One of the driver to propose cold H2 as a dark matter candidate is our increasing knowledge about evolution of galaxies along the Hubble sequence (e.g. Pfenniger, Combes & Martinet, 1994). Because of spiral waves and bars, galaxies progressively concentrate their mass towards the center, and the late-type galaxies evolve to early-types in the sequence. Besides, HI observations of rotation curves have shown that the fraction of dark matter in the total mass is larger in late-type galaxies: therefore, some of the dark matter must be transformed into stars during evolution (cf Pfenniger, this conference).
The quantity of baryons in the Universe (and more
precisely the fraction of the critical density in baryons
b) is constrained by
the primordial nucleosynthesis
to be b = 0.013
h-2, with h = H0 / (100 km/s/Mpc)
is the reduced Hubble constant. With h = 0.5,
b
is 0.09, and more generally
b is between 0.01 and 0.09
(Walker et al. 1991,
Smith et al. 1993),
while the visible matter corresponds to
* ~ 0.003 (M/L/5)
h-1 (+ 0.006 h-1.5 for hot gas). Therefore, most
of the baryons (90%) are dark.
In rich clusters of galaxies, the baryons are more visible,
under the form of hot gas, they constitute ~ 30% of the total mass
(White et al. 1993).
Since clusters must be representative
of the baryonic fraction of the Universe, this implies that the total
mass cannot be larger than 3 times the baryons mass
(or m < 0.3).
4.2. The smallest fragments
The existence of a large number of gas clumpuscules (of ~ 10
AU in size) in the Galaxy has already been invoked to
explain the observed ESE (Extreme Scattering Events) in front of
quasars, by Fiedler et al.
(1987,
1994).
About 300 QSOs were observed during a few years, 150 over 12 yrs.
More than 10 ESE events were detected, due to diffraction or
refraction by a region of high electronic density (ne).
From the duration of the events, sizes of ~ 10 AU are derived,
and from their frequency, the number of clumpuscules in the Galaxy
must be about 103 the number of stars. The neutral density
of these objects is still a matter of debate. Their stability is
best explained in the hypothesis that they are self-gravitating.
The mass of one clumpuscule is then of the order of 10-3
M
Dixon et al. (1998)
from EGRET observations have recently detected an excess of diffuse
4.4. Various models of H2 as dark matter
The first model proposes to prolonge the visible gaseous disk
towards large radii, with thickening and flaring, following the HI disk.
The cold and dark H2 component is supported by rotation,
exists only outside the optical disk, where it is required by rotation curves
(Pfenniger et al 1994,
Pfenniger & Combes
1994).
The gas is stabilised
through a constantly evolving fractal structure, experiencing
Jeans instabilities at all scales, in thermal equilibrium with the
cosmic background radiation at T = 2.7 (1 + z) K.
Other models distribute the dark molecular gas in a spherical or
flattened halo, with no hole within the optical disk. The molecular
gas is not so cold, and is associated with clusters of brown dwarfs or MACHOS
(de Paolis et al. 1995,
Gerhard & Silk 1996,
Shchekinov, this conference).
In the clumpuscule model, the HI gas can be considered as a tracer,
the interface between the molecular clumps and the extra-galactic
radiation field.
Beyond the HI disk, there could be an ionization front, and the interface might
become ionized hydrogen. In this context, there should exist a distribution
correlation between the dark matter and the HI gas. This is indeed the
case, as already remarked by
Bosma (1981),
Broeils (1992) or
Freeman (1993):
there is a constant ratio between the surface density of dark matter, as
deduced from the rotation curves, and the HI surface density,
Figure 3. Left: HI rotation curve of
NGC 1560 (dots +error-bars), with the
rotation curve due to the HI itself (dotted line) and the stellar
component (dash).
The full line is the resulting expected rotation curve, when the HI
mass has been
multiplied by 6.2. Right: The ratio of surface densities of dark
matter to HI
required to explain the rotation curve of galaxies, as a function of
type. Data from 23 galaxies have been taken from
Broeils (1992)
and references therein.
If there exists a transition region where the cold H2 is
mixed in part with evolved gas with enough metallicity and
dust, it might be possible to detect cold dust emission.
Encouraging results have been found by the COBE satellite,
concluding to the existence of a cold (4-7K) component with
column densities 10 times that of the warm component
(at 18K), and more confined to the outer Galaxy
(Reach et al. 1995).
Another promising tool for detection is H2 absorption in
the UV electronic lines.
The main problem is the low expected surface filling factor
of the cold gas (~ 1%).
H2 has already been detected in front of QSO
through intervening galaxies
(Foltz et al. 1988,
Ge et al 1997);
but these observations suffer from severe confusion problem in the
Ly
.
Walker & Wardle (1998)
have recently built models of self-gravitating clouds,
with envelopes ionised by the interstellar radiation field: they
found for the electronic density the right order of magnitude
to account for the observed ESE. This hypothesis is supported by
direct observations through HI VLBI in absorption in front of
remote radio sources
(Diamond et al. 1989,
Davis et al. 1996,
Faison et al. 1998);
large column densities (~ 1021 cm-2) are observed with
sizes of ~ 10 AU, leading to HI densities of 106
cm-3 or more.
-ray emission
in the galactic halo. This could be interpreted in several ways:
either coming from un-resolved sources associated to the Galaxy;
or being due to high-latitude inverse compton emission; or finally to extra
molecular gas in the halo, through cosmic ray/nucleon reaction
giving 
0 then
-rays. This has been developped
by
de Paolis et
al. (1999) and
Kalberla et al. (1999),
see also Shchekinov (this conference). Cosmic rays are stopped by thick
clumpuscules, that have enough column density
to be opaque for both cosmic rays and gamma rays.
Sciama (1999)
proposes that cosmic rays are fragmented in clouds, heat the clouds, and
are responsible for the their FIR emission
(Sciama 1999).
However, the absorbed energy is non negligeable, and since clouds in
these halo models are assumed to move through the optical plane
(in their z-ocillations), sweep up high-metallicity gas, and
therefore contain CO molecules, they should be visible through CO emission.
DM /
HI = 7-10 (cf
figure 3, and a recent work by
Hoekstra et al. 1999).
This ratio is constant with radius in a given galaxy, and varies
slightly from galaxy to galaxy, being larger in early-types.
However, the dark matter does not dominate the mass in the latter, and
therefore the estimate of its contribution is more uncertain.
The correlation is the most striking in dwarf galaxies, which are
dominated by dark matter. The observed velocity curve is
almost exactly proportional to the velocity curve expected from the HI
component alone. Figure 3 shows the example of
NGC 1560,
from Broeils (1992).
Let us note that dwarf galaxies represent a hard test for all models of
dark matter, since the stellar component does not dominate the mass.
They rule out cold dark matter (CDM) profiles
(Burkert & Silk 1997),
and hot dark matter (HDM) models are also unable to concentrate as much as
is observed
(Lake 1989,
1990,
Moore 1996).
Baryonic dark matter is thus required.
forest. With the Hubble
Space telescope, it was not possible
to observe the fundamental lines at zero redshift, but it will be possible
with FUSE. Let us remark that heavy lines of sight will be impossible
to observe, due to obscuration of the background source (e.g.
Combes & Pfenniger
1997).
Finally, observations of
the lowest pure rotational lines of H2 have suggested some clues
for the existence of large quantities of H2 in galaxies
(Valentijn & van der
Werf 1999)
and should be pursued
in external galaxies, at much further radius than was possible with ISO.