Vortex Generation in Protoplanetary Disks With An Embedded Giant Planet
Vortex Generation in Protoplanetary Disks With An Embedded Giant Planet
Vortex Generation in Protoplanetary Disks With An Embedded Giant Planet
&
Astrophysics
Context. Vortices in protoplanetary disks can capture solid particles and form planetary cores within shorter timescales than those
involved in the standard core-accretion model.
Aims. We investigate vortex generation in thin unmagnetized protoplanetary disks with an embedded giant planet with planet to star
mass ratio 104 and 103 .
Methods. Two-dimensional hydrodynamical simulations of a protoplanetary disk with a planet are performed using two different
numerical methods. The results of the non-linear simulations are compared with a time-resolved modal analysis of the azimuthally
averaged surface density profiles using linear perturbation theory.
Results. Finite-difference methods implemented in polar coordinates generate vortices moving along the gap created by Neptunemass to Jupiter-mass planets. The modal analysis shows that unstable modes are generated with growth rate of order 0.3 K for
azimuthal numbers m = 4, 5, 6, where K is the local Keplerian frequency. Shock-capturing Cartesian-grid codes do not generate very
much vorticity around a giant planet in a standard protoplanetary disk. Modal calculations confirm that the obtained radial profiles of
density are less susceptible to the growth of linear modes on timescales of several hundreds of orbital periods. Navier-Stokes viscosity
of the order = 105 (in units of a2 p ) is found to have a stabilizing effect and prevents the formation of vortices. This result holds
at high resolution runs and using different types of boundary conditions.
Conclusions. Giant protoplanets of Neptune-mass to Jupiter-mass can excite the Rossby wave instability and generate vortices in
thin disks. The presence of vortices in protoplanetary disks has implications for planet formation, orbital migration, and angular
momentum transport in disks.
Key words. planet and satellites: general accretion, accretion disks hydrodynamics instabilities methods: numerical
1. Introduction
The stability of differentially rotating disks has been considered analytically and numerically in the purely hydrodynamical case (Papaloizou & Pringle 1984, 1985; Goldreich et al.
1986;
Papaloizou & Pringle 1987) with applications to circumStability of rotationally supported gas disks is an area of active
stellar
disks and galactic disks. A rotating isentropic torus with
research, motivated among other reasons, by a need to undera
gradient
of specific angular momentum is found to be unstand the origin and stability of hydrodynamics turbulence underlying the so-called anomalous viscosity in accretion disks. stable to low-order non-axisymmetric perturbations due to the
The concept of -turbulence in accretion disks was introduced Papaloizou-Pringle instability. Several mechanisms have been
more than three decades ago by Shakura & Sunyaev (1973) to proposed that are able to sustain purely hydrodynamical turbuaccount for the angular momentum transfer and explain accre- lence and generate an anomalous -viscosity in accretion disks
tion onto the central object. The magnetorotational instability (Li et al. 2000; Klahr & Bodenheimer 2003; Mukhopadhyay
(MRI) has been proposed to explain the enhanced viscosity in et al. 2005). Dubrulle et al. (2005) studied non-axisymmetric inhot and sufficiently ionized accretion disks with a Keplerian an- stabilities in stratified Keplerian disks using numerical and analinear instability appears for Reynolds numgular velocity profile threaded by a weak magnetic field (Balbus lytical methods. A
3
& Hawley 1991; Balbus et al. 1996; Balbus & Hawley 1998). bers of order 10 and perturbations with characteristic scales
However, in the context of cold protoplanetary disks, the ioniza- smaller than the vertical scale of the disk, assuming the angular
tion by cosmic rays and stellar radiation is limited to the surface velocity decreases with radius. These results suggest that despite
layers of the disk while the so called dead zone in the vicinity the stabilizing effect of the Coriolis force, a Keplerian flow may
of the central plane is expected to have low ionization (Gammie undergo a transition to turbulence. Nevertheless, some of those
1996). In some astrophysical systems such as cataclysmic vari- mechanisms may depend on boundary or edge effects.
ables and outer regions of active galactic nuclei the coupling beRossby waves in thin Keplerian disks have been studied in
tween the magnetic field and the gas is also weak and MHD the linear approximation (Lovelace et al. 1999; Li et al. 2000)
and with fully non-linear numerical simulations (Tagger 2001).
effects may be negligible.
Article published by EDP Sciences and available at http://www.aanda.org or http://dx.doi.org/10.1051/0004-6361:20077169
1044
2. Modal analysis
We perform a modal analysis of analytical and numerically obtained density profiles, in order to see if there is agreement between the vortex generation in the simulations and growing unstable modes in the linear stability analysis. Linear perturbative
analysis provides a valuable tool to study the stability of disks.
The solution of the linearized Euler equations is treated along the
lines of the work of Lovelace et al. (1999) and Li et al. (2000),
where the stability can be evaluated solving a numerical eigenvalue problem for a given profile. Alternatively, the growth of the
initial perturbations can be determined by solving the equations
as an initial value problem.
We consider non-axisymmetric small perturbations sinusoidally varying in azimuth to the inviscid Euler equations
u) = 0
+ (
t
(1)
1
u
+ (u )u = P
t
(2)
(4)
RT
,
(5)
where R is the gas constant and is the mean atomic weight. The
initial unperturbed density is uniform in our simulations and the
temperature has a fixed profile r1 for the isothermal calculations, where r is the distance from the rotation axis. The sound
speed profile of the disk is that of a standard slightly flaring solar
nebula with constant aspect ratio
r
GM
(6)
cs = 0.05
r
where G is the gravitational constant and M the mass of the
central star.
The linearized equations can be reduced to a second order
differential equation for the enthalpy of the fluid, = P/,
in the general case when the pressure is a function of both
(7)
1 F 0 0
+
r
F
C(r) = k2
(8)
F0
2 2
2k
F
c2s
(9)
2 2
2
1 L0s B(r) + 4k / k cs / 1
Ls
Ls Lp
L2s
(10)
and
F (r) =
c2s /(Ls Lp )
(11)
k =
m
r
(12)
2 =
1 dl2
r3 dr
(13)
(r) = r + i m(r)
(14)
2
where is the epicyclic frequency, l = r is the angular momentum per unit mass, cs the adiabatic sound speed and Ls , Lp
are the equilibrium length scales of variation of entropy and
pressure changes, given by:
L1
s
1 dln(P/ )
=
dr
(15)
1 dln(P)
(16)
L1
p =
dr
Ls and Lp are calculated numerically from the averaged profiles
obtained in the simulations using Eq. (4).
The growth of the unstable modes can form vortices or
Rossby waves in the nonlinear regime (Li et al. 2001). The
Rayleigh criterion states that the disk will be stable to axisymmetric perturbations when the specific angular momentum increases with radial distance. For a Keplerian disk the epicyclic
frequency is always positive, 2 = 2K , where the Keplerian angular frequency is given by K = (GM /r3 )1/2 , and therefore
axisymmetric waves will be stable. However, when the pressure
effects are taken into account it is possible to have axisymmetric
instabilities for a sufficiently large pressure gradient according
to the Solberg-Hiland criterion,
2 (r) + Nr2 (r) 0
(17)
where
1 dP d P
ln
(18)
dr dr
1045
point, where continuity of the eigenfunction and its first derivative is required. Our implementation uses a leapfrog method to
integrate the equation from the boundaries to the fitting point.
The values of the entalphy and its derivative at the starting points
are specified based on several prescriptions using outgoing spiral
waves (Li et al. 2000) and vanishing eigenfuntion at the boundaries. We find that the obtained growth rates do not depend sensitively on the choice of boundary conditions. Several root finding
algorithms in the complex plane can be used to find the unstable
modes. The winding number theorem uses closed-path integrals
(e.g. Kargl & Marston 1989) to find the number of roots inside a
closed contour.
The winding number theorem states that for a complex analytic function f () defined inside a contour C
I 0
f ()
2i(N P) =
d
(19)
f ()
where N is the number of roots and P is the number of poles
inside the contour C, and the integral is defined in counterclockwise sense. We look for contours that comprise a single root and
are not located close to solutions or singularities on the real axis.
There are singularities in Eq. (7) at the corotation resonance
= 0 and when a modified form of the Lindblad resonance
for non-barotropic flow is satisfied, 2 2 c2s /(Ls Lp ) = 0.
It is important to avoid branch points close to the contours since
the winding number method would not give the right number of
roots. In the case of analytical density profiles given by hyperbolic trigonometric functions the branch points do not affect the
contours used to find solutions with positive growth rate.
The solutions are then calculated by
N
X
i=1
M
X
1
j =
i
2i
j=1
f 0 ()
d
f ()
(20)
where i are the zeros and j the poles of the complex function f (). The mode frequency can be determined in a contour
with a single root. A multidimensional Newton-Raphson method
(Press et al. 1992) is then employed to locate the roots with arbitrary accuracy.
Another approach to solve Eq. (7) involves discretizing the
equation on a finite grid and use appropriate boundary conditions to reduce the problem to finding numerically the roots of
the determinant of a complex tridiagonal matrix (e.g., Laughlin
et al. 1998; Li et al. 2000). We solved the determinant using the
previous root finding algorithm and obtain the radial profile for
the eigenfunction (r) and the perturbed variables.
We checked these two methods on the axisymmetric analytical step jump profiles in surface density studied by Li et al.
(2000). We considered azimuthal mode numbers from m = 1
to 10 and calculated the growth rates of the unstable modes and
the corresponding eigenfunctions. For analytical density profiles
with various shapes both our methods agree with the results of
Li et al. (2000). We tested the dependence of the solution on
the shape of the pressure profile and aspect ratio of the disk.
The eigenfunction for density profiles with a locally isothermal
equation of state is obtained by solving the discretized Eq. (7).
In Fig. 1, we show the real part of the mode frequency as
a function of the azimuthal number for the averaged density
profiles of a simulation (see Sect. 3) at nr n =
256 768 resolution after 10 orbits. Figure 2 shows the growth
rate as a function of the mode number for the same averaged
density profile, after 10 orbital periods when the threshold for
the excitation of the instability is reached. The density slope in
1046
Fig. 1. Real frequency of the most unstable modes for a gap opened by a
Jupiter-mass planet after 10 periods as a function of the azimuthal mode
number. The solid line shows the mode frequency at the outer edge of
the gap and the dashed line is the mode frequency at the inner edge
divided by mK (r0 ) where K (r0 ) is the Keplerian angular frequency at
the planet position.
Fig. 3. Radial profile of the eigenfunctions for the outer edge of the
gap at t = 10 periods and mode number m = 5. From top to bottom
the pressure perturbation and radial and azimuthal perturbed velocity
components are shown. The dotted and dashed lines are the real and
imaginary part of the eigenfunctions. The amplitude is shown by the
solid line which peaks at the position of the edge for the eigenfunction
of the perturbed pressure.
the modes correspond to a radial location just outside the corotating region where vortices are formed in the simulations after
about 10 orbits. The number of vortices in the numerical results
is consistent with the growth rates peaking at m = 46.
In Fig. 3 the real and imaginary parts of the radial eigenfunction of the perturbed variables at the outer edge of the gap
are plotted for azimuthal mode m = 5. The eigenfunction corresponds to a time when the gap becomes deep enough to generate
modes with positive growth rate. The middle and bottom panel
show the eigenfunctions of the perturbations of the velocity components. The radial eigenfunctions at the inner edge of the gap
after 10 orbits for azimuthal mode m = 5 are plotted in Fig. 4.
3. Numerical codes
We performed 2-dimensional hydrodynamical simulations using
two independent grid-based codes, implemented in cylindrical
and Cartesian coordinates. The simulations were run on a uniform grid for 100 orbital periods Different boundary conditions
were tested to avoid reflection of waves at the boundaries.
3.1. Initial setup
1047
(23)
Fig. 4. Radial profile of the eigenfunctions for the inner edge of the
gap at t = 10 periods and mode number m = 5. From top to bottom
the pressure perturbation and radial and azimuthal perturbed velocity
components are shown. The dotted and dashed lines are the real and
imaginary part of the eigenfunctions.
s2 + 2
(22)
we have used timescales between 510 orbital periods to introduce the gravity from the planet. Although the time when the
instability appears depend on how the gravity is started, there
is agreement between the modal analysis and numerical simulations for timescales of 5 and 10 orbits. The results presented in
Sect. 4 use a switch-on time of the gravity of 5 orbits.
In some of the calculations we use an initial gap profile derived under the WKB approximation (e.g., Lubow & DAngelo
2006)
!
f q2 a2 p a 3
,
(25)
0 (r) = exp
9 p
p
1048
3.2. NIRVANA
4. Results
Fig. 5. Real frequency and growth rates of the unstable modes with azimuthal number m = 5, as a function of time, for and polar
simulations with resolution nr n = 256 768. In the top
panel, the crosses represent the mode frequencies at the outer edge of
the gap and the dots are the mode frequencies at the inner edge for
. The circles are the mode frequencies at the outer edge of the
gap and the stars are the mode frequencies at the inner edge for polar . The frequencies are divided by mK (r0 ), where K is the
Keplerian angular frequency and r0 is the radius at the edge of the gap.
The crosses in the bottom panel are the growth rates at the outer gap
edge and the dots are the growth rates at the inner edge for a
calculation. The circles are the growth rates at the outer edge of the gap
and the stars are the growth rates at the inner edge for a polar
calculation. Growth rates are divided by the Keplerian frequency at the
edge of the gap.
1049
for the cylindrical schemes. The density peak in the inner disk is
about 50% of the peak at the outer gap edge. The overall jump
in density is larger at the inner edge despite the fact that the gap
is slightly deeper just outside the planets orbit. This produces a
larger growth rate inside the corotation region (see Fig. 2). In the
cylindrical simulation, the mass loss at the boundaries is
greater but this does not affect the results of the modal calculation. The Cartesian results have a deeper gap and smaller
density peaks at the edge of the gap. However, the growth rate
of their most unstable modes is about 50% compared with that
provided by the polar simulations.
The size of the peaks in the power spectrum of the torques
on a Jupiter-mass planet, from different regions in the disks, are
shown in Tables 1 and 2. The torques are calculated excluding
the material inside the Roche lobe of the planet, where the resolution may not be good enough to resolve the circumplanetary
disk. calculations have a larger peak, which correspond
to the amplitude of the vortices moving along the edge of the
gap. This is consistent with larger vortices being observed in the
density distributions of calculations in Fig. 6.
calculations have PDS peaks which are between one and two
orders of magnitude smaller. The frequencies in the corotating
frame are close to the local Keplerian angular frequency at the
edges of the gap. The difference in the peak amplitudes agrees
with the results from the upwind and Godunov schemes studied
by de Val-Borro et al. (2006). This correlation suggests that vortices can be formed by the non-linear evolution of Rossby waves
in protoplanetary disks.
The evolution of the vortensity in the simulation
with resolution nr n = 512 1536 is shown at several times
in Fig. 10. The vortensity and Bernoulli constant are conserved
for a barotropic inviscid fluid in the absence of discontinuities in
the flow. In our case, the vortensity is roughly conserved along
the streamlines in regions outside the Hill radius of the planet.
Shock dissipation close to the protoplanet can lead to vortensity
generation. As the gap is opened and strong trailing shocks are
formed, the vortensity grows at the edge of the gap and along the
spiral arms. The vortensity at 10 orbital periods shows small cavities outside the peaks at the edge of the gap. These depressions
break in 45 differentiated vortices when the growth rate of the
RWI becomes positive. The vortensity peak at the outer gap edge
and close to the outer spiral arm are corrugated while at the inner edge the vortensity is more stable. As explained before, this
is probably an artificial effect due to the inner edge of the gap
being closer to the inner boundary than the outer gap edge is
to the outer boundary. The minima of vortensity rotating along
the edge correspond to vortices observed in the density maps. In
the bottom right panel of Fig. 10 there is one single vortensity
depression which is associated with a vortex located at azimuth
after 100 orbits. The vortensity inside the corotating region
is considerably perturbed as the vortex moves along the gap.
In Figs. 11 and 12 the azimuthally averaged vortensity in
the inertial frame is shown. The initial vortensity profile for
a disk with uniform density, 0 r1.5 , has been subtracted.
The Cartesian code has a large vortensity excess in the
corotating region where the gap is more depleted than in the
cylindrical codes. The averaged vortensity in the outer disk is
also greater in our Cartesian model. calculation
shows vortensity peaks at the gap borders with a larger spike in
the inner disk.
The velocity fields plotted over the density contours in logarithmic scale at t = 100 orbits is shown in Fig. 13. The velocity
vectors are calculated in the corotating frame of the local maximum of pressure that coincides with the center of the vortex.
1050
The rotation in baroclinic sense is very clearly visible in the vortex close to the inner edge of the gap. The vortex in the outer
disk interacts with the spiral wake created by the planet and is
perturbed at this particular time. In Fig. 14, the streamlines are
shown in the frame corotating with the vortex core. They show
baroclinic rotation that is perturbed by interaction with the spiral
wakes created by the planet.
4.1. Dependence on physical viscosity
5. Discussion
We have studied vortex formation in protoplanetary disks with
an embedded giant planet, with mass ratios 104 and 103 , using numerical simulations and linear perturbation analysis. The
modal calculation is done following the strategy of Lovelace
et al. (1999) for a locally isothermal equation of state in a
vertically-averaged disk. Vortices are formed in the cylindrical
and 2-dimensional simulations in agreement
with the linear analysis of non-axisymmetric perturbations. The
growth rates calculated for and polar as a function of time agree within about 10%. The results of the linear
analysis are consistent with the absence of rapidly growing vortices near the edge of the gap in our Cartesian-grid PPM simulations, which is thus not due to artificial numerical damping of
1051
Frequency
0.568
0.531
1.611
Amplitude
2.81 105
8.96 108
2.19 107
Frequency
0.389
0.353
0.997
Amplitude
1.11 103
4.82 106
3.65 106
Table 3. The frequency and magnitude of the maxima of the PDS of the
torques from the inner disk for the Neptune case are shown sorted by
the magnitude of the PDS at the maximum. calculations have
larger maxima with frequencies close to the Keplerian frequencies at
the outer gap edge.
Code
NIRVANA
Flash polar
Flash Cart.
unstable modes. This type of code does not produce the necessary steepness of the surface density profile and does not support
growing non-axisymmetric perturbations.
Both our numerical schemes have very low numerical diffusion, which is estimated in Appendix A. Runs with an explicit Navier-Stokes viscosity were also performed. The unstable modes in the outer disk calculated for = 106 have growth
rates 0.25K at 100 orbits, which are 20% smaller than in the
inviscid calculations.
We speculate that the Cartesian-grid implementation may be
more diffusive for this problem than polar geometry (in which
the unperturbed Keplerian disk flows along the mesh structure)
hence damping the growth of Rossby waves. The linear theory
predicts that unstable modes will be present in the Cartesian
simulations after the gap is sufficiently deep but these modes
have smaller growth rate than those obtained from and
polar simulations. The growth rate for the inner edge of
the gap is larger than it is at the outer gap edge. This may be artificial because the inner boundary is closer to the planet position
that the outer boundary is to the outer gap edge.
Frequency
0.204
0.221
0.345
Amplitude
4.67 109
3.57 108
3.99 108
Table 4. The frequency and magnitude of the maxima of the PDS of the
torques from the outer disk for the Neptune case are shown sorted by
the magnitude of the PDS at the maximum.
Code
NIRVANA
Flash polar
Flash Cart
Frequency
0.389
0.204
3.997
Amplitude
1.11 107
5.27 108
5.411 107
We observe a correlation between the growth rate of the unstable modes in the linear analysis and the size of the peaks in the
power spectrum of the gravitational torque on the planet exerted
by the disk. This correlation suggests that vortices in protoplanetary disks can form close to the gap, produced by an embedded giant planet, from the collapse of Rossby waves. Vortices
may grow and be sustained for long timescales by interaction
with the planetary wake (Koller et al. 2003; Li et al. 2005). The
two-dimensional approximation for the disk flow is anticipated
to give qualitatively correct results although a three-dimensional
analysis is needed to understand heat dissipation in the vertical direction and refraction effects in radially propagating waves
(Lin et al. 1990). An important restriction of our simulations is
that the planet is kept on a fixed circular orbit. It would be of
interest to study how the vortices rotating along the edge of the
gap affect the migration rate of a freely moving protoplanet embedded in a 3-dimensional disk.
In summary, the linear analysis confirms that Rossby waves
are formed in a thin protoplanetary disk with a giant planet
with mass ratio between = 104 103 , within tens of orbital
1052
Fig. 9. Surface density distribution after 100 orbital periods for on the left hand side
and on the right hand side using the
same logarithmic color scale. Both models use
the same wave damping condition in the outer
disk between 2.12.5 a. while does not
have a damping condition in the inner boundary. has density enhancements close
to the gap opened by the protoplanet.
has a smooth density distribution and a larger
density peak at the planet position which is saturated in the image.
sufficiently ionized to sustain turbulence via the MRI instability. We conclude that vorticity generation in protoplanetary
disks with an embedded giant planet is a robust mechanism
that can lead to planet formation and radial transfer of angular
momentum.
Acknowledgements. M.d.V.B. was supported by a SAO predoctoral fellowship and a NOT/IAC scholarship. G.D. acknowledges support from the NASA
Postdoctoral Program. We thank Artur Gawryszczak for providing his numerical
code and enlightening discussions. The support of the RTN Planets funded
by the European Commission under agreement No. HPRN-CT-2002-0308 is
acknowledged during the course of this project. The code used in
this work is developed in part by the US Department of Energy under Grant
1053
500 tracer (massless) particles released in the disk. The equation of motion of each particle is integrated every hydrodynamics timestep by interpolating the velocity field at the particles
location and advancing in time its position by means of a secondorder Runge-Kutta method. The spatial interpolation of the velocity field is also second-order accurate. Hence, trajectories are
formally second-order accurate in both space and time.
We assume that the viscosity is constant in a radial interval
r = 0.1, which contains about 25 equally-spaced tracer particles and is orders of magnitude larger than their diffusion length
scale. We measure the averaged length travelled by the particles
in each radial interval, over about 50 orbits (at r = 1), and estimate the amount of numerical viscosity under the assumption
that the particle drift velocity is
dr = 3n
(A.1)
dt
2r
Experiments executed at resolutions nr n = 256 768 and
512 1536 provide very similar results.
The largest numerical diffusion is observed close to the inner
grid boundary, where n 107 (in code units, see Sect. 3.1) at
r 0.55 and 108 at r 0.65. In the radial domain between
r 0.75 and r 1.75, n lies between 1010 and 109 . In
the outer part of the simulated disk, the numerical viscosity is
comprised between 109 and 108 .
The numerical viscosity in is calibrated using a
2-dimensional local patch of a Keplerian disk with a massless
sink hole in the center of the domain The sink hole has radius
equal to the Roche radius
q 1/3
RR = a
(A.2)
3
for a protoplanet with mass ratio q = 103 . The dimensions of
the shearing box are 20RR 20RR . We use local Cartesian coordinates in the Hill approximation corotating with the sink hole.
The horizontal axis corresponds to the radial direction and the
vertical axis to the direction of motion of the flow. Initially, the
unperturbed surface density is uniform and the velocity of the
matter flowing into the computational domain has only a vertical component given by the linearized Keplerian velocity
3
vy = K x.
(A.3)
2
We implement periodic boundaries in the vertical direction. Our
units in the simulation are the Roche radius, RR , the initial surface density, 0 , and the Keplerian angular frequency, K , at the
center of the shearing box. A cavity is produced at the orbital
Appendix A: Calibration of numerical viscosity
radius of the sink hole while the gas in the vicinity of the hole is
The calculations presented in this paper are based on inviscid pulled into it on the viscous timescale.
Euler equations. However, even though neither physical nor arThe equation of viscous diffusion for a thin accretion disk
tificial viscosity terms enter these equations, each numerical can be obtained in the asymptotic limit assuming constant nuscheme has some intrinsic diffusivity that can be interpreted as a merical diffusivity (Pringle 1981; Bryden et al. 1999). Using the
numerical viscosity, n .
boundary conditions = 0 at x = 0, and /x = 0 at x = xout ,
To calibrate the numerical viscosity in each of the hydrody- where xout = 10RR is the distance from the center of the shearing
namics codes, we use the numerical setup described in Sect. 3.1 box to the outer radial boundary, the surface density distribution
with mass ratio q = 0. Tests are performed at two different res- for positive x is given by
olutions to check for consistency of the results. Numerical vis!
!
2
cosity is bound to depend on the flow properties. The values re- (x, t) = 0 exp 3 n t sin x
(A.4)
2xout
4x2out
ported in this Appendix apply to disks in quiescent conditions
and may therefore represent lower limits for numerical diffusion
The kinematic viscosity can be evaluated from the ratio of surin models with uniform initial density and fast gap formation.
face densities at different times
In the case of , a time-averaged measure of the
#
"
4x2out
(t1 )
numerical diffusivity is obtained over a 50 orbit period, as a
(A.5)
function of the radial position, by analysing the trajectories of n = 32 (t2 t1 ) ln (t2 )
No. B523820 to the Center for Astrophysical Thermonuclear Flashes at the
University of Chicago. Some of the calculations reported here were performed
on Columbia, operated by NASA Advanced Supercomputing Division, at NASA
Ames Research Center. We thank the anonymous referee for comments that improved the manuscript.
1054
Fig. 13. Velocity vectors in the corotating frame after 100 periods for the simulation at resolution n r n = 512 1536. The left panel
shows the vortex at the inner gap edge and the right panel shows the vortex at the outer gap edge.
References
viscosity is = 106 .
We estimate values of n between 108 and 107 using the surface density profiles averaged in the vertical direction at times
close to 50 orbits and grid resolutions of n x ny = 320 320 and
640 640.
1055