Central force motion/Kepler problem

This short note summarizes our discussion in the lectures of various aspects of the motion under
central force, in particular, the Kepler problem of inverse square-law (gravitational) force: for more
details, refer to the notes that you should have taken during lectures or GPS chapter 3.

1 Reducing 2-body motion to effective 1-body, that too

with 2 d.o.f and 1st order differential equations
(a). To begin with, we re-write the coordinates of the 2 bodies in terms of their relative coordinate
(denoted by r) and that of the center of mass (COM) (R). We assume that potential is a function
only of r or its time derivatives. Clearly, the COM then has a constant velocity, thus we neglect its
motion and focus on that of r, i.e., effectively 1-body (with reduced mass) moving around a fixed
center of force (chosen to be at the origin).
(b). Furthermore, we assume that potential depends only on r (i.e., magnitude of r), in which case
the angular momentum (denoted by L) is a constant of motion.
Using the direction of angular momentum being a constant of motion we deduce that motion is in
a plane, i.e., we reduce from 3 to 2 dimensions (D)/degrees of freedom (d.o.f.), for which we will
use polar coordinates (r and θ).
(c). Magnitude of angular momentum:

l = mr2 θ̇ (1)

is also constant.
(d). Kepler’s 2nd law (areal velocity of planets, = 12 r2 θ̇ is constant, for any central force) follows
from l being constant.
(e). Since force is conservative (Lagrangian is time-independent), it follows that energy (E) also
constant: it is given by sum of potential energy (PE) and kinetic energy (KE)
1 1
E = V (r) (“original” PE) + m ṙ2 (“radial” KE) + m r2 θ̇2 (angular KE) (2)
2 2
2
1 l
= V (r) + m ṙ2 + using Eq. (1) (3)
2 2 m r2
From the 2nd line of Eq. (3) just above, we have
r h
2 l2 i
ṙ = E − V (r) − (4)
m 2 m r2
Using Eqs. (3) and (1), i.e., the two constants of motion, we are thus down to first order differential
equations (DE’s) for r and θ (cf. 2nd order to begin with, i.e., either directly using Newton’s laws
or via Largrange’s equations)
(f). In particular, formal solution for t(r) – integrating Eq. (4) – is
Z r
dr0
t = r h i (5)
r0 2 0) − l2
m
E − V (r 2mr 0 2

where r0 is the value of r at t = 0.

2 Detour: equivalent 1D potential

(a). In fact, 2nd line of Eq. (3) suggests an equivalent 1D potential, i.e., for motion of r only, given
by

0 l2
V (r) ≡ V (r) + (6)
2 m r2
(with kinetic energy of this 1D motion being mṙ2 /2 to make up total energy).
(b). The advantage of V 0 is that we can readily (i.e., without actually solving EOM) deduce
qualitative features such as motion is bounded (unbounded) for E < (>)0 for a class of potentials
of the form V = −k/rp , with k > 0 and 0 < p < 2 (which includes inverse-square law force).
(c). In addition, energy and radius of circular orbit for inverse square law (p = 1) is given by
minimum of V 0 (r):
m k2
Ecircular = −
2 l2
l2
rcircular = (7)
mk

3 “Eliminating” time: orbit equation

(a). It is easier to compute (and useful to know) r(θ) by “canceling” dt between Eqs. (1) and (4)
and integrating:
Z r
dr
θ = q + constant (8)
r0 r 2 2mE 2 m V (r) 1
l2
− l2 − r 2

(b). For Kepler problem, we get conic sections for comets/planets (Kepler’s 1st law, valid only for
inverse-square law force):
1 m kh 0
i
= 1 + e cos (θ − θ ) (9)
r l2
with eccentricity given by
r
2 E l2
e = 1+ (10)
m k2
and θ0 corresponding to closest approach to focus (perihelion). Thus, we have hyperbolic (e > 1),
parabolic (e = 1), elliptical (0 < e < 1) and circular (e = 0) orbits respectively for E > 0, E = 0,
E < 0 and E = −mk 2 / (2 l2 ). All these conclusions match those obtained simply using equivalent
1D potential (i.e., without actually solving EOM, cf. approach just above); in particular, the last
result doing so quantitatively, i.e., being same as 1st line of Eq. (7).

4 More “playing around”

(a). A useful relation can be obtained between the semi-major axis of the ellipse, a (i.e., sum of
distances from focus to points of closet and farthest approaches or turning points) and E [based on
the quadratic equation for turning points obtained by setting ṙ = 0 in 2nd line of Eq. (3)]:
k
a = − (11)
2E
In turn, Eqs. (10) and (11) give
r
l2
e = 1− (12)
mka

(b). Kepler’s 3rd law (relating time period of orbit of planets, τ to size) can be obtained by equating
area from 2nd law to that of ellipse [in terms of a and e in Eq. (12)]:
r
3 m
τ = 2 π a2 (13)
k
For mplanet  mSun , we get reduced mass, m ≈ mplanet . Also, we have k = GN mplanet mSun so that
indeed
3
τ ∝ a2 (14)
i.e., proportionality constant is independent of planet.

5 Complete solution: r(t) and θ(t)

5.1 Parabola
In this case, the integral for t(θ) is easier as follows. In general, plugging r(θ) from Eq. (9) into
Eq. (1) (and rearranging/integrating), we have
Z θ
l3 dθ̃
t = 2  i2 + constant (15)
mk
h 
1 + e cos θ̃ − θ0

For parabola, i.e., e = 1, this simply gives (choosing perihelion to be at t = 0 and θ = 0, i.e., setting
θ0 = 0):
l3
 
θ 1 3 θ
t = tan + tan + constant (16)
2 m k2 2 3 2
which can be inverted to give θ(t) and plugging this into Eq. (9) can give r(t).
5.2 Ellipse
(a). For this case, we define an intermediate/auxiliary variable, ψ (called “eccentric anomaly”, θ
being true anomaly) by

r ≡ a (1 − e cos ψ) (17)

(b). We can get further insight into ψ by comparing r(ψ) in Eq. (17) to r(θ) in Eq. (9) re-written
using Eq. (12) (and assuming θ0 = 0), i.e.,
a (1 − e2 )
r = (18)
1 + e cos θ
This gives
r
θ 1+e ψ
tan = tan (19)
2 1−e 2
Clearly, as θ goes through a complete revolution/cycle (0 to 2π), so does ψ (justifying it as an
“angle”), although (in general) at a different (instantaneous) rate than does θ.
(c). The point of introducing a 2nd angle (ψ) is that evaluating integral in t(r) of Eq. (5) is easier
using r(ψ), i.e.,

m r r0 dr0
r Z
t = q + constant
2k 0 r0 2 a
r − 2 a − 2 (1 − e ) 2
r
ma3 ψ
Z
= (1 − e cos ψ 0 ) dψ 0 + constant (20)
k
where 1st line is obtained plugging V (r) = −k/r and Eqs. (12) and (11) into Eq. (5) and 2nd line
using Eq. (17).
Even though it is not really needed, we can introduce (angular) frequency of oscillation:
2π
ω ≡ (21)
τ
corresponding to a 3rd (!) angle, called mean anomaly

φ ≡ ωt (22)

which (obviously) goes from 0 to 2 π over one period, that too uniformly so [cf. θ and ψ are not (in
general) linear in time.]
We can easily evaluate the integral in 2nd line of Eq. (20) (asssuming ψ = 0 at t = 0) and use above
notation to re-write it as

φ(t)(= ωt) = ψ − e sin ψ (23)

which is the Kepler equation: solving this transcendental equation gives ψ(t), which when plugged
into Eqs. (19) and (17) finally gives us r(t) and θ(t).
6 Another constant of motion: Laplace-Runge-Lenz vector
(a). Clearly, direction of angular momentum, i.e., L̂, fixes which plane the orbit is in, whereas l
and E determine size and shape of orbit, e.g., via Eqs. (11) and (10) for ellipse.
(b). What remains is the (fixed) orientation of major axis of ellipse in plane: it is given by direction
of another constant of motion (only for inverse square-law force) called Laplace-Runge-Lenz vector
(denoted by A):
r
A = p×L−mk (24)
r
We can show using EOM of r that dA/dt = 0.
Also, A being constant can be used to (re-)derive that orbit is a conic section: this analysis shows
that A points in direction from focus to perihelion (point of closest approach), thus fulfilling its
role as above.
(c). |A| is also constant and is related to E and l

2 E l2
 
2 2
|A| = m k 1+ (25)
m k2