AN ANALYTIC SOLUTION TO THE CLASSICAL

TWO-BODY PROBLEM WITH DRAG

DON MITTLEMAN
Department of Mathematics, Oberlin College, Oberlin, 0H44074, U.S.A.

and
DON JEZEWSKI
Mission Planningand Analysis Division, NASA Johnson Space Center, Houston, TX 77058, U.S.A.

(Received 26 October, 1981; accepted 18 December, 1981)

Abstract. An analytic solution to the two-body problem with a specificdrag model is obtained. The model
treats drag as a force proportional to the vector velocity and inverselyproportional to the square of the
distance to the center of attraction. The solution is expressedin terms of known functions and is of a simple
and compact form. The time-of-flightis expressedas a quadrature in the 'true anomaly'.

I. Introduction

There are few solutions, either analytic or in closed-form, to the classical two-body
problem with drag. Brouwer and Hori (1961) developed a closed-form solution in
canonical variables of the motion of an artificial satellite in the gravitational field
of an oblate earth including first-order corrections due to a spherical atmosphere and
a velocity square law in drag acceleration.
Danby, (1962) proposed an alternative drag model in which the resistance is
proportional to the vector velocity and inversely proportional to the square of the
distance to the center of the attracting mass. This model leads to a scalar inhomogene-
ous 2nd-order linear differential equation. He uses a perturbation technique to obtain
the solution to the differential equation assuming that the constant of proportionality
in the resisting force is small and that terms involving powers of this constant greater
than one m a y be neglected.
Initially, we start with a non-linear 2nd-order vector differential equation with
arbitrary scalar coefficients. We then develop a 2nd-order vector differential equation
in which the dependent variable is a unit vector in the direction of the radius and the
independent variable is the 'true anomaly'. We show that this differential equation
includes as special cases the classical Kepler solution and Danby's model.
By analyzing the coefficients in this vector differential equation, we are able to
obtain an analytic solution to the model originally proposed by Danby. Our solution
is shown to include his when the resistive coefficient goes to zero.
Using this analytic solution, the vector differential equation for this drag model
reduces to the differential equation for the vector harmonic oscillator. The general
solution to the problem is then expressed in a simple compact form.

Celestial Mechanics28 (1982) 401-413. 0008-8714/82/0284-0401 $01.95.

Copyright ~ 1982 by D. Reidel Publishing Co., Dordrecht, Holland, and Boston, U.S.A.
402 DON MITTLEMANAND DON JEZEWSKI

We were not able to integrate the time-of-flight equation, the analogue of Kepler's
equation, in a finite number of terms of known functions. The time-of-flight, however,
is readily expressed in terms of a quadrature.
Details of the computational procedure and characteristics of the functions used
in the solution are included in appendices. Graphs illustrating the general characteris-
tic of these functions and the solutions are also included.

2. Statement of the Problem

The two-body problem may be expressed by the vector differential equation

i: +/3(r, r)r + y(r)r = 0, (1)
where r is the radius vector from the center of mass of the attracting body to the
particle,/3(r, ~) and 7(r) are arbitary scalar coefficients and dots denote differentiation
with respect to the independent variable, time. To retain as much generality as
possible, we refrain from specifying/3 and 7 at this time.
We first transform Equation (1) by introducing a polar angle, 0, which we shall
refer to as the true anomaly, as the independent variable.
/" = r'O, i~= r"/) 2 + r' O,

where the prime denotes differentiation with respect to the angle 0.

Equation (1) then becomes
OZr" + (0 +/30)r' + yr = O. (2)

Classically, it has been found useful to introduce the unit vector r = r/r, where
r = Ir [. With this definition, Equation (2) becomes:
r02~ " + (2/02 + rO + flrO)r + (r"02 + r'O + flr'O + ry)~ = 0, (3)
Paralleling the classical procedure, we let
1 U' RH (r
r = -, r' - r" . . . . t- 2 - -
U /./2' U2 U3
and Equation (3) becomes:
2u' ~ ' + /Tu' flu' ~ "~
u - r (4)

Equation (4) is written more simply if we let

fl 0 2u'
(5)
u

U' U" ; )
C+q'+ -v-- ~- ~=0 (6)
U U
 AN ANALYTIC SOLUTION TO THE CLASSICAL TWO-BODY PROBLEM WITH DRAG 403

3. Particular Solution

We are motivated at this point by the solution to the classical Keplerian problem,
fl = 0 and ~ = #/r 3 , and where # is the gravitational constant. For these values of fl
and 7, the conservation of angular momentum implies v = 0. We refrain from so
specifying fl and 7, but do explore the possibility that v = 0. Equation (5) becomes:

+ - U =~

and recalling that u = 1/r,

d(r20) + fir 2 dO = 0. (7)

(Note, if fl = 0, the Keplerian angular momentum r 2 0 = constant.) Equation (7)

readily admits another integral if fl = t~/r 2 where ~ is a constant. This integral is

r20 + ~tO = h o (8)

when h o is the constant of integration.

With v = 0, Equation (6) reduces to:

~"+ ~2 u ~=0 (9)

In retrospect, it is clear that v is equal to zero in at least two cases, (1) when no drag
is present and (2) when the drag is proportional to 1/r 2. It should be noted that any
perturbation having an effect directed along the position vector can be reflected in
our choice of 7. Thus, v = 0 removes any coupling between the unit vectors in the
direction of r and its derivative with respect to 0.
The simplist problem we can solve that includes a drag term is to choose ~ = #u 3.
For this choice, Equation (9) becomes, using Equation (8) to eliminate.0,

St u"~
~" + (h ~ + ,0)2 u - ~ - ) ~ = 0. (10)

Classically, when ~ = 0, i.e. no drag present, the coefficient of ~, when set equal to 1,
produces the differential equation for the Keplerian conics. With drag present, that
is ~ # 0, setting the coefficient of ~ equal to 1, produces the differential equations

~" + ~ = 0 (lla)
and

u" + u - /~ (llb)
2['ho _ ~2"

Equation (11b) is given by Danby (1962).

404 DON MITTLEMAN AND DON JEZEWSKI

4. Solution of Equation (lib)

Danby (1962) studied this equation with ~ ~ 0. His method of solution assumes that
aO/ho is small compared to unity; he expands the right-hand side of (llb) in powers
of a and neglects all but first order terms in ~. The solution to the equation, however,
for arbitrary ~ can be obtained in terms of known functions.
Knowing the solution of the homogeneous equation, u ' + u = 0, a particular
solution can be obtained using the method of variation of parameters. We chose,
however, to use the inverse Laplace transform (Churchill, 1958). Aside from simplify-
ing the algebraic manipulations, this method demonstrates an alternate procedure
for solving a class of differential equations. For convenience of notation, we temporari-
ly change the independent variable by letting z = ho/o:- O. Equation (llb) is now
rewritten as:
d2t/ //
dz 2 + u ~2z2. (12)
If u is the Laplace transform of U, then taking the inverse Laplace transform of
Equation (12), we obtain

z 2 U + U = 0~2
/~ 27 (13)

so that,
/2 "C
U-
0~2 ,~2 -t- 1"

The Laplace transform of this equation is

co

u(z)= ~ .] z +1
0
Using the standard notation (Abramowitz and Stegun, 1964: Chapter 5), u(z)=
= #/ct29(z) where
co
f te-Zt
9(z) = j ~ dt. (14)
0

For details concerning this function, see appendix (A).

The general solution for u(O) is:

u(O)=eoCoS(O-Oo)+~9(~-O), (15)

where e o and 0 o are constants of integration. Hence, the equation for the radius,
 AN ANALYTICSOLUTION TO THE CLASSICAL TWO-BODY PROBLEM WITH DRAG 405

r, is
r= P (16)
[ho"X 2 / h o

where e = eoh2/~ and p = h2/#.

It can be shown (see Appendix A) that

and this is recognized as the solution to the classic Keplerian two-body problem.
The solution for the position vector r, as a function of 0, is obtained from Equation
(lla).
r = r(a sin 0 + b cos 0), (17)
where a and b are arbitrary vector constants and r is given by Equation (16).
The solution, as given by Equations (16) and (17), involves nine constants of
integration, ho, e, 0o, and the vectors a and b. Since the original problem calls for the
solution of a second-order vector differential equation, only six of these constants are
independent. The relationships between them is developed in Appendix C.

5. The Time-of-Flight Equation

The only relationship involving time occurs in Equation (8), which indicates how the
magnitude of the angular momentum changes. Since r is expressible as function of 0
by Equation (16), the time-of-flight At between two values 0 = 01 and 0 = 02 is given
by the quadrature
02

At = -1 f (r(O))2 dO. (18)

Since we are dealing with a perturbed Keplerian problem, the standard transforma-
tion froln true anomaly to eccentric anomaly does not yield an integrand that is
intcgrab~e in terms of elementary functions. Any standard quadrature formula, for
example Romberg integration (Burden et al., 1978), may be used to evaluate the
integral.

6. Graphical Illustrations*

Figures (la) and (lb) illustrate the effect of drag, ~, on Equation (16). The values of the
constant h o and p were chosen equal to 1 and the constant e was taken as 0.01.
* We gratefully acknowledge the assistance of Eleta Malowitz and Jerry Yglesias of Barrios Technology
Inc., for their aid in obtaining the several graphs.
 .8

Fig. la. R a d i u s a s a f u n c t i o n o f t h e t a f o r e = 0.01 a n d ~ = 0.005.

-1.8 -~8 ~ -L8 -B.4 L2 8.4 I ~8 8.8 1.8

--~4

.8t
-1.8

Fig. lb. R a d i u s a s a f u n c t i o n o f t h e t a f o r e = 0.01 a n d ~ = 0.05.

 ~1.5i
Fig. 2a. R a d i u s as a f u n c t i o n o f t h e t a f o r e = 0.5 a n d c~= 0.005.

1.5 ~

: t I

1.8

-l~

-1.5

Fig. 2b. R a d i u s as a f u n c t i o n o f t h e t a for e = 0,5 a n d c~= 0.05,

408 DON MITTLEMAN AND DON JEZEWSKI

Comparing the two figures, it is apparent that the drag effect is more pronounced
as ~ goes from 0.005 to 0.05.
Figures (2a) and (2b) are similar to (la) and (lb) except that, in order to accentuate
the 'eccentricity', the constant e was increased to 0.5. The values of a and e were
chosen only to enhance the graphical display.

7. Conclusion

We have obtained a closed form, analytic solution to the two-body problem including
drag that generalized the classical Keplerian two-body problem and subsumes the
Danby perturbation model.
Two serendipitous results are: (1) the development of a vector differential equation
that permits the analysis of an infinite number of gravitational and drag models, and
(2) obtaining the solution of a linear differential equation using the inverse, rather
than the direct method of Laplace transforms.

Appendix A: The Function g (z)

The following material, pertinent to this problem, is taken from Abramowitz and
Stegun (1964; Chapter 5).
(A) In Equation (14), the function g(z) was given as
co

(' ze- *~
g(z) = / ~ dz. (A1)
3 . +1
0
This function may also be expressed in terms of sine and cosine integrals. If we denote
co

si(z)= - --dz
z
and

Ci(z) = - i COS"Cdz
z
then,
9(z) = - Ci(z) cos (z) - si(z) sin (z). (A2)
For 0 < z -N<1, the sine and cosine integrals may be expressed in terms of the following
expansions.
rt ~ ( - 1)"z 2"+1
si(2) =
n=O

( 1),z 2.
Ci(z) = Yo + In z +
 AN ANALYTICSOLUTIONTO THE CLASSICALTWO-BODYPROBLEMWITH DRAG 409

7o is the Euler constant: Yo = 0.5772 15664 90153 28606 ... Clearly the two infinite
series coverage for all z, and rapidly for [z] -%<1.
For 1 ~<z < oo, an asymptotic expansion is available.

g(Z) ~ - - -Z2
--t Z4 ... ,

Also for this domain of z, a rational approximation has been determined; for
example:
l~<z<oo
1 ( z s + a l z6 + a2 z4 -t- a3z 2 -~- a4~
g(z) = ~ . ~ + blZ6 + bzz4 + ~3z2~-~4 ] + s(z)

Is(z)[ < 3 x 10-7

al= 42.242855 bl= 48.196927

a2=302.757865 bz= 482.485984

a3=352.018498 b3=1114.978 885
a 4 = 21.821899 b 4 = 449.690326
The graph of g(z) is given in Figure (A1).
(B) It is known that g(z) satisfies the differential equation:
1
g"(z) + g(z) z2.

If we define

f ( z ) = i z~--~d~
e_Z~
0

then, f (z) satisfies the differential equation

1
if(z) + f(z) = -
Z
and,
1
g'(z) = - - +f(z)
Z

y ( z ) = - g(z).

Corresponding to (A2),
f(z) = Ci(z) sin (z) - si(z) cos (z).
 410 DON MITTLEMAN AND DON JEZEWSKI

2.000

1.5~0

9 (z) 1.000

0.500

o. o o 0 ~ _ ~ _ ~
Q
Q
Q
9 9
cJ o~ ,4 t~ u5 r< ~ o~

z (rQd)

Fig. A1. The graph of the function g(z).

An asymptotic formula forf(z) is:

f (z) ,.~ !( 1 - 71 + -~ .... )

For 1 ~< z < oo, a rational approximation t o f ( z ) is given by:

l~<z<oo

l ( zS + alz6 + a2z'~ + a3zZ + a4'~

f(z) = z \ ~ + bl z6 + b2 z4 + b3z 2 + b4,} + e(z)
le(z)l < 5 • 10 -7
 AN ANALYTIC SOLUTION TO THE CLASSICAL TWO-BODY PROBLEM WITH DRAG 411

2.00~

1.5~

Cz) 1. O~B

O. 5Oe

: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :

,d - eJ ~ 4 td Q6 t-Z ~ ~

z (r-od)
Fig. A2. The graph of the function f(z).

a 1 = 38.027 264 b1= 40.021433

a2 = 265.187 033 bz = 322.624 911

a3
= 335.677 320 b 3 = 570.236 280

a4 = 38.102 495 b 4 = 157.105 423

T h e g r a p h o f f ( z ) is given in F i g u r e (A2).
(C) W e n o w p r o v e the assertion m a d e in the text t h a t the

lim (h~176
F r o m the a s y m p t o t i c e x p a n s i o n for g(z),
1 3! R 6

0; -t )6'

where R 6 is b o u n d e d for small ~. Thus,

lim (h~176\ O~ R6()2 -1.

aoo+ \ ~ / ~o+
412 DON M/TTLEMAN AND DON JEZEWSKI

Appendix B: Computation of the constant

We propose three basic methods for computing ~. In all cases, we assume that there
exists a drag model given by D(r, i)i where D(r,/) is a scalar function of the vectors
r and i.
(1) The constant ~ may be computed from the initial conditions,
=r2D(r,t)
u F:
~O,F=F
. 9o"

(2) The second method uses an averaging technique.

02

1 fr2D(r,f)dO,
0~= 02 _ 0a
01

where r, and t are the position and velocity vectors determined from the Keplerian
orbit having the same initial conditions.
(3) A variation of (2) for longer flights would be an iteration on the trajectory.
The first step is to calculate an a~, assuming a Keplerian orbit. Using ~a in the drag
model, a drag orbit is determined. The values of r and t on this orbit are then used
to calculate ~2" The process is repeated.
(4) A third possibility is to segment the total trajectory and compute an 0t for each
segment using any of the above methods.

Appendix C: Computational sequence

Given initial position and velocity vectors r~ and v~ = ~a at a given t a and a resistive
constant ~, what steps are necessary to compute r 2 and v 2 at a time t 2 ?
The initial value of the magnitude of the angular momentum is Ira x v a I. Thus,
we may calculate (r20a) = Ira x v 1 [.
The angle 0 a is computed from the initial position and velocity vectors using the
Keplerian transformation to orbital elements. The constant h o is then determined from

h o = r20a + ~0a. (C1)

F r o m Equation (16) and its time derivative we may write:

~ sin(O - 0o) = - -

~cos(0_0o)= 1 ~ [ho )
By squaring and adding we get

e2=h4F(~--~g'(~-O))2+(~--~O(~-O))21 (C2)
 AN ANALYTIC SOLUTION TO THE CLASSICAL TWO-BODY PROBLEM WITH DRAG 413

and by dividing,

sin(O - 0o) _ r20

os,0 0o, 1 0)
r.~
where 1: = - -
?-

F r o m Equations (C2) and (C3), evaluated at r 1 and r l , the constants e and 0 o, 0 ~< 0 o <
2re are determined. F r o m Equations (17) and its time derivative, w e determine the
vector constants a and b.

a = r i s i n 0i rlil - - r l r l COS 01
2"
rl rio 1

b riri--riri rl
- 2" sin 01 + --cos 01 .
rio 1 rI

It m a y readily be shown that a and b are orthogonal unit vectors.

References

[1] Abramowitz, M. and Stegun, I. (eds): 1964, Handbook of Mathematical Functions with Formulas,
Graphs, and Mathematical Tables, National Bureau of Standards, U.S. Government Printing Office,
Washington, D.C., (Eighth Dover printing).
[2] Brouwer, D. and Hori, G. : 1961, Astron. J. 66, 193; (see also Brouwer, D. and Clemence, G. : 1961,
Methods of Celestial Mechanics, Academic Press, New York, Chapter 12).
[3] Burden, R. L., Faires, J. D., and Reynolds, A. C.: 1978, Numerical Analysis, Prindle, Weber, and
Schmidt, Boston, Mass., p. 210.
[4] Churchill, R. V. : 1958, OperationalMathematics, 2nd Edition, McGraw-Hill, New York, p. 183.
[5] Danby, J. M. A. : 1962, Fundamentals ofCelestialMeehanics, MacMillian Co., New York, p. 231.