Violent Relaxation of Indistinguishable Objects and Neutrino Hot Dark Matter in Clusters of Galaxies
Extracted source text ยท arXiv astro-ph/9606057
\documentstyle[12pt,aas2pp4,flushrt,epsf]{article} %preprint
%\documentstyle[12pt,aasms4]{article}
\begin{document}
\title{Violent Relaxation of Indistinguishable Objects and\\
Neutrino Hot Dark Matter in Clusters of Galaxies}
\author{Kull A., Treumann R. A., and B\"ohringer H.}
\affil{Max-Planck-Institut f\"ur Extraterrestrische Physik\\
D-85740 Garching, Germany}
\begin{abstract}
The statistical mechanical investigation of violent relaxation (Lynden-Bell
1967) is extended to indistinguishable objects. It is found that,
coincidentally, the equilibrium distribution is the same as that obtained for
classical objects. For massive neutrinos, the Tremaine \& Gunn (1979) phase
space bound is revisited and reinterpretated as the limit indicating the onset
of degeneracy related to the coarse-grained phase space distribution. In the
context of one of the currently most popular cosmological models, the Cold and
Hot Dark Matter (CHDM) model (Primack et al. 1995), the onset of degeneracy
may be of importance in the core region of clusters of galaxies. Degeneracy
allows the neutrino HDM density to exceed the limit imposed by the Tremaine
\& Gunn (1979) bound while accounting for the phase space bound.
\end{abstract}
\keywords{dark matter --- galaxies: clusters: general --- methods: statistical}
\section{Introduction}
Self-gravitating astrophysical systems having initial mass and velocity
distributions far from equilibrium are believed to evolve towards equilibrium
by violent relaxation, a collective, collisionless process mediated by the
fluctuating gravitational field. Compared to the time scale of the two-body
relaxation process, the time scale of violent relaxation is short.
The analytical investigation of violent relaxation in terms of statistical
mechanics was first introduced by Lynden-Bell (1967). Formally,
the equilibrium distribution resembles the Fermi-Dirac distribution.
For stellar and galactic systems it tends towards the
Maxwell-Boltzmann distribution (Lynden-Bell 1967, Shu 1978).
In this Letter we investigate violent relaxation of indistinguishable objects.
We show that, despite of the difference between the violent relaxation statistics of
indistinguishable and distinguishable objects, coincidentally, the equilibrium
distributions of both cases are of the same form as expected by Shu (1987).
When applied to massive neutrinos it turns out that the degeneracy implicit in the
equilibrium distribution plays an important role leading to a reinterpretation of the
Tremaine \& Gunn (1979) phase space bound as a limit indicating the onset of
degeneracy. In the context of one of the currently
most popular cosmological models based on initial conditions compatible with
inflation theory ($\Omega=1$), the Cold and Hot Dark Matter (CHDM) model
(Primack et al. 1995), the onset of degeneracy allows the neutrino HDM
density to exceed the limit imposed by the Tremaine \& Gunn (1979) bound
while accounting for the phase space bound.
\section{Violent Relaxation Statistics of Indistinguishable Objects}
Following Lynden-Bell (1967) we attempt to describe violent relaxation
by the use of a maximum entropy principle. The one-particle phase space
$\mu$ of the objects undergoing violent relaxation is divided into a
large number of microcells the dimension of which is chosen so as to yield the
mass density in the units mass per spatial and velocity volume, $\Delta^3x$
and $\Delta^3v$, respectively. The collisionless
nature of violent relaxation then imposes an additional constraint on the
time evolution of the $\mu$-space. Initially not overlapping phase elements
do never overlap. Therefore, a microscopic exclusion principle for phase
elements in $\mu$-space is established.
The volume of the microcells is chosen in such a way that the
microcells are occupied by at most one object each and that
the probability of finding occupied adjacent microcells
representing two-body encounters is small (Saslaw 1985). With $m$
the particle mass, the mass per microcell is either zero or $m$.
Let $\eta$ denote the corresponding mass density of an occupied microcell.
The initial phase space distribution before violent relaxation
may then be described by the set of occupied microcells.
At the macroscopic level, the microcells (or phase
elements) are grouped into macrocells containing a large number $\nu$ of
microcells. Let $N$ be the total number of occupied microcells corresponding
to $N$ objects. Suppose that there is a microstate in which there are $n_a$
occupied microcells in the $a$th macrocell. Due to the collisionless
interaction of the violent relaxation process, the occupation state of
each microcell does not change, i.e. there is no cohabitation. The
indistinguishable nature of the objects under consideration leads to a
total number
\begin{equation}
w_a={\nu!\over{n_a!(\nu-n_a)!}}
\end{equation}
of ways of assigning the $n_a$ occupied microcells to the $\nu$ microcells of
the $a$th macrocell as already realized by Lynden-Bell (1967) but not worked out
further. Since the elements are indistinguishable, there is just
one way of splitting the total of $N$ elements into groups $n_a$, i.e. there
are no permutations. Thus, for indistinguishable elements, the total number of
microstates is
\begin{equation}
W_i=\prod_a {\nu!\over{n_a!(\nu-n_a)!}}
\end{equation}
where, for distinguishable elements, the corresponding number is
(Lynden-Bell 1967)
\begin{equation}
W_d={N!\over{\prod_a n_a!}} \prod_a {\nu!\over{(\nu-n_a)!}} \,.
\end{equation}
The most probable state is found by the standard procedure of maximizing
log $W$ subject to the constraints of constant total energy and constant mass.
Since the expressions $W_i$ and $W_d$ differ only by a factor $N!$, the
maximization procedure performed in the continuum representation yields,
coincidentally, for both expressions log $W_i$ and log $W_d$ the same
coarse-grained phase space distribution
\begin{equation}
\bar{f}(v,x)={\eta \over{1+\mbox{exp}[\beta(\epsilon - \mu)]}} \,.
\label{lbdist}
\end{equation}
Here, $\epsilon=v^2/2+\Phi(x)$, with $\Phi(x)$ the gravitational
potential normalized as $\Phi(x \rightarrow \infty)=0$.
The Lagrange parameters $\beta$ and $\mu$ are chosen according to
the macroscopic constraints.
Thus, the most probable coarse-grained phase space distribution of
indistinguishable elements is the same as for distinguishable elements.
Shu (1978) reexamined the statistical mechanical discussion of violent relaxation
in terms of particles and obtained an expression equivalent to (\ref{lbdist})
determining the occupation number $n_a$
\begin{equation}
n_a={\nu \over{1+\mbox{exp}[\beta_p m (\epsilon_a - \mu_p)]}}
\label{shudist}
\end{equation}
of the $a$th macrocell where $\epsilon_a=v_a^2/2+\Phi(x_a)$, and the Lagrange
parameters, $\beta_p$ and $\mu_p$, now refer to the particle description. For
both the cases (\ref{lbdist}) and (\ref{shudist}), with $\bar{f} \ll \eta$ and
$\nu_a \ll \nu$, respectively, the phase space distribution tends towards
the Maxwell-Boltzmann distribution.
Despite of these agreements, the discussions by Shu (1978) and Lynden-Bell (1967)
differ what concerns degeneracy effects potentially present in (\ref{lbdist})
and (\ref{shudist}). While Lynden-Bell (1967) considers degeneracy to be eventually
important in central regions of galaxies, Shu (1978) concludes degeneracy to
indicate the onset of two-body encounters. However, if two-body encounters are
present, there is no microscopic exclusion principle any more. Thus the relevant
phase space distribution is of Maxwell-Boltzmann type. In any case,
for massive neutrinos, the discussion turns out to be irrelevant.
\section{The Case of Massive Neutrinos}
If the neutrino has a non-zero rest mass as predicted by some extensions of
the Standard Model (Gelmini \& Roulet, 1995), this may have important consequences
for both the solar neutrino and the cosmological dark matter problem. Massive
neutrinos with a rest mass of a few eV have been shown to account remarkably
well for both problems (Primack et al. 1995). The discussion in this section
is restricted to one neutrino flavor with $g_\nu=1$.
We now apply the result of the foregoing section to relict neutrinos of
non-zero rest mass trapped in a gravitational potential, e.g. related to a cluster
of galaxies. The initial phase space density of the unperturbed relict
neutrino background is a relativistic Fermi distribution
\begin{equation}
f_\nu(p)={m_\nu^4 \over{(2 \pi \hbar)^3}}
{1 \over {e^{(p c/{kT_\nu})}+1}}
\label{initdist}
\end{equation}
with $T_\nu=1.95$ K (for $z=0$ and in the units mass per spatial and
velocity volume $d^3x$ and $d^3v$, respectively). For $p=0$, the
maximum value $f_\nu(0)=(1/2)\,m_\nu^4/(2 \pi \hbar)^3$ is obtained.
If the neutrinos are massive, the phase space distribution $f$ of the
unperturbed relict neutrino background is changed due to gravitational
interactions. Violent relaxation considered here leads to a convolution
of the phase space structure. Regions of initially different
phase space densities end up entwined together. However, the
coarse-grained phase space density $\bar{f}$ cannot exceed the maximum
of the initial fine-grained phase space density $f_\nu$ since the
collisionless nature of violent relaxation implies conservation of the
fine-grained phase space density $f$.
Since neutrinos are elementary particles and their unperturbed, initial
phase space distribution is the relativistic Fermi distribution
(\ref{initdist}), the volume of the microcells is chosen to be the
elementary phase space volume $m_\nu^3/(2 \pi \hbar)^3$. This
choice guarantees the microcells to be occupied by at most one object
each. The corresponding phase space mass density of an occupied microcell
is $\eta=m_\nu^4/(2 \pi \hbar)^3$. Note, that the initial condition assumed
in the foregoing section is perfectly fulfilled by the phase space
distribution of the relict neutrinos.
Suppose now the relict massive neutrinos to be subject to violent relaxation.
According to the foregoing section, their coarse-grained phase space distribution
$\bar{f}$ tends towards the Fermi type distribution (\ref{lbdist}).
First we note, that, in general, the relation max$\{\bar{f}\}\le$ max$\{f\}$ must hold.
With the natural assumption $\beta > 0$, equation (\ref{lbdist}) implies
\begin{equation}
\left(\Phi - \mu\right) \ge 0 \,.
\label{inequ0}
\end{equation}
where
for a given $\beta$ equality corresponds to the highest possible phase space
density. If no degeneracy effects are present,
then $\bar{f} \ll \eta$ and equation (\ref{lbdist}) requires
\begin{equation}
\beta\left(\Phi - \mu\right) \gg 0 \,.
\label{inequ1}
\end{equation}
In terms of the mass density
\begin{equation}
\rho_\nu= 4\pi\eta\int\limits_{0}^{\infty}v^2
e^{-\beta\left[(v^2/2+\Phi)-\mu\right]}dv
\end{equation}
and the specified value of $\eta=m_\nu^4/(2 \pi \hbar)^3$,
inequality (\ref{inequ1}) leads to
\begin{equation}
\rho_\nu \ll {m_\nu^4 \sigma^3\over{\sqrt{8 \pi^3}\hbar^3}}
\label{inequ2}
\end{equation}
where $\sigma = \beta^{-1/2}$ is the one-dimensional velocity dispersion.
When solved for $m_\nu$ we get
\begin{equation}
m_\nu \gg \left( {\rho_\nu \sqrt{8 \pi^3}\hbar^3 \over{\sigma^3}} \right)^{1/4}
\doteq m_{d}
\label{inequ3}
\end{equation}
This expression is interpreted as a limit on the type of the
coarse-grained phase space density $\bar{f}$. For given velocity
dispersion $\sigma$, neutrino density $\rho_\nu$ and a neutrino mass $m_\nu$
well above $m_{d}$, $\bar{f}$ is the Maxwell-Boltzmann distribution. On the
other hand, if $m_\nu$ is of order or even below $m_{d}$, $\bar{f}$ differs
from the Maxwell-Boltzmann distribution. In this case, degeneracy effects
become important and the correct coarse-grained phase space density $\bar{f}$
is the Fermi-type distribution (\ref{lbdist}). We note, that the possible
onset of two-body encounters preventing Fermi-like degeneracy in the case
of stars or galaxies (Shu 1978) does not apply to neutrinos. The time scale
of gravitational two-body encounters for neutrinos is extremely long. Moreover,
even when strong two-body encounters would occur, the Fermionic nature of
the neutrinos prevents single microcells to become cohabitated. Thus,
for neutrinos, a microscopic exclusion principle applies in any case.
Inequality (\ref{inequ3}) is formally equivalent to the well known Tremaine \& Gunn
(1979) bound on the rest mass of neutrinos which are gravitationally bound in
galactic halos or galaxy clusters. However, the Tremaine \& Gunn (1979) limit
is derived from the assumption that $\bar{f}$ is the Maxwell-Boltzmann distribution
and the condition that the maximum of the coarse-grained phase space density
$\bar{f}$ cannot exceed the maximum of the initial fine-grained phase space density
$f$ given by (\ref{initdist}). Therefore, the Tremaine \& Gunn (1979) limit
implicitly relies on the assumption of $\bar{f}$ to be of Maxwell-Boltzmann type.
On the contrary, (\ref{inequ3}) refers to the type of
the phase space distribution. Therefore we stress,
that, from the point of view of (\ref{inequ3}), the limit implied should not
be interpreted as limit on the neutrino mass. Instead, (\ref{inequ3}) states
that the phase space distribution for $m_\nu$ near the Tremaine \& Gunn (1979)
bound differs significantly from the Maxwell-Boltzmann distribution.
In terms of the gravitational potential $\Phi$ and the neutrino mass
$m_\nu$ it is possible to formulate several limits on the neutrino rest
mass $m_\nu$ based on the phase space density $\bar{f}$ defined by
(\ref{lbdist}). We intend here to derive the most robust one.
According to (\ref{inequ0}) the mass density of the totally degenerate,
densest state is
\begin{equation}
\rho_\nu= 4\pi\eta/2\int\limits_{0}^{v_l} v^2 dv
\label{degstat}
\end{equation}
where $v_l$ is the limiting velocity below which the maximum number of
microcells is occupied ($\bar{f}=\eta/2$) while above $v_l$ all microcells
are empty ($\bar{f}=0$). Assuming only bound states to be relevant, i.e.
$\epsilon = v^2/2 + \Phi \le 0$, the most robust limit on the neutrino mass
is found for $v_l = \sqrt{|2 \Phi|}$ yielding
\begin{equation}
m_\nu \ge \left({12 \rho_\nu \pi^2 \hbar^3 \over{|2 \Phi|^{3/2}}}\right)^{1/4}.
\label{limit1}
\end{equation}
This limit is robust in the sense that it holds for every coarse
grained phase space distribution $\bar{f}$ representing a bound state including
those eventually not described by (\ref{lbdist}), as for example
anisotropic phase space distributions suggested by Madsen (1991) or
Ralstone \& Smith (1991).
In order to compare these limits to the Tremaine \& Gunn (1979) bound, the relation
$\sigma^2 \approx \Phi/3$ holding for a Maxwellian velocity distribution
related to a self-gravitating isothermal sphere is adopted.
The Tremaine \& Gunn (1979) bound on the neutrino rest mass then becomes
\begin{equation}
m_\nu \ge \rho_\nu^{1/4} \left({6 \pi \hbar^2 \over{|\Phi|}}\right)^{3/8}
\label{limit2}
\end{equation}
which is about a factor $\approx 1.18$ higher than (\ref{limit1}). In terms of
the neutrino density this leads to an augmentation of the density limit of about
a factor of $\approx 1.95$. As a consequence, the onset of degeneracy allows the
neutrino HDM density to exceed the limit imposed by the Tremaine \& Gunn (1979)
bound by a factor of $\approx 1.95$ while accounting for the phase space bound.
In comparing the two limits, one should take into account that the Maxwellian
velocity distribution related to (\ref{limit2}) has a tail of unbounded
neutrinos. For the probably more physical truncated Maxwellian distribution
consisting of bounded neutrinos only, the difference between the limits would
become bigger. On the other hand, the totally degenerate situation on which the
limit (\ref{limit1}) is based is unlikely to be realized by violent relaxation.
\section{Discussion and Conclusion}
This section considers violent relaxation of massive neutrinos in
clusters of galaxies in the context of the currently popular Cold and Hot Dark
Matter (CHDM) model. Extensive numerical simulations show these models to agree
well with observations on cosmological scales (e.g. Klypin et al. 1993,
Noltenius et al. 1994, Klypin \& Rhee 1994). In what follows we refer to the two most
promising neutrino mass schemes due to Primack et al. (1995). The first scheme divides
the neutrino HDM in two neutrino species each with a mass of 2.4 eV while the second
splits the HDM in three 1.6 eV neutrino species. The median value of the
initial phase space distribution $f_\nu$ of each family of the relict neutrinos
is only $0.06$ (Madsen \& Epstein 1984). As an approximation we therefore again
take $\eta=m_\nu^4/(2 \pi \hbar)^3$ where $m_\nu$ is 2.4 eV or 1.6 eV,
respectively, i.e. for our purposes the phase space elements occupied by two or
three neutrinos are neglected.
The left panel of figure \ref{fig1} presents the contour plot of density limits
derived from (\ref{limit1}) and (\ref{inequ2}) for the 1.6 eV neutrino mass scheme.
The right panel shows the density limits for the 2.4 eV mass scheme. The limits are
plotted as upper limits to the total matter densities $\rho$ consisting of HDM, CDM
and baryonic matter. They are functions of the gravitational potential depth,
indicated by the velocity dispersion of the corresponding self-gravitating isothermal
sphere $\sigma \approx \sqrt{\Phi/3}$, and the ratio of the neutrino
HDM density $\rho_\nu$ to the total matter density $\rho$. The solid contours represent
the most robust limit for bound states given by (\ref{limit1}). Along these contours
of specified total density $\rho$, the neutrino HDM is found in the totally degenerate,
densest state. Augmentation of $\rho$ is only possible if it is accompanied by a
change of the ratio $\rho_\nu/\rho$ or $\sigma$. On the other hand, the more $\rho_\nu/\rho$
is diminished while leaving $\rho$ and $\sigma$ unchanged, the more the coarse-grained
phase space $\bar{f}$ of the neutrino HDM becomes non-degenerate. The dashed contours
are the density limit obtained from (\ref{inequ2}). Well below this limit, $\bar{f}$
becomes a Maxwellian. In addition, this density limit by (\ref{limit2}) corresponds
to the Tremaine \& Gunn (1979) bound.
The parameter range of figure \ref{fig1} includes typical values of $\sigma$ and
$\rho_\nu/\rho$ expected for clusters of galaxies. The suppression of the HDM neutrino
density $\rho_\nu$ compared to the CDM and baryonic matter density due to its lower
density contrasts (e.g. Klypin et al. 1993, Noltenius et al. 1994)
is reflected by values of $\rho_\nu/\rho$ below the cosmological value $\Omega_\nu=0.2$
(Primack et al. 1995). For the 1.6 eV neutrino mass scheme (left panel of
figure \ref{fig1}), one observes a broad parameter region yielding upper total
density limits comparable to typical core densities of cluster of galaxies.
Adopting the 2.4 eV neutrino mass scheme (right panel of figure \ref{fig1}) shrinks
the relevant parameter region. However, since the coarse-grained phase space
distribution $\bar{f}$ of the neutrino HDM is expected to be a Maxwellian only well
below the limit indicated by the dashed contours, it is found that for both CHDM models
$\bar{f}$ may differ significantly form the Maxwell-type distribution. Accordingly,
$\bar{f}$ should be represented by the Fermi-type distribution (\ref{lbdist}). Due to
(\ref{inequ0}) it accounts for the phase space bound while it allows the neutrino density
to exceed the density limit imposed by the Tremaine \& Gunn (1979) bound
(indicated by the dashed contour lines, too) by a factor of $\approx 1.95$.
Violent relaxation eventually fades before the final state (\ref{lbdist})
is attained. Thus, (\ref{lbdist}) is unlikely to be reached throughout the whole
cluster, but it is reasonable to hold in the central region where violent
relaxation occurs most violently. In order to determine the global properties
of self-gravitating spheres based on coarse grained phase space distributions
(\ref{lbdist}), a careful investigation is needed. In particular, the onset of
degeneracy eventually related to core formation and not yet observed in simulations
(Bryan et al. 1994) appears interesting.
In summary, the statistical mechanical investigation of violent relaxation
has been extended to indistinguishable objects. It is found that the
equilibrium distribution is the same as that obtained for distinguishable
objects. However, in contrast to stellar and galactic systems, the onset of
degeneracy is not prevented by two-body encounters in the case of massive
neutrinos. Thus, the coarse-grained phase space density $\bar{f}$ is of
Fermi-type and, to some extent, degeneracy may be present. For currently
popular CHDM models, the onset of degeneracy is shown to be relevant in
the core region of clusters of galaxies.
The authors thank Anatoly Klypin for valuable criticism and discussion.
\newpage
\begin{references}
\reference{BRYAN94}Bryan G. L., Klypin A., Loken C., Norman M. L., \& Burns J. O. 1994,
\apjl, {\bf 437}, L5.
\reference{GELMINI95} Gelmini, G., Roulet, E. 1995, Rep. Prog. Phys., {\bf 58}, 1207.
\reference{KLYPIN93} Klypin, A., Holtzman, J., Primack, J., Regos, E. 1993,
\apj, {\bf416}, 1.
\reference{KLYPIN94} Klypin A., \& Rhee, G. 1994, \apjl, {\bf434}, L4
\reference{LBELL67} Lynden-Bell, D. 1967, \mnras, {\bf136}, 101.
\reference{MADSEN84} Madsen, J., \& Epstein, R. I. 1984, \apj, {\bf282}, 11.
%\reference{MADSEN87} Madsen, J. 1987, \apj, {\bf316}, 497.
\reference{MADSEN91} Madsen, J. 1991, \apj, {\bf367}, 507.
%\reference{MELOTT83} Melott, A. L. 1983, \apj, {\bf264}, 59.
\reference{NOLTENIUS94} Noltenius R., Klypin A., \& Primack J. R. 1994, \apjl, {\bf422}, L45
\reference{PRIMACK95} Primack, J. R., Holtzman, J., Klypin, A., \& Caldwell, D. O. 1995,
\prl, {\bf74}, 2160.
\reference{RALSTON91} Ralston, P. R., \& Smith, L. L. 1991, \apj, {\bf367}, 55.
\reference{SASLAW85} Saslaw, W. C. 1985, Gravitational Physics of Stellar and
Galactic Systems (Cambridge University Press: Cambridge).
\reference{SHU78} Shu F. H. 1978, \apj, {\bf225}, 83.
\reference{SHU87} Shu F. H. 1987, \apj, {\bf316}, 502.
\reference{TREMAINE79} Tremaine, S., \& Gunn, J. E. 1979, \prl, {\bf42}, 407.
\end{references}
\onecolumn
\newpage
\figcaption{Contour plot of upper limits to the total density $\rho$ (in kg m$^{-3}$) for
the 1.6 eV (left panel) and the 2.4 eV (right panel) neutrino mass scheme.
The limits are plotted as functions of the gravitational potential represented by
the one dimensional velocity dispersion $\sigma$ according to ${\sigma^2 \approx \Phi/3}$
and the ratio $\rho_\nu/\rho$. The solid contour lines represent the
most robust density limit for bound states. Dashed contour lines indicate the density limit related
to the type of the coarse-grained phase space distribution $\bar{f}$ of the neutrino HDM. Well
below this limit $\bar{f}$ is a Maxwellian.
\label{fig1}}
\begin{figure}
\plotone{fig1a.eps}
Fig. 1.- (left panel)
\end{figure}
\begin{figure}
\plotone{fig1b.eps}
Fig. 1.- (right panel)
\end{figure}
\end{document}