Supernova Neutrinos and the Tau-Neutrino Mass
Extracted source text ยท arXiv astro-ph/9701232
\documentstyle[art11,epsf]{article}
\hyphenpenalty 10000
\topmargin -50pt
\oddsidemargin -20pt
\textheight 680pt
\textwidth 490pt
\tolerance 10000
\def\baselinestretch{1.2}
\def\theequation{\thesection.\arabic{equation}}
\def\sez#1{\section{#1} \setcounter{equation}{0}}
\def\cap#1{\caption[#1]{\it #1}}
\def\sne{supernov\ae\hskip 0.2em}
\def\ev{{\it e\hskip-0.18em V}}
\def\kev{{\it ke\hskip-0.18em V}}
\def\mev{{\it Me\hskip-0.18em V}}
\def\gev{{\it Ge\hskip-0.18em V}}
\def\lapprox{\mathrel{\mathop
{\hbox{\lower0.5ex\hbox{$\sim$}\kern-0.8em\lower-0.7ex\hbox{$<$}}}}}
\def\gapprox{\mathrel{\mathop
{\hbox{\lower0.5ex\hbox{$\sim$}\kern-0.8em\lower-0.7ex\hbox{$>$}}}}}
\def\papprox{\mathrel{\mathop
{\hbox{\lower0.5ex\hbox{$\propto$}\kern-0.8em\lower-0.7ex\hbox{$\sim$}}}}}
\def\mb#1{\mbox{$#1$}}
\def\abs#1{\left| #1 \right|}
\def\nue{\nu_e}
\def\anue{\bar\nu_e}
\def\num{\nu_{\mu}}
\def\nut{\nu_{\tau}}
\def\nui{\nu_i}
\def\anui{\bar\nu_i}
\def\nux{\nu_x}
\def\anux{\bar\nu_x}
\def\mnut{m}
\def\anut{\alpha_{\nu_i}}
\def\eps{\epsilon _\nu}
\def\ee{\epsilon _e}
\def\emin{\epsilon _{{\it min}}}
\def\emax{\epsilon _{{\it max}}}
\def\thf{\vartheta _{{\it fw}}}
\def\tt{t_{{\it tr}}}
\begin{document}
\title{Supernova neutrinos and the $\nut$ mass.}
\author{Gianni {\sc Fiorentini}$^{1,2}$ and Camillo {\sc Acerbi}$^1$
\\[1.1em]
\small
$^1$ Istituto Nazionale di Fisica Nucleare - Sezione di Ferrara \\
\small
$^2$ Dipartimento di Fisica - Universit\`a di Ferrara$\hskip 45pt$ \\[0.5em]
\small 12, via Paradiso -- I-44100 Ferrara -- {\sc Italy} \\[1.0em]}
\date{December 23, 1996}
\maketitle
\begin{abstract}
We perform an extensive investigation of the sensitivity to non-vanishing
$\nut$ mass in a large water \v{C}erenkov detector, developing an analysis
method for neutrino events originated by a supernova explosion. This approach,
based on directional considerations, provides informations almost undepending
on the supernova model.
We analyze several theoretical models from numerical simulations and
phenomenological models based on $SN1987A$ data, and determine optimal values
of the analysis parameters so as to reach the highest sensitivity to a
non-vanishing $\nut$ mass. The minimal detectable mass is generally just above
the cosmologically interesting range, $m\sim 100\:\ev$, in the case of a
supernova explosion near the galactic center. For the case that no positive
signal is obtained, observation of a neutrino burst with Super-Kamiokande will
anyhow lower the present upper bound on $\nut$ mass to few hundred $\ev$.
\end{abstract}
\pagestyle{plain}
%\newpage
\sez{Introduction.}
Upper bounds on the electron (anti-) neutrino mass at the level of few
$\ev$ are obtained from $\beta$ decay experiments \cite{pdg}, and have
been confirmed by data on neutrinos from $SN1987A$, collected by Kamiokande-II
\cite{kII} and IMB \cite{imb} detectors, see e.g. \cite{lim_sn,l&l}.
Concerning the masses of $\num$ and $\nut$, the present experimental bounds
are much less compelling: \mb{m_{\num}\lapprox 170\: \kev},
\mb{\mnut\lapprox 24\: \mev} \cite{pdg}. Therefore, any approach
capable of significantly lowering these limits is highly
desirable. Particularly, an observation sensitive to neutrino masses in
the range \mb{10 - 100\: \ev} would be extremely important for cosmological
implications, see \cite{gershtein}.
Previous analyses \cite{seck,cline} claimed that this goal might be reached
with new generation larger detectors, for a supernova near the galactic
center.
The questions we address in this paper are the following:
\begin{enumerate}
\item Which assumptions about neutrino emission are critical for
observing the effects of a non-vanishing mass?
\item How the analysis procedure can be optimized in order to reach the highest
sensitivity to neutrino masses?
\item Which masses for $\num$ and $\nut$ can be actually explored with a
detector such as Super-Kamiokande?
\end{enumerate}
In general, the determination of $\nu$-masses by means of a supernova
observation is based on the comparison between the detected signal
and the expected one for the case $m=0$.
A non-vanishing mass, in fact, would reflect in an overall delay of the events
number distribution, and also in a spectral distortion of the signal, due to
the energy dependence of the flight-time for massive neutrinos:
\equation \Delta t\, (\eps) = {D\over 2c}\left( {m c^2 \over \eps}
\right) ^2 \endequation
where $D$ is the supernova distance, $c$ is the light speed, and $\eps$ is
the neutrino energy.
The greatest difficult in applying this procedure is in the dependence
on the supernova model assumed to calculate the expected signal.
Moreover, when interested in $\num$ and $\nut$ (hereafter we will refer to
these neutrinos, and their antiparticles, as $\nui$), there is an additional
problem: it is impossible to isolate their signal from that of $\nue$ and
$\anue$, because all $\nui$ detection processes (namely neutral current
reactions) are allowed for electron neutrinos too.
On the other hand, charged current interactions (only possible for $\nue$ and
$\anue$ at the energies of interest) might be used to clearly identify part of
the electron neutrino signal.
One can thus conceive the following strategy:
\begin{enumerate}
\item split, by use of a distinctive signature, the overall set
of events in two classes: the first one containing only $\nue$ and $\anue$,
the second one involving all neutrino species;
\item use the first class of events to derive the expected distribution
for $\nue$ and $\anue$.
In view of the strict limit on $m_{\nue}$, we can treat these neutrinos
as massless for our purpose (e.g. from \mb{m_{\nue}\lapprox 15\: \ev}
\cite{pdg} and equation 1.1, we get \mb{\Delta t \sim {\cal O}(1\, s)}
even for the lowest detectable energies, \mb{\eps\simeq 5\:\mev}, for a SN near
the galactic center);
\item use this distribution to infer that of $\num$ and $\nut$, again for
massless neutrinos;
\item compare the observed signal with that theoretically built in the previous
step and look for mass effects. We assume that only one neutrino (say $\nut$)
is massive and that flavor oscillations do not occur (see \cite{msw} and
references therein for search of $\nut$ mass by use of MSW effects).
The comparison should be performed in the second class of events
(\mb{\nue +\nui}): in such a way we deal with a subset which still
includes the whole $\nui$ signal, but only a part of the $\anue$ ``background''.
\end{enumerate}
The residual dependence on the supernova model is essentially in step 3: how
to connect the $\nue$ and $\anue$ emission features with those of
$\nui$. In this paper we extensively investigate this point, by
analyzing several theoretical and phenomenological SN models and their
implications for the extraction of a non-vanishing $\nut$ mass.
We also discuss how the analysis procedure can be optimized, by a proper
choice of the parameters, such as energy and time windows where to look
for mass effects.
Concerning the sensitivity to neutrino mass, two points are interesting: the
minimal detectable $\nut$ mass and the upper limit which can be established on
$\mnut$ when no positive signal comes out. Correspondingly, we shall address
the following questions:
\begin{list}{3\alph{enumi}.}{\usecounter{enumi}}
\item What is the minimal $\mnut$ value expected to give a signal
distinguishable from the statistical fluctuations of the massless case?
\item What is the minimal $\mnut$ value for which the signal expected for
massless neutrinos can be discriminated from the statistical fluctuations of
the massive case?
\end{list}
The paper is organized as follows: we develop a method of data analysis,
based on directional considerations (Section 2) which we use to determine the
minimal detectable $\nut$ mass for a set of theoretical supernova
models (Section 3). The same analysis is repeated for a set
of phenomenological models derived from $SN1987A$ data (Section 4).
We then compare our results with those of ref. \cite{krauss} (Section 5).
Finally, we calculate, both for theoretical and phenomenological models,
the mass upper bound obtainable through the same method (Section 6).
Detector characteristics, cross sections and expression for event rates are
collected in the Appendix.
\sez{The analysis method: isotropic vs. directional signal.}
Krauss {\it et al.} \cite{krauss} proposed in 1992 a beautiful procedure of
data analysis for \v{C}erenkov detectors, based on directional considerations.
In this section we briefly recall the idea and then we specify our approach to
this method in terms of: $i)$ the kinematical characterization of neutrino
events; $ii)$ the assumptions needed to connect the emission features of
electron and other flavors neutrinos; $iii)$ the quantitative criterion for
the evidence of a non-vanishing $\nut$ mass.
\subsection{Generalities.}
Let us consider Super-Kamiokande (SK hereafter), a water \v{C}erenkov detector
whose main characteristics are listed in the Appendix, see also \cite{url}.
Neutrino events are originated basically in processes of four kinds:
\begin{enumerate}
\item \it capture on a proton: \rm
\equation \anue + p\stackrel{W}{\longrightarrow} n + e^+ \endequation
\item \it scattering off an electron: \rm
\equation \stackrel{\mbox{{\tiny $(\!$}}-\mbox{{\tiny $\! )$}}}{\nue} +e^-
\stackrel{W,Z}{\longrightarrow}\; \stackrel{\mbox{{\tiny $(\!$}}-\mbox{{\tiny
$\! )$}}}{\nue} + e^- \endequation
\equation \nui +e^- \stackrel{Z}{\longrightarrow}\;\nui + e^- \endequation
\item \it capture on an oxygen nucleus: \rm
\equation \nue +\,^{16}\! O \stackrel{W}{\longrightarrow} \,^{16}\! F + e^-
\endequation
\equation \anue +\,^{16}\! O \stackrel{W}{\longrightarrow}\,^{16}\! N + e^+
\endequation
\item \it coherent scattering off an oxygen nucleus: \rm
\equation \hskip 60pt \nux +\,^{16}\! O \stackrel{Z}{\longrightarrow}
\nux +\,^{16}\! O^* \hskip 20pt x=e,\mu ,\tau\endequation
\end{enumerate}
Expressions for the cross sections are collected in the Appendix. The
approximate numbers of expected events in SK, for a $SN1987A$-like supernova
near the galactic center, are shown in Table 1, see also \cite{burr92}.
\begin{table}[htbp]
\begin{center}
\begin{tabular}{||rcl|c||}
\hline
\hline
& & & \\[-.8em]
$\anue p\!\!$ & $\rightarrow$ & $\!\! e^+ n$ & {\bf $\simeq$ 5000} \\
$\stackrel{\mbox{{\tiny $(\!$}}-\mbox{{\tiny $\! )$}}}{\nue}e^-\!\!$ &
$\rightarrow$ & $\!\!\stackrel{\mbox{{\tiny $(\!$}}-\mbox{{\tiny $\! )$}}}
{\nue}e^-$ & {\bf $\simeq$ 100} \\
$\nui e^-\!\!$ & $\rightarrow$ & $\!\!\nui e^-$ & $\bf{\simeq 2 \times 25}$\\
$\stackrel{\mbox{{\tiny $(\!$}}-\mbox{{\tiny $\! )$}}}{\nue}O\!\!$ &
$\rightarrow$ & $\!\!\stackrel{\mbox{{\tiny $(\!$}}-\mbox{{\tiny $\! )$}}}
{\nue}O$ & {\bf $\simeq$ 50} \\
$\nux O\!\!$ & $\rightarrow$ & $\!\!\nux O$ & {\bf $\simeq$ 200} \cite{lang}
\\[10pt]
\hline
\hline
\end{tabular}
\cap{Expected events in SK for a supernova at galactic center.}
\end{center}
\end{table}
Among the four processes, scattering off electrons is the only directional
one. For capture reactions and interactions with nuclei, the observable
particles ($e^\pm$ or $\gamma$) are emitted (almost) isotropically,
whereas in processes
(2.2-3) the recoil electron's angular distribution is strongly forward peaked.
Therefore one can define an appropriate observation angle $\thf$,
along the supernova direction, which includes all of the electrons from
$\nu - e$ scattering.
Here is the required signal splitting: outside the forward cone, nearly any
event is originated by $\nue$ and $\anue$. The only contribution of other
flavors neutrinos is due to (2.6), and can easily be excluded by raising the
minimal detected energy up to 7 -- 8 $\mev$ \cite{lang}. So we can regard
this signal as untouched by mass effects, and use it to derive the expected
distribution in the forward cone for the case \mb{\mnut = 0}.
The analysis procedure is sketched in Figure 1.
\begin{figure}[p]
\centering\leavevmode
\epsfxsize=5.5in\epsffile{method.eps}
%\smallskip
\cap{Sketch of the directional analysis method. Electron
(anti-)neutrinos coming from the supernova originate in the detector an
isotropic signal (on the right). This one is used to infer the directional
distribution for massless $\nut$ (dashed line), which is to be compared with
the observed signal (solid line), see bottom graph.}
\end{figure}
\subsection{Kinematics.}
A specification of the forward observation angle $\thf$ for the analysis.
When a neutrino with energy $\eps$ scatters off an electron at rest which
gets kinetic energy $\ee$, the electron scattering angle $\vartheta$ is
determined from energy-momentum conservation:
\equation \cos\, \vartheta (\eps,\ee) = \sqrt{{ \ee\over \ee + 2m_e}}
\:\left( 1+{m_e\over \eps} \right) \endequation
As $\vartheta$ is a decreasing function of $\ee$ and a rising
function of $\eps$, the maximum value $\vartheta
_{\it max}$ is obtained by taking in (2.7) the minimal detected energy $\emin$
and taking the limit $\eps\rightarrow \infty$.
In this way one has:
\equation \vartheta _{\it max} (\emin) = {\rm acos} \sqrt{{ \emin\over \emin
+ 2m_e}}\endequation
For the lowest value of $\emin$ (i.e. the energy threshold of SK, $\epsilon
_{th} = 5\:\mev$) this gives \mb{\vartheta _{\it max}\simeq 24^{\circ}}.
In Figure 2 we plot the angular distribution of the $\nut$ scattered electrons
along the supernova direction, for two significative values of the lower cut:
$\emin = 5\:\mev$ (SK threshold) and $\emin = 15\:\mev$ (the highest value
useful for the analysis, as we shall see). Although for these calculations we
used the ``reference'' model to be described in Section 3, very similar
distributions are obtained from other supernova models. In the same figure a
plot of \mb{\vartheta _{\it max}} is given as a function of the lower
energy cut.
\begin{figure}[p]
\centering\leavevmode
\epsfbox{kin.ps}
%\smallskip
\cap{The angular distribution of $\nut$ scattered electrons for
two values of $\emin$, and the maximum scattering angle
\mb{\vartheta _{\it max}} as a function of the lower cut. Calculations refer
to Burrows's reference model, see Section 3.1 for details.}
\end{figure}
\begin{figure}[p]
\centering\leavevmode
\epsfbox{det.ps}
%\smallskip
\caption[The probability to detect a diffused electron at an
angle $\alpha$ from the original trajectory, for two significative values of
$\ee$, and the experimental angular resolution as a function of the electron
energy in Kamiokande-II.]{\it The probability to detect an electron at an angle
$\alpha$ from the diffusion trajectory, for two values of
$\ee$, and the experimental angular resolution \mb{\langle \vartheta _d ^2
\rangle ^{1/2}} in Kamiokande-II, from \cite{koshiba}.}
\end{figure}
Besides this purely kinematic dispersion, a systematic angular uncertainty
is also present, due to the detector. This uncertainty
is a decreasing function of the scattered electron
energy too, for more energetic particles originate more \v{C}erenkov light,
and so one can point back to their directions with greater accuracy.
The probability to detect an electron at an angle $\alpha$ from the
scattering direction is assumed to be a gaussian function
\equation dP(\alpha,\ee) \propto e^{-{1\over 2}\left( {\alpha \over \sigma
_\alpha (\ee)}\right) ^2}d\alpha \endequation
where \mb{\sigma _\alpha = {1\over \sqrt{2}} \langle \vartheta _d ^2 \rangle
^{1/2}} and \mb{\langle \vartheta _d ^2 \rangle ^{1/2}} is the empirically
determined angular resolution of the detector for an electron with energy $\ee$.
In Figure 3 one finds the the probability to detect an electron at an
angle $\alpha$ from the original trajectory, as given by (2.9), for
$\ee=5,\, 15\:\mev$, and
a plot of the angular resolution dependence on the electron energy
for the \mbox{Kamiokande-II} detector \cite{koshiba}; a similar
performance is expected for SK. In the case of a 5 $\mev$ electron one has
\mb{\langle \vartheta _d ^2 \rangle ^{1/2} \simeq 38^{\circ}}.
Equation (2.9) tells us that one has a non-vanishing
probability to detect a scattered electron just backward the supernova
direction! This leads us to formulate a more
reasonable definition for the forward observation cone than ``the cone
containing {\it all} $\nu + e$ scattering events''.
We verified that the model independent definition
\equation \thf (\emin) \equiv \sqrt{\vartheta ^2 _{\it max}(\emin) + \langle
\vartheta ^2 _d \rangle(\emin)}\endequation
which can be read as the combination of the kinematic and detector
spreads just like two independent errors, gives $\thf$ values such that no
more than one $\nu +e$ event is missed.
For \mb{\emin =5\:\mev} one has \mb{\thf \simeq 45^{\circ}}.
The corresponding solid angle, \mb{\Omega _{fw} = 2\pi\,[1- \cos(\thf)]},
for \mb{\emin = 5\: \mev} includes a 95\% of the $\nut$ events,
but only a 15\% of the total $\anue$ signal.
In Figure 4 one has the angular distribution of the tau-neutrino events, with
respect to the supernova direction, considering both kinematic and detector
spreads, while in Figure 5 we plotted the forward observation angle $\thf
(\emin)$ as defined in (2.10).
\begin{figure}[p]
\centering\leavevmode
\epsfbox{kin_det.ps}
%\smallskip
\cap{The angular distribution of the $\nut +e$ signal with respect to the
supernova direction, including both kinematic and detector spreads, for two
significative values of $\emin$. Calculations refer to Burrows's reference
model (see Section 3.1).}
\end{figure}
\begin{figure}[p]
\centering\leavevmode
\epsfbox{thf.ps}
%\smallskip
\cap{The forward observation angle as a function of the lower energy cut.}
\end{figure}
\subsection{Supernova.}
Next, we have to describe how the $\nui$ signal is derived from the $\anue$
one. For this purpose, we use three assumptions, supported by the results of
several numerical simulations, (see \cite{sim}) which connect the production
features of electron and other flavor neutrinos {\it in the cooling phase}:
\begin{enumerate}
\item The neutrino energy distributions, at any time and for
each $\nu$ species, are (approximately) thermal, thus specified by the
temperatures of the neutrino-spheres $T_{\nue}(t)$, $T_{\anue}(t)$,
$T_{\nui}(t)$ and luminosities $L_{\nue} (t)$, $L_{\anue} (t)$,
$L_{\nui} (t)$. The emitted neutrino rate spectrum \mb{d^2 N_{\nux} /
(d\eps\, dt)} at time $t$ and energy $\eps$ is therefore:
\equation {d^2 N_{\nux}(\eps ,t) \over d\eps\, dt} = {L_{\nux}(t)\over {\cal F}
_3\: T_{\nux}^4 (t)}\, {\eps ^2 \over 1+e^{\eps / T_{\nux} (t)}}\endequation
where \mb{{\cal F}_3 \equiv \int _0 ^\infty dx\, x^3 (1+e^x )^{-1} \simeq 5.68}
\item Energy equipartition among the six neutrino species holds at
every time:
\equation L_{\nue} (t) = L_{\anue} (t) = L_{\nui} (t) \endequation
\item The temperatures of the six neutrino species are simply proportional at
any time:
\equation \left\{ \begin{array}{ccc} T_{\nue} (t) & = & \alpha _{\nue} \cdot
T_{\anue} (t) \\ T_{\nui} (t) & = & \anut \cdot T_{\anue} (t) \\
\end{array} \right. \endequation
\end{enumerate}
Arguments supporting the second assumption can be found in \cite{janka}.
Typical values of $\alpha _{\nue}$ and $\anut$ are:
\equation \begin{array}{ccc} \alpha _{\nue} & = & 0.5 \, -\, 0.9 \\
\anut & = & 1.1\, -\, 1.9 \\ \end{array}\endequation
In what follows, we fix the ratio between $\nue$- and $\anue$-sphere
temperatures at the central value \mb{\alpha _{\nue}=0.7}, while we'll
keep $\anut$ as a free parameter, in the range given by (2.14).
\subsection{Statistics.}
Finally, we introduce our criterion to estimate the evidence level of
a non-vanishing mass.
We define \mb{N_m ([t_1,t_2],[\emin,\emax])} the expected number of events in
the forward observation cone, in the time interval \mb{[t_1,t_2]}, for an
energy window \mb{[\emin,\emax]}, if the $\nut$ mass is $m$.
Let us assume that tau-neutrinos have mass $m\neq 0$: we expect the experiment
to detect $N_m$ events. The hypothesis that $\nut$ have a vanishing mass and
that the difference \mb{\abs{N_m - N_0}} is due to a statistical fluctuation
of the massless case can be tested by evaluating the quantity
\equation s(m,[t_1,t_2],[\emin,\emax]) = { \abs{N_m - N_0} \over
\sqrt{N_0}} \endequation
Clearly, one can vary \mb{[t_1,t_2]} and \mb{[\emin,\emax]} in order to
optimize the analysis. The best indicator of a non-vanishing mass is therefore
the quantity
\equation S(m) = \max _{[t_1,t_2],[\emin,\emax]}\hskip -6pt
\mbox{{\Large $\{$}} s(m,[t_1,t_2],[\emin,\emax])\mbox{{\Large $\}$}}
\endequation
where the time intervals are chosen arbitrarily within the observation window
(first 100 seconds after bounce), with the only request that at least an
event should be present in \mb{[t_1,t_2]}.
As an example, if \mb{S(100\: \ev) =3} one has a 3$\sigma$ evidence for
a mass \mb{m=100\: \ev}, i.e. the Confidence Level is 99.7\%.
For brevity, we shall define as ``detectable'' those masses such that
\mb{S(m) \ge 3}, and we denote the minimal detectable mass as
\mb{m_{3\sigma}\equiv \min \{ m\, :\, S(m)\ge 3 \}}.
Note that we are using only event counts, disregarding the possible
additional information of spectral deformations.
A sample of this statistical analysis is reported in Figure 6, where we
plot the expected signal in the forward observation cone for $m=0$ and
$m=200\:\ev$ (for details on the model used for the calculation, see Section
3). For example, in the time interval \mb{[t_1,t_2]=[25,35]\: s}, and for
\mb{[\emin,\emax]= [15,25]\:\mev}, one expects 1.3 events for the massless
case, and 4.2 events for massive neutrinos; the evidence index is therefore
\mb{s(200\ev ,\, [25,35]s,\, [15,25]\mev) = 2.5}. We conclude that,
in the time and energy intervals under examination, $m=200\:\ev$ is not
sufficient to produce a clear effect.
\sez{Results from theoretical models of supernov\ae.}
Till February 1987, numerical simulations were the only way to explore
supernova neutrinos features. Most of the analyses
just concerned the very initial phase of the emission (\mb{t\lapprox 1\:s}),
mainly for implications on explosion mechanism,
while only a few dealed with the long term cooling stage, which is relevant for
us. For a review of simulations see for instance \cite{suz_rev}.
In this section we use the simulations due to Burrows \cite{burr88}, which
describe the cooling phase, to evaluate the minimal mass $m$ whose effects
could be observed with the analysis method presented in Section 2.
We also investigate the dependence of the results on the free
parameters of the analysis.
\subsection{Models and general results.}
We performed a complete set of calculations using several supernova models
taken from Burrows. In ref. \cite{burr88} seventeen models are presented:
disregarding ``exotics'' and black holes, the remaining ten models look
consistent with data from $SN1987A$.
Burrows's data, that cover the first 20 seconds for $\anue$ luminosity and
temperature, were extrapolated to later times by use of power-laws:
$$A(t)=A_0 (1+{t\over \tau _A})^{-n_A}$$
where $A$ denotes here both $L_{\anue}$ and $T_{\anue}$, and the parameters
$A_0$, $\tau _A$ and $n_A$ are obtained from fits to late times behaviours
(\mb{t>10\: s}). We also introduced a time parameter $\tt$ beyond which
the neutrino luminosities vanish: this happens when the
proto-neutron star becomes transparent to neutrinos. Typical values for
$\tt$ are (40 -- 50) $s$ \cite{burr92}. Finally, for the other neutrino
species, luminosities and temperatures are determined through relations
(2.12-13).
Unless otherwise stated, all results reported in the text, tables and figures,
refer to the Super-Kamiokande detector for the following ``default'' set of
parameters:
\equation \begin{array}{lcll} \anut & = & 1.5 & \\ \tt & = & 45 & s \\
D & = & 10 & kpc \\ \end{array} \endequation
For each model, the main characteristics are listed in Table 2, together with
the results: the optimal choices for time and energy windows,
and the value of the minimal detectable mass, as defined in Section 2.4.
The $\anue$ event rate and the sensitivity to non-vanishing neutrino mass are
shown in Figures 7 and 8 for a few representative models: the {\it M}ost
{\it F}avourable case (\# 52), the {\it L}east {\it F}avourable case (\# 62)
and the {\it RE}ference model (\# 55), which very well accounts for $SN1987A$
data.
We notice that the minimal detectable mass lies anyhow outside the
cosmologically interesting region: even in the most favourable model
we have \mb{m_{3\sigma} \simeq 120\: \ev}. For the reference
model one has \mb{m_{3\sigma} \simeq 140\: \ev}.
When looking at the dependence on the model characteristics, we find that
in general: $i)$ for a given equation of state (EOS), mass effects are least
pronounced for more massive cores; $ii)$ for a given core mass, neutrino mass
effects are weaker for the soft EOS.
Those features can be understood, at least qualitatively. We remark that, for a
given EOS, the star radius decreases with mass, so that in more massive cores
matter is more compressed. Also, for a given mass, softer EOS allow higher
compression. Thus higher masses and/or softer EOS imply higher densities, and
consequentely higher temperatures and opacities. More energetic neutrinos are
produced, so that $\nut$-mass induced delays are shorter. Furthermore, and more
important, the longer cooling phase results in a higher background $\anue$
flux at later times.
The increase of the cooling time due to mass and EOS is shown in Figure 7:
MF and RE models have the same EOS but different masses, whereas LF and RE
models have the same mass but different EOS.
\begin{figure}[p]
\centering\leavevmode
\epsfbox{sample.ps}
%\smallskip
\cap{An example of the statistical analysis: the dashed line refers to the
expected distribution in the forward observation cone for the case $m=0$, while
the solid one is for $m=200\:\ev$, see text for explanations. Burrows's
``reference'' model described in Section 3 was used for this example.}
\end{figure}
\begin{table}[p]
\begin{center}
\begin{tabular}{||c||c|c|c|c||c|c|c||}
\hline
\hline
& & & & & & & \\[-.8em]
Model & EOS & $M_i\:[M _\odot]$ & $M_f\: [M _\odot]$ & $E_{tot}$ [{\it foe}] &
$[\emin,\emax]$ [$\mev$] & $[t_1,t_2]$ [$s$] & $m_{3\sigma}$ [\ev] \\[.4em]
\hline
\hline
52 & stiff & 1.2 & 1.2 & 152 & [10,20] & [25,85] & 120 \\
\hline
53 & stiff & 1.3 & 1.3 & 171 & [10,20] & [30,95] & 130 \\
\hline
54 & stiff & 1.4 & 1.4 & 191 & [10,20] & [30,100] & 130 \\
\hline
\bf 55 &\bf stiff &\bf 1.3 & 1.5 &\bf 228 &\bf [10,20] &\bf [40,95] &\bf 140 \\
\hline
56 & stiff & 1.3 & 1.6 & 262 & [10,20] & [45,100] & 130 \\
\hline
57 & stiff & 1.3 & 1.8 & 326 & [10,20] & [40,95] & 130 \\
\hline
59 & soft & 1.2 & 1.2 & 158 & [10,20] & [35,100] & 140 \\
\hline
60 & soft & 1.3 & 1.3 & 178 & [10,20] & [40,100] & 140 \\
\hline
61 & soft & 1.3 & 1.4 & 205 & [10,20] & [40,100] & 140 \\
\hline
62 & soft & 1.3 & 1.5 & 235 & [10,20] & [40,100] & 140 \\
\hline
\hline
\end{tabular}
\caption[Characteristics and results of Burrows's supernova models.]
{\it Characteristics and results for Burrows's supernova models: the columns
indicate the equation of state (EOS), the progenitor core mass $M_i$,
the proto-neutron star mass $M_f$ (after accretion), the total emitted
energy $E_{tot}$ (1 foe = \mb{10^{51}\: erg}), the optimal energy
$[\emin,\emax]$ and time $[t_1,t_2]$ windows in order to probe mass effects,
and the minimal detectable mass $m_{3\sigma}$.}
\end{center}
\end{table}
\begin{figure}[p]
\centering\leavevmode
\epsfbox{burr.ps}
%\smallskip
\cap{The isotropic $\anue$ signal in the whole SK detector, for the least/most
favourable model (LF/MF), and for the ``reference'' model (RE), see text.}
\end{figure}
\begin{figure}[p]
\centering\leavevmode
\epsfbox{mass.ps}
%\smallskip
\cap{Mass sensitivity for a few Burrows's models: same notations as in
Figure 7.}
\end{figure}
\subsection{Dependence on the model parameters.}
In order to investigate the dependence of the previous results on the
emission parameters, we repeated our calculations by varying one
parameter at a time, while keeping all the others constant at the values of
(3.1). The results quoted in the following are for the reference model,
but we obtained similar behaviours for the others.
We first varied the parameter $\anut$, defined in Section 2.3 as the ratio
of $\nui$-- to $\anue$--sphere temperatures, repeating the analysis
for five different values in the range given by (2.14).
The results of this search are summarized in Table 3, where we report, for
each value of the parameter, the minimal detectable mass, together with the
number of $\nut$ events and the characteristic delay time for \mb{m=150\:\ev}.
In Figure 9 we plot the evidence index $S(m)$ for the central value
and the two extrema of the $\anut$ range.
The minimal detectable mass depends rather
weakly on the $\nui$-sphere temperature. The point is that when varying
$\anut$ two competing effects arise. For example, when $\anut$ is increased:
$i)$ the $\nut$-sphere temperature is higher, leading to more energetic
neutrinos, and so to shorter delays;
$ii)$ the number of $\nut$ events increases, leading to a higher statistics.
Our calculations essentially show that these effects almost balance each other,
so that small variations of $\anut$ do not affect substantially the
sensitivity to a non-vanishing mass.
\begin{figure}[p]
\centering\leavevmode
\epsfbox{alpha.ps}
%\smallskip
\cap{Mass sensitivity for several values of the $\nui$-sphere temperature, in
Burrows's reference model.}
\end{figure}
\begin{table}[p]
\begin{center}
\begin{tabular}{||c|c|c||c|c|c||}
\hline
\hline
& & & & & \\[-.8em]
$\anut$ & $N_{\nut}$ & $\Delta t (\eps =\anut\!\cdot\! 15 \mev)$ [{\it s}] &
$[\emin,\emax]$ [$\mev$] & $[t_1,t_2]$ [$s$] & $m_{3\sigma}$ [\ev] \\[.4em]
\hline
\hline
1.0 & 4.93 & 52 & [10,20] & [35,95] & 140 \\
1.3 & 11.8 & 31 & [10,20] & [35,100] & 140 \\
\bf 1.5 & \bf 16.4 & \bf 23 & \bf [10,20] & \bf [40,100] & \bf 140 \\
1.7 & 16.8 & 18 & [10,20] & [40,100] & 140 \\
2.0 & 17.5 & 13 & [10,20] & [40,100] & 150\\
\hline
\hline
\end{tabular}
\cap{Mass sensitivity dependence on the $\nui$-sphere temperature in
Burrows's reference model, with $\tt = 45\: s$. $N_{\nut}$ is
the number of $\nut$ events in the forward cone and $\Delta t$ is the
characteristic delay time for a neutrino in the explorable mass range
($m=150\:\ev$) and for $D=10\, kpc$.}
\end{center}
\end{table}
Concerning the dependence on the transparency time,
we performed three calculations of the minimal detectable mass for $\tt$
around the value given by \cite{burr92}. Again the results are
essentially stable, see Figure 10, where we plotted the evidence index $S(m)$
for the inquired values.
Clearly, the more sudden the transparency occurs, the more evident the
delayed neutrinos will be: should we take the (maybe) unphysical
choice \mb{\tt =30\: s}, we find \mb{m_{3\sigma}\simeq 110\:\ev}.
\begin{figure}[p]
\centering\leavevmode
\epsfbox{transp.ps}
%\smallskip
\cap{Mass sensitivity for several values of the transparency time, in Burrows's
reference model.}
\end{figure}
\begin{table}[p]
\begin{center}
\begin{tabular}{||c|c||c|c|c||}
\hline
\hline
& & & & \\[-.8em]
$\tt$ [{\it s}] & $N_{\nut}(\tt\rightarrow t_{\it max})$ &
$[\emin,\emax]$ [$\mev$] & $[t_1,t_2]$ [$s$] & $m_{3\sigma}$ [\ev] \\[.4em]
\hline
\hline
30 & 6.2 & [10,20] & [30,70] & 110 \\
\hline
40 & 3.9 & [10,20] & [35,95] & 130 \\
\bf 45 & \bf 3.2 & \bf [10,20] & \bf [40,100] & \bf 140 \\
50 & 2.5 & [10,20] & [40,100] & 140 \\
\hline
\hline
\end{tabular}
\cap{Mass sensitivity dependence on the transparency time, in Burrows's
reference model, with $\anut = 1.5$.
\mb{N_{\nut}(\tt\rightarrow t_{\it max})} is the number of $\nut$ events
expected to occur after the transparency, in the case of $m=150\:\ev$ and for
$D=10\, kpc$.}
\end{center}
\end{table}
Finally,....
\begin{figure}[p]
\centering\leavevmode
\epsfbox{dist.ps}
%\smallskip
\cap{Mass sensitivity for several values of the supernova distance, in
Burrows's reference model.}
\end{figure}
\begin{table}[p]
\begin{center}
\begin{tabular}{||c|c|c||c|c|c||}
\hline
\hline
& & & & & \\[-.8em]
$D$ [{\it kpc}] & $N_{\nut}$ & $\Delta t (\eps =20\mev)$ [{\it s}] &
$[\emin,\emax]$ [$\mev$] & $[t_1,t_2]$ [$s$] & $m_{3\sigma}$ [\ev] \\[.4em]
\hline
\hline
8 & 25.5 & 14.4 & [10,20] & [40,100] & 140 \\
\bf 10 & \bf 16.4 &\bf 18.0 & \bf [10,20] & \bf [40,100] & \bf 140 \\
12 & 11.4 & 21.6 & [10,20] & [35,100] & 130 \\
\hline
50 & 1.3 & 90.0 & [5,100] & -- & -- \\
\hline
\hline
\end{tabular}
\cap{Mass sensitivity dependence on the supernova distance, in Burrows's
reference model, with $\anut = 1.5$ and $\tt=45\, s$.
$N_{\nut}$ is the number of $\nut$ events and $\Delta t$ is the typical
neutrino time-delay in the case of $m=150\:\ev$.}
\end{center}
\end{table}
\subsection{Best choices for the analysis parameters.}
These are particularly important dependences to investigate, because we deal
now with detection parameters, which we can change in order to optimize the
analysis.
It was already noted \cite{krauss} that an upper energy cut enhances the
mass sensitivity of the analysis method. This is due to the fact that the
isotropic signal occurs at a higher energy with respect to the directional one.
In fact in the dominant isotropic process, i.e. $\anue$ capture on a proton,
the emitted positron takes almost all of the neutrino energy, while in the
$\nut$ process, namely scattering off an electron, the average energy of the
outgoing particle is roughly $\ee \sim {1\over 2}\eps$.
We stress that by means of $\emax$ we cut high energy $\nut$ events, which have
short mass-induced delays, so that the loss in sensitivity is low.
We should be aware, anyway, that this dynamical effect is partially balanced
by the higher $\nut$ energy at the emission.
Concerning the lower energy cut that we introduced,
many competing effects are present, leading to a more complex situation.
Basically, at high values of $\emin$ the number of
$\nut$ events get small and, most of all, we deal only with
high energy neutrinos, for which the flight-time delay is shorter.
On the other hand, as we said in Section 2.2, the forward
observation angle is a decreasing function of the lower cut (see eq. 2.10).
Rising $\emin$ leads therefore to a narrower cone, in which the ratio
between directional and isotropic $\anue$ signal becomes more favourable.
In Table 5 we report the full results of our analysis on the reference model
for six different energy windows, while in Figure 11 we plot the mass
sensitivity for three significative choices.
As we can see from Figure 11, the best choice depends on the explored mass
region, because of the $m$ dependence in the delay formula (1.1): the larger
is $m$, the higher should be the lower cut. For our
explorable range \mb{m=100 - 200\: \ev} the choice \mb{\emin =10\:\mev;\:
\emax =20\:\mev} gives the best sensitivity for all models and for
all values of the emission parameters $\anut$ and $\tt$.
With regard to the time window, as we can see from Table 2 the best choice
again depends on the explored mass: larger $m$ mean more delayed
neutrinos, which are to be observed at later times. For our explorable range
the choice \mb{t_1 =40\: s;\: t_2 =100\: s} seems in general the most
appropriate.
\begin{figure}[p]
\centering\leavevmode
\epsfbox{energy.ps}
%\smallskip
\cap{Mass sensitivity for significative choices of the energy window
\mb{[\emin,\emax]}, in Burrows's reference model.}
\end{figure}
\begin{table}[p]
\begin{center}
\begin{tabular}{||c|c|c|c|c|c||c||}
\hline
\hline
& & & & & & \\[-.8em]
$[\emin,\emax]$ & $N_{\nut} / N_{tot}$ [\%] & $\thf$ & $N_{\nut}/N_{tot}$ [\%] &
$\Delta t _{\it max}$ [{\it s}] & $[t_1,t_2]$ & $m_{3\sigma}$ \\
$[\mev]$ & (all SK) & $[{\it deg}]$ & (in $\thf$) &
$\equiv \Delta t (\eps =\emin)$ & [$s$] & [\ev] \\[.4em]
\hline
\hline
$[7.5,15]$ & 0.91 & 37 & 6.4 & 210 & [40,100] & 140 \\
\hline
$\bf [10,20]$ & \bf 0.66 & \raisebox{-10pt}{\bf 32} & \bf 6.5 &
\raisebox{-10pt}{\bf 120} & \bf [40,100] & \bf 140 \\
$[10,100]$ & 0.52 & & 5.4 & & [30,100] & 190 \\
\hline
$[15,25]$ & 0.48 & \raisebox{-10pt}{27} & 7.0 & \raisebox{-10pt}{52} & [20,80]
& 160 \\
$[15,100]$ & 0.43 & & 6.4 & & [20,90] & 170 \\
\hline
$[20,100]$ & 0.39 & 24 & 6.8 & 29 & [20,90] & 180 \\
\hline
\hline
\end{tabular}
\cap{Mass sensitivity for significative energy windows, in Burrows's
reference model. Calculations are for $\tt = 45\: s$ and $\anut = 1.5$.
\mb{\Delta t _{\it max}} is the maximum delay for a neutrino in the
explorable mass range ($m=150\:\ev$), corresponding to the case of minimal
detected energy, for $D=10\, kpc$.}
\end{center}
\end{table}
\subsection{This section's conclusions.}
We conclude, for the case of a Burrows-like supernova near the galactic center,
that:
\begin{enumerate}
\item there is no way of getting evidence of \mb{\mnut \lapprox 120 \,\ev};
\item this conclusion is essentially independent of the value of the uncertain
emission parameters $\anut$ and $\tt$;
\item for the explorable mass range, the optimal energy window is
\mb{\ee \sim [10,20]\: \mev}.
\item for the explorable mass range, the optimal time window is
\mb{t\sim [40,100]\: s}.
\end{enumerate}
\sez{Results from phenomenological models of supernov\ae.}
So far we considered theoretical models for neutrino emissions, essentially
based on numerical simulations.
In order to investigate further the potentialities of the analysis method,
in this section we consider phenomenological models, built so as to reproduce
the data from $SN1987A$.
The mere 19 events of the 1987 supernova were not sufficient to allow a
detailed reconstruction of the emission features, and several models were
obtained by fitting the data with different analytical functions. We selected
two of these models, belonging to significative classes, to explore a
wider range of possibilities.
\begin{figure}[p]
\centering\leavevmode
\epsfbox{phenom.ps}
%\smallskip
\cap{Total signals in the whole detector for exponential cooling
(dotted), power-law cooling (dashed), and Burrows's reference model (solid).}
\end{figure}
\subsection{Power-law cooling.}
In this framework, we first consider a power-law parametrization,
defined by the relations:
\equation \left\{ \begin{array}{lll} T(t) & = & T^0 \,\left( 1+{t\over \tau}
\right) ^{-n}\\ L(t) & = & L^0 \,\left( 1+ {t\over \tau} \right) ^{-4n} \\
\end{array} \right. \endequation
Note that, being \mb{L(t)\propto T^4 (t)}, in this model the radii of the
neutrinospheres are assumed as constant.
We use as best fit parameters for $SN1987A$ those obtained by Bludman \&
Schinder \cite{b&s}:
\equation \left\{ \begin{array}{llll} T_{\anue}^0 & \simeq & 4.20 & \mev \\
L_{\anue}^0 & \simeq & 19.8 & foe/s \\ \tau & \simeq & 2.78 & s \\
n & \simeq & 0.4 & \\ \end{array} \right. \endequation
Again we used relations (2.12-13) to derive luminosities and temperatures
for the other neutrinos species.
As we can see from Figure 12, where we plotted the $\anue$ event rates in SK
for (4.1) and for Burrows's reference
model, the behaviour of the power-law cooling is similar to that of
theoretical models of Section 3.
With the values of (4.2) and for the default set of parameters (3.1),
we obtained the minimal detectable mass \mb{m_{3\sigma}\simeq 110\:\ev}.
This result is consistent with what found for Burrows theoretical models,
and is again outside the cosmologically interesting range.
%Concerning the dependence on parameters...
\subsection{Exponential cooling.}
A simpler analytic parametrization is obtained assuming that the neutrinosphere
temperatures fall exponentially, while the radii are again kept constant.
One has this way
\equation \left\{ \begin{array}{lll} T (t) & = & T^0 \, e^{-t/4\tau} \\
L(t) & = & L^0 \, e^{-t/\tau}\\ \end{array} \right. \endequation
The best fit parameters for $SN1987A$ are taken from Loredo \& Lamb
\cite{l&l}:
\equation \left\{ \begin{array}{llll} T_{\anue}^0 & \simeq & 4.16 & \mev
\\ L_{\anue}^0 & \simeq & 12.9 & foe/s \\ \tau & \simeq & 4.6 &
s \\ \end{array} \right. \endequation
With these choices one obtains a minimal detectable mass \mb{m_{3\sigma}\simeq
70\:\ev}.
The exponential cooling, although not clearly distinguishable from the
power-law on the grounds of the $SN1987A$ data,
predicts at later times (\mb{t\gapprox 20\: s}) a substantially reduced
$\anue$ flux (see again Figure 12). This behaviour can account for the
sensitivity to smaller masses: in fact, with a lower $\anue$ background,
the presence of delayed massive neutrinos is much more evident.
Actually, there is no physical reason for assuming an exponential law, and
theoretical models generally predict a slower decrease of neutrino luminosity
at late times \footnote{We remind that cooling of a degenerate gas by
black-body radiation yields a power-law decrease of temperature: \mb{T(t)=T^0
\left( 1+{t\over \tau}\right) ^{-1/2}}.}. Furthermore, when compared with
$SN1987A$ data, exponential cooling models yield fits which are worse than
those of power-law models \cite{b&s}. In conclusion, it seems to us that
exponential cooling models are not very reliable for predicting late times
neutrino luminosities.
\sez{Comparison with Krauss's results.}
Krauss {\it et al.} \cite{krauss}, once developed their
analysis method, also evaluated the minimal detectable mass. They found
that a mass as small as 50 $\ev$ is detectable at 99\% C.L.
with Super-Kamiokande, for a medium luminosity burst from a SN at $D=10\, kpc$.
This result is more optimistic than we found by using Burrows's
theoretical models and/or $SN1987A$ phenomenological models (we recall, e.g.,
that for Burrows's most favourable model one has \mb{m_{3\sigma}= 120\:\ev} at
99.7\% C.L.).
This discrepancy can't be ascribed to the differences in the analysis
procedures. We have verified that, with our approach, one obtains
for the middle model of \cite{krauss} (n. 17) a minimal detectable mass
\mb{m_{3\sigma}\simeq 48\:\ev}.
Actually, in the supernova models of \cite{krauss},
after an initial accretion phase (\mb{t\sim 0.5\: s}), the
luminosities of all neutrino flavors are assumed to decay {\it exponentially
with a very fast rate} (\mb{\tau _L\sim 1\: s}). This results in a quite
short tail of the event distribution, and so the signal of delayed $\nut$ is
much more evident, as we already noted for the $SN1987A$ exponential model.
This feature is evident if we compare a typical model from
\cite{krauss} with one of those we used in the foregoing.
In Figure 13, for instance, we plotted the isotropic (scaled to $\thf$) and
directional event rates both in Krauss's model 17 and in Burrows's model 54,
which give fairly the same event number. As we can see, with just
\mb{m=100\,\ev}, in Krauss's model the directional signal outruns the
isotropic ``background'' after a few seconds, while in Burrows's model this
occurs only at a very later times (\mb{t>30\: s}).
It looks to us that the fast luminosity decay assumed in \cite{krauss} is not
justified since: $i)$ phenomenological models with exponential cooling yield
longer decay times when adapted to fit $SN1987A$ data, see \cite{l&l,b&s};
$ii)$ also theoretical models predict longer decay times, see \cite{burr88}
and the previous discussion.
\begin{figure}[p]
\centering\leavevmode
\epsfbox{krauss.ps}
%\smallskip
\cap{Isotropic (scaled to $\thf$) and directional signals
in Krauss's model n. 17 and in Burrows's model n. 54 (same event number);
a $m =100\,\ev$ was assumed.}
\end{figure}
\sez{Upper bounds.}
The situation may look somehow discouraging; nevertheless we should
remember that the present experimental limit to $\nut$ mass is tens $\mev$.
Therefore even in the most probable case of non-observation of mass effects,
we will however significantly lower this limit.
In order to obtain actual upper bounds, anyway, we have to introduce a more
stringent statistical criterion than the one we used to define our
``detectable'' mass.
With our definition of Section 2.4, in fact, we calculated the minimal value
of $m$ expected to produce a $3\sigma$ effect, checking against statistical
fluctuations of the massless case. We should now take into account
fluctuations of both massive and massless case.
Therefore, the new best indicator is the quantity
\equation \bar S (m) \equiv \max _{[t_1,t_2],[\emin,\emax]}
{\abs{N_m - N_0} \over \sqrt{N_m + N_0}} \endequation
As an example, if \mb{\bar S(100\: \ev) =3} we can claim the upper
bound $m\le 100\: \ev$ at a 3$\sigma$ level, i.e. with a 99.7\%
Confidence Level.
In practice, $N_m$ is generally (much) larger than $N_0$, so that (6.1) can be
replaced to a good approximation with:
\equation \bar S (m) \equiv \max _{[t_1,t_2],[\emin,\emax]}
{\abs{N_m - N_0}\over \sqrt{N_m}} \endequation
This can be read very much like (2.15), i.e. assuming that neutrinos are
massless, the observer will detect $N_0$ events (or a number close to it).
Eq. (6.2) tells then how many sigma's the result is out of the expectation
for massive neutrinos.
\begin{table}[h]
\begin{center}
\begin{tabular}{||l|c||c|c|c||}
\hline
\hline
& & & & \\[-.8em]
Model & $m_{3\sigma}$ [\ev] & $[\emin,\emax]$ [$\mev$] & $[t_1,t_2]$ [$s$] &
$\bar m_{3\sigma}$ [\ev]\\[.4em]
\hline
\hline
Burrows RE & 140 & [15,100] & [15,100] & 200 \\
Burrows MF & 120 & [10,20] & [15,95] & 140 \\
Power-law & 110 & [10,20] & [10,85] & 140 \\
Exponential & 70 & [10,20] & [10,60] & 90 \\
\hline
Krauss 17 & 50 & [10,20] & [5,15] & 70 \\
\hline
\hline
\end{tabular}
\cap{The minimal excludable mass for a few significative models of supernova.}
\end{center}
\end{table}
The results for two significative theoretical models (Burrow's reference and
most favourable models - see Section 3) and two phenomenological models
(exponential and power-law cooling - see Section 4) are listed in Table 6.
Using the same notation of Section 2.3, we define the minimal excludable mass
as \mb{\bar m_{3\sigma}\equiv \min \{ m\, :\, \bar S(m)\ge 3\} }.
Clearly a larger mass is required to probe an upper bound rather than a
simply detectable effect, for one has \mb{N_m > N_0 \:\Rightarrow\:
\bar S(m) < S(m)}.
Anyway, one can see that even in the case of non-observation of
mass effects, we will able to lower the present upper bound on $\nut$ mass by
5 orders of magnitude!
One should be aware, anyway, that the bounds of Table 6 (and also the previous
results on the minimal detectable mass) may not hold if the
$\nut$ mass is too large. In this case, in fact, one has a too
broad event distribution, see for istance \cite{wolf}.
A very crude estimate of the maximum mass $m_{\it max}$ explorable by
delay-based analyses can be obtained by requiring that the
rate of delayed $\nut$ events keeps greater than the background rate.
For SK one finds roughly \mb{m_{\it max}\sim {\cal O}(10\:\kev)}. Our results
should be read therefore as lower extrema of an excluded mass region: in
the case of non-observation of mass effects one could exclude $m$ in the
window \mb{200\: \ev - 20\:\kev}.
\sez{Conclusions.}
Let us summarize the main points of this paper:
\begin{list}{\roman{enumi})}{\usecounter{enumi}}
\item We performed an extensive investigation of the sensitivity to
non-vanishing $\nut$ mass in a large water \v{C}erenkov detector, developing
the analysis method introduced by Krauss {\it et al.} \cite{krauss}.
As the most important point is the dependence of the results on the supernova
model, we analysed several theoretical models from numerical simulations of
Burrows \cite{burr88} and phenomenological models based on $SN1987A$ data
\cite{b&s,l&l}.
\item We determined optimal values of the analysis parameters (energy cuts and
time window) so as to reach the highest sensitivity to a non-vanishing $\nut$
mass.
\item Our conclusion is that the minimal detectable mass is generally just
above the cosmologically interesting range, $m\sim 100\:\ev$, see Table 2
for details.
\item Differences with respect to \cite{krauss}, which claimed sensitivity to
significantly smaller masses, are ascribed to the shorter decay time of
neutrino luminosities assumed in \cite{krauss}, contrary to theoretical models
as well to $SN1987A$ phenomenological models.
\item For the case that no positive signal is obtained, observation of a
neutrino burst with SK will anyhow lower the present upper bound on $\nut$ mass
(\mb{m\lapprox 24\:\mev} \cite{pdg}) to few hundred $\ev$, see Table 6.
\end{list}
\appendix
\sez{Appendix}
\subsection{Super-Kamiokande characteristics.}
In Table 7 we summarize the main characteristics of the Super-Kamiokande
detector (the fiducial mass, the threshold energy and detection efficiency, and
the background rate spectrum) as we used in our analysis.
\begin{table}[h]
\begin{center}
\begin{tabular}{|cclr|}
\hline
\hline
& & & \\[-.8em]
$M_{det}^\dagger$ & = & $32\; kton$ & \\
$\epsilon _{th}$ & = & $5\; \mev$ & \\
$\eta(\ee)$ & $\simeq$ & $\vartheta (\ee - \epsilon _{th})\cdot
\left[ 0.93-e^{-(\ee\: /\: 9\, \mev)^{2.5}}\right]$ & \cite{burr88} \\
$b(\epsilon)$ & $\simeq$ & $M_{det}\cdot {0.082\over \sqrt{2\pi}\, 0.87}\,
e^{-{1\over 2}\left( {\epsilon - 6.2 \over 0.87}\right) ^2}\: {\#\over \mev\,
s}\;$ & elab. \cite{l&l} \\[.4em]
\hline
\hline
\multicolumn{4}{c}{$^\dagger$ We refer to the fiducial mass for supernova
events.}
\end{tabular}
\end{center}
\cap{Super Kamiokande characteristics.}
\end{table}
\subsection{Cross sections.}
\begin{enumerate}
\item \it capture on a proton: \rm
\equation {d\sigma _{\anue p}\over d\ee}={1\over 4}\sigma _0 \left( {\eps \over
m_e c^2}\right) ^2 \: (1 + 3\alpha ^2)(1+\delta _{wm}) \left( 1 -{Q \over \eps}
\right) \left[\left(1-{Q \over \eps} \right)^2 -\left({m_e c^2 \over\eps}\right)
^2 \right]^{1/2} \cdot \delta(\eps - \ee -Q) \endequation
where $\alpha \simeq -1.26$ is the axial-vector coupling constant, $Q\simeq
1.293\,\mev$ is the neutron-proton mass difference, \mb{\delta _{wm}\simeq
-3.3\cdot 10^{-3}(\eps -Q/2)/\mev} is the weak-magnetism correction,
and
$$\sigma _0 \equiv {4\over\pi}{1\over(c\hbar)^4} G_F^2 (m_e c^2)^2 \simeq 1.76
\cdot 10^{-44} \; cm^2$$
\item \it scattering off an electron: \rm
\equation {d\sigma _{\nu_x e} \over d\ee} = {1\over 2}{\sigma _0 \over m_e c^2}
\left[ A_x + B_x \left( 1-{\ee \over \eps}\right) ^2 -{m_e c^2 \ee \over\eps ^2}
\sqrt{A_x B_x} \right] \endequation
where the constants $A$ and $B$ have values:
$$ A_x = \left\{ \begin{array}{lllll} \left( {1\over 2}+\sin ^2 \theta _W
\right) ^2 & \simeq & 0.536 &\; , \hskip 10pt & \nu_e \\ \sin ^4 \theta _W &
\simeq & 0.0538 &\; , \hskip 10pt & \bar \nu_e \\ \left( -{1\over 2}+\sin ^2
\theta _W \right) ^2 & \simeq & 0.0719 &\; , \hskip 10pt & \nu_i \\ \sin ^4
\theta _W & \simeq & 0.0538 &\; , \hskip 10pt & \bar \nu_i \\ \end{array}
\right. $$
$$ B_x = \left\{ \begin{array}{lllll} \sin ^4 \theta _W & \simeq & 0.0538 & ,
\hskip 10pt & \nu_e \\ \left({1\over 2} +\sin ^2 \theta _W \right) ^2 & \simeq
& 0.536 & , \hskip 10pt & \bar \nu_e \\ \sin ^4 \theta _W & \simeq & 0.0538 &
, \hskip 10pt & \nu_i \\ \left(-{1\over 2} +\sin ^2 \theta _W \right) ^2 &
\simeq & 0.0719 & , \hskip 10pt & \bar \nu_i \\ \end{array} \right. $$
\item \it capture of a $\nue$ on an oxygen nucleus: \rm \\
The cross section of this process can be roughly approximated to \cite{nue_ox}
\equation {d\sigma _{\nue O} \over d\ee} \simeq 0.16\sigma _0 \left( {\eps -
\tilde \epsilon \over m_e c^2}\right) ^2 \theta (\eps - \tilde \epsilon)\,\cdot
\, \delta (\eps - \tilde \epsilon -\ee + m_e)\endequation
where $\tilde \epsilon \sim 13\,\mev$ is an effective reaction threshold.
\item \it capture of a $\anue$ on an oxygen nucleus: \rm \\
We use the approximation given by \cite{krauss}
\equation {d\sigma _{\anue O} \over d\ee} \simeq 0.074\sigma _0 \left(
{\eps - \tilde \epsilon \over m_e c^2}\right) ^2 \theta (\eps -\tilde \epsilon)
\,\cdot\, \delta (\eps - \tilde \epsilon -\ee + m_e)\endequation
where the reaction threshold is again $\tilde \epsilon \sim 13\,\mev$.
\end{enumerate}
\subsection{From emission to detection.}
Being \mb{d^2 N_{\nux} / d\eps\, dt'} the rate spectrum of $\nux$ emitted
at time $t'$ with energy $\eps$, as defined in Section 2.3, the corresponding
differential flux at the detector at time $t$ is
\equation {d^2 \Phi_{\nux}(\eps ,t) \over d\eps\, dt}={1\over 4\pi D^2}\int dt'
\, {d^2 N_{\nux}(\eps ,t') \over d\eps \, dt'}\,\cdot\,\delta (t-t'-\Delta
t(\eps, m))\endequation
For each detection process one obtains therefore the event rate $dN_b /dt$ as
\equation {dN_b (t)\over dt} = \int _{\emin} ^{\emax} d\ee \,\eta
(\ee) \int _0 ^\infty d\eps {d^2 \Phi_{\nux}\over d\eps \, dt}\, n_b\,
{d \sigma _{\nu b} (\eps ,\ee ) \over d\ee}\endequation
where \mb{d\sigma _{\nu b} / d\ee} is the differential cross section
of the reaction under examination, $n_b$ is the target number, and
\mb{\eta (\ee)} is the electron/positron detection efficiency.
For the natural background one has instead the time-independent rate
\equation B(\emin,\emax) =\int _{\emin} ^{\emax} d\ee\, b(\ee)\, \eta (\ee)
\endequation
where $b(\ee)$ is the empirically determinated background rate spectrum.
%\newpage
\begin{thebibliography}{\hspace{1.5cm}}
\bibitem{pdg} Particle Data Group, {\it Phys. Rev.} {\bf D 54} (1996) 1
(and references therein)
\bibitem{lim_sn} W.D. Arnett and J.L. Rosner, \it Phys. Rev. Lett. \bf 58 \rm
(1987) 1906 \\
D.N. Spergel and J.N. Bahcall, \it Phys. Lett. \bf B 200
\rm (1987) 366
\bibitem{l&l} T.J. Loredo and D.Q. Lamb, in: Proc. 14th Texas Symp. on
Relativistic Astrophysics, {\it Ann. N.Y. Ac. Sc.} (1988) 601
\bibitem{kII} K.S. Hirata \it et al., Phys. Rev. Lett. \bf 58 \rm (1987) 1490
\bibitem{imb} R.M. Bionta \it et al., Phys. Rev. Lett. \bf 58 \rm (1987) 1494
\bibitem{gershtein} S.S. Gershtein and Y.B. Zeldovich, \it JETP Lett. \bf 4
\rm (1966) 120 [\it Zh. Eksp. Teor. Fiz. Pis'ma Red. \bf 4
\rm (1966) 174]
\bibitem{seck} D. Seckel, G. Steigman, and T.P. Walker, \it Nucl. Phys. \bf B
366 \rm (1991) 233
\bibitem{cline} D.B. Cline, G.M. Fuller, W.P. Hong, B. Meyer, and J. Wilson,
\it Phys. Rev. \bf D 50 \rm (1994) 720
\bibitem{msw} H. Minakata and H. Nunokawa, {\it Phys. Rev.} {\bf D 41}
(1990) 2976
\bibitem{krauss} L.M. Krauss, P. Romanelli, D. Schramm, and R. Lehrer,
{\it Nucl. Phys.} {\bf B 380} (1992) 507
\bibitem{url} For a review on SK visit the URL
{\it http://www-sk.icrr.u-tokyo.ac.jp/}
\bibitem{burr92} A. Burrows, D. Klein, and R. Gandhi, {\it Phys. Rev.}
{\bf D 45} (1992) 3361
\bibitem{lang} K. Langanke, P. Vogel, and E. Kolbe, {\it Phys. Rev. Lett.}
\bf 76 \rm (1996) 2629
\bibitem{koshiba} M. Koshiba, \it Phys. Report \bf 220 \rm (1992) 231
\bibitem{sim} R. Mayle, J.R. Wilson, and D.N. Schramm, \it Astrophys. J.
\bf 318 \rm (1987) 288 \\
S.W. Bruenn, \it Phys. Rev. Lett. \bf 59 \rm (1987) 938
\bibitem{janka} T.-H. Janka, {\it Astropart. Phys.} {\bf 3} (1995) 377
\bibitem{suz_rev} H. Suzuki, in: {\it Physics and Astrophysics of Neutrinos},
eds. M. Fukugita and A. Suzuki (Springer-Verlag, Tokyo, 1994)
763
\bibitem{burr88} A. Burrows, {\it Astrophys. J.} {\bf 334} (1988) 891
\bibitem{b&s} S.A. Bludman and P.J. Schinder, {\it Astrophys. J.} {\bf 326}
(1988) 265
\bibitem{wolf} J.M. Soares and L. Wolfenstein, {\it Phys. Rev} {\bf D 40}
(1989) 3666
\bibitem{nue_ox} J. Arafune and M. Fukugita, {\it Phys. Rev. Lett.} {\bf 59}
(1987) 367
\end{thebibliography}
\end{document}
\sez{Photons vs. positrons: the two-bands method.}
Recently, Langanke, Vogel, and Kolbe \cite{lang} showed that, handling data
in a proper way, coherent scattering off oxygen nuclei (2.6) can provide a
more evident contribute to $\nui$ signal than scattering off electrons (2.3).
\equation
\stackrel{(-)}{\nu_i}+ ^{12}C\stackrel{Z}{\longrightarrow}\; \stackrel{(-
)}{\nu _i} + ^{12}\parbox[t]{1.6in}{$C^*$ \\ $^| \!\!\! \rightarrow \: ^{12}C
+ \gamma \: _{15.1 \: MeV}$}
\endequation
\item \it Extended exponential cooling: \rm \\
\equation \left\{ \begin{array}{lll} R(t) & = & R_0 e^{-t\, p/4\tau} \\
T(t) & = & T_0 e^{-t/4\tau} \\ \Rightarrow L(t) & = & 4\pi A R_0^2 T_0^4
e^{-t\, (p+2)/2\tau}\\ \end{array} \right. \endequation
\equation \left\{ \begin{array}{llll} p & = & 2 & \\ T_{\anue}^0 &
\simeq & 3.9 & \mev \\ \tau_{\anue} & \simeq & 8.2 & s \\ R_{\anue} & \simeq &
33 & km \\ \end{array} \right. \endequation
VECCHIA CONCLUSIONE
In this paper we investigated the possibility of determinating the $\nut$ mass
by a proper analysis of the data obtained from a large water \v{C}erenkov
detector in the case of a supernova explosion near the galactic center.
In particular we dealed with an analysis method, already introduced by Krauss
{\it et al.}, which, by means of directional considerations on neutrino events,
provides informations almost undepending on the supernova model.
Basically, the idea is to compare the isotropic signal (only due to ``massless''
$\anue$, via capture on a proton) with the distribution in a small cone
along the SN direction: this one contains also the $\nut$ signal due to
scattering off electrons. In the case of non-vanishing $\mnut$,
the event time distribution inside the cone would be somehow delayed
with respect to the isotropic one.
We extensively developed this method: first we introduced a definition for
the forward observation cone, as the one containing up to 95\% of the $\nut$
signal, and calculated it for several values of the energy cut.
Concerning the dependence on SN model, we showed that if the supernova emission
features in the cooling phase are independent of neutrino flavor
(i.e. one has luminosities equipartition and temperatures proportionality),
the only important residual dependence is the ratio between $\anue$- and
$\nut$-sphere temperatures.
We also introduced our approach to the data statistical analysis, showing how
the sensitiveness to a non-vanishing $\nut$ mass can be improved by a proper
choice of the analysis parameters, such us the energy cut and the time
observation window, and we defined a quantitative criterion to evaluate the
evidence of mass effects.
Then we used the above analysis method to derive the minimal detectable mass
for a set of theoretical supernova model.
STIMA DELLA MASSIMA MASSA ESPLORABILE
Given two neutrinos of mass $m$ and energy $\epsilon _1$ and $\epsilon _2$, the
temporal spread at detection, as caused by the flight-time dependence on
$\eps$, is
$$ \Delta t_{1,2} (m) = {D\over 2c} m^2 \abs{{1\over \epsilon _1^2}-
{1\over \epsilon _2^2}}$$
The maximum duration of the $\nut$ signal is obtained by using the two extrema
of the detected neutrino spectrum, i.e. $\epsilon _1 =\ecut$ and $\epsilon _2
\rightarrow \infty$. One has this way:
\equation \Delta t_{max}(m,\ecut) = {D\over 2c}\left( {m\over \ecut}\right) ^2
\endequation
The average rate of $\nut$ events is therefore
\equation \langle {dN_{\nut}\over dt}\rangle (m,\ecut) = {N_{\nut}(\ecut)\over
\Delta t_{max}(m,\ecut)} \endequation
where $N_{\nut}(\ecut)$ is the number of detected events if the energy cut
is $\ecut$.
By requiring that the rate (6.4) keeps greater than the background rate
$B(\ecut)$ above the lower energy cut, one has the following
expression for the maximum explorable mass
\equation m_{max} (\ecut) = \ecut\,\sqrt{{2c\over D}\:{N_{\nut}(\ecut)\over
B(\ecut)}}\endequation