A&A 443, 201-210 (2005)
DOI: 10.1051/0004-6361:20053236
A. Marek1 - H.-Th. Janka1 - R. Buras2 -
M. Liebendörfer3 - M. Rampp1,
1 - Max-Planck-Institut für Astrophysik,
Karl-Schwarzschild-Str. 1, 85741 Garching, Germany
2 -
Max-Planck-Institut für Physik (Werner-Heisenberg-Institut),
Föhringer Ring 6, 80805 München, Germany
3 -
CITA, University of Toronto, Toronto, Ontario M5S 3H8, Canada
Received 13 April 2005 / Accepted 19 June 2005
Abstract
The role of ion-ion correlations in suppressing neutrino-nucleus
elastic scattering during stellar core collapse is reinvestigated,
using two different equations of state.
We test the improved description by Itoh et al. against the
treatment suggested by Horowitz and find that the stronger
cross-section reduction for small momentum transfer in the former
case does not lead to noticeable changes of the core deleptonization
and entropy increase during collapse, because the improvements are
relevant below neutrino trapping conditions only for very low
neutrino energies, corresponding to a very small phase-space volume.
Treating screening effects for ionic mixtures by the linear mixing
rule applied to the collection of representative
heavy nucleus,
particles, and free nucleons, which is assumed
to characterize the composition in nuclear statistical equilibrium,
we cannot determine mentionable differences during
stellar collapse, because
particles are not sufficiently
abundant, and their coherent scattering opacity is too small.
Key words: stars: supernovae: general - neutrinos - radiative transfer - hydrodynamics
In the medium of the supernova core, nuclei are coupled strongly
with each other by Coulomb forces. They thus form a highly
correlated
plasma, in which the interactions of neutrinos with
wavelengths larger than the average ion-ion separation
(corresponding to neutrino energies
MeV)
are reduced by phase interference effects (Itoh 1975). The
corresponding ion screening was more recently calculated
by Horowitz (1997) and Itoh et al. (2004), and investigated in
its effects on stellar core collapse by Bruenn & Mezzacappa (1997). The
last authors employed
the correction factor for neutrino-nucleus
scattering cross sections as given by Horowitz (1997).
Itoh et al. (2004), however, pointed out that the Monte
Carlo calculations, which Horowitz's fit was based on,
did not allow him to accurately represent the cross-section
reduction for low neutrino energies, i.e., for energies
MeV, thereby
underestimating the importance of ion-ion correlation
effects. Itoh et al. (2004) provided a more accurate analytic
fitting formula by
using the correct behavior of the liquid structure factor for
small momentum transfer in neutrino-nucleus scattering.
The investigations presented in this work have two goals.
On the one hand, we aim at studying the differences in
stellar core collapse and the formation of the supernova
shock, which arise from the improved description of ion
screening as suggested by Itoh et al. (2004), compared to a
treatment using the formulae of Horowitz (1997). On the
other hand, we intend to explore the sensitivity of the
evolution to ion screening effects associated with the
ionic mixture of nuclei and nucleons that are present
during core collapse. In accordance with the treatment of
NSE in current equations of state (EoSs) for
supernova simulations, we consider the nuclear components
to be free neutrons, free protons,
particles, and
one kind of heavy nucleus which is considered as representative
of the NSE distribution of nuclei beyond 4He.
Two different nuclear EoSs with largely different
mass fractions during core collapse are employed. The first
EoS ("L&S''), provided by Lattimer & Swesty (1991), is based on a
compressible liquid drop model and uses a Skyrme force for the
nucleon interaction (Lattimer et al. 1985). Our choice of the
compressibility modulus of bulk nuclear matter is 180 MeV,
and the symmetry energy parameter 29.3 MeV,
but the differences in the supernova evolution caused by other
values of the compressibility were shown to be
minor (Thompson et al. 2003; Swesty et al. 1994). The second
EoS used here ("Shen'') is the new relativistic mean field EoS of
Shen et al. (1998a,b) with a compressibility of nuclear matter
of 281 MeV and a symmetry energy of 36.9 MeV.
The paper is organized as follows. In Sect. 2 the prescriptions for the ion-ion correlation factor provided by Horowitz (1997) and Itoh et al. (2004) are briefly summarized. In Sect. 3 the input (code, initial stellar model) in our simulations is described, while in Sect. 4 the results are presented and finish with conclusions in Sect. 5.
In this study we make use of two different fitting formulae for the
angle-averaged correlation factor
,
which describes the reduction of the neutral-current scattering
of neutrinos off nuclei by ion-ion correlation effects. It is
used as a multiplicative correction to the neutrino-nucleus
isoenergetic scattering opacity (cf. Bruenn & Mezzacappa 1997; Rampp & Janka 2002; Horowitz 1997).
The first formula is provided by Horowitz (1997) and is based on Monte
Carlo results. The second one is given by Itoh et al. (2004) and
was obtained from data calculated with the improved
hypernetted-chain method
(see Itoh et al. 1983, and references therein) for a classical
one-component plasma. In all simulations with ion screening,
we also take into account (the rather small) electron screening
effects according to Horowitz (1997) by applying the additional
correction factor of his Eq. (19) to the rates of coherent
neutrino-nucleus scattering.
![]() |
Figure 1:
The ratio of the angle-averaged ion-ion correlation factor
as calculated with the fitting formula of Horowitz (1997)
(see Fig. 9 in Bruenn & Mezzacappa 1997) relative
to the one obtained from the prescription provided by Itoh et al.
(2004, see their Fig. 2) as a function of the variable ![]() ![]() |
Open with DEXTER |
If the stellar plasma consists of only one nuclear
species of ions
,
the ion sphere radius
which gives the mean interion distance is defined as
Horowitz (1997) provided the following fitting formula for the
angle-averaged suppression factor
The simulations presented here intend to study the
dynamical consequences of these differences during supernova
core collapse. Up to now we have considered a stellar medium
consisting of only one nuclear species and referred to
formulae derived for a classical one-component plasma.
However, in the collapsing core of a massive star a mixture
of nuclei besides free neutrons and protons is present, usually
approximated by
particles plus one representative, heavy
nucleus. To deal with that we decided, in the simplest approach,
to calculate the suppression factor
for
particles
and for the heavy nucleus independently, using the values
for the average distance
between ions of the
same kind, derived from the number densities
or nA,
respectively. This assumes that different ionic components
coexist without collectively
affecting the screening of neutrino-nucleus interactions.
Alpha particles thus change the ion screening
for heavy nuclei only by the fact that their presence may reduce
the number density of heavier nuclei.
For a liquid mixture of different ions (Zj,Aj) including free protons, Itoh et al. (2004) suggest a modified treatment, referring to earlier work by Itoh et al. (1979). Employing the so-called linear mixing rule, one can extend the calculations of neutrino-nucleus scattering cross sections, obtained for a one-component ion liquid, to the case of multi-component fluids.
The ion sphere radius for an ion j in the mix is now given by
Applying the naive procedure of
Sect. 2.1 for the conditions in
a supernova core shows that usually for
particles, because
's are less abundant
than heavy nuclei in the central part of the core and during
most phases of the collapse.
Their
is then less than unity. Both factors
diminish ion screening for
's to a negligible level.
In contrast, following the description in this section,
the presence of
's can also affect
the ion screening of heavy nuclei by reducing
the interion separation (Eq. (6))
and thus
(Eq. (8)). Moreover,
for heavy nuclei according to Eq. (9)
might become larger than in Eq. (2).
Therefore the presence of
's has an indirect
influence on neutrino scattering off heavy nuclei and thus
on ion-ion correlations during stellar core collapse, despite
the fact that the screening effects for
particles
are still small because their
is usually below
unity; following Horowitz (1997),
is then set to unity
for evaluating the angle-averaged cross section suppression
factor.
While our procedure for treating the effects of particles in an ionic mixture with heavy nuclei and free
nucleons adopts the recipe of Itoh et al. (2004),
Sawyer (2005) has recently discussed an alternative approach
to the problem of multi-component fluids by applying
the Debye-Hückel approximation.
He points out that in multi-component plasmas the ion-ion
correlation effects might be greatly reduced, leading
to much larger neutrino opacities than for a one-component
plasma, even if the constituent ions have only a small
range of N/Z ratios. Electron density fluctuations for
an ionic mixture enhance this tendency. Sawyer's calculations
therefore yield a result which is opposite to our
application of the linear mixing rule for calculating
multi-component plasma
parameters. They tend to bring one back closer to
the case with ion-ion correlations being ignored, simulations
of which will be presented below, too. Thus we provide a set
of models with the intention of encompassing and bracketing
the "extreme'' possibilities discussed in the literature.
The core collapse simulations performed in this work were carried out with the neutrino-hydrodynamics code VERTEX in spherical symmetry. This code is described in detail in Rampp & Janka (2002).
VERTEX employs the PROMETHEUS hydrodynamics code (Fryxell et al. 1989), which is a direct Eulerian implementation of the Piecewise Parabolic Method. It uses an explicit Godunov-type scheme with second-order temporal and third-order spatial accuracy, which allows for the simulation of high Mach number, self-gravitating flows with discontinuities, and it can handle complicated equations of state. The VERTEX transport solver employs a variable Eddington factor algorithm in spherical coordinates. Angular moments of the O(v/c) comoving frame transport equation form a system of lower-dimensional moment equations, which is closed with a variable Eddington factor. The latter is determined from a (simplified) "model'' Boltzmann equation. General relativistic gravity and corresponding effects in the transport are approximately taken into account. Details about this can be found in Rampp & Janka (2002) and Marek et al. (2005). A comparison with fully relativistic calculations reveals excellent agreement during core collapse and shock formation (Liebendöerfer et al. 2005).
Table 1:
Overview of neutrino-matter and neutrino-neutrino
interactions included in our simulations. For each process, we provide
reference(s) where more information can be found about physics and
approximations employed in the rate calculations. The numerical
implementation is described in detail in Rampp & Janka (2002) and
Buras et al. (2003). The symbol
represents any of the neutrinos
,
the
symbols e-, e+, n, p, and A denote
electrons, positrons, free neutrons and protons, and heavy nuclei,
respectively. The symbol N means neutrons or protons.
The neutrino interactions used in the simulations of the
present work are summarized in Table 1.
Note that neutral-current scatterings of neutrinos off nucleons
and charged-current -processes include the effects of
nucleon recoil, thermal motions, phase space blocking,
nucleon correlations in dense media (Burrows & Sawyer 1999,1998),
corrections due to the weak magnetism
of nucleons (Horowitz 2002), the possible quenching of the
axial-vector coupling in nuclear matter (Carter & Prakash 2002), and
the reduction of the effective nucleon mass at high densities
(Reddy et al. 1999). Electron captures on nuclei are
implemented according to the improved treatment
of Langanke et al. (2003) in regions where NSE holds, taking
into account the collective
e-captures of a large sample of nuclei in NSE with rates
determined from shell model Monte Carlo calculations;
the prescription of Bruenn (1985) is used in regions which are
out of NSE. With this
input, the production of
by nuclei dominates
the one by protons during core collapse (Langanke et al. 2003).
The core collapse simulations presented in this paper were
started from the 15
progenitor s15a28 from Heger et al. (2001).
For describing the thermodynamics and composition of the stellar
plasma, the EoS of Lattimer & Swesty (1991) and the one of
Shen et al. (1998a,b) are applied at high densities
(
g cm-3
or
g cm-3, respectively).
At lower densities the EoS contains a mixture of electrons,
positrons, photons, nucleons, and nuclei, with the nuclear
composition
described by a simple approximation to a four-species
NSE for temperatures
above about 0.5 MeV. Below that temperature, the composition
is adopted from the progenitor star and modified if nuclear
burning plays a role during collapse (for details, see
Appendix B in Rampp & Janka 2002). The two EoSs show major differences
in the abundances of
particles, which can be
larger by up to a factor of
10 in the case of the
Shen et al. (1998a,b) EoS. This is visible in
Fig. 2, where the number fractions of free
neutrons, protons,
's, and of the representative heavy
nucleus (whose mass and charge numbers typically grow with
density until nuclei disappear at the phase transition to
nuclear matter) are displayed as functions of increasing
density during collapse both at the stellar center and at an
enclosed mass of 0.5
.
Although their mass fraction
is much lower,
particles in the Shen et al. (1998a,b) EoS
can become equally or even more abundant (by a
factor up to about two) than heavy nuclei in the outer
layers of the collapsing core, in particular outside of
-0.6
.
![]() |
Figure 2:
Composition as a function of density at the center
( top) and at an enclosed mass of 0.5![]() ![]() |
Open with DEXTER |
For each of the employed EoSs, four core collapse simulations were
performed, all starting from the onset of gravitational instability
and carried on until the moment of shock formation.
Calculations with ion-ion correlation (and electron screening)
effects in neutrino-nucleus scattering
switched off, i.e. for
(and
,
instead of Eq. (19) of
Horowitz 1997),
are denoted with "EoS_ion_off'', where EoS stands
for "L&S'' or "Shen''. They are compared to (i) simulations
(models "EoS_ion_Hor'') where the ion-ion correlation factor of
Horowitz (1997, Eqs. (3), (4)) is
used and (ii) models in which ion-ion
correlations are described according to Itoh et al. (2004) (models
"EoS_ion_Itoh''). Finally, the sensitivity of stellar core
collapse to the treatment of ion screening for ionic mixtures
is investigated by simulations
(models "EoS_ion_mix'') in which the correction
factors
are
calculated from the Itoh et al. (2004) formulae with
and
,
as given in Sect. 2.2.
Figures 3 and
4 show electron fraction ,
lepton
fraction
,
and (gas) entropy s, respectively,
at the core center during collapse simulations with the
Shen et al. (1998a,b) EoS. For both EoSs employed in this work,
the same relative changes are found when models without ion
screening are compared with
calculations with ion-ion correlations according to Horowitz (1997)
(see Models Shen_ion_Hor and Shen_ion_off in
Figs. 3 and 4). We shall
concentrate here mostly on the results obtained with the
Shen et al. (1998a,b) EoS, because
particles are
much more abundant there (see Fig. 2)
and many aspects of ion screening in simulations with the
EoS of Lattimer & Swesty (1991) have already been discussed
by Bruenn & Mezzacappa (1997). Our results agree qualitatively with those of
the latter paper. Quantitative differences compared to
Bruenn & Mezzacappa (1997) are caused by the inclusion of improved electron
capture rates on nuclei in our work, which significantly increases
electron captures above a few 1010 g cm-3so that lower values of
and s result after
trapping (cf. Martínez-Pinedo et al. 2005; Langanke et al. 2003).
![]() |
Figure 3:
Central electron fraction ![]() ![]() |
Open with DEXTER |
As explained in detail by Bruenn & Mezzacappa (1997), the screened cross
section for neutrino-nucleus scattering reduces the transport
optical depth of low-energy neutrinos (cf. Fig. 3 in Bruenn & Mezzacappa 1997)
and allows them to escape from the core more
easily. This is obvious from a
flux enhancement of neutrinos at energies
MeV in Fig. 6, where the
situation is displayed at a density of
g cm-3.
Ion-ion correlations thus cause a decrease in
and
that is stronger by about 0.02 until neutrino
trapping sets in (Fig. 3).
The homologously collapsing stellar
core correspondingly shrinks, and the shock forms at a somewhat
smaller enclosed mass (Fig. 5).
The shock formation is defined by the moment when the
postshock entropy first reaches a value of
3
per nucleon. A part of the effect visible in
Fig. 5 might, therefore, be a consequence of
the slightly higher core entropy after neutrino trapping in models
with ion screening
(see Figs. 4 and 7).
![]() |
Figure 4: Same as Fig. 3 but for the central (matter) entropy s. |
Open with DEXTER |
![]() |
Figure 5:
Gas entropy vs. enclosed mass at the moment of
shock formation in the
models of Figs. 3 and 4.
The shock formation is defined by the instant when the
entropy in the core first reaches a value of 3
![]() |
Open with DEXTER |
Neutrino-electron scattering is very efficient
in downscattering neutrinos from the high energies, where they are
created mostly by electron captures, to lower energy states.
Therefore the phase space at low energies is quickly refilled.
Figure 8 shows the source term for
energy redistribution by neutrino scatterings off electrons for
two density values below trapping conditions. The downscattering
of high-energy neutrinos explains why the local energy
spectra,
with
being the neutrino energy density,
are essentially the same in Models Shen_ion_off and
Shen_ion_Hor, despite clear differences between the energy
flux spectra of both runs (Fig. 6).
On their way out, escaping neutrinos transfer a part of their
energy to electrons in collisions, thus heating the stellar
medium (Bruenn 1986). Since ion screening of neutrino-nucleus
scatterings reduces the transport opacity and, therefore, the
effective optical depth for energy
exchange with the stellar background mainly for low-energy
neutrinos, but hardly changes the downscattering probability
of high-energy neutrinos (Fig. 9);
the larger loss of lepton number
leads to an increase of the central entropy by about
0.12
per nucleon (Fig. 4).
In Fig. 9 the optical depth
at
the center is calculated as
![]() |
(10) |
![]() |
(11) |
Figure 10 displays the ion screening
factors
superimposed on the normalized neutrino energy spectra
at the stellar center for densities of
1011 g cm-3, 1012 g cm-3,
1013 g cm-3, and 1014 g cm-3.
Figure 10 once more demonstrates
that the reduction of neutrino-nucleus scattering
mostly affects neutrinos at energies below the spectral
maximum for all plotted cases. In combination with
Fig. 9, it also shows that the trapping
conditions for the bulk of the neutrino spectrum are not
influenced strongly by ion screening.
This was identified by Bruenn & Mezzacappa (1997)
as the reason ion-ion correlations have no dramatic
effect on the core deleptonization.
From Fig. 1 it is clear that
differences between the treatments of ion-ion correlations
by Horowitz (1997) and Itoh et al. (2004) are largest for
.
The improvements by Itoh et al. (2004) are,
therefore, most important for the lowest neutrino energies
in the energy window affected by ion screening.
Since the phase space
available at such low energies is small, one cannot expect
large quantitative consequences for stellar core collapse.
This is confirmed by
Figs. 3-5 and 7.
Itoh et al.'s (2004) description (in Model Shen_ion_Itoh)
leads to values of
,
,
and s after
trapping which are essentially indistinguishable from those
obtained with Horowitz's (1997) formulae, consistent
with the insignificant differences
between Models Shen_ion_Hor and Shen_ion_Itoh seen
in the other plots. Note that the crossing of the
-,
- and s-profiles
for simulations with and without ion screening
at 0.45
(Fig. 7)
was also present in the results of Bruenn & Mezzacappa (1997).
Finally testing the sensitivity of the core collapse
evolution to the treatment of ion screening for the ionic
mixture of free protons,
particles, and a representative
heavy nucleus, we also could not discover any differences in
relevance.
A more detailed analysis reveals the reasons for this
insensitivity, which are valid for both employed EoSs.
Below the neutrino trapping regime (i.e., for
g cm-3) even for low-energy
neutrinos (
MeV), the parameter
is larger than unity or around unity, except for neutrinos
interacting with
particles in an ionic mix when
is computed from
Eqs. (6)-(8).
Moreover,
holds
at the same time, implying that the ion-ion correlation
factors
and
for neutrino scattering off heavy nuclei
are essentially the same (see Fig. 1).
Only at densities above the trapping density, the value of
drops significantly below unity and
exceeds 50,
causing visible (typically, factors 2-3 for
MeV; Fig. 10) differences in the
ion-ion suppression factors
.
At these densities, however, the
exact value of the neutrino-nucleus scattering cross section
has no noticeable influence on the evolution of the core
properties and on the neutrino transport.
![]() |
Figure 6:
Spectra of neutrino energy density (solid line)
and energy flux (dashed)
for Model Shen_ion_off ( left) and Model Shen_ion_Hor ( right),
when a density of 1012 g cm-3 is reached at an
enclosed mass of 0.3![]() |
Open with DEXTER |
![]() |
Figure 7:
Profiles of electron fraction ![]() ![]() |
Open with DEXTER |
![]() |
Figure 8: The energy source term for neutrino-electron scattering for densities of a) 1011 g cm-3 and b) 1012 g cm-3at the stellar center in the collapse models with the Shen et al. (1998a,b) EoS. Negative values mean that neutrino energy is "absorbed'' (net scattering out of the corresponding energy bin), and positive values mean "emission'' of neutrino energy (i.e., net scattering of neutrinos into the energy bin). |
Open with DEXTER |
![]() |
Figure 9: The optical depth for energy exchange between neutrinos and stellar plasma as a function of the neutrino energy at the center of the iron core for densities of a) 1011 g cm-3, b) 1012 g cm-3, c) 1013 g cm-3, and d) 1014 g cm-3. The results were taken from collapse calculations with the Shen et al. (1998a,b) EoS. The left panels show enlargements of the low-energy window where ion-ion correlations have the largest effect. |
Open with DEXTER |
![]() |
Figure 10:
The cross-section suppression factor
![]() |
Open with DEXTER |
Alpha particles in the inner core do not become
sufficiently abundant to cause mentionable differences. Their
indirect effect on ion-ion correlations of heavy nuclei
in an ion mixture by reducing aj (and thus )
and increasing
(Sect. 2.2)
for the heavier nuclei is essentially negligible, because their
contribution to the sum in Eq. (7) is
diminished by their number density
being multiplied with a factor
.
Moreover,
particles do not account
for a significant contribution to the total neutral-current
scattering opacity, because the opacity for coherent scattering
of neutrinos by nuclei (Z,N,A) scales roughly with N2/A
and, therefore, is much smaller for
particles than for
heavy nuclei. For this reason, the direct influence of
particles, and thus of the suppression of their coherent
(elastic) scattering cross section for low neutrino
energies, is miniscule, despite the fact that
drops
below unity already at densities
g cm-3in a mixture with heavy nuclei.
On the other hand,
turns out to be always
less than unity and, following Horowitz (1997), is therefore
set to
for evaluating the angle-averaged
ion screening correction factor.
In this paper we presented results from simulations of stellar core collapse with the aim of investigating the consequences of ion-ion correlations in neutrino-nucleus scattering, comparing the improved description of Itoh et al. (2004) with an older one by Horowitz (1997). We employed the EoS of Shen et al. (1998a,b), in addition to Lattimer & Swesty's (1991) EoS, and treated electron captures on heavy nuclei according to Langanke et al. (2003), making nuclei dominant over protons in producing neutrinos up to the density of the phase transition to nuclear matter.
Despite these differences in the input physics, our models essentially confirmed the previous calculations by Bruenn & Mezzacappa (1997), who followed Horowitz (1997) in their description of ion screening. Because ion screening is effective only in a low-energy window where the available phase space is rather small, the influence of ion-ion correlations during stellar core collapse and on the formation of the supernova shock is moderate (Bruenn & Mezzacappa 1997).
We found that the improvement by Itoh et al. (2004) does not
lead to any noticeable differences, because it affects only
neutrinos of very low energies (5 MeV) before
trapping densities (
g cm-3)
are reached. Differences at larger neutrino energies occur
only at higher densities and thus do not affect the
deleptonization and entropy evolution.
Effects due to the ionic mixture of free protons,
particles, and a representative heavy nucleus - using the
linear mixing rule as suggested by Itoh et al. (2004) - were found
to be negligibly small, too, mainly because the abundance
of
particles in the inner regions of the
collapsing stellar core is too low to affect the ion
screening of heavy nuclei indirectly
(see Sect. 2.2). Alpha particles
do not contribute to the total opacity for
elastic neutrino-nucleus scattering on a level where
their ion screening, which becomes sizable only
when the mixture effects of Sect. 2.2
are accounted for, might be relevant.
Improving the description of ion-ion correlations for the complex mix of heavy nuclei with a large variety of components, alpha particles, and free nucleons in the supernova core, however, is desirable. Referring to multi-component calculations based on the Debye-Hückel approximation in the limit of small momentum transfer, Sawyer (2005) argues that a range of N/Z ratios in ionic mixtures can protect against the strong ion screening suppression of neutrino-nuclei scattering predicted by the effective averages of one-component plasma parameters applied in the current literature and in this work. Moreover, the description of nearly free nucleons and nuclei in NSE is expected to hold only up to a density of about 1013 g cm-3. Above this density and below the normal nuclear matter saturation density, a pasta phase may develop with nucleons clustered in subtle and complex shapes. Correlation effects for coherent neutrino scattering can then not be treated within the single heavy nucleus approximation (e.g., Watanabe et al. 2004; Horowitz et al. 2004b,a).
Acknowledgements
We thank Naoki Itoh for providing us with a subroutine to compute his ion-ion correlation factor and for helpful discussions about the treatment of ionic mixtures. We are grateful to the anonymous referee for pointing out an inaccurate use of an average ion sphere radius in the case of ionic mixtures. We also thank K. Langanke, G. Martínez-Pinedo and J. M. Sampaio for their table of electron capture rates on nuclei, which was calculated by employing a Saha-like NSE code for the abundances written by W. R. Hix. Support by the Sonderforschungsbereich 375 "Astro-Particle Physics'' of the Deutsche Forschungsgemeinschaft is acknowledged.