeV because of the required extrapolations of the structure functions to small x ≪
10−4 . We have used the charged current cross section of Ref. [65]. For back-of-the-envelope
calculations Pν (Eν , 0) can be approximated by two power laws (shown by the straight lines
for the 1 GeV case in Fig. 2):

Pν→µ ≃ 1.3 × 10−6 E 2.2 for E = 10−3 –1 TeV

(22)
≃ 1.3 × 10−6 E 0.8 for E = 1–103 TeV

where E is in TeV. The two energy regimes directly reflect the energy dependence of the
neutrino cross section and the effective muon range in Eq. (19). σν ∼ E at low energy followed
by a slower energy dependence above 1 TeV reflecting the effect of the W propagator. The
muon range makes a transition from increasing linearly to constant behavior in a similar
energy range.

3.2.2 Atmospheric neutrinos above 1 GeV

The atmospheric neutrino spectrum is quite steep at high energy, approaching E −(γ+2) ≈
E −3.7 for E ≫ 1 TeV (see Eq. 9). For this reason, despite the increase of Pν with energy,
the contribution to the upward muon flux from atmospheric neutrinos with E > 10 TeV is
small. We illustrate this in Fig. 3, which shows the distribution of energies of atmospheric
neutrinos that gives rise to upward muons. In this example, “throughgoing” is defined as
Eµ > 4 GeV and “stopping” as 1 < Eµ < 4, and the fluxes are averaged over all angles below
the horizontal. The response curve for contained interactions is shown for comparison.
The upward going neutrino rate has been measured with significant statistical accuracy
by three experiments—IMB [66,67], Baksan [68] and Kamiokande [69]. Since the experi-
mental arrangements, biases and effective areas as a function of the zenith angle are quite
different, it is impossible to compare these results exactly. One can, however, scale the
quoted experimental Eµmin to a common value of 3 GeV, which is the effective threshold at

11
Kamiokande.1 The data are converted using Eq. 21 for each angular bin. A comparison
between angular distribution of upward muons, measured by the three experiments is shown
in Fig. 4. The quoted muon energy thresholds, measured rates averaged over the upward
hemisphere and the rates converted to a common threshold are given in Table 2. The units
are 10−13 cm−2 s−1 sr−1 . The numbers in the last column of Table 2 are to be compared with a
calculation of the upward muon flux due to atmospheric neutrinos. Calculated values range
from 1.95 to 2.36, depending on the neutrino cross section and muon flux used for the calcu-
lation [70]. Given the experimental errors and uncertainties in the input to the calculations,
the observed rates are consistent with expectation.

Table 2: Measured upward muon rates (data) shifted to a common threshold.

Experiment Eµmin (GeV) data shifted
Baksan [68] 1.0 2.77 ± 0.17 2.08
IMB [66,67] 2.0 2.26 ± 0.11 1.92
Kamioka [69] 3.0 2.04 ± 0.13 2.04

The question then arises whether the agreement between calculation and measurement
of upward muons is good enough to rule out some possible explanations of the contained
event anomaly in terms of neutrino oscillations. We first note that the “allowed region”
[11] of parameter space for explanation of the contained event anomaly in terms of νµ ↔ ντ
oscillations should lead to a significant depletion of the neutrino-induced muon flux.
For example, for δm2 = 8 × 10−3 eV2 , and L = 104 km (a typical propagation distance
for an upward neutrino that interacts below the detector) the first node of

Lkm
 
2 2 2
1 − Pνµ →νµ = Pνµ →ντ = sin 2θ sin 1.27 δm (23)
EGeV
is at E = 65 GeV. This energy is in the middle of the energy range important for generation
of neutrino-induced upward muons (see Fig. 3). Thus, for large mixing angles the upward
muon flux would be significantly suppressed, approaching the level of 1 − 12 sin2 2θ for
L δm2 / E >> π/2.
Zenith angle dependence of the neutrino signal probes oscillation lengths L varying from
∼10 (downwards) to 104 km (upward). For small values of E and large values of L the
value of the second sin2 in Eq. (23) will average to 1/2 and the zenith angle dependence will
not be observable. This is the case for the contained events. Explicit variation with zenith
angle may be observable for higher energy neutrinos if the contained event anomaly is due
to neutrino oscillations.2
1
Kamiokande does not use a sharp threshold of Eµmin = 3 GeV. Rather, the definition of a throughgoing
muon is any muon with a projected trajectory > 7 m inside the detector which actually exits from the
detector.
2
There is a hint of such behavior in the recent Kamiokande preprint [71], received after this manuscript
was prepared.

12
 The major sources of uncertainty in Eq. (21) are the neutrino flux and, to a lesser extent,
the charged current cross section in the relevant energy range from ∼1 to ∼104 GeV. Various
calculated values of the neutrino flux are listed in Table 3 at three characteristic energies.
In the most important region for upward, throughgoing muons the calculations differ by as
much as 17%. Different standard parametrization of the charged current cross sections also
differ by as much as 13%.

dN
Table 3: (νµ + ν̄µ , cm−2 s−1 sr−1 )
d ln Eν
10 GeV 100 GeV 1000 GeV
Volkova [72] 6.0 × 10−4 6.1 × 10−6 4.5 × 10−8
Mitsui [73] 6.3 6.2 4.1
Butkevich [74] 7.3 6.9 4.2
Bartol [75] 6.9 7.2 4.7

Both IMB [66] and Kamiokande [69] used low values of the cross section [76] and the
neutrino flux [72] as central values for comparison with their data. In Ref. [70] it was
shown that, with this input the Kamiokande data on throughgoing muons rule out a region
of parameter space for νµ ↔ ντ oscillations for sin2 2θ > 0.4 and δm2 ∼ > 5 × 10−3 . This
is very similar to the excluded region obtained by the IMB group starting from the same
assumptions.
When a better representation of the cross section [77] and a higher neutrino flux [74]
are used, however, the region excluded by upward, throughgoing muons is much smaller.
In particular, much of the region “allowed” by the Kamiokande contained events [11] is
also allowed by the Kamiokande measurement of upward muons. In the further numerical
examples below the neutrino cross section and flux used for calculations correspond to these
higher inputs [77,74].
Although consistent conclusions seem to emerge from interpretation of the Kamiokande
and IMB data on upward muons, the same cannot be said for Baksan. The Baksan group still
exclude most of the Kamiokande “allowed” region even when they use the highest neutrino
flux calculation [74]. The Baksan limit on νµ ↔ ντ oscillations is based on upward events
near the vertical, specifically zenith angles in the interval −1 < cos θ < −0.6.
They observe 161 upward muons in this angular range during a live time of 7.15 years.
With the low neutrino flux [72] they expect 142, as compared to 163 with the high neutrino
flux [74]. In the comparable angular region the Kamiokande measurement is (1.42 ± 0.18) ×
10−13 upward muons per (cm2 s sr). The calculated result with high flux is 1.76 and with low
flux 1.55. The prediction of Kamiokande with the high neutrino flux but assuming νµ ↔ ντ
oscillations is 1.34 (instead of 1.76), which is completely consistent with the experimental
value of 1.42 ± 0.18. This is for sin2 2θ = 0.5 and 10−2 < δm2 < 10−1 eV2 . The preliminary
result from MACRO [78] is similar to the Kamiokande result: they measure 74±9±8 events
as compared to a calculated number for the same exposure of 101 ± 15.
In summary, therefore, we conclude that the present data on upward muons do not rule

13
out a νµ ↔ ντ oscillation at a level sufficient to explain the contained event anomaly. A
tenfold increase in statistics would help resolve the situation, particularly if the full zenith
angle range can be measured accurately. For example, with δm2 ∼ 0.1 eV2 the flux would
be significantly suppressed near the horizontal as well as for cos θ < −0.2, but for δm2 ∼
0.01 eV2 the near horizontal flux would not be suppressed.
The IMB group points out that the fraction of upward muons that stop in the detector
is relatively insensitive to uncertainties in the calculation because the flux normalization
cancels. They find [66] that the measured fraction of stopping muons rules out a portion of
the (δm2 , sin2 θ) plane for 3×10−4 < δm2 < 10−2 eV2 and large mixing angle. The constraint
on the δm2 parameter from the fraction of stopping muons comes from the absence of a
distortion of the muon energy spectrum. The relevant neutrino energies are illustrated in
Fig. 3. For example, if δm2 ∼ 10−3 eV2 the transition probability (23) is relatively large for
Eν ∼ 10 GeV (and L ∼ 104 km) but negligible for Eν ∼ 100 GeV. In this case one would
have a significant distortion of the energy spectrum of upward muons and hence an anomaly
in the stopping fraction provided the mixing angle is sufficiently large. On the other hand,
if δm2 ∼> 10−2 then both the high and low energy portions will be affected similarly and no
distortion of the stopping/throughgoing ratio would occur.
It should be mentioned that the IMB constraint from the stopping fraction is based on
use of a single neutrino flux calculation [72]. It therefore reflects a particular assumption
about the slope of the primary cosmic ray spectrum and other factors that could affect the
shape of the neutrino spectrum. One example of such a factor is the uncertainty in the
production of kaons, because kaons contribute about 50% of the throughgoing signal but
only about 25% of the stopping muons. Nevertheless, as Table 3 illustrates, the uncertainty
in shape is less than the uncertainty in normalization.

3.3 Neutrinos and Muons from Charm production

So far in discussing atmospheric neutrinos we have considered only those which come from
decay of pions, kaons and muons. The (semi)-leptonic decay of charmed particles, produced
in the interactions of cosmic rays in the atmosphere, is also a source of atmospheric muons
and neutrinos—the “prompt” leptons. These prompt leptons are also described by Eq. (9),
but with a much larger value of ǫcharm ∼ 104 TeV, which is a consequence of the short
lifetime of charmed particles. Charmed particles almost always decay before they interact
in the atmosphere.
Thus, whereas the spectrum of conventional muons and neutrinos becomes one power
of energy steeper than the primary spectrum for E ∼ > 1 TeV, the prompt leptons continue
with the same power as the primary spectrum to much higher energy. Prompt leptons
eventually dominate the atmospheric spectra at energies above 10–100 TeV as a result of
their flatter energy spectrum. For this reason, an estimate of prompt neutrinos is important
for estimating the background for astrophysical neutrinos. Prompt neutrinos and muons
have a characteristic isotropic angular distribution, in contrast to the characteristic sec θ
dependence of the decay products of pions and kaons above ∼ 1 TeV (see Eq. 9).
Searching for an isotropic component of the atmospheric muon spectrum with deep un-
derground detectors is a traditional technique to look for a prompt component and hence to
estimate the charm cross section. Underground detectors exploit the depth-intensity rela-

14
tions for muons in order to obtain different muon thresholds. There is some weak evidence
[79,80] for a prompt muon component at a level corresponding to one prompt muon for every
1000 pions produced at the same energy. Because of the small probability for pion decay at
high energy, this would actually correspond to a rather large charm cross section. The raw
measurements are, however, difficult to interpret [81]. This is not surprising as there are
many problems in detecting prompt muons with underground experiments [82].
It is also difficult to predict accurately the flux of prompt leptons because the exper-
imental status of charm production at laboratory energies is rather confusing. Especially
large uncertainties are associated with the production of mesons and baryons in the Feyn-
man xf → 1 region. As is the case for strange particles, the (forward) xf → 1 behavior
is expected to vary strongly with the nature of the produced particle. Neutrino and muon
fluxes are rather sensitive to the behavior of the cross section at large xf because of the steep
parent cosmic ray spectrum. Theory cannot come to the rescue. Perturbative calculations
are unreliable at low energies because of their sensitivity to the assumed quark mass and
renormalization scale. Because charm is predominantly produced by the fusion of gluons at
high energies, the cross section critically depends on the low-x behavior of the gluon struc-
ture function which is poorly or totally unknown depending on the energy[83]. A calculation
of the high-energy charm cross section is actually beyond the scope of perturbative QCD
because it requires the resummation of large logarithms of 1/x [84].
Prompt neutrino (or muon) fluxes corresponding to five parametrizations of the high
energy behavior of the charm production cross section [82] are shown in Fig. 5 (solid lines).
For comparison, we also show the conventional muon flux from the decay of pions and
kaons. The left dashed line is for vertical muons and the right for horizontal. The prompt
curves represent both neutrino and muon fluxes which are very nearly identical for two-body
decays of massive particles into light leptons. The prompt muon flux from charm decay is
independent of zenith angle up to 108 GeV.
In contrast, the conventional muon and neutrino fluxes have a strong dependence on
zenith angle, with a ratio of vertical to horizontal fluxes approaching an order of magnitude
at high energy.
Prompt production dominates above an energy whose value depends on zenith angle
and, of course, on the assumption for the high-energy charm cross section. This cross-over
energy is lower for neutrinos than for muons because the conventional muon flux exceeds the
conventional neutrino flux. The five models for the hadroproduction of charm are discussed in
detail in reference [82]. As an extreme guess on the high side they assume a charm production
cross section which is 10% of the total inelastic cross section, σin [85]. It behaves like log2 (s)
at high energies. The model is inspired by the fact that a high energy gluon fragments
10% of the time into charm particles, and its applicability is limited to energies above a few
TeV since at lower energies it violates charm cross-section measurements from accelerator
experiments [86]. As a lower limit they evaluated the charm cross section from leading order
perturbation theory [87] using relatively hard parton distributions. Some interesting results
have been obtained by an X-ray chamber array measuring the vertical muon spectrum up to
50 TeV [88]. These data rule out the highest parametrization shown in Fig. 5, which is not
totally surprising as it also overestimates the accelerator data at low energy. We therefore
consider the curve labelled B in Fig. 5 to be the upper limit for the atmospheric background
of neutrinos above 100 TeV.

15
3.4 Neutrinos and Muons in Horizontal Air Showers
Cosmic rays with energy of order 100 TeV and higher initiate air showers which penetrate
deeply enough to be studied with particle detectors at ground level. The detected flux is
a steeply falling function of zenith angle because the depth of atmosphere traversed by a
cascade reaching the ground rises rapidly from 1030 to 36000 g cm−2 as the zenith angle varies
from zero to 90 degrees. Thus near the horizon (90◦ ), close to a thousand radiation lengths of
matter separate the interaction from the detector, and the configuration of the experiment
is analogous to that of any accelerator-based beam dump experiment. Most secondaries
such as pions and kaons are absorbed in the dump and only penetrating particles, such as
muons and neutrinos produced in the initial interaction, reach the detector. High energy
muons will traverse the atmosphere and occasionally lose energy by catastrophic photon
bremsstrahlung. If the photon shower is produced close to the detector it will be recorded
and is referred to as a horizontal air shower. Because horizontal air showers are a signature
for penetrating particles in general, they can also be used to search for cosmic neutrinos,
so we discuss them in some detail. For comparable muon and neutrino spectra, muons will
dominate the horizontal air shower flux as a result of their larger interaction cross section.
Therefore energetic atmospheric muons constitute the main background for horizontal air
showers from cosmic neutrinos.
High energy muons produced in cosmic ray interactions in the first layers of the atmo-
sphere, will radiate hard bremsstrahlung photons that initiate electromagnetic cascades. If
the bremsstrahlung interaction occurs close to the particle detector the cascade will be reg-
istered. Neutrinos can similarly interact deep in the atmosphere and deposit a fraction y
of their energy in particles which produce an electromagnetic shower close to the detector.
The horizontal shower rate φsh is determined by convolution of i) the flux of parent muons
or neutrinos φpar and ii) the y-differential interaction cross section of the parent particles
integrated over the atmospheric depth t at a given zenith angle,
!
tmax 1 dy E(Ne , t) dσ dE
Z Z
φsh (Ne ) = dt φpar Epar = , (24)
0 0 y y dy dNe

using the following notation for the differential and integral shower size spectra

d4 Nsh
φsh (Ne ) = , (25)
dNe dt dA dΩ
Z ∞
Φsh (Ne ) = dNe∗ φsh (Ne∗ ) . (26)
Ne

The notation for the parent spectrum φpar is the same. For a given shower size and depth
of first interaction the energy of the shower is fixed, on average, by the depth development
of the cascade. This is the meaning of E(Ne , t) in Eq. (24). This development is somewhat
different for purely electromagnetic, hadronic or mixed showers. As a rule of thumb the
shower size at maximum is roughly the half the energy in GeV units.
Calculation of the rates of horizontal cascades with fixed shower size Ne are discussed
in detail in Refs. [89,82]. Both analytic and Monte Carlo calculations can be made [96].
Starting from the horizontal rate as given in Eq. 24, the number of observed showers above

16
a given size and angle is given by
Z ∞ Z Ω0
Nshower (Ne > N0 , θ > θ0 ) = T dNe A(Ne ) dΩ φsh (Ne , θ) (27)
Ne0 0

where A(Ne ) is the effective area of the array and T is the observation time. The method
described above has been used to calculate the horizontal air shower rates, shown in Fig. 18,
associated with the muon fluxes of conventional and charm origin from Fig. 5.
Horizontal shower sizes in the range Ne = 103 –105 have been extensively studied by the
University of Tokyo [90], and their observed rate of showers with zenith angle θ > 70◦ is
consistent with production by conventional atmospheric muons [91]. The EASTOP group
have also reported very recently a measurement of horizontal air showers with θ > 75◦ that
is consistent with a muonic origin [92].
For shower sizes above 105 the AKENO group has published an upper bound on muon-
poor, air showers with zenith angle greater than 60◦ [93]. The bound is obtained by selec-
tion of muon poor showers in order to isolate purely electromagnetic showers initiated by
bremsstrahlung photons from muons [94]. The high energy muon fluxes from charm decay
are therefore bounded by this data which provide us with indirect, but relevant information
on the charm production cross section. This is illustrated in Fig. 6. It is clear that the
largest charm cross sections are ruled out by both the data of the University of Tokyo or by
the AKENO bound.
It is interesting to point out that the differential muon spectrum from prompt decays at
energy Eµ reflects pp interactions of average lab energy roughly a factor 10 (8) times the
muon energy, for a 2.7 (3) spectral index [95,96]. The measurements for largest Eµ [88] thus
correspond to an average proton lab energy around 500 TeV. Measurements of horizontal
shower of sizes above 105 correspond to a primary photon energy of around 100 TeV, which
has been on average produced by a primary p-air collision of lab energy of order 1.7 PeV just
beyond the reach of TEVATRON. The AKENO bound extends the average energy reach
for prompt production even further, to close to 100 PeV of laboratory energy, although the
information should be taken with care. Firstly, there is some inconsistency between the rate
of horizontal showers measured by the university of Tokyo and the bound [93]. Secondly,
the measurement relies on selecting muon poor showers at zenith angles above 60◦, and
contamination from ordinary cosmic ray showers can be a problem. The already difficult
selection of muon poor showers is further complicated by the presence of the original muon
that gave rise to the bremsstrahlung photon. For a more detailed discussion of the relevance
of this data to the charm cross section we refer the reader to reference [96].

4 Gamma ray astronomy

Many authors have discussed candidate point sources of ∼ > TeV neutrinos, especially in con-
nection with reported observations of air showers from point sources. Possible sources include
accreting X-ray binaries [97,98,99], compact binary systems with interacting winds [100], a
neutron star engulfed by a giant companion [98,101] and young supernova remnants [102,103,104,105].
There are two approaches to computing the signal expected from such sources. The first
starts from observations (or limits) on photon signals from candidate sources. If the photons

17
are products of decay of neutral pions, then one expects a comparable flux of νµ . Photons
can also be produced by electrons through synchrotron radiation, bremsstrahlung and inverse
Compton scattering. To be as certain as possible that one is dealing with π 0 γ-rays it is
therefore desirable to look in an energy range above that accessible to electrons, which are
typically limited to energies in the TeV range by ambient magnetic fields. We use limits of
observations on photons in the 100 TeV range, i.e. from air shower experiments. Typical
limits for steady emission from point sources are in the range [106,107,108,109,110,111]

dNγ < 10−13 cm−2 s−1

= Eγ φγ ∼ (28)
d ln Eγ

for Eγ ∼ 100 TeV. Limits on certain Northern hemisphere sources obtained using the muon
rejection technique are somewhat lower, approaching 10−14 cm−2 s−1 [112].
To find the implied limit on the corresponding neutrino flux requires knowing the relation
between photons and neutrinos at production and the fraction of the produced photons that
is absorbed in the source. The latter is highly uncertain. For a spectrum of photons from
π 0 -decay of the form
φγ = Const × Eγ−α (29)
the corresponding spectrum of neutrinos from decay of π ± is
1
φν = Const × (1 − rπ )α−1 × , (30)
1 − Aγ

where Aγ is the fraction of photons absorbed at the source (which may in general be energy-
dependent). For α ∼ > 2 the kinematic factor is approximately 1 . When νµ from decay of
2
muons are added, the summed flux of νµ + ν̄µ is approximately equal to the photon flux at
production [24].
We can convolve the neutrino flux (30) corresponding to the limit (28) with the charged
current neutrino cross section and muon range (see Eq. 21) to obtain a limit on the upward
muon rate. For spectral index γ in the range 1.1 to 1.3, we find

< 2 events (Eµ > 1 TeV) ×

Flux(↑ µ) ∼
1
. (31)
5 2
10 m yr (1 − Aγ )

The total number of upward muon events with Eµ > 1 GeV is only a factor of two larger for
such a flat spectrum.
A source producing at this limit and having Aγ ∼ > 0.98, i.e. a factor 50 enhancement of
the neutrino relative to the photon flux, would be detectable in DUMAND in the sense of
giving ∼20 events per year in 2 × 104 m2 with Eµ > 1 TeV. (The exact number depends on
“details” such as the location of the source relative to the detector, which determines the
fraction of the time it is sufficiently below the horizon to produce a signal.)
The photon absorption factor Aγ could be much larger (e.g. in the case of the neutron star
swallowed by a giant star), but this would be at the expense of requiring still greater power
at the source.3 This leads to the other approach to estimating likely neutrino fluxes from
3
It is shown in Ref. [113] that the power of a hidden source cannot be increased indefinitely without
making the object so hot and bright as to violate observations.

18
various sources. It is straightforward to calculate the power in accelerated protons required
to give a detectable signal of neutrino-induced muons independent of any model of photon
reabsorption. The result depends on the distance to the source, the assumed spectral index
and the fraction of the accelerated proton beam that interacts. Estimates [113,114] show
that a power of about 1039 to 1040 erg/s of accelerated protons is required for a source at
the distance of the Galactic radius to produce a detectable signal in DUMAND, assuming a
fully absorbed proton beam. This could be a young supernova remnant or a young pulsar. A
system accelerating particles with a power of 1038 erg/s (e.g. an X-ray binary accreting at the
Eddington limit for a solar mass star with a large efficiency for converting accretion energy
into high energy particles) would have to be relatively nearby (∼1 kpc) to be detectable.
Recently the Whipple collaboration reported the observation of TeV (1012 eV) photons
from the giant elliptical galaxy Markarian 421 [115], an observation which might be directly
relevant to our quest for sources of high energy neutrinos. With a signal in excess of 6 stan-
dard deviations, this is the first convincing observation of TeV gamma rays from outside
our Galaxy. That a distant source like Markarian 421 can be observed at all implies that
its luminosity exceeds that of galactic cosmic accelerators such as the Crab, the only source
observed by the same instrument with comparable statistical significance, by close to 10 or-
ders of magnitude. The Whipple observation implies a Mkn 421 photon luminosity in excess
of 1043 ergs per second. It is interesting that these sources have roughly the same flux of
energy per logarithmic energy interval in the TeV region as in the GeV region.
Why Markarian 421? Whipple obviously zoomed in on the Compton Observatory cat-
alogue of active galaxies (AGN) known to emit GeV photons. Markarian, at a distance of
barely over 100 Mpc, is the closest blazar on the list. Stecker et al. [116] recently pointed
out that TeV gamma rays are efficiently absorbed on infra-red starlight, anticipating that
TeV astronomers will have a hard time observing 3C279 at a redshift of 0.54. Production
of e+ e− pairs by TeV gamma rays interacting with IR background photons is the origin of
the absorption. The absorption is, however, minimal for Mkn 421 with z = 0.03, a distance
close enough to see through the IR fog.
This observation was not totally unanticipated. Many theorists [117] argue that blazars
such as Mkn 421 may be powerful cosmic accelerators producing beams of very high energy
photons and neutrinos. Acceleration of particles is by shocks in the jets which are a charac-
teristic feature of these radio-loud active galaxies. Many arguments have been given for the
acceleration of protons as well as electrons. Inevitably beams of gamma rays and neutrinos
from the decay of pions appear along the jets. The pions are photoproduced by accelerated
protons on the dense target of optical and UV photons in the galaxy. The latter are the
product of synchrotron radiation by electrons accelerated along with the protons. There are
of course no neutrinos without proton acceleration. The arguments that protons are indeed
accelerated in AGN are rather compelling. They provide a “natural” mechanism for i) the
energy transfer from the central engine over distances as large as 1 parsec, ii) the heating
of the dusty disc over distances of several hundred parsecs and iii) the near-infrared cut-off
of the synchrotron emission in the jet. Protons, unlike electrons, efficiently transfer energy
in the presence of the magnetic field in the jet. A detailed case for proton acceleration in
active galaxies is made in reference [118].
Other possible models for the emission of gamma radiation up to 1 TeV from Mkn
421 involves inverse Compton scattering by accelerated electrons, rather than π 0 γ-rays

19
[119,120,121]. Such models therefore do not predict a corresponding flux of high energy
neutrinos.

5 Guaranteed Sources of High Energy Neutrinos: the Galactic

Plane and the Sun
By their very existence, high-energy cosmic rays do guarantee the existence of a definite
source of high energy cosmic neutrinos [122]. Cosmic rays interact with the interstellar gas
in our galaxy and are therefore inevitably accompanied by a flux of diffuse photons and
neutrinos which are the decay products of the pions produced in these interactions. A rough
estimate of the diffuse fluxes of gamma rays and neutrinos from the galactic disk can be
obtained by convoluting the observed cosmic ray flux with interstellar gas with a nominal
density of 1 particle per cm3 . The target material is concentrated in the disk of the galaxy
and so will be the secondary photon flux. Its observation would reveal “point sources”
associated with molecular clouds and the spiral arm of the galaxy.
An estimate of the expected fluxes at TeV and PeV energy can be easily performed under
the assumption of a constant cosmic ray density in the Galaxy. Imagine a concentration of
matter of density ρ and linear dimension R. For example, the flux at Earth of photons
generated by pions produced in cosmic ray interactions with this matter is given by
σinel 2ZN π0
 
Φγ(ν) = ΦCR fA [ρR] , (32)
mN γ+1

where ΦCR ≈ 1.8 E −2.7 cm−2 sr−1 s−1 GeV−1 is the cosmic ray intensity, σinel is the total
inelastic pp cross section, mN is the nucleon mass and ρR is the column density of the source.
The quantity ZN π = 1/σ dx xγ dσ/dx is the spectrum-weighted moment for production of
R

pions by nucleons with differential energy spectrum E −(γ+1) , and fA (≃ 1.22) is a correction
factor to account for the fact that some primaries and targets are nuclei [123]. As noted in
the previous section, the differential spectrum of ν + µ + ν̄µ is very nearly equal to the that
of photons after accounting for muon decay.
In a detailed calculation, as recently performed by Berezinsky et al. [124], one must
explicitly account for the energy dependence of the inelastic cross-section, of the particle
physics parameters and of the spectral index γ. Nevertheless, Eq. (32), which neglects
these energy dependences, can be used to make adequate first order estimates of the γ-ray
and neutrino fluxes. Equation (32) states that the photon to cosmic ray ratio is directly
proportional to the linear matter density ρR. The ratio is of order 6 × 10−5 for a column
density of 0.1 grams/cm2 . Maps of the galactic linear column density can therefore be
directly translated into photon or neutrino fluxes with the assumption that the cosmic ray
density in the Galaxy is constant at its local value. The predicted flux is of order 10−5 of
the cosmic ray flux in the PeV energy range, with the different estimates varying within “a
factor”. Observation of a photon flux at this level has turned out to be challenging. The
best experimental limits [125] are still an order of magnitude higher than expectation [124].
It is clear that a roughly equal diffuse neutrino flux is produced by the decay of charged
pion secondaries in the same collisions. For example, from Eq. (32), and assuming a threshold
of Eµ > 1 TeV, we can estimate the number of neutrino events from within 10 degrees of the

20
disc as 5 events per year for a 105 m2 detector at the South Pole which views 1.1 steradian
of the outer Galaxy with an average density of 0.013 grams/cm2 . This would be increased
to 15 events per year if the spectral index in the outer galaxy were indeed as small as 1.4,
as suggested by an analysis [26] of the GeV γ-ray results of the COS-B satellite. Another
interesting example is the direction of Orion, a molecular cloud with a column density of
0.04 g/cm2 and an angular width of 0.07 sr [126]. For this case we estimate 0.3 neutrino-
induced muons per year in a 105 m2 detector in this angular bin. There are several gas
concentrations of similar or smaller density in the galaxy. These numbers account for the
fact that neutrinos are produced by the decay of muons as well as pions. Although these
rates are significantly below the atmospheric background, the source is guaranteed and the
event rate might be significantly higher if there are regions of the galaxy where the spectrum
is flatter than the local spectrum.
The above estimates assume a cosmic ray intensity that is constant throughout the disk
of the Galaxy and equal to that measured at Earth. A recent COMPTEL [126] observation
of 3 to 7 MeV fluxes from the region of Orion may suggest, however, that the cosmic ray
energy density could be significantly higher in that region. The observed γ-ray line intensities
(from excited 12 C and 16 O nuclei) would correspond to cosmic ray energy density of more
than 50 eV cm−3 , a factor of 100 higher than in the vicinity of the Earth. Although the
lines are generated by cosmic ray nuclei of kinetic energy around 10 MeV, which are subject
to strong solar modulation inside the solar system, it is quite possible that the cosmic ray
density in a wider energy range is significantly higher in this active star-formation region.
The limits derived from the COS-B data[127] allow a cosmic ray intensity and respectively
neutrino fluxes higher by factor of 5 in the region of Orion.
The other guaranteed extraterrestrial source of high energy neutrinos is the Sun. The
production process is exactly the same as for atmospheric neutrinos—cosmic ray interactions
in the solar atmosphere. Neutrino production is enhanced because the atmosphere of the
Sun is much more tenuous—the scaleheight of the chromosphere is ∼115 km, compared with
6.3 km for the upper atmosphere. As a result ǫπ is higher by a factor of ∼20, as is the energy
where the slope of the neutrino spectrum increases. The difference is even larger for cosmic
rays that enter the Sun at large angles and never reach atmospheric densities higher than
10−7 g/cm3 . A detailed calculation of the neutrino production by cosmic rays in the solar
atmosphere [128] shows a neutrino spectrum higher than the angle averaged atmospheric
flux by a factor of ∼2 at 10 GeV and a factor of ∼3 at 1000 GeV.
The decisive factor for the observability of this neutrino source is the small solid angle
(6.8×10−5 sr) of the Sun. Although the rate of the neutrino induced upward going muons
is higher than the atmospheric emission from the same solid angle by a factor of ∼5, the
rate of muons of energy above 10 GeV in a 105 m2 detector is only 5 per year. Taking into
account the diffusion of the cosmic rays in the solar wind, which decreases significantly the
value of the flux for energies below one TeV, cuts this event rate further by a factor of 3.
Folded with a realistic angular resolution of 1 degree, observation of such an event rate also
requires a 1 km2 detector.

21
6 Possible Galactic Neutrino Sources
In §2 we noted that the galactic cosmic radiation above ∼100 TeV might be accelerated
in compact sources, and that, if so, these would be likely point sources of photons and
neutrinos. In §4 it was shown that present limits on ∼100 TeV gamma rays from potential
point sources make it unlikely that there will be prolific galactic point sources of neutrinos.
Before discussing several possible types of point sources in more detail, we give an example
to set the scale for what may be the maximum reasonable expectation for neutrinos from
galactic point sources. We consider a two-component model of the cosmic radiation as
described in §2 in which the low energy component steepens around 100 TeV and a high
energy component dominates at much higher energies. For this illustration we assume that
the high energy component is produced in compact sources scattered in the disk of the
Galaxy.
We start by estimating the power that would be needed to supply such a high energy
component. In a two-component picture [129] the cosmic ray energy density in energy range
between 100 and 1000 TeV is about 10−15 erg/cm3 , of which about half would be from the
high energy component. The local energy density in this component is thus estimated as

ρE = 5 × 10−16 erg cm−3 . (33)

Assuming this is typical of the energy density elsewhere in the Galaxy, the luminosity of the
galaxy in such particles is then
Vgal ρE
ǫ × Lp = = 1.5 × 1038 erg s−1 , (34)
τ
where Vgal is the volume of the galaxy and τ the confinement or lifetime of PeV cosmic rays in
the galaxy. The values of these parameters are very uncertain and also depend on the model
of propagation and escape of cosmic rays from the Galaxy. The numerical result in Eq. (34)
is obtained for Vgal = 5 × 1066 cm3 (the volume of the galactic disk) and τ = 5 × 105 year
as an estimate of the time cosmic rays spend in the disk. The parameter ǫ < 1 in Eq. (34)
is the fraction of the accelerator power required just for the decade above 100 TeV. If we
assume that these accelerators produce a hard spectrum with equal energy per logarithmic
interval, then the estimate of the total power needed to maintain the steady observed PeV
cosmic ray flux is Lp ∼ 1039 erg/sec. This source will resupply the galaxy and compensate
for the loss of cosmic rays resulting from their limited confinement time [130].
Let us next assume that a comparable amount of energy is absorbed by collisions of
protons with gas near the high energy accelerators. Then the total power in neutrinos is
related to the neutrino flux at Earth from all compact sources by
dNi,ν
Z
(4πd2i )
X
Lp Dp→νµ Dν = dE E (35)
i dE

Here Dp→νµ (≃ 0.3) is the fraction of energy in an accelerated proton spectrum ∝ E −2 that
goes into νµ (ν̄µ ), Dν (≃ 1) describes the fraction of neutrinos that escape from the sources
and di is the distance to the ith source. If there are n such sources distributed throughout
√
the galactic plane, then an estimate of the distance to the nearest source is d ∼ Rg / n

22
where Rg ∼ 10 kpc is a nominal radius of the disk in which the sources are concentrated.
If each source has a particle luminosity of Lp /n, then number of sources cancels in the
relation between the total cosmic ray luminosity Lp and the neutrino luminosity of a “nearest
neighbor” source [131]. One has
dNi,ν Lp Dp→π Dν /n
Z
dE E = , (36)
dE 4πRg2 /n
from which we estimate the neutrino flux from a nearby source as
dNν 100 TeV
E = 2 × 10−11 cm−2 s−1 , (37)
dE E
assuming an E −2 spectrum. Such a flux of high energy νµ + ν̄µ would give some 300 events
per year in 105 m2 .
This is a very high flux and should be considered an extreme upper limit for a neutrino-
induced signal from a galactic point source. To avoid the existing limits on photons from
point sources (28), one would need a factor Aγ ∼ > 100 absorption of high energy photons
in the source. Absorption arguments depend strongly on the exact mechanism. If photon
absorption is due to interactions and cascading on the ambient matter at source, the acceler-
ated protons will also be absorbed, which weakens the original motivation for this estimate.
It is, however, not only possible, but likely [132] that the high energy γ-rays would be ab-
sorbed in γγ → e+ e− collisions on the strong radiation field at the source. In this case the
protons will not be absorbed since the photoproduction threshold is (mπ /me )2 higher and
the protons lose very little energy in pγ → e+ e− collisions.
The preceding argument is based on an assumed random distribution of point sources in
the disk of the galaxy. Because of the cancellation of the number n of sources in Eq. (36), a
similar estimate can be made of the neutrino flux from a single source at the galactic center.
We note that the upper bound on the > 100 TeV gamma flux from the galactic center is
[109] ∼ 2 × 10−13 cm−2 s−1 . There could, however, be significant absorption of a high energy
photon source from the galactic center. Thus, as pointed out by Berezinsky [133], not seeing
high energy neutrino emission from the center of the galaxy would be an interesting result.
We now look in somewhat more detail at two possible classes of galactic point sources of
neutrinos.

6.1 X-ray Binary Systems.

The interest in X-ray binaries was initiated by the reports [134] of detection of UHE (>
1014 eV) γ-rays from Cygnus X-3. Such γ-rays would most likely be produced in inelastic
hadronic interactions, in which case they would be accompanied by high energy neutrinos.
Current upper limits on steady emission from Cygnus X-3 are about an order of magnitude
lower than the level implied by the original report, which referred to showers with energies
above 2 × 1015 eV. Nevertheless, it is still interesting to consider X-ray binary systems
as potential accelerators of UHE cosmic rays and to ask at what level one might expect
accompanying neutrino fluxes.
X-ray binaries consist of a compact object (neutron star or a black hole) and a non-
compact companion star. Such systems are dynamically complicated, involving mass transfer

23
from the companion onto the compact object through an accretion disk. Neutron stars
are known to have very strong (1012 G) surface magnetic fields and sometimes millisecond
periods. Both the accretion and the magnetic dipole radiation are possible energy sources.
The existence of high magnetic fields and plasma flows creates the environment necessary
for the formation of strong shocks, and corresponding particle acceleration. The companion
star itself, the accretion flow, or the heavy stellar winds might be targets for inelastic nucleon
interactions and neutrino production.
Calculations of the neutrino flux expected from Cygnus X-3 were done independently
by different authors [135], and the results agree to better than a factor of two for similar
assumptions about the input parameters and the configuration of the accelerated beam and
beam dump. The total upward going muon flux for a distance of 10 kpc is 2–3×10−15
cm−2 s−1 , i.e. 50–100 upward going muons per 105 m2 per year for a fully efficient detector.
Fluxes at this level are well above the atmospheric background for Eµ ∼ > 100 GeV [64]. Such
39
a large flux corresponds to a proton luminosity at the source of 2×10 erg/s, comparable to
the generic estimate in the introduction to this section. The estimated flux is a factor ∼10
lower here, however, because the models typically assume a 10% duty cycle for the beam to
intercept the target mass (e.g. the companion star).
Models motivated by the original Cygnus X-3 observations in which the high energy
gamma rays emerge from source obviously cannot be correct for Cygnus X-3 in view of the
current limits from air shower observations. Emission of neutrinos in other models in which
the photons are absorbed can, however, be obtained by scaling from these calculations by
an assumed proton luminosity and the distance of any potential source.
The crucial question then is the luminosity that might be expected from such systems.
For accretion powered sources the luminosity is limited by the Eddington luminosity

LEdd = 4πGMmp /σT erg/s , (38)

which is the maximum X-ray luminosity that will not prevent accretion. Since the proton
inelastic cross-section is lower than the Thomson cross section σT , technically the proton
luminosity can exceed LEdd . On the other hand LEdd can only be achieved at the surface
of the neutron star and a realistic luminosity limit depends on the ratio of the neutron
star radius to the shock radius Rns /Rs . Thus a reasonably optimistic limit for the proton
luminosity will be
Lmax
p = LEdd × (Rns /Rs ) × (σ inel /σT ) , (39)
i.e. of the order of or lower than LEdd = 1.4 × 1038 × M/M⊙ erg/s, corresponding to a rate
of upward TeV muons < 50 events per year in a 105 m2 detector for a source at 10 kpc.
Another potential source of energy in an X-ray binary is pulsar rotation. Harding &
Gaisser [100] have studied proton acceleration at X-ray binaries powered by the pulsar
through a pulsar wind shock. An absolute upper bound on the energy is the power released
by magnetic dipole radiation,

Ld = 4 × 1043 B12 Pms

−4
erg/s , (40)

where B12 is the pulsar surface magnetic field strength in 1012 G and Pms is the pulsar period
in milliseconds. Discussing different X-ray systems, however, they end up with a maximum

24
proton acceleration luminosity of 6 × 1038 erg/s for Cygnus X-3 with a pulsar period of
12.8 ms [136]. This is still factor of 2 smaller than the luminosities used in the Cygnus X-3
estimates above and illustrates that an X-ray binary has to put almost all of its energy in
high energy protons to be detectable in neutrinos. For a more modest X-ray binary at the
galactic center that accelerates 1/10 LEdd in high-energy protons the actual upward muon
rate will be ∼3 events per 105 m2 per year.

6.2 Young Supernova Remnants

Young supernova remnants are another candidate for production of observable neutrino fluxes
[137,138]. If protons are accelerated inside a young supernova remnant, they will interact
with the material of the expanding shell and produce γ-rays and neutrinos until the particle
adiabatic losses exceeds the collision loss. In the approximation of a uniform density shell of
mass M expanding with velocity v = 109 cm/s this occurs at
3Mc σpp 1/2 M 1/2
τa = ( 3
) = 1.3 × 107 ( ) s. (41)
4πmH v M⊙
The active time during which the production is significant is of order 1 year. Two modifica-
tions of this idea [139] were motivated by the explosion of SN1987A and the proliferation of
detailed supernova models that followed. If one accounts correctly for the velocity distribu-
tion of the supernova shell, the γ and ν emission time increases by a factor of three. Also,
if the accelerated particles are contained within the shell as it expands the pathlength for
collisions will increase and the duration of the signal will be extended for a period estimated
in Ref. [139] as ∼ 10 years, with a gradual decrease in intensity. If the accelerated protons
are not confined in the shell, the duration of the signal would be 1–2 years [140].
All these considerations concern the target for inelastic interactions. The other ingredi-
ent is the abundance of accelerated protons at this stage of remnant evolution. The pulsar
wind model [139] utilizes the pulsar spin down energy to create a shock inside the contact
discontinuity of the shell. The proton luminosity is bounded by the magnetic dipole lumi-
nosity of the pulsar given in Eq. (40). The efficiency for producing a signal in such a model
depends on the efficiency for accelerating protons and on the degree of mixing between the
accelerated particles and the expanding shell. The latter depends on mixing the pulsar wind
region with the shell through Rayleigh-Taylor instabilities.
Although it is clear now that SN1987A does not contain a strong pulsar, it is still of inter-
est to discuss the signal that could be expected from a a young galactic supernova (∼10 kpc)
with a rapidly spinning, strongly magnetized pulsar. The answer is extremely sensitive to
the magnetic field and pulsar period assumed. Both parameters enter into the pulsar power
and into the maximum energy. For example, for P = 10 ms and Bsurface = 1012 Gauss
and a 25% efficiency for particle acceleration and interaction, the model of Ref. [139] gives
1039 erg/sec and Epmax ≈ 105 TeV. The corresponding neutrino luminosity would be sufficient
to produce a signal of ∼ 100 upward muons in 105 m2 for several years. For a longer pulsar
period and/or a smaller surface magnetic field, both the maximum energy and the available
power rapidly decrease.
Berezinsky & Ptuskin [141] argued that acceleration at the supernova blast wave could
also produce an observable signal, even though in this case the accelerated particles are not

25
deep inside the expanding shell. When a supernova expands into the surrounding medium
it drives a blast wave ahead. There is also a reverse shock in the supernova ejecta. Particle
acceleration occurs at both shocks, with the accelerated particles injected into the respective
downstream regions, which are contained between the two shocks. The kinetic energy of the
expanding shell is huge (of order 1051 erg/s) but the rate at which it is dissipated is limited
by the rate at which matter is swept up by the expanding shell. Thus the luminosity from
accelerated particles in this region is quite sensitive (quadratically [141]) to the mass loss
rate of the progenitor star, which was relatively low for SN1987A. For what is considered
a “typical” mass loss rate of 10−5 M⊙ [141], the estimated neutrino flux for a supernova at
10 kpc corresponds to several hundred upward muons with Eµ > 100GeV in the first 100
days [141]. The rate falls off slightly faster than 1/t.
The big disadvantage of young supernova remnants as potential neutrino sources is, of
course, that supernova explosions are rare events. The one that we were lucky to observe,
SN1987A, was not only quite distant, in the LMC, but also shows no signs of pulsar activity
at a level above ∼1037 erg/s.

7 Possible Extragalactic Sources

Active galactic nuclei are the most luminous objects in the Universe and have long been
recognized as possible sources of high energy signals [142]. These first estimates were mostly
based on the total AGN power and number density. More recent calculations [143,144]
developed the idea in two important ways. They first identified the potential importance
of hadrons (especially neutrons) for transporting energy in active galactic nuclei. Secondly,
shock acceleration models were at least crudely incorporated into the AGN models, and the
photoproduction process was shown to be the most important one for proton energy loss.
This led to estimates of the maximum proton energy achievable in acceleration at AGN
shocks and to the prediction of high energy neutrino fluxes.
Active galactic nuclei have luminosities ranging from 1042 to 1048 erg/s, which corresponds
to black hole masses from 104 to 1010 M⊙ [145] on the natural assumption that they are
powered by Eddington limited accretion onto a black hole. AGN’s have generally flat emission
spectra with a luminosity up to ∼3×1046 erg/s per decade of energy. In the IR band a steady
dust emission is observed, most probably coming from a large region far away from the core.
The main thermal feature is the UV bump, which is variable on a timescale of days and
weeks[146].
Its energy source is either X-ray heating[147] or viscous heating of the accretion disk[146],
either of which would be closely related to the central engine. X-rays have a hard, nonther-
mal spectrum, variable on even shorter timescales[148], which often cuts off at several MeV.
AGN’s have been extensively studied at radio frequencies, where the most general iden-
tification is as radio-loud or radio-quiet, depending on the fraction of energy in the radio
portion of the spectrum [149]. Roughly 10% of all observed AGN’s are classified as radio-
loud [150]. Blandford [151] suggests that radio-loud AGN’s have rapidly spinning black holes
and therefore also strong jets. The UV bump is not always easy to see in radioloud AGN’s.
Two possible sources within AGN’s of intense, high energy neutrino fluxes have been
identified. The first is associated with the central engine and the second with production in

26
jets associated with blazars, which are radio-load AGN’s in which the observer is illuminated
by the beam of a jet. We first discuss central emission.

7.1 Generic AGN

To introduce most of the parameters important for the production of neutrinos, we briefly
describe the spherical accretion model used in most of the calculations of the neutrino pro-
duction in central regions of AGN’s [144,36,152,38]. Some of the limitations of this model
will be mention in §7.4 below. The model is based on work performed by Kazanas, Protheroe
and Ellison [33,34]. They assume that close to the black hole the accretion flow becomes
spherical and a shock is formed where the ram pressure of the accretion flow is balanced by
radiation pressure near the black hole. The shock radius is parameterized by R = x1 × RS ,
where RS is the gravitational (Schwarzschild) radius of the black hole, and x1 is estimated
to be in the range 10 to 100 [38]. The continuous emission is dominated by the ultraviolet
and X-ray radiation, which are assumed to emanate from inside the radius enclosed by the
shock. Since the region inside the shock is optically thick, the radiation density at the shock
can be estimated from the surface brightness of the AGN. This leads to the relation

Urad ≃ L × (πR2 c)−1 (42)

between luminosity and radiation density in the central region. Since R = x1 × RS ∝

LEddington , it follows from Eq. (42) that Urad ∝ L−1 . Numerically,
2
1 30

Urad ∼ 2 × 106 erg/s × × , (43)
L45 x1

where L45 is the luminosity divided by 1045 erg/s. The radiation energy density also defines
the magnetic field value B at the shock under the assumption of equipartition of the radiation
1
and magnetic energy. For the numerical example above B ∼ 7000 Gauss × (L45 )− 2 × x301 .
Acceleration of protons is assumed to occur by the first order diffusive Fermi mechanism
at the shock, resulting in an E −2 differential spectrum that extends up to Emax . Energy loss
processes occur during acceleration, including pγ → Nπ and pγ → p + e+ + e− in the dense
radiation fields as well as pp collisions in the gas. All three processes contribute an energetic
electromagnetic component, either through π 0 → γγ or by production of electrons. Both
photo-meson production and pp collisions also give rise to neutrinos via the π ± → µ± → e±
decay chain. In the astrophysical environment all unstable particles (except quasi-stable
neutrons) decay practically without energy loss. An important detail is that photoproduction
of charged pions by protons is dominated by the nπ + channel [153].
Although high energy neutrinos escape directly from the core, the electromagnetic com-
ponent does not. The core is optically thick to photons with energies greater than ∼ 5 MeV.
All γ-rays generated in the dense photon field immediately lose energy in γγ → e+ e− col-
lisions. Inverse Compton/pair-production cascades downscatter all electrons and photons
to X-ray and lower energies. The essential ingredient of these models is that the observed
X-ray spectrum is produced as the end product of the electromagnetic cascades initiated by
high energy photons and electrons produced by the accelerated protons. Thus, estimates of

27
expected neutrino fluxes from individual AGN’s are normalized through the model to their
observed X-ray luminosities.
The proton density at the shock, np (R), can be estimated from the accretion rate needed
to support the black hole luminosity, and from the radius and accretion velocity at the shock.
It is
1/2
np ≃ 1.3 × 108 x1 R−1.5 L1/2 Q−1 cm3 , (44)
where Q is the efficiency for converting accretion power into accelerated particles at the
shock. Such proton densities are not only a good injection source for proton acceleration,
but also a possible target for pp interactions.
The proton energy loss, however, is dominated at high energy by the photoproduction
process pγ → nπ + (pπ 0 ) simply because the target photon density nph is much higher than
np . For thermal radiation with temperature T◦ K the density ratio would be
np 3/2
≃ 2.5 × 10−13 x1 T Q−1 . (45)
nph

The high cross-section pair production process (pγ → pe+ e− ) is relatively unimportant
because of the low proton energy loss per collision. The thermal radiation corresponds to
photon energies in the range 1 to 40 eV.
The relative importance of the different energy-loss mechanisms depends in detail on
the energy-dependence of the various cross sections and on the intensity and spectral shape
of the target radiation field. The detailed calculations [36,38,144,152] use numerical and/or
Monte Carlo techniques to follow the production, propagation and cascading of the secondary
particles inside the central region and to determine the fluxes of neutrinos, nucleons and
X-rays that emerge. Without going into such detail, it is still possible to describe in a
semi-quantitative way the basic results, especially the shape and upper limit of the neutrino
spectra. To do this, we make use of the approximate form of the radiation field given by
Stecker et al. [36].
The minimum energy of a target photon for photoproduction by a proton of energy Ep is

∆2 − m2p
ǫ ≈ (46)
2Ep

where ∆ = 1.232 GeV is the mass of the (3, 3) resonance. The collision length of a nucleon
against photoproduction is Z
−1
ℓ = σ(ǫ) n(ǫ)dǫ, (47)

where n(ǫ) is the number density of photons (differential in energy). Using a resonance
approximation for the cross section gives
πΓ∆
ℓ−1 ≈ ǫ n(ǫ) × σpeak , (48)
∆2 − m2p

where Γ ≈ 115 MeV is the width of the ∆ resonance and σpeak ≈ 5 × 10−28 cm2 . Since R ∝ L
and n(ǫ) ∝ Urad ∝ L−1 , the ratio R/ℓ is independent of luminosity in the model. From
the spectrum of Ref. [36] one finds that R/ℓ < 1 above the UV bump, i.e. for ǫ > 40 eV.

28
From Eq. 46 this corresponds to Ecrit ≈ 8 × 106 GeV. Thus nucleons with energy less than
∼ 1016 eV can escape from the central region (r < R) if they propagate rectilinearly.
Nucleons that escape no longer contribute to the production of secondary photons and
neutrinos. This has little effect on the predicted down-scattering into the X-ray region since
most of the energy has already been dumped by nucleons with higher energy (provided the
accelerated spectrum extends to Emax ≫ 1016 eV). The assumption made about propagation
of protons does, however, have a crucial effect on the predicted neutrino spectrum. If, as
assumed by Stecker et al. [36], protons travel in straight lines inside the central region, then
the neutrino spectrum will follow the proton spectrum only down to an energy roughly
Eν
h i × 1016 eV ∼ 5 × 105 GeV. (49)
Ep
At lower energy the neutrino spectrum dNν /dEν will be constant, reflecting the flat momen-
tum distribution of neutrinos produced in a pγ collision. If, as seems more likely, protons
remain confined in the central region by the same turbulent magnetic fields necessary for
the diffusive shock acceleration to work, then the neutrino spectrum will follow the proton
spectrum down to much lower energy. Both Refs. [38] and [152] assume that protons will be
confined to the central region.
Figure 7 illustrates the difference the assumption of proton confinement makes. It shows
the model neutrino spectra (νµ + ν̄µ ) for the extragalactic source 3C273. The thin lines show
several of the models of Protheroe and Szabo [38], who performed their calculation for x1
values from 10 to 100 and two different photon target spectra. The thick line represents
the model of Stecker et al.[36]. For both sources the neutrino spectrum continues to follow
the E −2 proton spectrum down to low energy in the calculation where the protons remain
confined in the central region. Because of the large neutrino flux in the important region
around 1 TeV, the models of Szabo and Protheroe generate significantly more upward going
muons than predicted by Stecker et al. [36].
The slight dip in the spectra of Ref. [38] around 104 –105 GeV is caused by proton energy
loss to e+ e− pair production, which dominates proton energy losses for proton energies
between ∼30 and ∼3000 TeV. This feature is much more prominent in the calculation of
Ref. [152] than in Ref. [38]. As a consequence of the larger relative contribution of pair
production, the predicted neutrino-induced signal of Ref. [152] is somewhat smaller than
that of Ref. ([38]), as shown with a dash line on Fig. 7. Sikora and Begelman [152] give only
the spectral shape for a generic source. We have normalized their neutrino spectrum to the
3C273 luminosity. The exact position of the maximum neutrino energy is thus uncertain,
because it depends on the parameters of the particular source.
Neutrons are not confined by magnetic scattering in the inner region. Thus neutrons
with E < Ecrit escape from the central region inside R provided they do not decay first.
For the relevant range of parameters, neutrons with energy above a TeV will usually escape.
These energetic neutrons decay at distances ranging from ∼0.01 to ∼100 parsec, and their
decay products can have a profound effect on energy transport in AGN’s, for example driving
winds [154] and producing radio emission [155] far from the core.
As far as neutrinos are concerned, the escape of neutrons from the core has an interesting
consequence for the shape of the spectrum of electron antineutrinos. The dominant channels
for photoproduction of charged mesons by nucleons are pγ → nπ + and nγ → pπ − . The

29
kinematics of the π + → µ+ → e decay chain is such that the flux of νµ from pion decay is
approximately equal to the flux of ν̄µ from muon decay, and vice versa for decay of π − . Thus
from protons one has roughly equal fluxes of νµ , ν̄µ and νe from the π + . The neutron chain
leads to ν̄e instead of νe . Thus for E > Ecrit , when both neutrons and protons interact inside
the core region, the flux of ν̄e is nearly equal to the flux of νe (only slightly suppressed by
the small energy loss of the nucleon in pγ → nπ + ). At lower energies, the neutrons escape
before interacting. One then gets ν̄e from n → p e− ν̄e , with, however, a strong kinematic
suppression because of the very small energy transfer to the leptons in β-decay of the neutron.
For ν̄e from neutron β-decay, the the spectrum of ν̄e is a factor ∼ 5 × 10−4 lower than the
parent neutron spectrum, as compared to a factor of about 5 × 10−2 when E > Ecrit and the
process n + γ → π − → µ− → ν̄e can occur.
As it turns out, the photoproduction in the UV bump also limits the maximum proton
acceleration energy Epmax , and hence the maximum neutrino energy. This differs from the
situation in a more diffuse environment, such as acceleration by a supernova blast wave
expanding into the interstellar medium. In that case the upper limit is determined by the
characteristic lifetime of the shock. (See the discussion of Eq. 11 above.) Epmax is roughly
1
proportional to L 2 , reaching a value of 1017 eV for L = 1045 erg/s, with at least a factor of
two uncertainty [38].
This maximum energy can be estimated by comparing the acceleration rate (11) to the
energy loss rate, Kinel Ep c/ℓ. The acceleration rate in this case is

dE u2 30 1
∼ 0.1 eB ≈ 2 × 105 GeV s−1 × × (L45 )− 2 , (50)
dt c x1
where we have used the equipartition estimate of the magnetic field from Eq. (43). Since
(see Eq. 47) ℓ−1 ∝ n(ǫ) ∝ L−1 , we estimate
1
Epmax ∝ L 2 . (51)

Using once again the radiation spectrum of Stecker et al., one finds that the energy loss
rate becomes comparable to the acceleration rate at the peak of the UV bump, where Ep ≈
3 × 108 GeV. For Epmax above ∼1019 eV, which in this model can be achieved only in AGN
with total luminosity ≃1048 erg/s, proton synchrotron radiation becomes the most important
energy loss channel.
The results of the calculation of Szabo & Protheroe [38] can be summarized by the
following approximate formula [156], which gives the neutrino flux at Earth from an AGN
of given X-ray flux and Epmax in [cm2 .s.TeV]−1

Fν Eν ≃ 0.25FX exp(−20Eν /Epmax ) × Eν−2 , (52)

where FX is the 2–10 KeV X-ray flux (erg cm−2 s−1 ) and Eν is the neutrino energy in TeV.
The generic AGN model, described above, is a first order approximation of the physical
processes that may take place in active galactic nuclei. The assumption of spherical accretion
could be an adequate representation of the accretion flow inside a thick accretion disk.
An attempt to use a different geometry in the center of AGN’s was made in the model
due to Nellen, Mannheim & Biermann [157]. In this model it is assumed that protons are

30
accelerated somewhere near the central region, perhaps in the bases of the jets, and that
both X-rays and neutrinos are generated in collisions of the accelerated protons in the inner
regions of the disk. This model is less specific than those of Refs. [38,36,152], and the main
production process is assumed to be quite different. Nevertheless, the predicted neutrino
fluxes are rather similar to those of Refs. [38] and [152]. This is because the intensity is
normalized to the X-ray luminosity and the protons are assumed to be confined to the
central region until they lose all their energy through collisions.
Individual radioquiet AGN will be difficult to detect, although the atmospheric back-
ground in a 1◦ radius around the source is extremely small, ∼1.6 × 10−6 m−2 yr−1 muons
above 1 TeV. Even with optimistic luminosities, the number of such events from individual
sources is less than 2–3 yr−1 in a 105 m2 detector. For example, the estimated rates from the
models of Ref. [38] are ∼1 per 105 m2 yrs for 3C273 and ∼ 3 in the same units for NGC4151.

7.2 Diffuse AGN neutrino flux

In their pioneering paper Stecker et al [36] integrated the neutrino fluxes from single generic
AGN’s to obtain a diffuse flux of neutrinos from all cosmological AGN. The integration has
to account for the AGN density and luminosity distribution, as well as for the neutrino
adiabatic energy loss due to the expansion of the Universe. This procedure is identical to
the integration used to calculate the value of the diffuse X-ray background. In fact, it uses
the AGN luminosity function derived from X-ray observations [158,159] and assumes that
the neutrinos and the X-rays have a common source.
The AGN luminosity function as a function of the redshift can be expressed as
!
g(z) LX
ρ(LX , z) = R03 ρ0 , (53)
f (z) f (z)

where ρ0 comes from measurements of the AGN luminosity, R0 is the present scale size of the
Universe and g(z) and f (z) describe the number density and luminosity evolution of AGN
in the co-moving volume. Any AGN induced background, including X-ray and neutrino, will
then have energy spectrum [38]

dI 1 c 1 Zmax dL
Z Z
= dLX dz × ρ(LX , z)(1 + z)−α {E(1 + z), LX }, (54)
dE 4π H0 ER03 0 dE

where L is the appropriate differential luminosity and α=5/2 for the Einstein-de Sitter
cosmological model.
Figure 8 shows the current estimates of the isotropic ν background (νµ + ν̄µ ). The esti-
mates of Szabo & Protheroe are made with different values of x1 and photon target spectra,
and they are integrated using two independent sets of luminosity functions [158,159]. The
resulting ν flux extends to very high energy, where it dominates the atmospheric neutrino
background by several orders of magnitude. Because of the isotropic nature of the back-
ground flux, its major feature is the extremely flat energy spectrum. The thick line shows
the corrected prediction of Stecker et al. [36]. While the ν spectra are now in very reasonable
agreement at the higher energy end, the biggest difference occurs at energies below 3×105
GeV, where Stecker et al spectrum becomes flat while the spectrum of Protheroe & Szabo

31
follows the primary proton spectrum, for the reason described in the previous subsection.
The difference reaches 2.5–3 orders of magnitude at Eν = 104 GeV, which makes a crucial
difference in the estimate of the νµ -induced upward muon signal.
Figure 9 shows the muon fluxes generated by the isotropic neutrino background as in
the bracketing high and low models of Szabo & Protheroe [38] and by Stecker et al. [36].
These have been calculated [160] as described in § 3.2.1 for comparison with the Frejus
measurement [161], which gives a 90% C.L. upper limit of 2.3 events with Eµ < 2 TeV for
the range of zenith angles −0.3 < cos θ < 0.3 [161]. The corresponding upper limit [162]
is shown in Fig. 9. The muon flux generated by atmospheric neutrinos averaged over the
same angular interval is shown for comparison. The AGN background dominates at muon
energies above 1 TeV.
Although the diffuse neutrino flux from AGN’s is isotropic, the produced muons will be
suppressed in the vertical direction (from below) by an amount that depends on the relative
importance of high energy neutrinos in the spectrum. The interaction length of neutrinos
in the Earth is less than an Earth radius when σν ∼ 10−33 cm2 , i.e. for Eµ ∼ 107 GeV [65].
Accounting for absorption in the Earth, the predicted TeV muon rate in a downward looking
detector with acceptance of 105 m2 sr will be 160 to 800 per year for Ref. [38] and ∼40 for
Ref. [36] over an atmospheric background of ∼140 events. The higher range of predictions
of Ref. [38], however, are already ruled out by the Frejus limit.
So far we have discussed signals generated by muon neutrinos and antineutrinos. Elec-
tron neutrinos of sufficiently high energy can generate air showers which could be observable
above the background or ordinary showers near the horizontal because of the great pen-
etrating power of neutrinos. Limits on horizontal showers have been given by the Akeno
air shower experiment, as discussed in §3.4 above. The Fly’s Eye detector has searched
for upward-going showers, which would be generated by electron neutrinos from below that
interact near enough to the surface so the resulting electromagnetic cascade emerges from
the ground before it is absorbed. For Eν ≫ 107 GeV, these events would be mostly near the
horizontal since the Earth would absorb the more vertical high energy neutrinos. The Fly’s
Eye limits [164] apply for E > 108 GeV and are discussed further below in connection with
cosmological neutrinos.
Electromagnetic cascades generated by charged current interactions of electron neutrinos
can also be detected when they occur inside the volume of a Cherenkov detector. The rates
are then simply the convolution of flux, cross section and fiducial volume. Figure 10 [163]
shows the rates predicted for the νe + ν̄e spectra of Refs. [36,38]. The dotted line shows the
background of atmospheric electron neutrinos. Since the atmospheric neutrino spectrum is
steeper for νe than for νµ , the flux of AGN νe crosses the atmospheric background at lower
energy than for νµ . In the examples given in Fig. 10, there are ∼ 10 interactions per 1000 kt
years of electron neutrinos with Eν > 1 TeV in a typical model from Ref. [38] as compared to
0.5 in Ref. [36] and 0.2 atmospheric above the same energy threshold. The rates plotted in
Fig. 10 are integrated over all directions. For Eν ∼ 107 GeV, absorption of upward neutrinos
by the Earth begins to be significant, suppressing the quoted rates of νe from AGN slightly.
The spike at 6 × 106 GeV in Fig. 10 represents the interaction of ν̄e at the “Glashow

32
resonance” [165]. The resonance cross section for ν̄e + e− → W − → ν̄e + e− is [166]

G2 s 4
" #
MW
σ(ν̄e e ) = F ×
−
2 2
(55)
3π (s − MW ) + Γ2W MW
2

where ΓW ≈ 2.1 GeV is the width of the W and resonance occurs for Eν = E0 = s/(2 me ) ≈
6.4 × 106 GeV. The peak cross section value is

1 G2F MW4
σ(E0 ) = ≈ 9 × 5.2 × 10−32 cm2 ≈ 4.7 × 10−31 cm2 (56)
BW →ν̄e e− 3πΓ2W

for a total of nine (3 leptonic and 6 hadronic) W− decay channels. Integrating Eq. 55 gives
the rate per electron as
2
!
π σ(E0 ) ΓW MW MW
Rate = φν̄e ≈ φν̄e (E0 ) × (2.4 × 10−25 GeV cm2 ). (57)
2 me 2 me

Detection of other exotic phenomena, such as multi-W production [168], is also possible.

7.3 AGN Jets

The recent observations of GeV γ-rays from a large number of extragalactic objects by
the EGRET instrument[170] on GRO has stimulated intense interest in models of particle
acceleration in jets with relativistic bulk flow. This is because most, if not all, of the
EGRET sources are radio-loud AGN, which are thought to be AGN’s viewed from a position
illuminated by the cone of a relativistic jet. Jets carry a sizeable fraction of the AGN
luminosity. Moreover, the apparent luminosity to an observer looking at a small angle to the
jet axis is increased by a factor of up to 104 for a jet Lorentz factor of 10. This is a consequence
[171] of the fact that I(ν)/ν 3 is a relativistic invariant, so that I(ν) = Γ3 I ∗ (νΓ−1 ), where I ∗
is the photon intensity (erg s−1 cm−2 sr−1 Hz−1 ) seen by an observer moving with the gas in
the jet and Γ is the Lorentz factor of the jet averaged over the cone of the jet relative to the
line of sight.
The interest intensified still further with the discovery of ∼TeV photons from the nearby
AGN Markarian 421 [115]. What is the origin of such high energy photons?
Proposed explanations can be divided into two classes. The traditional approach to the
production of very high energy photons is based on inverse Compton (IC) scattering of
accelerated electrons on a seed photon field. The photon field could be either external, i.e.
generated outside the electron acceleration region, or due to the synchrotron radiation of the
electrons (synchrotron-self-Compton). Examples in this category are Refs. [119,120,121].
An alternative approach is that of Mannheim et al. [37,172,173,176]. Following the
arguments of Ref. [143], it is assumed that protons also are accelerated in the jets. These
protons lose there energy by collisions on the synchrotron photons. In the process they
dump energy into photons, electrons and neutrinos via production of π 0 and π + . (The
jets are sufficiently diffuse so that high energy neutrons escape and production of π − is
therefore greatly suppressed.) The photons and electrons are reprocessed, and cascade to
form an E −2 power law differential photon spectrum down to the energy below which the

33
accelerating region becomes transparent. These photons which originate from interactions
of accelerated protons dominate the high energy signal in this picture. At low frequency
< 1015 Hz) synchrotron radiation from the electrons dominates.
(ν ∼
Both pictures have some difficulties to overcome. For example, the models that do not
involve nucleons generally require a higher bulk Lorentz factor of the jet. The radiation
target density has to be high enough for IC scattering and, at the same time, low enough
for the generated γ-rays not to be absorbed by γγ collisions. This is difficult to arrange
for, especially when IC scattering in the Klein-Nishina regime is the relevant process. In
addition, the energy densities in soft photons (IR to X-rays) and γ-rays are comparable,
which requires that the two types of radiation are generated in different locations [121]. To
prevent the electrons from losing too much energy to synchrotron radiation, the energy of
the magnetic field in the jet has to be of order 5% of the radiation density[120], far from the
0th order assumption of equipartition.
There is a corresponding set of problems that the models of hadronic origin have to
overcome. Because of the smaller rate of energy loss by protons, the jet Lorentz factor
is no longer a big limitation. The seed photon density, however, has to be high enough
for photoproduction to occur, and, as a consequence, γ-ray absorption in the source is a
problem. Similarly, there are several other free parameters of the model, such as the ratio
of power in protons and in electrons. These are fixed from multi-frequency observations. In
Fig. 11 we compare results of quasi-simultaneous observations of 3C273 in the optical, X-ray
and γ-ray bands [177] with the predictions of a synchrotron-self-Compton model [119] (solid
line) and a hadronic model [176]. The hadronic model gives correctly the spectral shape at
X-ray and γ-ray energy. The turn-up in the >10 GeV region can not be detected at Earth,
because of the absorption on the IR/optical background on propagation from the distant
source (z = 0.158 for 3C273).
The detection of neutrinos from blazars would confirm the hadronic model, since there
is no source of neutrinos in the electromagnetic models. In this connection, the uncertainty
in optical depth for photons in the source is particularly problematic because it introduces
extra model-dependence in the relation between the observed photon flux and the predicted
neutrino flux. This problem, of course, disappears once the neutrinos are detected. In the
remainder of this section we review the estimates of the predicted neutrino signals in the
hadronic model of production of high energy photons in jets of AGN.
A calculation of the neutrino production in AGN jets was first published by Biermann
and Mannheim[37,143,172]. The produced neutrino flux reflects the physical conditions in
the AGN jet. Because of the low photon density protons can achieve quite high energy at
acceleration (Epmax ∼ 3 × 1010 GeV in the frame moving with the bulk flow of gas in the jet
(jet frame).
In addition both the acceleration and proton interactions proceed in the jet frame, so the
neutrinos are boosted to high energy (blueshifted) with a Doppler factor of order 10. In the
case of 3C273 the flux is shown by a dotted line in Fig. 7.
This model [173] of the proton acceleration and interactions envisions a bulk flow of
magnetized plasma, streaming from the base of the jet towards its end (the hot spot). The
acceleration occurs in a sheets of radial dimension R and thickness D moving out with the
jet. In order to explain the variability of 3C279 on timescale of a day the radiating sheet
thickness should be D ∼ 1015 cm, much smaller than its radius (∼ 1018 cm). The acceleration

34
of protons (and possibly nuclei) proceeds via first order Fermi acceleration in the frame of
the flowing plasma. As usual, the acceleration rate is given by (compare Eq. 50)

1 dE eB c
∝ . (58)
E dt E
In this model, protons collide with the synchrotron photons generated by the accelerated elec-
trons, and the photon spectrum is approximated as a power law with integral index γ ≈ −1.
The threshold photon energy for production of the ∆ resonance obeys Eγ (threshold) ∝ Ep−1 ,
so the density of photons at the resonance is proportional to the proton energy. Since
equipartition is assumed between the energy in electrons and the magnetic field, the normal-
ization of the photon field is proportional to B 2 . Thus the loss rate for protons depends on
magnetic field and proton energy as

(tp )−1 ∝ Ep B 2 . (59)

Combining Eqs. 58 and 59 using the numerical values of Ref. [143], leads to an estimate of
the maximum proton energy in the jet frame of

Epmax,∗ ≤ 2 × 1010 B −1/2 GeV, (60)

where B is the magnetic field strength in Gauss.
The fact that the dominant energy loss process for high energy protons in the hadronic
model of AGN jets is photoproduction at threshold in collisions on a power law spectrum of
photons (rather than a thermal spectrum) has an interesting consequence also for the shape
of the produced spectrum of secondary pions. If, as expected in first order shock acceleration,
the differential proton spectrum is E −2 , then the differential pion production spectrum will
be harder by one power of energy, i.e. E −1 . Occasionally energetic protons also collide with
thermal gas in the jet, which leads to an E −2 spectrum of pions. Thus the characteristic
shape of the production spectrum of pions is E −2 at low energy, flattening to E −1 at high
energy up to some characteristic maximum energy.
The spectrum of neutrinos from decay of π + and µ+ follows the pion production spectrum,
shifted down in energy by appropriate kinematic factors. All energies are boosted by the
bulk Lorentz factor Γ relative to an external observer. The observed (boosted) maximum
neutrino energy is estimated by Mannheim [173] as Eνmax ∼ 109 GeV. The boosted energy
at which the observed neutrino spectrum flattens from E −2 to E −1 is in the range 105 for
3C273 [176] to 107 GeV for Mkn 421 and 3C279 [173]. Photons are even more complicated
than neutrinos because their spectra at production (from π 0 → γγ and from radiation by
electrons) are reprocessed by pair cascading in the ambient photon and magnetic fields. The
total photon luminosity eventually emerges from the source at lower energy (but boosted by
the bulk Lorentz factor). Approximately equal amounts of energy are carried by the four
leptons that result from the decay chain

π + → νµ µ+ → e+ νe ν̄µ .

In addition,
pγ → p π 0 ≈ 2 × pγ → n π +

35
at the ∆ resonance. Thus 3/4 of the energy lost to photoproduction ends up in the elec-
tromagnetic cascade and 1/4 goes to neutrinos. In addition, some of the energy of the
accelerated protons is lost to direct pair production (p + γ → e+ e− p). Thus
1
Lν ≤ Lγ (61)
4
Equation (61) can be used to relate an observed photon spectrum to a predicted neutrino
> TeV)
flux in the model. The relation is further complicated by attenuation of high energy (∼
photons during propagation from the source [116,174].
The spectra of both γ-rays and neutrinos are generated in pγ and, to a lesser extent, pp
interactions. For photoproduction the energy carried by neutrinos is directly related through
kinematics to the γ-ray luminosity as
3 3
Lν = Lπ+ = Lγ , (62)
4 13
where Lγ includes a contribution from e+ e− pairs. The predicted flux for 3C273 [176] is
shown in Fig. 7. It would give a rate of upward muons with Eµ > 1 TeV of ∼0.1 per year in
a 105 m2 detector.
The simultaneous observation of the BLLac source Mkn 421 by EGRET [178] and the
Whipple observatory [115] is especially valuable for the understanding of the physics of AGN
jets because of the large range of energy for detected photons. The two observations define an
energy spectrum with γ = 2.06 ± 0.04 over more than four decades in energy [179]. Mkn 421
is the closest source observed by EGRET at a redshift of 0.031. This is significant because
of the absorption on propagation. The exact energy dependence of the absorption feature is
uncertain because the magnitude and the energy spectra of the IR and optical background(s)
are not well known. Within a factor of 2, however, 3 TeV γ-rays emitted at the distance
of Mkn 421 will already start being absorbed and will show an apparent steepening of the
spectrum independently of the production spectrum.
One should appreciate that weakly interacting neutrinos will make their way to our
detectors unattenuated by ambient matter in the source or by IR light. So, while high energy
photons are absorbed on intergalactic IR photons for AGNs much further than Mkn 421,
neutrinos are not and sources should be detected with no counterpart in high energy photons.
Halzen and Vasquez [180] scale the Mkn 421 γ-ray flux to neutrino flux, making a range of
assumptions for the γ-ray absorption at source, expressed in terms of the magnetic field
value B of the jet. The VHE γ-ray flux from Mkn 421 is [179]
" #
Z
EdNγ
dE = 1.5 × 10−11 cm−2 s−1 . (63)
1/2 TeV dE

It is assumed [179] that the production spectrum (before attenuation) is given by a power law
with integral spectral index γ. The corresponding neutrino flux that would be expected if
the observed photon spectrum is to be understood in the hadronic model is quite uncertain.
This is because the amount of absorption in the source is not well determined. What is
needed is the optical depth of the source, i.e. the jet. Biermann [181] gives the following

36
(admittedly model dependent) estimate:
1/2 
B Eγ
 
τoptical = 2 . (64)
1 Gauss 1 TeV
The photon flux will be attenuated for energies above which τoptical = 1. According to (64)
the optical depth of the source is unity for the 0.5 TeV photons observed by Whipple for a
1 G field. The true value of the B-field in the jet is a guess which ranges from 10−4 to 104 G.
The gamma ray flux can be computed inside the source by correcting the observed flux (63)
for absorption in the jet. The answer depends critically on the magnitude of the B-field.
Once the unattenuated photon spectrum is established, the neutrino flux is estimated simply
by assuming one neutrino per gamma ray, as appropriate for pion decay.
Table 4 [179] shows the results for a range of assumptions for B and the spectral index
γ. We conclude that in this scenario Mkn 421 should produce a handful of upcoming muon
events per year in a generic 105 m2 detector. Within the framework of the model the magnetic
field would be limited to B ∼ < 1 Gauss because the observed γ-ray spectrum extends above
1 TeV. The situation could be different in other potential sources.

Table 4: Number of upcoming muons (N) per 105 m2 per y for the different scenarios. Lγ
is in 1043 erg/s.

B(Gauss) γ Epmax Eγ for τopt = 1 Lγ N

1 30 2
10−4 0.8 2 × 1022 eV 50 TeV 500 11
0.4 106 450
1 1 2 × 1020 eV 500 GeV 200 13

Stecker et al. [182] use the γ-ray absorption on propagation to normalize the expectations
from other GRO sources. They find a flux of neutrinos from the 3C273 jet sufficient to
generate ∼0.1 muons above 1 TeV in 105 m2 per year. The corresponding flux from the
3C273 core is 40% smaller. The quiescent state of 3C279 would generate 5 muons, while
the highest observed γ-ray flux from 3C279 would correspond to 25 such muons. The 3C279
core contribution is only 0.1 event. These are the bracketing values for the expectations of
neutrino fluxes from active galactic nuclei, it the γ-rays are indeed generated in hadronic
interactions by accelerated nuclei.
Analogously to the background from generic, radio-quiet AGN, one can integrate the
emission of all AGN jets to obtain an isotropic ultra high energy neutrino background. Two
estimates of this background, due to Stecker[183] and Mannheim[173], are shown in Fig. 8.
Although the overall normalization of the isotropic neutrino background from AGN jets is
lower that that of generic AGN, it extends to higher energy, and crosses it over at some very
high energy. The cross-over is explicit in the flux of Mannheim, while Stecker et al. give
only the slope (γ = 2) and the normalization. In any case, the normalization is somewhat
uncertain because of the variability of the blazars. The normalization of Mannheim’s diffuse

37
flux in Fig. 8 corresponds to an assumption of a 15% duty factor for blazars to be in a high
state [184].

7.4 AGN Neutrinos: Discussion and conclusions

Although the spherical accretion model is very useful for estimating the neutrino fluxes that
might be expected from cores of AGN, it is subject to criticism from various points of view.
The model is only applicable to accretion disks with thickness comparable to or exceeding the
dimension of the shock radius. It has to be constructed in such a way that there is no leakage
of the generated γ-rays before their energy is downscattered to X-ray and longer wavelengths.
Any significant leakage would exceed the experimental limits on diffuse extragalactic γ-rays.
At the same time, the radiation density cannot be so high as to prevent the acceleration
processes from occurring. Some authors [185] estimate the source efficiency Q to be lower
than 1/3 and ask if the conditions in the AGN nucleus are suitable for shock formation
at all. Others [186] show the danger of overproducing background radiation through pair
production and synchrotron radiation, which could lead to shock instability and drastically
decrease Epmax . Some of these problems might be avoided by placing the shocks in the bases
of the AGN jets [157].
The calculations of the neutrino production in AGN jets are no less difficult. To model
correctly all the jet physics one has to follow in some detail all the processes involved in the
frame of the relativistic plasma flow, including particle acceleration and reacceleration at
multiple shocks, γ-ray production, multiplication and absorption in electromagnetic cascades
in a non stationary fashion. This is very complicated problem that involves many free
parameters. The simple scaling of the neutrino fluxes with the γ-ray luminosity for individual
sources may not be exact, since the conditions at the source are poorly known. The ratio
of the magnetic to radiation energy density, for example, which is essential for the γ-ray
absorption at the source, can vary within at least one order of magnitude. The sources are
also highly variable, and many might have been observed during the peak of their activity.
The big question is the fraction of the AGN luminosity that goes through the nucleonic
channel. Although it has been pointed out [35,143,144] that hadrons have suitable interaction
cross sections and are a natural vehicle for the energy transport throughout the AGN disk,
nucleons are not strictly necessary for the solution of this problem. Since the models of the
non-nucleonic origin of the Mkn 421 γ-rays are already struggling to extend the theory to
γ-rays above 1 TeV, a possible observation of, say, 10 TeV γ-rays would be a confirmation
of their π 0 origin.
This is hardly possible, however, because of the absorption on the IR/optical background,
even if the production spectrum reaches much higher energy. There is only a slight chance
that [174], for very low values of the extragalactic magnetic field, the cascading on this
background will flatten considerably the spectrum observed in the GeV/TeV region. Such
flattening would reveal the extension of the production spectrum to much higher energy and
correspondingly confirm the π 0 origin of the γ-ray flux.
The criticism above does not imply that the current predictions are not reliable. They
are results of the first generation of research, which will become more exact in the near fu-
ture. The differences between various estimates reflects the uncertainties of the calculations.
Conclusions are that the expected fluxes of source neutrinos are well below the sensitivity

38
of the currently active deep underground detectors with effective area less than 1000 m2 .
They are, however, tantalizingly close to being detectable by the new generation of detectors
especially designed for neutrino astronomy.

8 Cosmological Neutrinos
Another possible source of extremely energetic diffuse neutrinos could be the interactions
of ultra high energy cosmic rays on the microwave background. The importance of such
interactions was noted by Greisen [187] and independently by Zatsepin and Kuzmin [188]
soon after the discovery of the background radiation. These early papers stated the existence
of an universal cut-off of the cosmic ray spectrum due to photopion production. The question
of the production of neutrinos and γ-rays was developed later in works by Wdowczyk et
al [189], Stecker [190], Hill & Scramm [191], Berezinsky and Grigorieva [192], Halzen et
al [193] and others, and in a recent paper of Yoshida & Teshima [194], who perform a
detailed Monte Carlo calculation of the proton propagation in the microwave background
and the generation of neutrinos.
The major source of energy loss is photoproduction, as described in §7.1. Here the target
is the microwave background, with a density of ∼ 400 photons/cm3 and an average energy
ǫ ≃ 7 × 10−4 eV, corresponding to the temperature of the background radiation. For cosmic
rays exceeding
∆2 − m2p 5 × 1020
Ep ≈ ≈ eV , (65)
2(1 − cos θ)ǫ (1 − cos θ)
where θ is the angle between the proton and photon directions, the photopion cross-section
grows very rapidly to reach a maximum of 540 µb at the ∆+ resonance (s = 1.52 GeV2 ). The
∆+ decays to pπ 0 with probability of 2/3, and to nπ + with probability 1/3. Neutral pions
give rise to ultra-high-energy γ-ray fluxes, and charged pions—to neutrino fluxes through
the decay channels of Eq. (12). Decay kinematics is such that all three neutrinos take
approximately 1/4 of the parent pion energy. In addition the neutrons also decay and
produce a small flux of ν̄e at much lower energy.
Because of the width of the thermal photon distribution, and the isotropic nature of
the microwave background, there is some phase space for photopion production at proton
energies as low as 1019 eV. These are only possible in head to head interactions on the high
energy tail of the microwave background spectrum (or on the infrared/optical background).
Most of the proton energy loss in this energy range, however, is on direct pair production
(pγ → pe+ e− ), which has a lower threshold but does not contribute to the neutrino fluxes.
Photopion production starts dominating at energy above 3 × 1019 eV and the cross-section
reaches maximum at ∼ 5 × 1020 eV , where the proton mean free path λp = (σpγ nγ )−1 is
≃ 5×1024 cm (∼2 Mpc). Since protons lose on the average 1/5 of their energy per interaction
the proton attenuation length Λp comes to ∼ 10 Mpc, a number that the exact calculation
of Ref. [194] shows is reached for proton energies above 1021 eV.
The magnitude and intensity of the cosmological neutrino fluxes is than determined by
the maximum injection energy of the ultra-high-energy cosmic rays and by the distribution
of the sources. If the sources are relatively close by, at distances measured in tens of Mpc,
and the maximum injection energy is not much greater than the highest observed cosmic ray

39
energy (few ×1020 eV ), the generated neutrino fluxes are negligible. If, however, the highest
energy cosmic rays are generated at many sources at large redshift, then a large fraction
of their injection energy would be presently contained in γ-ray and neutrino fluxes. The
most important reason is that the energy density of the microwave radiation, and the proton
photopion production cross-section, scales with (1 + z)4 . The effect is even stronger if the
source luminosity were increasing with z, i.e. cosmic ray sources were more active at large
redshifts—‘bright phase’ models.
The neutrino flux is given by an integral identical to Eq. (54), where g(z) and f (z) now
correspond to number density and the luminosity function of the cosmic ray sources. In
general, the cosmic ray sources are defined by their injection spectra, luminosity and cosmo-
logical evolution. The normalization comes from the requirement that the ultra-high-energy
cosmic rays after propagation in the microwave background match the observed spectra. The
loss resulting in γ-ray fluxes, downscattered on the microwave background, should not vio-
late the experimental limits on isotropic extragalactic γ-rays [195]. Yoshida & Teshima [194]
give the muon and electron neutrinos separately for different injection models characterized
by the zmax value, maximum injection energy Emax and different source evolution functions
of the form η(z) = η0 (1 + z)m . Fig. 15 below shows these fluxes for Emax = 1022 eV and two
extreme sets of evolution parameters: m = 0, zmax = 2 (low) and m = 4, zmax = 4 (high).
It is important to remember that such drastically different source evolution models can fit
equally well the observed cosmic ray spectrum.
There are specific models that identify the sources of the extragalactic cosmic rays.
Rachen & Biermann [196] propose that hot spots of Fanaroff-Riley class II radio galax-
ies, being the largest and most powerful shock waves in the Universe, are the dominant
sources of cosmic rays of energy above 1018 eV. In this case g(z) and f (z) are the num-
ber and luminosity density functions of FR-II galaxies, derived from radio observations at
particular radio frequencies. They do not calculate the neutrino fluxes generated in cosmic
ray interactions of the microwave background, but since their models match the observed
cosmic ray flux at energies around 1018 eV, such a calculation should be close to the results
of Ref. [194] for similar source evolution functions, i.e. to be bracketed by the extreme fluxes
shown on Fig. 15.
Independently of the specific model of the sources of the highest energy cosmic rays,
the associated neutrino fluxes can only dominate the highest energy region, above Eν =
1017 –1018 eV. This would only happen in the case that cosmic rays of energy above 1018
eV are indeed accelerated at numerous high redshift astrophysical sources. At lower energy
the neutrino background is dominated by the neutrinos generated in interactions at active
galactic nuclei
An intriguing possibility is (see Refs. [197,198]) that the highest energy cosmic rays are
produced by energy loss of superconducting cosmic strings. The strings lose energy in the
form of massive fermions (MF ∝ 1015 GeV) that decay into baryons and fermions. The
spectrum is modified by interactions on the 3K background but extends up to the Planck
mass. This, plus the contribution from strings at large red-shift, increases the cosmic ray
energy loss in the > 1019 eV range. A neutrino flux a factor of 30 higher than the proton
flux is created. Only nucleons (not nuclei) can be generated through this channel.
More recently the neutrino emission from cosmic strings has been discussed in the context
of ‘cusp annihilation’ [199,200,201]. The total energy of the string in the region of the cusp

40
is released in the form of massive scalar and gauge particles that decay rapidly into particle
jets. Equal numbers of particles and antiparticles are generated by conservation of quantum
numbers. Neutrinos come mostly from the ordinary pion decay channels. The shape of the
jet fragmentation function is crucial for the number of generated neutrinos and their energy
spectrum. The assumptions used require an extrapolation of the observed jet fragmentation
function up to the Planck scale. Another crucial parameter is µ, the string mass per unit
length, which defines to total luminosity of the string. For Gµ/c2 ≃ 10−6 , consistent with
large-scale structure formation and the observed anisotropy of the microwave background,
the generated neutrino fluxes are smaller than the predictions from ultra-high-energy proton
propagation. Maximum neutrino fluxes are obtained for Gµ/c2 ≃ 10−15 . In this case,
however, the cosmic strings would not have other cosmological implications.

9 Search for Dark Matter

It is believed that most of our Universe is made of cold dark matter particles. In the context
of big bang cosmology, these particles have interactions of order the weak scale and masses
of order MW , i.e. they are WIMPs [202]. From rotation curve measurements we also know
their density and average velocity in the galactic halo. This information is the basis for
estimating the annihilation rate of WIMPS into high energy neutrinos.
Galactic WIMPs, scattering off protons in the sun, lose energy. They may fall below
escape velocity and be gravitationally trapped. Trapped dark matter particles eventually
come to equilibrium temperature, and therefore to rest at the center of the sun. While
the WIMP density builds up, their annihilation rate into lighter particles increases until
equilibrium is achieved where the annihilation rate equals half of the capture rate. The sun
has thus become a reservoir of WIMPs which annihilate into any open fermion, gauge boson
or Higgs channels. The leptonic decays from annihilation channels such as bb̄ heavy quark
pairs and W + W − turn the sun into a source of high energy neutrinos. Their energies are
in the GeV to TeV range, rather than in the keV to MeV range familiar from its nuclear
burning. These neutrinos can be detected in deep underground experiments.
We will illustrate the power of neutrino telescopes as dark matter detectors using as an
example the search for a 500 GeV WIMP with a mass outside the reach of present accelerator
and future LHC experiments. A quantitative estimate of the rate of high energy muons of
WIMP origin triggering a detector can be made in 5 easy steps. An exact quantitative
calculation requires a complex code [203].
Step 1: The halo neutralino flux φχ . It is given by their number density and average velocity.
The cold dark matter density implied by the observed galactic rotation curves is ρχ = 0.4
GeV/cm3 . The galactic halo is believed to be an isothermal sphere of WIMPs with average
velocity vχ = 300 km/sec. The number density is then
" #
−4 500 GeV
nχ = 8 × 10 cm−3 (66)
mχ

and therefore " #

4 500 GeV
φχ = nχ vχ = 2 × 10 cm−2 s−1 . (67)
mχ

41
Step 2: Cross section σsun for the capture of neutralinos by the sun. The probability that a
WIMP is captured is proportional to the number of target hydrogen nuclei in the sun (i.e.
the solar mass divided by the nucleon mass) and the WIMP-nucleon scattering cross section.
2
From dimensional analysis σ(χN) ∼ (GF m2N ) /m2Z which we can envisage as the exchange
of a neutral weak boson between the WIMP and a quark in the nucleon. The main point
is that the WIMP is known to be weakly interacting. We obtain for the solar capture cross
section
Msun h ih i
Σsun = nσ = σ(χN) = 1.2 × 1057 10−41 cm2 . (68)
mN
Step 3: Capture rate Ncap of neutralinos by the sun. Ncap is determined by the WIMP flux
(67) and the sun’s capture cross section (68) obtained in the first 2 steps:

Ncap = φχ Σsun = 3 × 1020 s−1 for mχ = 500 GeV. (69)

Step 4: Number of solar neutrinos of dark matter origin One can check that the sun comes to
a steady state where capture and annihilation of WIMPs are in equilibrium. For a 500 GeV
WIMP the dominant annihilation rate is into weak bosons; each produces muon-neutrinos
with a leptonic branching ratio which is roughly 10%:

χχ̄ → W W → µνµ . (70)

Therefore, as we get 2 W ’s for each capture, the number of neutrinos generated in the sun is
1
Nν = Ncap (71)
5
and the corresponding neutrino flux at Earth is given by
Nν
φν = = 2 × 10−8 cm−2 s−1 , (72)
4πd2
where the distance d is 1 astronomical unit.
Step 5: Event rate in a high energy neutrino telescope. For (70) the W -energy is approx-
imately mχ and the neutrino energy half that by 2-body kinematics. The energy of the
detected muon is given by
1 1
Eµ ≃ Eν ≃ mχ . (73)
2 4
where we used the fact that, in this energy range, roughly half of the neutrino energy is
transferred to the muon. For the neutrino flux given by (72) we obtain

# events/year = 105 × φν × ρH2 O × σν→µ × Rµ ≃ 100 (74)

for a 105 m2 water cherenkov detector, where Rµ is the muon range and φν × ρH2 O × σν→µ is
the simple analog of Eq. (19).
This exercise illustrates that present high energy neutrino telescopes (of area ∼104 m2 )
are powerful devices in the search for dark matter and supersymmetry. They are complemen-
tary to present and future accelerator searches in the sense that they are naturally sensitive

42
to heavier WIMP’s because the underground high energy neutrino detectors have been op-
timized to be sensitive in the energy region where the neutrino interaction cross section and
the range of the muon are large. Also, for high energy neutrinos the muon and neutrino are
nicely aligned along a direction pointing back to the sun with good angular resolution. In
addition, in the estimate given above we neglected some other decay channels that contribute
neutrinos, and we did not include the signal from annihilation of WIMP’s trapped in the
center of the Earth [205]. Direct searches for dark matter are clearly highly desirable. To
achieve comparable sensitivity to these indirect searches requires a sensitivity at the level of
0.05 events/kg day [204].
An elegant way to extend the Standard Model is to make it supersymmetric [207]. If
supersymmetry is indeed Nature’s extension of the Standard Model it must produce new
phenomena at or below the TeV scale. An attractive feature of supersymmetry is that
it provides cosmology with a natural dark matter candidate in form of a stable, lightest
supersymmetric particle [202]. There are a priori six candidates: the (s)neutrino, axi(o)n(o),
gravitino and neutralino. These are, in fact, the only candidates because supersymmetry
completes the Standard Model all the way to the GUT scale where its forces apparently
unify. Because supersymmetry logically completes the Standard Model with no other new
physics threshold up to the GUT-scale, it must supply the dark matter. Here we will focus
on the neutralino, which, along with the axion, is for various reasons the most attractive
WIMP candidate [206].
The supersymmetric partners of the photon, neutral weak boson and the two Higgs
particles form four neutral states, the lightest of which is the stable neutralino

χ = z11 W̃3 + z12 B̃ + z13 H̃1 + z14 H̃2 . (75)

In the minimal supersymmetric model (MSSM) [207] down- and up-quarks acquire mass by
coupling to different Higgs particles, usually denoted by H1 and H2 , the lightest of which
is required to have a mass of order the Z-mass. Although the MSSM provides us with a
definite calculational framework, its parameters are many. For the present discussion we
only have to focus on the following terms in the MSSM lagrangian
1 1 1 1
L = · · · µH̃1 H̃2 − M1 B̃ B̃ − M2 W̃3 W̃3 − √ gv1 H̃1 W̃3 − √ gv2 H̃2 W̃3 + · · · , (76)
2 2 2 2
which introduce the (unphysical) masses M1 , M2 and µ associated with the neutral gauge
bosons and Higgs particles, respectively. M1 and M2 are related by the Weinberg angle.
The lagrangian introduces two Higgs vacuum expectation values v1,2 ; the coupling g is the
known Standard Model SU(2) coupling. Although the parameter space of the MSSM is more
complex, a first discussion of dark matter uses just 3 parameters:

µ, M2 , and tan β = v2 /v1 . (77)

Further parameters which can also be varied include the masses of top, Higgs, squarks, etc.
Neutralino masses less than a few tens of GeV have been excluded by unsuccessful collider
searches. For supersymmetry to resolve the hierarchy problem of the Standard Model the
masses of supersymmetric particles must be of order the weak scale and therefore, in practice,
at the TeV scale or below. Also, if neutralinos have masses of order a few TeV and above,

43
they overclose the Universe. Despite its rich parameter space supersymmetry has therefore
been framed inside a well defined GeV–TeV mass window.
Assuming supersymmetry we can fill in some factors in the “back-of-the-envelope” es-
timates in the previous section. In supersymmetry, heavy WIMPs annihilate preferentially
into weak bosons. Other important annihilation channels include [208]

χ + χ̄ → b + b̄. (78)

Heavy quark decays dominate neutralino annihilation below the W W -threshold. Also the
dimensional estimate of the neutralino-nucleon interaction cross section σ(χN) can be re-
placed by an explicit calculation. It supports the dimensional estimate in the previous
section. σ(χN) receives contributions from 2 classes of diagrams: the exchange of Higgses
and weak bosons, and the exchange of squarks. The result is often dominated by the large
coherent cross section associated with the exchange of the lightest Higgs particle H2 and is
of the form
m2χ m2Z
σ = αH (GF m2N )2 (79)
(mN + mχ )2 m4H
or, for large mχ
 2 m2Z
σ = αH GF m2N . (80)
m4H
The proportionality parameter αH is of order unity, but can become as small as 10−2 in
some regions of the MSSM parameter space. This is illustrated in Fig. 12 where the MSSM
parameter space is parametrized in terms of the unphysical masses M(µ) of the unmixed
wino(Higgsino). (The ratio of the vacuum expectation values associated with the two Higgs
particles v2 /v1 (= 2) is here fixed to some arbitrary value.) The relation of these parameters to
the neutralino mass is shown in the figure. The full lines show fixed values of the neutralino
mass mχ . The lines labelled by squares trace fixed values of the “coupling” αH . The
dashed area indicates M, µ values which are excluded by cosmological considerations. In
standard big bang cosmology neutralinos with the corresponding parameters will overclose
the Universe.
Note that for a given χ mass there are two possible states with the same αH value. One
of them will preferentially annihilate into weak bosons, the other into fermions. Therefore,
their neutrino signature is provided by W, Z decay and semi-leptonic heavy quark decays,
respectively. Fig. 12 illustrates that, for heavy neutralinos, which can only be searched for by
the indirect methods discussed here and are therefore of prime interest, any detector which
can study dark matter with αH as small as 0.1 can exclude the bulk of the phase space
currently available to MSSM dark matter candidates.
Our main conclusions are summarized in Fig. 13 which exhibits, as a function of the
neutralino mass, the detector area required to observe one event per year. The detailed
calculation confirms our previous estimate of 100 events per 105 m2 per year for a 500
GeV neutralino. The two branches in this and the following figures correspond to the two
solutions for a fixed neutralino mass; see also Fig. 12. Various annihilation thresholds are
clearly visible, most noticeable is the threshold associated with the W, Z mass near 100 GeV.
The graphs confirm that large detectors are required to study the full neutralino mass range.
It is clear from Fig. 13, however, that even detectors of more modest size significantly extend

44
the range explored by accelerators [209,210]. Neutralinos of 1 TeV mass are observable in
a detector of area a few times 103 m2 . The energy of the produced neutrinos is typically “a
fraction” of the neutralino mass, e.g. 1/2 for neutralino annihilation into a W followed by a
leptonic eν decay. For lower masses the event rates are small because the detection efficiency
for low energy neutrinos is reduced. This mass range has, however, already been excluded
by accelerator experiments. For very high masses the number density of neutralinos, and
therefore the event rate, becomes small. This is not a problem as problematically large
masses are excluded by theoretical arguments as previously discussed. The same results
are shown in Fig. 14(a) as contours in the M, µ plane which denote the neutrino detection
area required for observation of 1 event per year. Clearly the 105 m2 contour covers the
parameter space. The problematic large µ, M2 -region does not really represent a problem
as its parameters lead to values of the matter density Ω exceeding unity as shown in the
accompanying Fig. 14(b).
A realistic evaluation of the reach of an underground detector requires more than counting
events per year. Realistic simulations of statistics and systematics must be done. Also a
more complete mapping of the MSSM parameter space is required. For those interested we
refer to Ref. [203].

10 Event Rates in a Generic 0.1 km2 Detector: Synthesis

In the preceding sections we have attempted to summarize the new limits that will be set
and the most likely observations that may me made by the next generation of high energy
neutrino telescopes. If past history is a guide, however, the most important discoveries that
occur when a new window is opened may be completely unanticipated.
We summarize some of the estimated event rates in Table 5. The corresponding neutrino
fluxes are presented on Figs. 15 and 16. We remind the reader that a 0.1 km2 detector is
2500 times larger than IMB, 100 times MACRO or LVD, but only “a factor” larger than
many of the detectors under consideration or construction [211], e.g. AMANDA, BAIKAL,
DUMAND and NESTOR. A list of operating and proposed underground detectors having
the capability to detect high energy neutrinos is given in Table 6.
Table 5 gives the rates of upward going neutrino induced muons of atmospheric and
extraterrestrial origin. The absorption in the Earth becomes important for the flatter ex-
traterrestrial neutrino fluxes and the event rates are given both with and without absorption.
The event rates expected from astrophysical neutrino sources are estimated with an account
for absorption.
Some of the event rates in Table 5 predicted for the same type of source differ from each
other by two orders of magnitude. This reflects the degree of uncertainty of our knowledge
about the conditions and the role of different physical processes for the energetics of the
source. The low event rates from the diffuse AGN background come from the revised calcu-
lations of Stecker et al. [36], while the high rates reflect the highest neutrino background of
Protheroe and Szabo [38]. These highest rates are, however, in contradiction with limits set
by the Frejus experiment [160,161,162]. The highest rate of diffuse TeV muons allowed by
the Frejus limit is ∼200 per 105 m2 per year.
The estimated event rates from galactic sources come from the considerations presented

45
 Table 5:

EVENTS PER YEAR IN 0.1 KM2

• ATMOSPHERIC (angle averaged, per steradian)
muon energy [72] [75]
> 1 GeV 7800 8300
> 1 TeV 129 104
• ATMOSPHERIC in 1◦ circle, Ref. [75]
muon energy cos θ = 0.05 cos θ = 0.95
> 1 GeV 12.6 5.6
> 1 TeV 0.21 0.05
• EXTRATERRESTRIAL FLUXES (angle averaged)
φν = 2.7 × 10−5 (Eν /GeV )−1.7 cm−2 s−1
muon energy no abs. with abs.
> 1 GeV 32.7 32.0
> 1 TeV 4.3 3.8
φν = 4.0 × 10−8 (Eν /GeV )−1 cm−2 s−1
muon energy no abs. with abs.
> 1 GeV 8.8 6.6
> 1 TeV 5.0 3.3
• ASTROPHYSICAL DIFFUSE FLUXES (per steradian)
muon energy plane of galaxy AGN
> 1 GeV 12–20 80–1600
> 1 TeV 1.5–3.0 40–800
also νe (6.3 PeV) + e → W − 0.3 per 1000 kton
• ASTROPHYSICAL POINT SOURCES (Eµ > 1 TeV)
Galactic source (Eq. 35)/100 2.6
Extragalactic source (3C273)
J 0.1–25
• 500 GeV WIMPS from 100

in §6. The event rate generated by the flux of Eq. (35) is an extreme upper limit, which would
be difficult to reconcile with the observational limits from VHE/UHE γ-ray observations. As
a conservative estimate we quote a rate corresponding to a neutrino flux smaller by two orders
of magnitude. The least certain rate is the one expected for single AGN, given in Table 5
for the source 3C273. The smaller rate (0.1/yr) comes from the estimates for emission from
the jets by Mannheim [176] and Stecker et al [182]. The highest event rate (25/yr) actually
comes to the neutrino flux corresponding to the ‘high state’ γ-ray flux of 3C279 [182]. It is
not likely that such a high luminosity could be maintained at the source for periods as long
as a whole year. The atmospheric neutrino background relevant for source searches is given
in two directions: close to the zenith and close to the horizon.
Even the smallest predicted event rates for the diffuse AGN background are easily de-
tectable by a 105 m2 neutrino telescope. The expected ratio of signal to background for
TeV muons is from 0.3 to more than 2. The background, atmospheric neutrino rate, is large
and allows calibration and continuous monitoring of the detector. The observations of the
diffuse neutrino flux from the plane of the galaxy are much more difficult, although the likely

46
 Table 6:
OPERATING DETECTORS WITH HIGH ENERGY NEUTRINO
DETECTION CAPABILITY
Detector Location Area (m2 ) ∗ Technique
NUSEX Mont Blanc 10 streamer tubes/Fe
KGF India 20 streamer tubes, very deep
SOUDAN II USA 100 drift tubes/concrete
KAMIOKANDE Japan 120 water Cherenkov
BAKSAN Caucasus 250 liquid scintillator tanks
IMB USA 400 water Cherenkov
LVD Gran Sasso 300 † liquid scintillator, streamer tubes
MACRO Gran Sasso 850 liquid scintillator, streamer tubes

FUTURE INITIATIVES (partial list)

Detector Location Area (m2 ) Technique
SNO Canada 600 D2 O
SUPERKAMIOKANDE Japan 740 water Cherenkov
BAIKAL Baikal 2000 water Cherenkov
GRANDE type USA, Italy, Japan ∼30000 water Cherenkov
DUMAND Hawaii 20000+ water Cherenkov
AMANDA South Pole 20000+ Cherenkov in deep ice
RAMAND Antarctica 106 m2 ? microwave detection

concentration of the excess events in the direction of the galactic plane should be of some
help.
The atmospheric background in point source searches is generally small. For energies of a
TeV or more the neutrino direction can be reconstructed to 1 degree or better. We therefore
expect less than one event per year in a 1◦ bin from the combined atmospheric and diffuse
AGN backgrounds. It is then quite likely that one or more sources will be discovered by a
105 m2 detector, provided that hadronic processes play an important role in the energetics
of powerful astrophysical objects.
With an even larger 1 km2 detector we could then begin to study neutrino sources in
some detail. It may be possible not only to count sources, but also to observe a multiplicity
of sources with enough statistics to begin extracting information from energy spectra and
temporal behaviour, particularly in comparison with photon observations. Comparisons of
the neutrino and γ-ray spectra contain information about the photon absorption at source
1
The total detector area for vertical upward upward going neutrinos is given for all existing detectors. The
effective area for source searches depends on the detector and source location and on the detector efficiency
for different zenith angles.
2
In operation

47
and during propagation to Earth, i.e. about important physical properties of the sources and
the intergalactic medium. We should be able to observe episodic flux increases and maybe
even the periodicity of the neutrino emission from binary sources. It is unlikely that such
detailed observations can be carried out with detectors smaller than 1 km2 .
Finally, we mention the possibility that such a neutrino telescope can carry out Earth
tomography [212], employing the attenuation of ultra high energy neutrinos, and making a
direct density profile of the Earth (seismic measurements only yield velocity profiles, and
moreover give little information on the Earth’s core).
Halzen & Learned [213] have presented arguments such as these for doing neutrino astron-
omy on the scale of 1 kilometer. In order to achieve large area it is unfortunately necessary
to abandon the low MeV thresholds of detectors such as IMB and Kamiokande. One focuses
on high energies where: i) neutrino cross sections are large, ii) the muon range is increased,
iii) the angle between the muon and parent neutrino is less than 1 degree and, iv) the at-
mospheric neutrino background is small. The accelerator physicist’s method for building a
neutrino detector uses absorber, 3 chambers with x, y wires with associated electronics with
a price of 104 US dollars per m2 . Such a 1 km2 detector would cost 10 billion dollars. It
is therefore a high priority to find methods which are more cost-effective to be able even-
tually to commission neutrino telescopes with effective area of order 1 km2 . Obviously the
proven technique developed by IMB, Kamiokande and others cannot be extrapolated to the
1 km scale. All present proposals do however exploit the Cherenkov technique well-proven
by these experiments. The direction of the neutrino is inferred from the muon direction
which is reconstructed by mapping the Cherenkov cone of the muon travelling through the
detector. The arrival times and amplitudes of the Cherenkov photons, recorded by a grid of
detectors, are used to reconstruct the track of the radiating muon.
Detectors presently under construction have a nominal effective area of 104 m2 . Baikal
is presently operating 36 optical modules and the South Pole AMANDA experiment started
operating 4 strings with 20 optical modules each in January 94. The first generation tele-
scopes [214] will consist of roughly 200 optical modules (OM). The experimental advantages
and challenges are different for each experiment and, in this sense, they nicely complement
one another. Briefly,

• AMANDA is operating in deep clear ice with an attenuation length in excess of 60 m,

which is similar to that of the clearest water used in the Kamiokande and IMB de-
tectors. Although residual bubbles are found at depth as large as 1 km, their density
decreases rapidly with depth. Ice at the South Pole should be bubble-free below 1100-
1300 m as it is in other polar regions [215]. The ice provides a convenient mechanical
support for the detector. The immediate advantage is that all electronics can be posi-
tioned at the surface. Only the optical modules are deployed into the deep ice. Polar
ice is a sterile medium with a concentration of radioactive elements reduced by more
than 10−4 compared to sea or lake water. The low background results in an improved
sensitivity which allows for the detection of high energy muons with very simple trig-
ger schemes which are implemented by off-the-shelf electronics. Being positioned under
only 1 km of ice it is operating in a high cosmic ray muon background. The challenge is
to reject the down-going muon background relative to the up-coming neutrino-induced
muons by a factor larger than 106 . The group claims to be able to meet this challenge

48
 with an up/down rejection which is at present similar to that of the deeper detectors.
The task is, of course, facilitated by the low background noise. The polar environment
is difficult as well, with restricted access and one-shot deployment of photomultiplier
strings. The technology has, however, been satisfactorily demonstrated with the de-
ployment of the first 4 strings. It is now clear that the hot water drilling technique can
be used to deploy OM’s larger than the 8 inch photomultiplier tubes now used to any
depth in the 2.8 km deep ice cover.

• BAIKAL shares the shallow depth with AMANDA, and has half its optical modules
pointing up, half down. The depth of the lake is 1.4 km , so the experiment cannot
expand downwards and will have to grow horizontally. Optical backgrounds similar in
magnitude to ocean water have been discovered in Lake Baikal. The Baikal group has
been operating for one year an array with 18 down-looking Quasar photomultiplier (a
Russian-made 15 inch tube) units in April 1993, and may well count the first neutrinos
in a natural water Cherenkov detector.

• DUMAND will be positioned under 4.5 km of ocean water, below most biological
activity and well shielded from cosmic ray muon backgrounds, which are a factor of
100 lower than for the shallower detectors. One nuisance of the ocean is the background
light resulting from radioactive decays, mostly K40 , plus some bioluminescence, yielding
a noise rate of 60 kHz in a single OM. Deep ocean water is, on the other hand very
clear, with an attenuation length of order 40 m in the blue. The deep ocean is a difficult
location for access and service. Detection equipment must be built to high reliability
standards, and the data must be transmitted to the shore station for processing. It has
required years to develop the necessary technology and learn to work in an environment
foreign to high-energy physics experimentation. The DUMAND group has successfully
analysed data on cosmic ray muons from the deployment of a test string [19]. The
power and signal cables from the detector location to shore (length 25 km) and the
junction box are already installed. The group will proceed with the deployment of
three strings in 1995.

• NESTOR is similar to DUMAND, being placed in the deep ocean (the Mediterranean),
except for two critical differences. Half of its optical modules will point up as in
BAIKAL. The angular response of the detector is being tuned to be much more
isotropic than either AMANDA or DUMAND, which will give it advantages in, for
instance, the study of neutrino oscillations. Secondly, NESTOR will have a higher
density of photocathode (in some substantial volume) than the other detectors, and
will be able to make local coincidences on lower energy events, even perhaps down to
the supernova energy range (tens of MeV).

Other detectors have been proposed for near surface lakes or ponds (e.g.
GRANDE, LENA, NET, PAN and the Blue Lake Project), but at this time none is in
construction [214]. These detectors all would have the great advantage of accessibility and
ability for dual use as extensive air shower detectors, but suffer from the 1010 –1011 down-to-
up ratio of muons, and face great civil engineering costs (for water systems and light-tight

49
containers). Even if any of these are built it would seem that the costs may be too large to
contemplate a full, kilometer-scale detector.
In summary, there are four major experiments proceeding with construction, each of
which has different strengths and faces different challenges. For the construction of a 1 km
scale detector one can imagine any of the above detectors being the basic building block for
the ultimate 1 km3 telescope. The redesigned AMANDA detector (with spacings optimized
to the attenuation length in excess of 60 m), for example, consists of 5 strings on a circle
with 60 meter radius around a string at the center (referred to as a 1 + 5 configuration).
Each string contains 13 OMs separated by 15-20 m. Its effective volume is just below 107 m3 .
Imagine AMANDA “supermodules” which are obtained by extending the basic string length
(and module count per string) by a factor close to 4. Supermodules would then consist of
1 + 5 strings with 51 OMs separated by 20 meters on each string, for a total length of 1 km.
A 1 km scale detector then might consist of a 1 + 7 + 7 configuration of supermodules, with
the 7 supermodules distributed on a circle of radius 250 m and 7 more on a circle of 500 m.
The full detector then contains 4590 phototubes, which is less than the 7000 used in the
SNO detector. Such a detector (see Fig. 17) can be operated in a dual mode:
• it obviously consists of roughly 4 × 15 the presently designed AMANDA array, leading
to an effective volume of ∼ 6 × 108 m3 . Importantly, the characteristics of the detector,
including threshold in the GeV-energy range, are the same as those of the AMANDA
array module.

• the 1 + 7 + 7 supermodule configuration, looked at as a whole, instruments a 1 km3

cylinder with diameter and height of 1000 m with optical modules. High-energy muons
will be superbly reconstructed as they can produce triggers in 2 or more of the super-
modules spaced by large distance. Reaching more than one supermodule (range of
250 m) requires energy of 50 GeV. We note that this is the energy for which a neu-
trino telescope has optimal sensitivity to a typical E −2 source (background falls with
threshold energy, and until about 1 TeV little signal is lost).
Alternate methods to reach the 1 km scale have been discussed by Learned and Roberts [216].
How realistic are the construction costs for such a detector? AMANDA’s strings (with
10 OMs) cost $150,000 including deployment. By naive scaling the final cost of the pos-
tulated 1 + 7 + 7 array of supermodules is of order $75 million, comparable to that of
Superkamiokande [217] (with 11,200 × 20 inch photomultiplier tubes in a 40 m diameter by
40 m high stainless steel tank in a deep mine). It is clear that the naive estimate makes
several approximations over- and underestimating the actual cost.
In the next and final section of this paper we will briefly review alternate ideas for making
a kilometer-scale neutrino detector in a cost-effective manner.

11 Alternative Methods for Neutrino Detection

11.1 Radio Detection

Over 30 years ago the suggestion was first made that radio antennas might be able to detect
microwave emission from neutrino-induced cascades [218]. This method can in principle lead

50
to the construction of relatively inexpensive neutrino telescopes. Until recently only rough
estimates have been made of the power emitted [219,220]. They indicated that the detection
threshold is quite high, so high that the technique is probably insensitive to the atmospheric
neutrinos.
That the radio signals emitted by neutrino-induced electromagnetic cascades are even
close to observability is the result of interesting physics. According to the Frank-Tamm
formula the power radiated by a particle with charge ze travelling a pathlength l in a medium
of refractive index n is given by [221]

4π 2 h̄
" # " #
dW 1
= α z2ν 1 − 2 2 l , (81)
dν c β n

where ν is the frequency and α = (137)−1. Naively one might expect the power generated in a
shower of N charged particles to be proportional to Nhli. This is not correct. If the emitted
wavelength is large compared to the physical dimensions of the shower, or equivalently,
the electric pulse generated by the shower is short compared to the period of the waves
observed, then the emission by the shower particles is coherent and the power is of order
(∆q · N)2 hli [218,222]. Here ∆q is the excess negative charge in the shower

N(e− ) − N(e+ )
∆q = , (82)
N(e− ) + N(e+ )
which enters in the coherent case because the electric fields from opposite charges cancel
for the same hli. ∆q is positive mainly because Compton scattering of shower photons
on atomic electrons creates an excess of negative charges in the shower. Coherence thus
implies an enhancement (∆q)2 N which can compensate the loss in power associated with
the ν dependence of Eq. (81). Roughly, from visible light to GHz radiowaves the suppres-
sion associated with the factor ν in the Frank-Tamm relation is a factor 106 which can be
compensated by coherence because N is of order 106 for PeV showers.
The technique has been successfully tested in experiments measuring radio emission by air
showers observed in coincidence with particle arrays [223]. Whereas atmospheric fluctuations
make the systematics of the radioemission difficult to handle, this is not a problem in denser
material, like ice. The physical dimension of the shower is reduced because the radiation
length is only 39 cm. The coherence is retained to higher frequencies, where more energy
is available. Determination of the precise threshold for observation in a medium like ice
depends on the details of the cascade and one has to perform a real time numerical simulation
of electromagnetic cascades in ice. The calculations [224] show that to a very good precision
the enhancement factor from coherence (∆q)2 N is proportional to the primary energy, and
therefore the power of the radioemission is proportional to the square of the cascade energy.
The critical parameter is the energy threshold for detection which not only depends on
the power generated but also on the absorption in the ice and the background noise from the
apparatus and its environment. Absorption at 1 GHz depends critically on temperature and
therefore on the location of the experiment. Determination of in situ background noise is a
complex problem. Experiments indicate that thermal noise of temperature 300K represents
an adequate guess of the background [225]. For this assumption the amplitude of the noise
spectrum rises linearly with frequency, i.e. it exhibits the same dependence as the signal below

51
∼ 1 GHz [223]. In a detector of bandwidth ∆ν the noise varies as ∆ν 1/2 , and therefore the
bandwidth enhances the signal to noise ratio by the square root of the bandwidth. Neglecting
absorption and assuming ∆ν = 1 GHz, a signal to noise ratio of unity is achieved for a
detection threshold linearly proportional to the distance r from the shower,
Eth (PeV) ≃ 5 × 103 r (m) . (83)
E.g. only neutrinos of 5 PeV and above can be sampled in 1 km of ice. This is a very
high threshold. Existing experiments already set limits on high neutrino fluxes which imply
extremely low event rates above 5 PeV. This threshold estimate is based on a signal to noise
argument. The power in the signal is about an order of magnitude smaller than the result
quoted by Zeleznykh and collaborators [219].

11.2 Acoustic Detection

High detection threshold may also be the main shortcoming of the idea of acoustic detection
of neutrinos, originally suggested by Bowen [226]. Acoustic waves are produced by the
heating of the medium by the ionization loss of the cascade particles. Shower direction can
also be determined by the timing and amplitude of the signal in different detectors [227]. The
beam pattern is coherent and reinforcing in a plane perpendicular to the cascade direction.
One can envisage a detector consisting of pressure transducers placed in an array. The
optimal depth is not as yet clear, but it would have to be deep enough to avoid the problem
of acoustic wave scattering, and perhaps also to escape high frequency surface noise. The
acoustic losses set the size scale of the detector.
Laboratory data [228] indicate that pure ice has a sound velocity (of compressional waves)
of v = 3200 m/s and a Q of perhaps 1000 at low temperatures. South Pole ice near the surface
is below −50◦ C. With an attenuation length of about 100 m, one can imagine detectors in a
lattice of roughly 100 m spacing. Using Bowen’s [226] figure of merit, we estimate a signal-
to-noise ratio of unity in a single detector at 100 m distance from a 6 PeV cascade, with the
signal-to-noise ratio scaling as the square of energy/distance. The frequency maximum is
about 20k Hz. Despite the dauntingly high detection threshold in comparison to the optical
technique, acoustic detection remains interesting enough for further investigation for several
reasons. First, acoustic sensors are compact, pressure tolerant, inexpensive (piezoelectric
sensors are cheap compared to photomultipliers), and could be installed relatively easily. It
does not seem outrageous to imagine a 30 × 30 lattice of strings extending downwards for
several kilometers thus covering a volume of the order of 30 billion tons.

11.3 Horizontal Air Showers Revisited

We already discussed in section 3.3 the implications of horizontal air shower measurements
for atmospheric muons and neutrinos, especially the high energy component of charm origin.
We discuss here the use of horizontal air showers to search for cosmic neutrinos. Given a
flux φν (Eν ), the expected event rate for neutrino-induced horizontal showers is calculated as
described in §3.4, Eqs. (24) and (27).
The atmospheric muon spectrum falls rapidly with energy (γ ∼ 3.7 for π, K decay
muons; ∼ 2.7 for muons from charm decay), while neutrino cross sections rise with energy.

52
For large enough fluxes of astrophysical neutrinos, neutrino-induced showers can dominate
at sufficiently high energies. In addition to charge current νµ (ν̄µ ) interactions, there will
be a contribution from νe (ν̄e ). For flat spectra and νµ and νe fluxes of a similar order
of magnitude, the resonant W − production dominates the rate of ν-induced horizontal air
showers for shower sizes in the narrow range [2 × 106 to 4 × 106 ].
Horizontal shower measurements [90] have been compared to expectations from diffuse
AGN fluxes in Ref. [89], where the potential of air shower arrays for detection of diffuse neu-
trino fluxes was demonstrated. Similar limits have been obtained recently by the EASTOP
Collaboration [92]. Muon-poor showers are selected in the Tokyo data [90], so that the limit
is stronger for νe (ν̄e ) fluxes, which actually dominate the neutrino-induced horizontal air
shower signal.
Figure 18 shows the rate of (muon-poor) horizontal showers associated with the neutrino
emission by active galaxies. Also shown is the background from atmospheric muons. The
signal-to-background increases with shower size and exceeds unity for shower sizes exceeding
106 for the prediction of reference [38]. The figure illustrates the kind of sensitivity air shower
arrays can reach in the detection of cosmic neutrinos. An improved detection method would
be to select all, not only muon-poor, showers that are very close to the horizon. This would
avoid any contamination from cosmic rays without rejecting hadron-like neutrino induced
showers. Unless prompt muon production by charm turns out to be unexpectedly large,
the larger of the range of AGN isotropic backgrounds of Szabo and Protheroe [38] should
be detectable by a horizontal shower measurement with enough exposure to detect showers
with Ne > 106 . A flux peaked at energies exceeding 1 PeV, will be strongly attenuated in
the earth and only detectable in upward going muons close to horizontal angles. Under these
circumstances, neutrino detection via horizontal air showers may be a useful alternative de-
tection method.

Acknowledgements. We are grateful for helpful discussions with many colleagues, includ-
ing Karl Mannheim, Ray Protheroe, David Seckel, Floyd Stecker and Enrique Zas. This
research is supported in part by DOE Grants DE-FG-91ER40626 (TKG & TS) and DE-
AC02-76ER00881(FH). The work of FH is also supported by the University of Wisconsin
Research Committee with funds granted by The Wisconsin Alumni Research Foundation.

References
[1] Y. Totsuka, Rep. Prog. Phys. 55 (1982) 377.
[2] J.N. Bahcall & M.H. Pinsonneault, Revs. Mod. Phys. 64 (1992) 885.
[3] J.N. Bahcall, Neutrino Astrophysics (Cambridge University Press, 1989).
[4] S. Turck Chieze et al., Ap. J. 355 (1988) 415.
[5] K. Greisen, Ann. Rev. Nucl. Sci 10, (1960) 63.
[6] M.A. Markov & I.M. Zheleznykh, Nucl. Phys. 27 (1961) 385. See also M.A. Markov
in Proc. 1960 Annual International Conference on High Energy Physics at Rochester
(ed. E.C.G. Sudarshan, J.H. Tinlot & A.C. Melissinos) (1960).

53
 [7] C.V. Achar et al., Phys. Letters 18, 196 and 19 (1965) 78.

[8] M.G.K. Menon et al. Proc. Roy. Soc. A301, (1967) 137.

[9] F. Reines et al., Phys. Rev. Letters 15, (1965) 429.

[10] F. Reines, Proc. Roy Soc. A301, (1967) 125.

[11] K.S. Hirata et al., (Kam-II Collaboration), Phys. Letters, B280 (1992) 146.

[12] R. Becker-Szendy et al., (IMB Collaboration), Phys. Rev. D46 (1992) 3720. See also
D. Casper et al. Phys. Rev. Letters, 66 (1991) 2561.

[13] M.R. Krishnaswami et al., Nuovo Cimento, C9 (1986) 167; Phys.Letters, 106B, 339
(1981) and 115B (1982) 349; Proc. 6th Workshop on Grand Unification (World Scien-
tific, Singapore, 1986) eds. S. Rudaz and T. Walsh, p. 97.

[14] Ch. Berger et al., Physics Letters B245 (1990) 305 and 227 (1989) 489.

[15] M. Aglietta et al., Europhysics Letters 8 (1989) 611; Physics Letters, B280 (1992) 146.

[16] M.C. Goodman (Soudan 2 Collaboration), in Proc. DPF ’92 Meeting, eds. C. H. Al-
bright et al., (World Scientific, Singapore, 1993) p. 1300.

[17] BAIKAL Collaboration, I. Sokalski and C. Spiering, “The BAIKAL Neutrino Telescope
NT-200,” BAIKAL 92-03 (1992).

[18] I.A. Belolaptikov et al., Nucl. Phys. B (Proc. Suppl.) 35 (1994) 290.

[19] J. Babson et al. (DUMAND Collaboration), Phys. Rev. D42 (1990) 3613.

[20] S. Barwick, et al., J. Phys. G: Nucl. Part. Phys. 18 (1992) 225.

[21] D. Lowder et al., Nature 353 (1991) 331; T. Miller et al., Proceedings of the 23rd Inter-
national Cosmic Ray Conference, Calgary, Canada (1993); S. Tilav et al., Proceedings
of the 23rd International Cosmic Ray Conference, Calgary, Canada (1993).

[22] Proceedings of the NESTOR Workshop at Pylos, Greece, ed. L. K. Resvanis, October
1992, U. Athens (1993).

[23] J.J. Engelmann et al., Astronomy & Astrophys. 233 (1990) 96.

[24] T.K. Gaisser, Cosmic Rays and Particle Physics, Cambridge University Press (1990).

[25] Paolo Lipari, Astroparticle Physics 1 (1993) 195.

[26] H. Bloemen, Ann. Revs. Astron. Astrophys. 27 (1989) 469.

[27] V.S. Berezinsky, T.K. Gaisser, F. Halzen & Todor Stanev, Astroparticle Physics 1
(1993) 281. See also T.K. Gaisser, T. Stanev and F. Halzen, in Proc. 22nd Int. Cosmic
Ray Conf. (Dublin Inst. Adv. Studies, 1991) Vol. 1, p. 564.

54
[28] P.O. Lagage and C.J. Cesarsky, Astron. Astrophys. 125 (1983) 249.

[29] B. Peters, Nuovo Cimento 22 (1961) 800.

[30] W. I. Axford,in Proceedings of 1990 Kofu Symposium on “Astrophysical Aspects of

the Most Energetic Cosmic Rays” (eds. M. Nagano and F.Takahara, World Scientific)
p. 406 (1991).

[31] H.J. Völk and P.L. Biermann, Ap. J. Lett. 333 (1988) L65.

[32] J. R. Jokipii, Astro. Journal 313(1987) 842.

[33] R.J. Protheroe & D. Kazanas, Ap. J. 265 (1983) 620.

[34] D. Kazanas & D. Ellison, Ap. J. 304 (1986) 178.

[35] M. Sikora et al., Ap. J. Lett. 320 (1987) L81.

[36] F.W. Stecker, C. Done, M.H. Salamon and P. Sommers, Phys. Rev. Lett. 66 (1991)
2697 and 69 2738(E) (1992).

[37] K. Mannheim and P.L. Biermann, Astron. Astrophys. 253 (1992) L21.

[38] A.P. Szabo & R.J. Protheroe, in Proc. High Energy Neutrino Astrophysics Workshop
(Univ. of Hawaii, March 1992, ed. V.J. Stenger, J.G. Learned, S. Pakvasa & X. Tata,
World Scientific, Singapore) and Astroparticle Physics (to be published).

[39] G. Barr, T.K. Gaisser & Todor Stanev, Phys. Rev. D39 (1989) 3532.

[40] H. Lee & Y.S. Koh, Nuovo Cimento B 105 (1990) 883.

[41] M. Honda, K. Kasahara, K. Hidaka & S. Midorikawa, Physics Letters B 248 (1990)
193.

[42] E.V. Bugaev & V.A. Naumov, Physics Letters B 232 (1989) 391. The published fluxes
do not account for muon polarization. The results of a new calculation including the
effect are available now (V.A. Naumov, private communication).

[43] T.K. Gaisser Phil. Trans. R. Soc. Lond. A346 (1994) 75.

[44] T. Kajita, in Frontiers of Neutrino Astrophysics (Takayama, ed. Y. Suzuki & K. Naka-
mura, Universal Academy Press, Tokyo, (1992) p. 293.

[45] E.W. Beier et al., Physics Letters B 283 (1992) 446.

[46] D. Casper et al., Phys. Rev. Letters 66 (1991) 2561.

[47] T.K. Gaisser & J.S. O’Connell, Phys. Rev. D34 (1986) 822.

[48] E. Beier & E. Frank, Phil. Trans. R. Soc. Lond. A346 (1994) 63.

[49] M. Takita, Ph.D. Thesis, Univ. of Tokyo, February 1989, ICR-Report-186-89-3.

55
[50] David W. Casper, Ph.D. Thesis, Univ. of Michigan, 1990.

[51] J. Engel, E. Kolbe, K. Langanke & P. Vogel, Phys. Rev. D48 (1993) 3048.

[52] R. Merenyi et al., Phys. Rev. D45 (1992) 743.

[53] D.D. Koetke et al., Phys. Rev. C46 (1992) 2554.

[54] M. Honda, K. Kasahara & S. Midorikawa in Frontiers of Neutrino Astrophysics

(Takayama, ed. Y. Suzuki & K. Nakamura, Universal Academy Press, Tokyo, 1992)
p. 309.

[55] G.L. Fogli, E. Lisi & D. Montanino, Phys. Rev. D49 (1994) 3628.

[56] W.A. Mann, T. Kafka & W. Leeson, Physics Letters B291 (1992) 200.

[57] T.K. Gaisser, Nucl. Phys. B (Proc. Suppl.) 35 (1994) 209.

[58] M. Conversi, Phys. Rev. 79 (1950) 749.

[59] E.A. Bogomolov et al. as quoted by E.V. Bugaev, G.V. Domogatsky & V.A. Naumov
in Proc. Japan-U.S. Seminar on Cosmic Ray Muon and Neutrino Physics/Astrophysics
(ed. Y. Ohashi & V.Z. Peterson, 1986) p. 232.

[60] L.T. Baradzei et al., Zh. Eksp. Teor. Fiz. 36 (1959) 1617.

[61] M. Circella et al., Proc. 23rd Int. Cosmic Ray Conf. (Calgary) 4 (1993) 503.

[62] W. Lohmann, R. Kopp & R. Voss, CERN Yellow Report No. EP/85-03 (unpublished).

[63] Paolo Lipari & Todor Stanev, Phys. Rev. D44 (1991) 3543.

[64] T.K. Gaisser & A.F. Grillo, Phys. Rev. D36 (1987) 2752. See also T.K. Gaisser &
Todor Stanev, Phys. Rev. D31 (1985) 2770.

[65] C. Quigg, M.H. Reno and T.P. Walker, Phys. Rev. Lett. 57 (1986) 774; also see
M.H. Reno and C. Quigg, Phys. Rev. D37 (1988) 657.

[66] R. Becker-Szendy et al. (IMB Collaboration) Phys. Rev. Lett. 69 (1992) 1010.

[67] R. Becker-Szendy et al., in Proceedings of the XXVth International Conference on

High Energy Physics, Singapore, 1990, edited by K.K. Phua and Y. Yamaguchi (World
Scientific, Singapore, 1991) p. 662.

[68] M.M. Boliev et al., in Proc. 3rd Int. Workshop on Neutrino Telescopes (ed. M. Baldo
Ceolin) (1991) 235.

[69] M. Mori et al., Phys. Letters B210 (1991) 89.

[70] W. Frati, T.K. Gaisser, A.K. Mann & T. Stanev, Phys. Rev. D48 (1993) 1140.

56
[71] KAMIOKANDE Collaboration (Y. Fukuda et al., ICRR-Report-321-94-16, submitted
to Physics Letters B.

[72] L.V. Volkova, Yad. Fiz. 31 (1980) 784 (Sov. J. Nucl. Phys. 31 (1980) 1510).

[73] K. Mitsui, Y. Minorikawa & H. Komori, Nuovo Cimento 9C (1986) 995.

[74] A.V. Butkevich, L.G. Dedenko & I.M. Zheleznykh, Yad. Fiz. 50 (1989) 142 (Sov. J.
NuclṖhysics 50 (1989) 90).

[75] V. Agrawal, T.K. Gaisser, Paolo Lipari & T. Stanev, in preparation.

[76] E. Eichten, I. Hinchcliffe, K.Lane & C. Quigg, Revs. Mod. Phys. 56 (1984) 578 (Erra-
tum 58, 1065 (1986)).

[77] J.F. Owens, Physics Letters B266 (1991) 126.

[78] MACRO Collaboration, talk presented by D. Michael, Snowmass, July 1994.

[79] H. Adarkav et al., Proc. of the 21st Int. Cosmic Ray Conference, ed. R. Protheroe,
Adelaide (1990), Vol. 9, p. 310.

[80] Yu M. Andreyev et al., Proc. of the 21st Int. Cosmic Ray Conference, ed. R. Protheroe,
Adelaide (1990), Vol. 9, p. 301.

[81] R.P. Kokoulin and A.A Petrukhin, Proc. of the 22nd Int. Cosmic Ray Conference, ed.
M. Cawley et al., Dublin (1991) Vol. 4, p. 537.

[82] F. Halzen, R. Vázquez and E. Zas, Astroparticle Physics 1 (1993) 297.

[83] J.P. Guillet, P. Nason and H. Plothow-Besch, Proc. of the Large Hadron Collider
Workshop, Aachen 1990, Vol. II, p. 116.

[84] J.C. Collins and R.K. Ellis, Nucl. Phys. B380 (1991) 3.

[85] See Review of Particles Properties, J.J. Hernández et al. Phys. Lett. B239 (1990) 1.

[86] R. Ammar et al., Phys. Lett. B183 (1987) 110.

[87] P. Nason, S. Dawson and R.K. Ellis, Nucl. Phys. B303 (1988) 607.

[88] T.N. Afanasieva et al., Proc. of the 20th Int. Cosmic Ray Conference, (Moscow) Vol. 9,
p. 161 (1987).

[89] F. Halzen and E. Zas, Phys. Lett. B289 (1992) 184.

[90] M. Nagano et al., J. Phys. Soc. Japan 30 (1971) 33.

[91] S. Mikamo et al., Lett. al Nuovo Cimento 34 (1982) 273.

[92] M. Aglietta et al. (EASTOP Collaboration) LNGS preprint 94/96, May 1994.

57
 [93] M. Nagano et al., J. Phys. G: Nucl. Phys. 12 (1986) 69.
[94] T.K. Gaisser, T. Stanev, F. Halzen, W.F. Long and E. Zas, Phys. Rev. D43 (1991)
314.
[95] E. Zas, F. Halzen and R.A. Vázquez, Proc. of the International Symposium on Mul-
tiparticle Dynamics, Santiago de Compostela 1992, ed. C. Pajares, World Scientific,
Singapore (1992).
[96] C. Gonzales-Garcia, F. Halzen, R. Vázquez and E. Zas, Phys. Rev. D49 (1994) 2310.
[97] D. Eichler, Nature 275 (1978) 725. See also W.Y. Vestrand & D. Eichler, Ap. J. 261
(1982) 251.
[98] V.S. Berezinsky, Proc. 1979 DUMAND Workshop (ed. J. Learned) (1980) 674.
[99] T.K. Gaisser & Todor Stanev, Phys. Rev. Letters 54 (1985) 2265.
[100] A.K. Harding & T.K. Gaisser, Ap. J. 358 (1990) 561.
[101] V.S. Berezinsky, C. Castagnoli & P. Galeotti, Nuovo Cimento 8C (1985) 185.
[102] V.S. Berezinsky and O.F. Prilutsky, Astron. Astrophys. 66 (1978) 325.
[103] H. Sato, Prog. Theor. Phys. 58 (1977) 549.
[104] M.M. Shapiro & R. Silberberg, in Relativity, Quanta and Cosmology, (ed. DeFinis,
New York: Johnson Reprint) Vol. 2, p. 745 (1979).
[105] T.K. Gaisser & Todor Stanev, Phys. Rev. Letters, 58 (1987) 1695. See also T.K.
Gaisser, Alice K. Harding & Todor Stanev, Ap. J. 345 (1989) 423.
[106] S. Biller et al. Ap. J. 423 (1994). See also D.E. Alexandreas et al. Ap. J. 383 (1991)
L53.
[107] T.A. McKay et al., Ap. J. 417 (1993) 742.
[108] J. van Stekelenborg et al., Phys. Rev. D48 (1993) 4504.
[109] W.H. Allen et al., Phys. Rev. D48 (1993) 466. See also W.H. Allen et al., Astroparticle
Physics 1 (1993) 269.
[110] M. Amenomori et al., Phys. Rev. Letters 69 (1992) 2468.
[111] J.W. Cronin, K.G. Gibbs and T.C. Weekes, Ann. Revs. Nucl. Part. Phys. 43 (1993)
883.
[112] A. Borione et al. Proc. 23rd Int. Cosmic Ray Conf. (Calgary) Vol. 1 p. 385 (1993).
[113] V.S. Berezinsky & John G. Learned, in Proc. High Energy Neutrino Astrophysics
Workshop (Univ. of Hawaii, 1992, ed. V.J. Stenger, J.G. Learned, S. Pakvasa & X.
Tata, World Scientific, Singapore).

58
[114] T.K. Gaisser, Proc. 2nd Int. Workshop on Neutrino Telescopes (ed. Milla Baldo-Ceolin)
(1990) p. 397.
[115] M. Punch et al., Nature 358 (1992) 477.
[116] F.W. Stecker, O.C. De Jager and M.H. Salamon, Ap. J. 390 (1992) L49.
[117] Talks by P. L. Biermann, K. Mannheim, R. J. Protheroe and F. W. Stecker in High
Energy Neutrino Astronomy, V. Stenger, J. Learned, S . Pakvasa and X. Tata, Editors,
1992 (World Scientific, Singapore).
[118] P. Biermann, in Proceedings of 1990 Kofu Symposium on “Astrophysical Aspects of
the Most Energetic Cosmic Rays” (eds. M. Nagano and F.Takahara, World Scientific)
p. 301 (1991).
[119] C.D. Dermer, R. Schlickeiser and A. Mastichiadis, Astron. Astrophys. 256 (1992) L27.
[120] A.A. Zdiarski & J.H. Krolik, Ap. J. 409 (1993) L33.
[121] M. Sikora, M.C. Begelman and M.J. Rees, Ap. J. 421 (1994) 153.
[122] F. W. Stecker, Astro. Journal 228 (1979) 919.
[123] T.K. Gaisser & R.S. Schaefer, Ap. J. 394 (1992) 174.
[124] V. S. Berezinsky, T. K. Gaisser, F. Halzen and T. Stanev, Astropart. Phys. 1 (1993)
281. See also V. S. Berezinsky and V. A. Kudryavtsev, Sov. Astron. Lett. 14 (1988)
873.
[125] J. Matthews et al. Ap. J. 375 (1991) 202.
[126] H. Bloemen et al., COMPTEL Preprint No. 10 (1993).
[127] H. Bloemen et al., Astron. Astrophys. 139 (1984) 37.
[128] D. Seckel, Todor Stanev & T.K. Gaisser, Ap. J. 382 (1991) 652.
[129] T.K. Gaisser, in Proceedings of 1990 Kofu Symposium on “Astrophysical Aspects of
the Most Energetic Cosmic Rays” (eds. M. Nagano and F. Takahara, World Scientific)
p. 146 (1991).
[130] A.M. Hillas, Nature 312 (1984) 50.
[131] R. Cowsik & T.K. Gaisser, Proc. 17th Int. Cosmic Ray Conf. (Paris) 2 (1981) 218.
[132] R.J. Protheroe and T. Stanev, Ap. J. 322 (1987) 838.
[133] V. S. Berezinsky, Proc. of the Third International Symposium on Neutrino Telescopes,
Venice (1991), ed. by M. Baldo-Ceolin, p. 125.
[134] M. Samorski and W. Stamm, Ap. J. Lett. 268 (1983) L17; J. Lloyd-Evans et al. Nature
305 (1983)784.

59
[135] T.K. Gaisser and T. Stanev, Phys. Rev. Lett. 54 (1985) 2265; E.W. Kolb, M.S. Turner
and T.P. Walker, Phys. Rev D32 (1985) 1145; V.S. Berezinsky, C. Castagnoli and P.
Galeotti, Il Nuovo Cim. 8C (1985) 185.

[136] P.M. Chadwick et al., Nature 318 (1985) 642.

[137] V.S. Berezinsky and O.F. Prilutsky Astron. Astrophys. 66 (1978) 325.

[138] H. Sato Prog. Theor. Phys. 58 (1978) 549x.

[139] T.K. Gaisser, Alice K. Harding & Todor Stanev, Ap. J. 345 (1989) 423.

[140] Y. Yamada, T. Nakamura, K. Kasahara & H. Sato, Prog. Theor. Phys. 79 (1987) 426.

[141] V. S. Berezinsky and V.S. Ptuskin, Astron. Astrophys. 215 (1989) 399.

[142] V. S. Berezinsky, in Proc. Neutrino 77 (Nauka, Moscow, 1977) 1, p. 177; D. Eichler,

Ap. J. 232 (1979) 106; R. Silberberg and M.M. Shapiro, in Proc. 16th Int. Cosmic Ray
Conf. Vol. 10 (1979) p. 357; V.S. Berezinsky and V.L. Ginzbusrg, Mon. Not. Royal
Astr. Soc. 194 (1981) 3.

[143] P.L. Biermann and P.A. Strittmatter, Ap. J. 322 (1987) 643.

[144] M. C. Begelman, B. Rudak and M. Sikora, Ap. J. 362 (1990) 38.

[145] M.J. Rees, Ann. Rev. Astron. Astrophys. 22 (1984) 471.

[146] J.H. Krolik et al., Ap. J. 371 (1991) 541.

[147] J. Clavel et al., Ap. J. 393 (1992) 113.

[148] R.F. Mushotzky, C. Done and K.A. Pounds, Ann. Rev. Astron. Astrophys. 31 (1993)
717.

[149] D.B. Sanders et al., Ap. J. 347 (1989) 29.

[150] K.I. Kellerman et al., Astron. J. 98 (1989) 1195.

[151] R.D. Blandford in Active Galactic Nuclei (Springer-Verlag, Berlin, ed. R.D. Blandford,
H. Netzer & L. Woltjer) pp. 161-275.

[152] M. Sikora and M.C. Begelman, in High Energy Neutrino Astronomy, eds. V.J. Stenger
et al. (World Scientific, Singapore, 1992), p. 114.

[153] F.W. Stecker, Phys. Rev. Lett. 21 (1968) 1016.

[154] M. Sikora, M.C. Begelman & B. Rudak, Ap. J. Lett. 341 (1968) L33.

[155] D. Kazanas & P.M. Giovanoni in Proc. Workshop on High Energy Neutrino Astronomy,
(ed. V.J. Stenger, J.G. Learned, S. Pakvasa & X. Tata, World Scientific, 1992) p. 94.

60
[156] R.J. Protheroe and T. Stanev, in High Energy Neutrino Astronomy, eds. V.J. Stenger
et al. (World Scientific, Singapore, 1992), p. 40.

[157] L. Nellen, K. Mannheim and P.L. Biermann, Phys. Rev. D47 (1993) 5270.

[158] T. Maccacaro et al. Ap. J. 374 (1991) 117.

[159] K. Morisawa et al. Astron. Astrophys. 236 (1990) 299.

[160] T.K. Gaisser, Nucl. Phys. B (Proc. Suppl.) 31 (1993) 399.

[161] H. Meyer, in 4th Int. Workshop on Neutrino Telescopes, ed. M. Baldo-Ceolin, p. 213.

[162] W. Rhode, preprint (University of Wuppertal, 1994).

[163] T.K. Gaisser, in Proc. 1994 Venice Workshop on Neutrino Telescopes (in press).

[164] R.M. Baltrusaitis et al. Phys. Rev. D31 (1985) 2192.

[165] S.L. Glashow Phys. Rev. 118B (1960) 316.

[166] Francis Halzen & Alan D. Martin, Quarks & Leptons, John Wiley & Sons, New York
(1984) pp. 269 and 374.

[167] J.G. Learned and T. Stanev, in Proc. 3rd Int. Workshop on Neutrino Telescopes, ed.
M. Baldo-Ceolin, p. 473.

[168] D.A. Morris and A. Ringwald, in Proc. 23rd ICRC, eds. D.A. Leahy et al. 4, p. 407.

[169] R.C. Hartman et al. Ap. J. Lett. 385 (1992) L1.

[170] C.E. Fichtel et al. Preliminary EGRET Source Catalog, in Proc. Compton Symposium,
in print.

[171] J. Frank, A.R. King & D.J. Raine, Accretion Power in Astrophysics, Cambridge Uni-
versity Press (1985) p. 193.

[172] K. Mannheim and P.L. Biermann, Astron. Astrophys. 22 (1989) 211.

[173] K. Mannheim, Astron. Astrophys. 269 (1993) 67.

[174] R.J. Protheroe and T. Stanev, Mon. Not. Royal A. Soc. 264 (1993) 191.

[175] K. Mannheim, in High Energy Neutrino Astronomy, eds. V.J. Stenger et al. (World
Scientific, Singapore, 1992), p. 105.

[176] K. Mannheim, Phys. Rev. D48 (1993) 2408.

[177] G.G. Lichti et al., to appear in Proc. 2nd Compton Symposium, College Park, 1993.

[178] Y.C. Lin et al. Ap. J. Lett. 401 (1992) L61.

61
[179] G. Mohanty et al. in Proc. 23rd ICRC, eds. D.A. Leahy et al. 1, p. 440.

[180] F. Halzen and R. A. Vazquez, in Proc. 23rd ICRC, eds. D.A. Leahy et al. 1, p. 447.

[181] P.L. Biermann, in Cosmic Gamma Rays, Neutrinos and Related Astrophysics, eds.
M.M. Shapiro and J.P. Wefel (Kluwer, Dordrecht, 1988) p. 21.

[182] F.W. Stecker et al. in High Energy Neutrino Astronomy, eds. V.J. Stenger et al. (World
Scientific, Singapore, 1992), p. 105.

[183] F.W. Stecker, in Proceedings of the Fifth International Workshop on Neutrino Tele-
scopes, Venice, 1993, ed. M. Baldo-Ceolin, p. 443.

[184] K. Mannheim, private communication.

[185] V.S. Berezinsky and J.G. Learned, in High Energy Neutrino Astronomy, eds. V.J.
Stenger et al. (World Scientific, Singapore, 1992), p. 43.

[186] A. Mastichiadis and J.G. Kirk, in High Energy Neutrino Astronomy, eds. V.J. Stenger
et al. (World Scientific, Singapore, 1992), p. 63.

[187] K. Greisen, Phys. Rev. Lett 16 (1966) 748.

[188] G.T. Zatsepin and V.A. Kuzmin, Pis’ma Zh. Eksp. Teor. Fiz. 4 (1966) 53 (JETP Lett.
4 (1966) 78).

[189] J. Wdowczyk, W. Tkaczyk and A.W. Wolfendale, J. Phys. A 5 (1972) 1419.

[190] F.W. Stecker, Astrophys. Space Sci. 20 (1973) 47.

[191] C.T. Hill and D.N. Schramm, Phys. Rev. D31 (1985) 564.

[192] V.S. Berezinsky and S.I. Grigor’eva, Astron. Astrophys. 199 (1988) 1.

[193] F. Halzen, R.J. Protheroe, T. Stanev and H.P. Vankov, Phys. Rev. D41 (1990) 342.

[194] S. Yoshida and M. Teshima, Progr. Theor. Phys. 89 (1993) 833.

[195] J. Wdowczyk and A.W. Wolfendale, Ann. Rev. Nucl. Part. Sci. (1989).

[196] J. P. Rachen and P. L. Biermann, Astron. Astrophys. 272 (1993) 161.

[197] J. Ostriker, G. Thompson and E. Witten, Phys. Lett. B180 (1986) 231.

[198] C.T. Hill, D.N. Schramm and T.P. Walker, Phys. Rev. D36 (1987) 1007.

[199] P. Bhattacharjee, Phys. Rev. D40 (1989) 3968.

[200] J. MacGibbon and R. Brandenberger, Nucl. Phys. B331 (1990) 153.

[201] P. Bhattacharjee and N. Rana, Phys. Lett. 246B (1990) 365.

62
[202] J. R. Primack, B. Sadoulet, and D. Seckel, Ann. Rev. Nucl. Part. Sci. B38 (1988) 751.

[203] F. Halzen, M. Kamionkowski, and T. Stelzer, Phys. Rev. D45 (1992) 4439.

[204] A. Bottino et al. Mod. Phys. Lett. A7 (1992) 733.

[205] A. Gould, Ap. J. 321 (1987) 571; Ap. J. 368 (1991) 610.

[206] V. S. Berezinsky, in Proc. of the Fourth International Symposium on Neutrino Tele-

scopes, Venice (1992), ed. M. Baldo-Ceolin.

[207] H. E. Haber and G. L. Kane, Phys. Rep. 117 (1985) 75.

[208] M. Drees and M. M. Nojiri, Phys. Rev. D47 (1993) 376.

[209] M. Mori et al., KEK Preprint 91-62; N. Sato et al., Phys. Rev. D44 (1991) 2220.

[210] J. M. LoSecco et al. (IMB Collaboration), Phys. Lett. B188 (1987) 388; R. Svoboda
et al., Ap. J. 315 (1987) 420.

[211] J. G. Learned, Proceedings of the Venice Workshop on Neutrino Telescopes, (ed.

M. Baldo-Ceolin, February 1990) p. 103.

[212] A. DeRujula, S.L. Glashow, R.R. Wilson and G. Charpak, Phys. Reports, 99 (1983)
341.

[213] F. Halzen and J.G. Learned, Proc. of the Fifth International Symposium on Neutrino
Telescopes, Venice (1993), ed. M. Baldo-Ceolin, p. 483.

[214] See recent reports on these experiments in the following reference, and a critical sum-
mary of the various projects in J. G. Learned, Proceedings of the European Cosmic Ray
Symposium, Geneva, Switzerland, July 1992, ed. by P. Grieder and B. Pattison, Nucl.
Phys. B (1993); see also presentations in Proceedings of the High Energy Neutrino As-
trophysics Workshop, ed. by V. J. Stenger, J. G. Learned, S. Pakvasa, and X. Tata,
World Scientific, Singapore (1992).

[215] AMANDA-collaboration, submitted to Nature.

[216] J. G. Learned and A. Roberts, Proceedings of the 23rd International Cosmic Ray
Conference, Calgary, Canada (1993), Vol. 4, p. 579.

[217] Y. Suzuki, Proceedings of the 3rd International Workshop on Neutrino Telescopes,

Venice, March 1992, ed. M. Baldo-Ceolin, Venice (1992) p. 237.

[218] G. A. Askar’yan, Soviet Physics JETP 14 (1962) 441–443; 48, (1965) 988–990.

[219] M. A. Markov and I. M. Zeleznykh, Nucl. Inst. Methods A 248 (1986) 242–251.

[220] I. M. Zeleznykh, Proc. XXIst International Cosmic Ray Conference (Adelaide, 1989),
Vol. 6, p. 528–533; J. P. Ralston and D. M. McKay, Astrophysics in Antarctica, Amer-
ican Institute for Physics Conference Proceedings 198 (1990) 24.

63
[221] I. Frank and I. Tamm, Proc. Acad. Sc. USSR 14 (1937) 109.

[222] M. Fujii and J. Nishimura, Proc. XIth International Cosmic Ray Conference, (Bu-
dapest 1969), pp. 709–715.

[223] For a review see, H. R. Allan, Progress in Elementary Particles and Cosmic Ray Physics
10 (1971) 171 (North Holland Publ. Co.).

[224] F. Halzen, T. Stanev, and E. Zas, Phys. Lett. 257B (1991) 432 and Phys. Rev. D45
(1992) 362.

[225] I. N. Boldyrev, G. A. Gusev, M. A. Markov, A. L. Provorov and I. M. Zeleznykh, Proc.

XXth International Cosmic Ray Conference, (Moscow, 1987) Vol. 6, p. 472.

[226] T. Bowen, Proceedings of the 16th International Cosmic Ray Conference, (Kyoto,
1979), Vol. 11, p. 184, T6-1.

[227] J. G. Learned, Phys. Rev. D19 (1979) 3293.

[228] H. Spetzler and D. L. Anderson, J. Geophys. Res. 73 (1968) 6051.

64
Figure Captions
Fig.1. A schematic presentation of neutrino fluxes of different origin. (1) atmospheric neu-
trinos; (2) neutrinos from cosmic ray interactions on galactic matter; (3) source neutrinos.

Fig. 2. Pν for two values of muon threshold energy, 1 GeV and 1 TeV. The solid lines are
for ν and the dashed lines for ν̄. The dotted lines show the power law approximations.

Fig. 3. Neutrino energies giving rise to contained events, stopping and throughgoing muons.

Fig. 4. Upward going muon fluxes detected by IMB [66,67](squares), Baksan [68](stars) and
Kamiokande [69](circles), converted to a common muon threshold energy of 3 GeV.

Fig. 5. Prompt neutrino ( or muon) fluxes corresponding to five parametrizations of the

high energy behavior of the charm production cross section. The vertical and horizontal
fluxes of muons from the decay of pions and kaons are shown with dash lines. See text and
Ref. [82] for details.

Fig. 6. Rates of horizontal showers, initiated by the muon fluxes shown in Fig. 5. See text.
The data points are from Ref. [90] and the dotted line shows the shower rate expected from
muons form π and K decay.

Fig. 7. Neutrino fluxes from 3C273 predicted in different AGN models. Thick solid line
shows the flux of Stecker et al [36]; thin solid lines show several models due to Szabo and
Protheroe [38]; and the dash line is from Ref. [152] (see text). These three models are for
neutrinos from the AGN nucleus. The dotted line shows the calculation of Mannheim for
the neutrino emission from the AGN jet [176].

Fig. 8. The isotropic neutrino background from AGN. The thick line is the prediction
of Stecker et al [36] for generic AGN, and the thin lines represent models of Szabo and
Protheroe [38]. The backgrounds from AGN jets are calculated by Mannheim [173] (dotted
line) and Stecker et al [183] (dash-dot line). The shaded area shows the angle averaged flux
of atmospheric neutrinos.

Fig. 9. Horizontal (−0.3 < cos θ < 0.3) muon fluxes generated by the isotropic neutrino
background as in the bracketing high and low models of Szabo & Protheroe [38] (thin solid
lines) and by Stecker et al. [36] (thick solid line). The 90% C.L. upper limit of the Frejus
experiment [161] is shown for muon threshold energy of 2 TeV.

Fig. 10. Differential interaction rate of νe + ν̄e as a function of Eν . Dotted line: atmo-
spheric; thin solid line: Ref. [38]; thick solid line: Ref. [36]. The numbers by the curves are
number of interactions per 1000 kT per year, calculated for Eν > 1 TeV.

Fig. 11. Comparison of quasi-simultaneous observations of 3C273 in the optical, X-ray and
γ-ray bands [177] with the predictions of a synchrotron-self-Compton model [119] (solid line)
and a hadronic model [176]. See Ref. [177] for references to all experimental data.

65
Fig. 12. Contours in the M, µ plane of constant αH = 1.0, 0.1, 0.01 (boxes) and constant
neutralino mass Mχ = 30, 100, 500 and 1000 GeV (solid). The shaded region is excluded by
cosmological considerations.

Fig. 13. As a function of the neutralino mass we show the telescope size required to be
sensitive at the one event per year level. We fix tan β = 2, αH = 0.1. The two branches
correspond to the two solutions for fixed αH .

Fig. 14. In the M2 , µ plane for Mq̃ = ∞, (a) contours of constant detection rate (events
m−2 yr−1 ) and (b) regions of Ωχ h2 > 1 and Ωχ h2 < 0.02 which are ruled out by cosmological
considerations.

Fig. 15. Summary of isotropic neutrino fluxes of energy above 1 GeV. (1) atmospheric
neutrinos; (2) diffuse galactic neutrinos; (3) diffuse extragalactic neutrinos—maximum and
minimum predictions of Ref. [38]; (4) cosmological neutrinos—maximum and minimum pre-
dictions of Ref. [194].

Fig. 16. Summary of source neutrino fluxes of energy above 1 GeV. (1) neutrinos generated
by cosmic rays in the Sun [128]; (2) a galactic neutrino source; (3) extragalactic neutrino
source (3C273) [38]; (4) AGN jet emission (3C279 high state) [37]. The atmospheric neutrino
background within 1◦ is shown with a dash line.

Fig. 17. A possible configuration of a 1 km neutrino detector, based on AMANDA-design

supermodules.

Fig. 18. Rate of (muon-poor) horizontal showers associated with the neutrino emission by
active galaxies (thin lines are the upper and lower bound of Ref. [38], the thick line shows
the background of Ref. [36]) compared to the background from atmospheric muons (dotted
line) and the experimental results of Ref. [90].

66
This figure "fig1-1.png" is available in "png" format from:

http://arXiv.org/ps/hep-ph/9410384v1
This figure "fig1-2.png" is available in "png" format from:

http://arXiv.org/ps/hep-ph/9410384v1
This figure "fig1-3.png" is available in "png" format from:

http://arXiv.org/ps/hep-ph/9410384v1
This figure "fig1-4.png" is available in "png" format from:

http://arXiv.org/ps/hep-ph/9410384v1
This figure "fig1-5.png" is available in "png" format from:

http://arXiv.org/ps/hep-ph/9410384v1
This figure "fig1-6.png" is available in "png" format from:

http://arXiv.org/ps/hep-ph/9410384v1
This figure "fig1-7.png" is available in "png" format from:

http://arXiv.org/ps/hep-ph/9410384v1
This figure "fig1-8.png" is available in "png" format from:

http://arXiv.org/ps/hep-ph/9410384v1
This figure "fig1-9.png" is available in "png" format from:

http://arXiv.org/ps/hep-ph/9410384v1
This figure "fig1-10.png" is available in "png" format from:

http://arXiv.org/ps/hep-ph/9410384v1
This figure "fig1-11.png" is available in "png" format from:

http://arXiv.org/ps/hep-ph/9410384v1
This figure "fig1-12.png" is available in "png" format from:

http://arXiv.org/ps/hep-ph/9410384v1
This figure "fig1-13.png" is available in "png" format from:

http://arXiv.org/ps/hep-ph/9410384v1
This figure "fig1-14.png" is available in "png" format from:

http://arXiv.org/ps/hep-ph/9410384v1
This figure "fig1-15.png" is available in "png" format from:

http://arXiv.org/ps/hep-ph/9410384v1
This figure "fig1-16.png" is available in "png" format from:

http://arXiv.org/ps/hep-ph/9410384v1
This figure "fig1-17.png" is available in "png" format from:

http://arXiv.org/ps/hep-ph/9410384v1
This figure "fig1-18.png" is available in "png" format from:

http://arXiv.org/ps/hep-ph/9410384v1
3
2
1

Fig. 1
 -2

-4

-6

Fig. 2
1 GeV

-8

-10
1 TeV

-12
0 1 2 3 4 5 6 7 8
50

1
40
3

30
2

Fig. 3
20

10

0
.1 1 10 100 1000 10000
5

Fig. 4
2

0
-1 -.8 -.6 -.4 -.2 0
0 A

B
-1

Fig. 5
C
-2

-3

-4

2 4 6 8 10
-12

A
-14

Fig. 6
-16 B
E

-18

-20
4 4.5 5 5.5 6 6.5 7
-10

-12

Fig. 7
-14

-16

-18
3 4 5 6 7 8 9
 -8

-10

-12

Fig. 8
-14

-16

-18

-20
3 4 5 6 7 8 9 10
-12.5

-13

-13.5

Fig. 9
-14

-14.5

-15

-15.5
1 1.5 2 2.5 3 3.5 4 4.5 5
2

Fig. 10
-2

-4

-6
1 2 3 4 5 6 7 8
 Fig. 11
14 16 18 20 22 24
104

103

102

102 103 104

Fig. 12
Fig. 13
 (a) M squark = oo
4
10

M (GeV)
3
10 −3 −5
10 10

2
10

2 3 4
10
µ 10
(GeV) 10

(b) M squark = oo
4
10

Ω χ h 2 < .02 Ω χ h 2 > 1.

M (GeV)

3
10

2
10

2 3 4
10
µ 10
(GeV) 10

Fig. 14
 0

-5

-10

Fig. 15
3

-15

-20
10 15 20
 -5

-10
2

Fig. 16
3

-15
4

10 15 20
 1+7+7

supermodule
super-
1+5 modules

OM
m
60
250 m

250 m

1 km
superstring
4590 OMs

20 m

Fig. 17
-15

Fig. 18
-20
5