Though the result obtained in the previous section is simple and in
good agreement with the experimental result, the mechanism which
determines the value of 
is not at all clear. In this section,
we try to determine 
with more approximate approaches and to
obtain qualitative understanding. First,
we obtain the growth rate and growth mode of a typical
planetesimal, using the collision probability obtained in section
2, and then derive the stationary distribution from
them.
The growth rate of a planetesimal is obtained by
integrating equation (5) over the range 
as
Safronov1969,BargePellat1991 e.g.
By substituting equations (5) through (8)
into equation (20) and neglecting the dependence on 
of 
and r and that on m of 
, we have
For the power-law mass distribution of equation (1), equation (21) is reduced to
where 
is the cut-off mass of the distribution function at the
low-mass end. These two cases correspond to two different modes of
accretion. If 
, the main mode of the growth is to eat
small planetesimals, since the growth rate is determined by the low
end of the integration over mass. If 
, on the other hand,
the growth is driven by the collision with particles with similar
masses. In other words, the growth is hierarchical.
Both the numerical results and the result obtained in the previous
section suggest that 
. So we consider the case
. The stationary state is realized if the mass flux
does not depend on either the time or the mass, which
is expressed as
Note that here the condition of the stationary distribution is the constant flux of the mass nm, and not the constant flux of the number of planetesimals n. This is because the dominant mode of the growth of the planetesimals is the hierarchical growth. The number of planetesimals would become smaller as they grow, since growth is driven by collisions of planetesimals of similar sizes.
If collisions are dominantly hierarchical, strictly speaking it does not make sense to use the growth equation expressed as the differential equation (20), since the change of the mass of planetesimals is not continuous. However, in this case we can use the mass doubling timescale
and
consider the discrete mass bins with the ranges 
,
, 
, ... 
,
... In this picture, collisions take place only between
particles in the same mass bins. The product of the collision
moves to the next mass bin.
In this case, the stationary state is realized when the incoming and outgoing flux for each mass bin are equal. For mass bin k, the total mass of planetesimals lost per unit time is
where the subscript k denotes the mass bin k and 
is the
total mass of planetesimals in the mass bin k. Note that 
, since the width of the mass bin is proportional to m itself.
The incoming mass to bin k is what is lost from bin k-1. Thus, the condition for the stationary state is that the outgoing mass is the same for all mass bins,
Equations (24) through (26) leads to equation (23) as the condition for the stationary state. Thus, whether we assume continuous growth of the mass or discontinuous jumps in the mass, the condition for the stationary state is the same.
From equation (1), (22), and (23), we obtain
and therefore
This value is the same as we obtained in section 2.
If we assume a constant flux of the number of
planetesimals, we obtain 
, which is significantly
smaller.
The reason why 
has a large negative value is now clear. As can
be seen in equation (22), the growth rate of planetesimals
has a strong positive dependence on their mass. This is, of course,
why the runaway growth occurs [].
This strong
dependence requires that the number of planetesimal goes down quickly
as its mass increases, if a stationary distribution is to be
realized. The strong dependence of the growth rate on the mass comes
from the size, scale height and the effect of gravitational focusing,
all of which enhance the growth rate for larger mass.
From equations (22) and (28), we obtain
which implies the mass of a planetesimal can reach infinity in a finite time. Of course, this cannot happen in reality, since the number of planetesimals with this infinite mass would go below unity. In addition, at a certain point the most massive particle grows so massive that its effect dominates the velocity dispersion of less massive particles, resulting in the slowing down of the growth []. This slowing down causes the massive planetesimals of similar mass to be formed in roughly equal orbital separation. Thus, these massive particles cannot collide with each other []. Nevertheless, equation (29) implies that the stationary distribution of the mass for a wide range of masses can be realized in a short time.
It should be noted that the growth rate of equation (29)
is larger than what has been used for the timescale estimate for the
runaway stage (see . e.g., Ida and Makino 1993). In previous
estimates, uniform background distribution of planetesimals was
assumed. Under this assumption, the power of m in equation
(29) becomes 
.
In the case of the two-dimensional space, equation (5) becomes
which gives 
. Therefore. from equation
(16) we obtain
However, with this formula, the total mass diverges if the high-mass end is taken infinite. Therefore, for the total mass to conserve, there must be a cutoff at the high-mass end, and no stationary solution is possible. In other words, the growth would be an orderly one.
The value of 
for two-dimensional case coincides with that for
collisional cascade []. This is simply because the
value of 
is the same.
Note that the existence of the stationary solution is directly related to the occurrence of the runaway growth. The runaway implies that the heavy particles can become heavier by themselves, without being affected by lighter particles. Thus, the power-law behavior is observed because there is no characteristic scale. In the case of the two-dimensional simulation, the evolution of the heaviest particle is slower than that of lighter particles, and therefore all particles tend to have a mass similar to that of the heaviest particle. The power-law behavior is not observed in this case.
