
2
they prohibitively require starting from already large par-
ticle sizes [12].
Besides planet formation, the origin and nature of tur-
bulence in protoplanetary disks, and how it couples to
particle dynamics, is key to several other astrophysical
questions. For instance, disks are known to have a rather
short lifetime (of the order of a few millions years). The
dissipation of angular momentum by the gas molecular
viscosity is orders of magnitude too low to explain this
fast accretion onto the central star. Yet turbulence is
an excellent candidate to transport matter on shorter
timescales, hence providing estimates on the expected
turbulent intensity of disks. Transport mechanisms in
protoplanetary disks, possible instabilities and the de-
velopment of a turbulent state are still the topic of in-
tense work (see, e.g., [13, 14] for a review). Whether it
originates from magneto-rotational effects or purely com-
pressible or incompressible hydrodynamical instabilities,
disk turbulence is confined, in rotation, stratified, and
thus generally displays fairly strong two-dimensional fea-
tures, as for instance long-lived vortical structures [15].
The presence of such vortices might clearly impact dust
accretion and consequently the formation of planetesi-
mals. While heavy inertial particles are usually expelled
from rotating regions by Maxey’s centrifugal effect [16],
the presence of shear and rotation in protoplanetary disks
actually has an opposite influence. Numerical [17, 18],
analytical [19, 20] and experimental [21] studies indeed
give evidence that dust particles cluster in some of the
structures of the flow. More specifically, solid particles
experience a Coriolis force that can either accelerate their
ejection from cyclones (vortices with the same sign as
global rotation) or, when strong enough, overcome cen-
trifugal forces in anticyclones. This pushes the particles
toward the core of anticyclonic vortices, possibly creating
extremely dense point clusters that are excellent candi-
dates for gravitational instabilities. These mechanisms
have been known for years, but their study remained es-
sentially qualitative or limited to model flows. Under-
standing further what is the impact of vortex clustering
on planetesimal formation requires more quantitative in-
sights, such as estimates on involved timescales in the
presence of turbulent fluctuations, dependence on phys-
ical parameters, nature of the associated mass distribu-
tion, etc. The present work aims providing insights on
such questions.
We present in this paper 2D direct numerical simula-
tions with the idea of shedding some lights on the com-
plex problem of dust dynamics in turbulent protoplan-
etary disks. We investigate the clustering properties of
solid particles in a forced, developed turbulent incom-
pressible flow subject to Keplerian rotation and shear.
We follow the “shearing box” approach, which consists
in solving locally the Navier–Stokes equation with a mean
constant shear and with periodic boundary conditions on
a domain that is distorted in the direction of the mean
velocity. A number of simulations are performed vary-
ing systematically the two main physical parameters of
the problem: the particle response time τp, which mea-
sures their inertia and their lag on the gas flow, and the
rotation rate Ω, which prescribes the mean shear. We
then analyse quantitatively various dynamical and sta-
tistical features of solid particles, and in particular their
clustering properties when flow structures, typical of a
turbulent flow in rotation, are present. Tools borrowed
from the study of dynamical systems, such as Lyapunov
exponents and fractal dimensions, are used to character-
ize and quantify particle clusters.
Before focusing on particles, we find that the mean
shear has noticeable impacts on the properties of the tur-
bulent flow. Both energy and enstrophy budgets, as well
as energy and enstrophy spectra, change substantially
with increasing rotation. Strong rotation also leads to
pronounced anisotropies in the flow and thus to increased
skewness and kurtosis of the vorticity distribution. Shear
is found to cause the preferential formation and survival
of anticyclonic vortices. Concerning particles, we find
that at specific values of the physical parameters, they
can form point clusters located inside anticyclones. This
strong clustering requires large-enough values of the ro-
tation rate Ω and intermediate values of the particle re-
sponse time τp. These clusters are found to form in a hi-
erarchical manner and to merge one with the other when
time increases. We propose a simple kinetic model that
catches most aspects of their evolution and distribution.
Outside this extreme regime, solid particles can never-
theless concentrate on dynamically evolving fractal sets
whose dimensions non-trivially depend on the rotation
rate. Evidence is obtained that the mass distribution of
particles is then multifractal, with a dimension that de-
creases and saturates at large orders. Such a behavior is
key to quantify the probability to get a large local den-
sity of solids and to trigger self-gravitating instabilities.
In the strong clustering regime, such instabilities would
occur during transients.
This paper is organised as follows. We first introduce
and characterise in section II the flow used for the nu-
merical simulations and we explain in section III the con-
sidered dynamics of particles. In Section IV we show
the results of the simulations, highlighting the different
behaviors observed for the distribution of inertial parti-
cles. In section V and section VI the mass distribution of
particles is inspected in the regime of fractal and strong
clustering, respectively. Finally, in section VII we con-
clude the paper summarizing the results of this study and
connecting it back to the astrophysical context.
II. THE TWO-DIMENSIONAL SHEARING BOX
Dust and gas dynamics in protoplanetary disks, as in
most astrophysical situations, involve gravity and rota-
tion. The balance between gravity and centrifugal forces
defines the Keplerian angular velocity, which is given by
Ω(r) = pGM?/r3(1)