TWO-BODY PROBLEMS WITH DRAG OR THRUST:
QUALITATIVE RESULTS
FLORIN DIACU
Department of Mathematics and Statistics, University of Victoria, PO Box 3045, Victoria, B.C.,
Canada, V8W 3P4, e-mail: diacu@[Link]
(Received: 26 October 1998; accepted: 22 April 1999)
Abstract. We consider two-body problems in which the drag is proportional to the velocity divided
by the square of the distance and whose radial and tangential components have distinct coeffi-
cients. For all parameters, we study the flow of the system obtained by suitable coordinate and time
transformations and draw conclusions about the qualitative behavior of solutions. In each case, we
examine the existence of collisionejection, collisionescape, capturecollision, captureescape, and
oscillatory rectilinear orbits, study the motion near collision, and show that if periodic orbits exist
they must be limit cycles.
Key words: radiation pressure, generalized Stokess drag, qualitative dynamics
1. Introduction
Two-body problems with drag appeared in the literature at the beginning of the
20th century. They model the motion of two point masses subject to Newtonian
gravitation and a drag force that involves the velocity (e.g. some radiation or atmo-
spheric pressure). Such problems have been tackled by Poynting (1903), Plummer
(1905, 1918), Robertson (1937), Whyat and Whipple (1950), Brower and Hori
(1961), Mittleman and Jezewski (1982), Leach (1987), Mavraganis (1991), Mav-
raganis and Michalakis (1994), and several others. More details on the history of
the problem can be found in Burns et al. (1979) and Mignard (1992). Alternatively,
thrust forces can appear when studying the motion of microcosmic particles in the
neighborhood of large bodies.
As in the recent paper by Breiter and Jackson (1998), we consider a generalized
drag or thrust force, which is proportional to the velocity vector divided by the
square of the distance and such that different proportionality constants, 6 = 0
and 6 = 0, are assigned to the radial and tangential components of the velo-
city, respectively. The Stokess drag, for example, is encountered in cosmogonical
and astrophysical problems (see Ferraz-Mello, 1992). The generalized Stokess
drag covers several drag forces studied in the literature: the PoyntingPlummer
drag ( = 4, see Plummer, 1906), the PoyntingRobertson drag ( = 2, see
Robertson, 1937), and the PoyntingPlummerDanby drag ( = , see Poynting,
1903; Plummer, 1905; Danby, (1962).
Celestial Mechanics and Dynamical Astronomy 75: 115, 1999.
c 1999 Kluwer Academic Publishers. Printed in the Netherlands.
2 FLORIN DIACU
In this note we perform a qualitative study of the equations of motion that de-
scribe the two-body problem with a generalized drag or thrust force. Using suitable
space and time-rescaling transformations, we blow-up the collision singularity into
a plane (which we will call collision plane) and paste this plane to the phase space.
We first study the invariant set of rectilinear orbits and see that four cases arise,
depending on the signs of the parameters involved. These forces are: attractive
with radial drag (, > 0), attractive with radial thrust ( > 0, < 0), repelling
with radial drag ( < 0, > 0), and repelling with radial thrust (, < 0). In
each case, we completely describe the invariant set of rectilinear solutions. We then
return to the collision plane and show that the dynamics on it is given by a simple
linear system with constant coefficients. Though the orbits in this plane have no
physical correspondent, the continuity of the solutions with respect to initial data
provides information on the behavior of physical orbits near collision. The study
of the flow in the invariant collision plane allows us to draw conclusions about
the qualitative behavior of collision and near-collision orbits in each of the above-
mentioned cases. The transformed equations of motion also allow us to see that for
+ > 0 periodic orbits (if they exist) must be limit cycles.
2. Equations of Motion
Assume that the motion of two point masses (a particle and a primary) is subject
to a perturbation force of the type F = r 2 (vr + vt ), where r is the distance
between bodies, vr and vt are the radial and the tangential velocity, respectively,
and , are real, nonzero constants. With respect to a frame whose origin is in the
primary, the equations of motion in polar coordinates (r, ) are given by the system
d 2
r r 2 = r 2 rr 2 , (r ) = , (1)
dt
where combines several physical constants (mass, gravitational constant, charge,
etc.). Unlike in the purely Newtonian case, the total energy and the angular mo-
mentum are not conserved quantities.
All possible choices of signs for , , and present practical interest, from
the one modeling the motion of small satellites near the earth to the one involving
microcosmic particles under radiation pressure. Therefore, we will further discuss
the qualitative aspects of each case. Notice that > 0 corresponds to an attractive
force, < 0 represents a repelling one, while > 0 implies the presence of a
radial drag and < 0 that of a radial thrust.
Using the substitutions r = u and = , we transform Equations (1) into the
first-order system
r = u, = , u = r 2 r 2 ur 2 ,
= r 2 2ur 1 .
TWO-BODY PROBLEMS WITH DRAG OR THRUST: QUALITATIVE RESULTS 3
Since the vector field of this system is independent of , we can drop the second
equation and thus obtain the simpler form
r = u, u = r 2 r 2 ur 2 , = r 2 2ur 1 ,
which under the time-rescaling analytic diffeomorphism d = r 2 dt is converted
into the system
r 0 = r 2 u, u0 = u + r 3 2 , 0 = 2ur. (2)
Equations (2) are defined for (r, u, ) [0, ) R2 and 0 represents the derivative
with respect to the new fictitious time-variable . Notice that though Equations (1)
were undefined at r = 0, system (2) extends to this set, that is, the plane r = 0 is
pasted to the three-dimensional phase space of the variables (r, u, ). The collision
plane r = 0 is thus an invariant manifold for the Equations (2). (Recall that an
invariant manifold is a union of orbits, that is, initial data in the set constrain the
whole orbit to the set.)
Since system (2) has only one equilibrium solution at (0, /, 0), we will
shift the origin of the frame into it with the help of the transformation v = u+/.
Thus, system (2) becomes
r0 = r2 v , v 0 = v + r 3 2 ,
0 = 2 v r, (3)
and will be the main object of our further qualitative investigations. System (3) has
the unique equilibrium solution (r, v, ) = (0, 0, 0). Due to the imbedding of the
collision manifold into the phase space, system (3) is well suited for understanding
the behavior of orbits near collision.
We will also study an alternative system, suitable for understanding the behavior
of orbits at infinity. To obtain it, we transform system (3) via the change of variable
x = 1/r, which is an analytic diffeomorphism that brings infinity into the phase
space. The new equations are
x0 = v , v 0 = v + x 3 2 ,
1
0 = 2 v x . (4)
The time-rescaling transformation
ds = x 3 d,
which is an analytic diffeomorphism, changes system (4) into the regularized
system
x , = x 3 v , v , = x 3 x 3 v + 2 ,
2
, = x 3 2 v x , (5)
4 FLORIN DIACU
where , denotes differentiation with respect to the new fictitious time variable s.
The invariant plane x = 0, called the infinity manifold, represents the physical
points at infinity. In other words, a solution tending to this plane reaches infinity in
physical space. On the other hand, a solution of system (5) that tends to infinity in
phase space reaches a collision in physical space. Equations (5) form the alternative
system we will further study.
3. Rectilinear Orbits
From the last equation of (3) or (5), we can conclude that the plane = 0 is an
invariant manifold for each of these systems. Indeed, if (0) = 0 then 0 (0) = 0,
and since the last equation of (3) or (5) is linear, it follows that = 0 wherever it
is defined. This plane represents all rectilinear solutions, that is, those moving on
a straight line, which correspond to some constant angle . Restricting (3) to this
invariant plane we obtain a simple integrable system
r0 = r2 v , v 0 = v, (6)
whose solution is
c 1
1
r( ) = e + + c2 , v( ) = c1 e ,
where c1 and c2 are constants. However, the flow in the plane = 0 is easier to
draw if we formally divide the first equation of (6) by the second and then solve r
for v. This leads to the separable equation
dr 1
=r 2
,
dv 2v
whose solution is
1
1
r(v) = v 2 ln |v| + c . (7)
As expected, a rectilinear solution depends only on the coefficients and and is
independent of , that is, it matters whether the force is attractive or repulsive, with
radial drag or with radial thrust, but the motion is not influenced by the tangential
velocity.
Alternatively we restrict Equations (5) to the invariant plane = 0 and obtain
the system
x 0 = x 3 v , v 0 = x 3 v. (8)
Formally dividing the first equation of (8) by the second, we obtain the equation
dx 1
= 2 ,
dv v
TWO-BODY PROBLEMS WITH DRAG OR THRUST: QUALITATIVE RESULTS 5
whose solution is
1
x(v) = v 2 ln |v| + c, (9)
where c is the same constant as in (7). The family of curves defined by (9) will help
us understand some aspects of the flow of system (5).
Before separately studying each of the four cases described above, let us cla-
rify the terminology. Although we will describe the notions with respect to sys-
tem (3), it is easy to translate the definitions below in terms of the variables defining
Equations (5).
DEFINITIONS. A solution of system (3) is called a capture orbit if r tends to
infinity in the past, an escape orbit if r tends to infinity in the future, an ejection
orbit if r tends to 0 in the past, and a collision orbit if r tends to 0 in the future.
An oscillatory orbit is one for which r becomes unbounded but does not tend to
infinity, in the past or in the future (for example, if lim sup r( ) = + and
lim inf r( ) < +).
4. Attractive Force with Drag
In this section we will study the case , > 0, which corresponds to an attractive
force with radial drag. The flow in the plane = 0, given by (7), depends on the
solutions v of the equation
1
v + c = 2 ln |v|,
which correspond to curves with vertical asymptotes in the (v, r)-plane. This equa-
tion can have one, two, or three solutions, depending on the value of the constant
c. However, for certain values of and , some solutions correspond to negative
values of r, which have no physical meaning. The flow given by (7) is depicted in
Figure 1(b). To understand this, let us first have a look at Figure 1(a), which shows
the half-plane r > 0 divided into four regions, A, B, C, and D, by v = 0 (i.e. the
r-axis) and by the two curves graphing the solution of (7) for c = c, , where
c, = 2 ln 1 .
These two curves have the line v = / as vertical asymptote when r .
For each c > c, the graph of the corresponding function in (7) is formed by two
curves, one in A and one in C. Each curve in A has a vertical asymptote when
r . For each c < c, the graph of the corresponding function in (7) is also
formed by two curves, disconnected this time, one in B and one in D. Each of
these curves has its own vertical asymptote as r . (The curves in between
6 FLORIN DIACU
Figure 1. (a) The regions A, B, C, and D in which the flow of system (6) is divided. (b) The flow of
system (6) for > 0, > 0.
the asymptotes appear in the half-plane r < 0, so we did not include them in the
picture.)
From the physical point of view the curves in regions A and B correspond to
capturecollision orbits, the ones in C are ejectioncollision orbits, and the ones in
D are ejectionescape orbits. Notice that the family of curves in C is unbounded,
that is the closer a curve to the one corresponding to c = c, , the larger the
corresponding r. This means that independently of and , rectilinear orbits with
initial conditions in C can depart as far as possible from the primary and still return
to end in a collision.
Alternatively we can visualize all these orbits in terms of system (8) and its
corresponding solution (9). Observe that every point of the x-axis is an equilibrium
solution of system (8), which represents a point at infinity in physical space (see
Figure 2).
The typical picture obtained by drawing the family the curves in (8) is as shown
in Figure 2, i.e., concave up. Indeed, since dx/dv = (v )/ 2 v, it follows
that dx/dv < 0 for v (0, /) and dx/dv > 0 for v (, 0) (/, ).
Also, d2 x/dv 2 = / 2 v 2 > 0. We are, of course, interested only in solutions for
which x > 0, so the flow is as shown in Figure 2. Solutions that come from infinity
and reach equilibria on the v-axis represent ejectionescape orbits, solutions that
leave equilibria and tend to infinity represent capturecollision orbits, and solutions
that come from infinity and tend back to infinity without reaching an equilibrium
represent ejectioncollision orbits. We can summarize the above conclusions in the
following statements.
THEOREM 1. In the two-body problem given by an attractive force with drag,
defined by system (3) with given constants > 0, > 0, and , the set of
rectilinear solutions is formed by ejectioncollision, ejectionescape, and capture
collision orbits. Moreover, there is no upper bound for the distance between bodies
in the set of rectilinear ejectioncollision orbits.
TWO-BODY PROBLEMS WITH DRAG OR THRUST: QUALITATIVE RESULTS 7
Figure 2. The flow of system (8) for > 0, > 0.
COROLLARY 2. In the two-body problem given by an attractive force with drag,
every rectilinear solution encounters an ejection or a collision, so there are no
captureescape rectilinear orbits.
5. Repelling Force with Drag
In this section, we will study the case < 0, > 0 of a repelling force with
radial drag. The flow in the plane = 0, given by Equation (7), is depicted in
Figure 3. For v < 0, the concave up curves, in which all have v = 0 as vertical
asymptote to the right and each has its own vertical asymptote to the left, represent
captureescape orbits that never encounter collisions. However, there is no lower
bound for the distance between bodies in the set containing these orbits. The curves
with v > 0 represent ejectionescape orbits. Each such curve has its own vertical
asymptote to the left.
Alternatively, solution (9) of system (8) is represented in Figure 4. The concave
down curves with v < 0 connect distinct equilibria and represent captureescape
orbits, whereas the ones with v > 0 that tend to the equilibria represent ejection
escape trajectories. The above conclusions can be summarized in the following
statements.
THEOREM 3. In the two-body problem given by a repelling force with drag,
defined by system (3) with given constants < 0, > 0, and , the set of rectilin-
ear solutions is formed by captureescape and ejectionescape orbits. Moreover,
8 FLORIN DIACU
Figure 3. The flow of system (6) for < 0, > 0.
Figure 4. The flow of system (8) for < 0, > 0.
there is no lower bound for the distance between bodies in the set of rectilinear
captureescape orbits.
COROLLARY 4. In the two-body problem given by a repelling force with drag,
rectilinear orbits do not encounter collisions.
TWO-BODY PROBLEMS WITH DRAG OR THRUST: QUALITATIVE RESULTS 9
6. Attractive Force with Thrust
In this section, we will study the case > 0, < 0 of an attractive force with radial
thrust. The flow in the plane = 0, given by Equation (7), is depicted in Figure 5,
and we can see that it is quite similar to the one in Figure 1(b). Again, the curves
corresponding to c = c, divide the half-plane in four distinct regions. After an
analysis similar to that of Section 4 we find capturecollision, ejectioncollision,
and ejectionescape orbits.
Alternatively, the solution (9) of system (8) is represented in Figure 6.
The capturecollision, ejectioncollision, and some of the ejectionescape
orbits appear for v < 0, whereas the remaining ejectionescape orbits appear
for v > 0. The above conclusions can be summarized in the following
statements.
THEOREM 5. In the two-body problem given by an attractive force with thrust,
defined by system (3) with given constants > 0, < 0, and , the set of rec-
tilinear solutions is formed by capturecollision, ejectioncollision, and ejection
escape orbits. Moreover, there is no upper bound for the distance between bodies
in the set of rectilinear ejectioncollision orbits.
COROLLARY 6. In the two-body problem given by an attractive force with thrust,
every rectilinear solution encounters an ejection or a collision, so there are no
rectilinear captureescape orbits.
Figure 5. The flow of system (6) for > 0, < 0.
10 FLORIN DIACU
Figure 6. The flow of system (8) for > 0, < 0.
7. Repelling Force with Thrust
In this section, we will study the case , < 0 of a repelling force with radial
thrust. The flow in the plane = 0, given by Equation (7), is depicted in Figure 7.
Indeed, for c 6 c, , the graph of each of the corresponding functions in (7) with
r > 0 is formed by one curve, increasing with respect to v and concave up, such
that r(v) 0 when v ; also, for each c 6 c, , r along a vertical
asymptote. For c > c, , the graph of each corresponding function with r > 0 is
formed by two disconnected curves one that behaves as in the previous case and
the other one concave up decreasing with respect to v between 0 and / and
increasing for v > / (see Figure 7). For each curve, r when v approaches
0 with positive values and also along a vertical asymptote for v > /, specific
for each curve. None of the curves ever reaches the v-axis. The signs of r 0 and v 0
in system (6) show that the direction of the flow is as shown in Figure 7.
Alternatively we recover the same types of orbits if analysing system (8), whose
flow for x > 0 is depicted in Figure 8. For v < 0, all orbits are concave up and leave
the equilibria of the v axis, tending to infinity when the fictitious time variable s
tends to infinity. For v > 0, the orbits are represented by concave down curves that
eject from the equilibria of the v-axis and then tend to other equilibria as shown in
Figure 8. From all these we can now draw the following conclusions.
THEOREM 7. In the two-body problem given by a repelling force with thrust,
defined by system (3) with given constants < 0, < 0, and , the set of
rectilinear solutions is formed by capturecollision and captureescape orbits.
TWO-BODY PROBLEMS WITH DRAG OR THRUST: QUALITATIVE RESULTS 11
Figure 7. The flow of system (6) for < 0, < 0.
Figure 8. The flow of system (8) for < 0, < 0.
Moreover, there is no lower bound for the distance between bodies in the set of
captureescape orbits.
All the above cases lead us to the following conclusion regarding rectilinear
oscillatory orbits.
12 FLORIN DIACU
COROLLARY 8. Independently on the signs of , , and , two-body problems
with drag or thrust have no rectilinear oscillatory orbits.
8. Motion Near Collision
In this section, we will further analyse system (3) and present some results con-
cerning the motion near collision, that is near the plane r = 0 in general and
near the equilibrium (0, 0, 0) in particular. As we mentioned in Section 3, from the
first equation of system (3) we see that the plane r = 0 is an invariant manifold.
Although the flow on this manifold has no correspondent into the physical reality,
the continuity of the solutions with respect to initial data provides information on
the physical motion near collision. System (3) restricted to the invariant plane r = 0
takes the form
v 0 = v, 0 = , (10)
a decoupled linear homogeneous system having the solution
v( ) = c1 e , ( ) = c2 e ,
where c1 and c2 are constants. The equilibrium (0, 0) is a sink if , > 0, a source
if , < 0, and a saddle if > 0, < 0 or < 0, > 0 (see Figures 9(ad)).
If we linearize system (3) around (0, 0, 0), we obtain the eigenvalues r =
0, v = , and = , which show that the origin is a non-hyperbolic
equilibrium, so we cannot draw any direct conclusion about the stability of this
equilibrium only from linearization alone. However, since we already know how
the flow acts in the invariant planes = 0 and r = 0, we can describe its beha-
vior near (0, 0, 0) for all choices of parameters , , and . We thus obtain the
following results.
THEOREM 9. In the two-body problem with drag defined by system (3) with
given constants > 0, > 0, and > 0, there is a set of initial conditions
of positive measure that leads to collisions. Except for the rectilinear ones, all
Figure 9. The flow of system (10) for (a) , > 0, (b) , < 0, (c) > 0, < 0, (d) < 0, > 0.
TWO-BODY PROBLEMS WITH DRAG OR THRUST: QUALITATIVE RESULTS 13
orbits with such initial conditions spiral towards collision but are asymptotically
rectilinear.
The first part of the theorem follows from analyzing the flows depicted in Fig-
ures 1(a) and 9(a). The second part, already derived in a different way by Breiter
and Jackson (1998), follows from the fact that 0 at the collision instant but
is nonzero before that. Using the properties of the flow near the origin, depicted in
Figures 3 and 9(b), we obtain the following result.
THEOREM 10. In the two-body problem with drag defined by system (3) with given
constants > 0, , and , with < 0, there are no collision orbits, but orbits
can come arbitrarily close to collisions.
From the properties of the flow depicted in Figures 5, 7, 9(b), and 9(d), we can
draw the following final conclusion.
THEOREM 11. In the two-body problem with thrust defined by system (3), with
given constants < 0, , and , there are no collisions that take place at the
equilibrium (r, v, ) = (0, 0, 0), but there are orbits that lead to collisions with
rapidly decreasing distance (i.e. with r 0 at the collision instant).
9. Periodic Orbits
Another important question that appears in celestial mechanics in general and in
two-body problems with drag or thrust, in particular, is that of the existence of
periodic orbits. Let us remark that for + > 0 system (3) is dissipative, that is,
the divergence of the vector field is negative (which implies that the flow makes
the volume shrink in phase space). It is well known that dissipative systems have
attractors, which can include a limit point, a torus, a homoclinic orbit, and a limit
cycle (i.e. a periodic orbit that is the limit of a sequence of orbits). So if periodic
orbits exist, they must be limit cycles. The divergence of the vector field in (3) is
indeed negative for + > 0 since
h 2 i
r v + (v + r 3 2 ) +
r v
h i
+ 2 v r = < 0.
We can summarize the above property in the following statement.
THEOREM 12. In a two-body problem with drag or thrust, defined by system (3),
in which the coefficients and are such that + > 0, if periodic orbits exist,
then they are limit cycles.
14 FLORIN DIACU
10. Conclusions
The geometric study of system (3), which describes the attractive or repelling
motion of two-body problems with drag or thrust, has led us to several qualitat-
ive results. In case of an attracting force with radial drag (, > 0) we have
seen that the set of rectilinear solutions is formed by ejectioncollision, ejection
escape, and capturecollision orbits. Moreover, there is no upper bound for the
distance between bodies in the set of rectilinear ejectioncollision orbits, i.e. this
distance can become arbitrarily large. In case of a repelling force with radial drag
( < 0, > 0) we have proved that the set of rectilinear solutions is formed
by captureescape and ejectionescape orbits. Moreover, there is no lower bound
for the distance between bodies in the set of rectilinear captureescape orbits,
that is this distance can become arbitrarily small though no collisions occur. In
case of an attracting force with radial thrust ( > 0, < 0) we proved that
the set of rectilinear solutions is formed by capturecollision, ejectioncollision,
and ejectionescape orbits. Also, there is no upper bound for the distance between
bodies in the set of rectilinear ejectioncollision orbits. In case of a repelling force
with thrust (, < 0) we showed that the set of rectilinear solutions is formed
by capturecollision and captureescape orbits. Moreover, there is no lower bound
for the distance between bodies in the set of rectilinear captureescape orbits. An
aspect concerning all the above cases is that independently on the values of the
constants there exist no rectilinear oscillatory orbits. We also proved that in the case
of an attractive force with drag for which > 0, the set of initial conditions leading
to collisions has positive measure. Moreover, except for rectilinear solutions, all
other collision orbits spiral towards collision but are asymptotically rectilinear. In
case of a force with radial drag ( > 0) for which the other forces are such that
< 0, we have proved that though there are no collisions, orbits can come
infinitesimally close to collision. In case of a radial thrust ( < 0), independently
of the signs of and , there can appear only collisions with rapidly decreasing
distance (i.e. such that r 0 at the collision instant). Finally we have shown
that, for + > 0, if periodic orbits exist they must be limit cycles.
References
Breiter, S. and Jackson, A. A.: 1998, Mon. Not. R. Astron. Soc. 229, 237.
Brower, D. and Hori, G.: 1961, Astron. J. 66, 193.
Burns, J. A., Lamy, P. L. and Soter, S.: 1979, Icarus 40, 1.
Danby, J. M. A.: 1962, Fundamentals of Celestial Mechanics, MacMillan, New York.
Ferraz-Mello, S.: 1992, In: D. Benest, C. Froeschle (eds) Interrelations between Physics and
Dynamics for Minor Bodies in the Solar System, Editions Frontieres, Gif-sur-Yvette, p. 45.
Leach, P. G. L.: 1987, J. Phys. A: Math. Gen. 20, 1997.
Mavraganis, A. G.: 1991, Celest. Mech. 51, 395.
Mavraganis, A. G. and Michalakis, D. G.: 1994, Celest. Mech. 58, 393.
Mittleman, D. and Jezewski, D.: 1982, Celest. Mech. 28, 401.
TWO-BODY PROBLEMS WITH DRAG OR THRUST: QUALITATIVE RESULTS 15
Mignard, F.: 1992, In: D. Benest and C. Froeschle (eds), Interrelations Between Physics and
Dynamics for Minor Bodies in the Solar System, Editions Frontieres, Gif-sur-Yvette, p. 447.
Plummer, H. C.: 1905, Mon. Not. R. Astron. Soc. 65, 229.
Plummer, H. C.: 1906, Mon. Not. R. Astron. Soc. 67, 63.
Poynting, J. H.: 1903, Phil. Trans. R. Soc. A 202, 525.
Robertson, H. P.: 1937, Mon. Not. R. Astron. Soc. 97, 423.
Wyatt, S. P. and Whipple, F. L.: 1950, Astrophys. J. 111, 558.