Prime Integrals in Relativistic

Celestial Mechanics
Vittorio Banfi*

In relativistic celestial mechanics it is possible to use the Schwarzschild solu-

tion, for a fixed spherical central body, in order to describe the orbital motion
of a particle around it. This is the astronomic case of a planet’s orbital motion
around the Sun or of a satellite’s orbital motion around the mother planet. This
paper reports the study of two prime integrals of the motion.

1. Generality on relativistic problem of motion

The track of a particle moving freely in the space-time with metrics
 rg  2 1
ds 2 = c 2  1 −  dt − dr 2 − r 2 dθ 2 − r 2 sin θ 2 d ϕ 2 ,
 r  rg
1−
r
in which rg = 2GM/c2 is the Schwarzschild radius, r,θ and ϕ the polar coordinates, t the time
coordinate and also G = gravitational constant, M = mass of central body and c = velocity of
light, is determined by the equations of a geodesics
d 2 xα i
α dx dx
j
+ Γ ij = 0, (1)
ds 2 ds ds
being the space-time coordinates x1=r, x2=θ, x3=ϕ, x4=ct; so α, i, j are varying 1,2,3,4. We start
with α = 2, putting the only Christoffel Γij2 symbols which are not zero, and we obtain

d 2 x2 1
2 dx dx
2 2
2 dx dx
1 3
2 dx dx
3
+ Γ12 + Γ 21 + Γ 33 = 0.
ds 2 ds ds ds ds ds ds
The foregoing equation is rewritten so:
2
d 2θ 2 dr dθ  dϕ 
+ − sin θ cosθ   = 0, (2)
ds 2 r ds ds  ds 
i.e., the differential equation for θ=θ(s). Because of the spherical symmetry, there is no loss of
generality in confining our attention to particles moving in the “equatorial plane” given by
θ=π/2 (the point-mass M is regarded at rest in the origin of the reference frame in polar coordi-
nates). With this constant value for θ, eq. (2) is satisfied. Putting α=1 we have
d 2 x1 dx1 dx1 dx 2 dx 2 dx3 dx 3 dx 4 dx 4
2
+ Γ111 + Γ122 + Γ133 + Γ144 = 0.
ds ds ds ds ds ds ds ds ds

*
Centro di Astrodinamica “G. Colombo” Politecnico di Torino - Corso Duca degli Abruzzi, 24 - Torino -
ITALY (Postal Address: Prof. V. Banfi - Corso Sempione, 82 - 20154 - Milano - ITALY)
APEIRON Vol. 7 Nr. 3-4, July-October, 2000 Page 133
 Since θ =constant=π/2 and the only surviving Christoffel symbols are
rg rg
2  rg  2
Γ111 = − , Γ1 = −r 1 −  , Γ144 = ,
2
rg  33  r  2
rg 
r 1 −  r 1 − 
 r   r 
the foregoing equation becomes
rg rg
2 2 2
d 2r 2  dr   rg   d ϕ  2  dt 
− − r 1 −    + c 2   = 0. (3)
ds 2 2
rg   ds   r   ds  2
rg   ds 
r 1 −  r 1 − 
 r   r 
Then for α=3
d 2 x3 2
3 dx dx
3 3
3 dx dx
2 1
3 dx dx
3 3
3 dx dx
1
+ Γ 23 + Γ 32 + Γ13 + Γ 31 = 0.
ds 2 ds ds ds ds ds ds ds ds
Since, as before, θ=constant and the only surviving Christoffel symbol is Γ13
3
= Γ31
3
=1 r ,
we obtain
d 2ϕ 2 dr dϕ
+ = 0. (4)
ds 2 r ds ds
Finally for α=4
d 2 x4 1
4 dx dx
4
dx 4 dx1
2
+ Γ14 + Γ 441 = 0,
ds ds ds ds ds
rg
the only surviving Christoffel symbol is Γ = Γ = 4 4 2 and then we have
14 41
 rg 
r 1 − 
2

 r 
d 2t rg dr dt
+ = 0. (5)
ds 2  rg  ds ds
r 2 1 − 
 r 
Eqs. (4) and (5) are readily integrated. Multiplying by r2 eq. (4) we obtain
d 2ϕ dr d ϕ
r2 2
+ 2r = 0,
ds ds ds
or
d  2 dϕ 
r  = 0,
ds  ds 
which is integrated giving
dϕ
r2 = h1 = integration constant. (6)
ds
It is easy to demonstrate that the expression

Page 134 APEIRON Vol. 7 Nr. 3-4, July-October, 2000

 dt k
= , [k = constant] (7)
ds rg
1−
r
satisfies eq. (5). Therefore from eqs. (6) and (7) we deduce
d ϕ h1  rg   rg  h1
r2 = 1 −  = L 1 −  [L= = constant]. (8)
dt k r   r  k
Eq. (8) is the first prime integral. Let us come back to Schwarzschild metrics rewritten so
2 2 2
1  dr  2  dϕ   rg  2  dt 
−1 =   +r   − 1 − c   . (9)
 rg   ds   ds   r   ds 
1 − 
 r 
Putting, into eq. (9), eqs. (6) and (7) we have
2
h2 c2 k 2 1  h1 dr 
−1 = 12 − +   ,
r  rg   rg   r 2 dϕ 
1 −  1 − 
 r   r 
or
2
rg  h dr  h12  rg  2 2
− 1 =  12  + 2 1 − −c k ,
r  r dϕ  r  r 
and then
2
 h12 dr  h12 rg h12 rg
 2  + 2 = + c2 k 2 − 1 + ;
 r dϕ  r
2
r r r
finally we get
2
 1 dr  1 rg 1 c 2 k 2 − 1 rg
 2  + = + + 2.
 r dϕ  r
2
r r2 h12 rh1
1 du 1 dr
With substitution u = , =− 2 , we have
r dϕ r dϕ
2
 du  c2k 2 − 1 rg
  + u 2
= rg u 3
+ + 2 2 u, (10)
 dϕ 
2 2
Lk Lk
since h1=Lk. Differentiating eq. (10) with respect to ϕ
du d 2 u du du rg du
2 + 2u = 3rg u 2 + 2 2 ,
d ϕ dϕ 2
dϕ dϕ L k dϕ
du
and removing the factor 2
dϕ

d 2u 3rg u 2 rg 3rg u 2 GM
+u = + 2 2 = + 2 , (11)
dϕ 2 2 2L k 2 C0
with Lkc = C0.

APEIRON Vol. 7 Nr. 3-4, July-October, 2000 Page 135

 Both eqs. (10) and (11) govern the particle relativistic motion, giving its trajectory in
Schwarzschild field. It is very easy to show that the connection with newtonian theory is very
close, as of course it must be. Starting from
r
dv GM µ
= − 2 rˆ = − 2 rˆ [µ=GM] (12)
dt r r
of newtonian paradigm ( r̂ being the unit vector of r direction and ϕ̂ the unit vector of trans-
verse direction), the vector velocity is
r
& ˆ + rϕϕ
v = rr &ˆ (13)
Scalar multiplication of eqs. (12) and (13) provides
r
r dv d  1 2  µ
v⋅ =  v  = − 2 r&
dt dt  2  r
which is readily integrated so
1 2 µ
v = + integration constant.
2 r
With a = semi-major of an ellipse, in which point-mass M occupies one of two foci, we
have also
1 1 
v2 = 2µ  −  . (14)
 r 2a 
Owing to the central force we deduce
dϕ
r 2ϕ& = r 2 = C0 = µ p , (15)
dt
with p = semi-latus rectum = a(1-e2) and e = eccentricity of ellipse (Finley-Freundlich, 1958).
From eqs. (14) and (15) we get
2
 d  1  1 1 
v = r& 2 + r 2ϕ& 2 = r 4ϕ& 2 
2
   + r ϕ& = 2 µ  − 
2 2

 d ϕ r
   r 2a 
and also
2
 d  1  1 2µ  1 1 
 d ϕ  r   + 2 = 2  r − 2a  .
   r C0  
1
Putting u = we have
r
2
 du  2µ µ
 +u = 2 u −
2
 . (16)
 dϕ  C0 aC02
Differentiating eq. (16) with respect to ϕ gives
du d 2 u du 2 µ du
2 + 2u = ,
d ϕ dϕ 2 d ϕ C02 d ϕ
and finally

Page 136 APEIRON Vol. 7 Nr. 3-4, July-October, 2000

 d 2u µ
+u = 2 . (17)
dϕ 2 C0
This equation is the same as eq. (11) if rg=0, which implies L→C0.

2. Analytical development of the relativistic model

In item 1 we have shown the prime integral, i.e.
 rg 
r 2ϕ& = L 1 −  . (8)
 r 
Now we will provide another prime integral, say, a generalized total energy integral.
In order to show this we start from Eddington’s treatise “The mathematical theory of Rela-
tivity” (Eddington, 1952). As shown in item 1 we are able to work out the planetary orbits from
Einstein’s law independently of newtonian paradigm. Nevertheless, in this context, Eddington
is looking for a “perturbation theory,” i.e. he wants to introduce a vectorial perturbing force (for
r r µ
unit mass) Fp = Fpr rˆ + Fpϕ ϕˆ superimposing to the newtonian one Fn = − 2 rˆ .
r
The problem now is to determine Fpr and Fpϕ on the basis of GRT (general relativity the-
ory), or better using eqs. (3), (4), (5) of item 1.
After some analytical treatments, Eddington provided the following expressions of radial
and transverse components (see Appendix)
3 rg 2 rg µ 
Fpr = r& − rg ϕ& 2 − 3 
2r 2
r 
 (18)
rg 
Fpϕ = r&ϕ&
r 
It should be noted that, strictly speaking, the net force (per unit mass, or acceleration) on the
particle is not a central force.
L  rg 
Substitution of ϕ& = 2  1 −  from eq. (8) into (18) gives
r  r 

L2  rg  rg µ 
2
3 rg 2
Fpr = r& − rg 4  1 −  − 3 
2 r2 r  r  r 
 (19)
rg  rg  
& 1 − 
Fpϕ = 3 rL 
r  r  
Differential equation (12) becomes
r r
dv µ
= − 2 rˆ + Fp . (20)
dt r
r
Again with scalar multiplication and using Fp = Fpr rˆ + Fpϕ ϕˆ we have
r
r dv d  1 2  µ
v⋅ =  v  = − 2 r& + rF & pr + rϕ& Fpϕ ,
dt dt  2  r
and then using eq. (19)

APEIRON Vol. 7 Nr. 3-4, July-October, 2000 Page 137

 2
d 1 2 µ 3 rg 3 L2  rg  rg µ
 v  = − 2
&
r + 2
&
r − rg 4 
1 −  r& − 3 r& +
dt  2  r 2r r  r  r
2
rg  rg  µ 3 rg 3 rg µ
+ 4 L2  1 −  r& = − 2 r& + r& − 3 r&.
r  r  r 2 r2 r
Integration with respect to time, between t0 and t, gives
t
µ µ
1 2
2
( 2

r r0 t0
)
v − v0 = − + ∫ Γdt , (21)

with v(t0)=v0, v(t)=v and also r(t0)=r0, r(t)=r. Since

dr dr dϕ dr L  rg  dr
r& = = = ϕ& = 2 1 −  ,
dt d ϕ dt d ϕ r  r  dϕ
we obtain
t
3 rg 3
t r
rg µ
∫ Γdt = ∫ 2
&
r dt − ∫ 3
dr ,
t0 t0
2r r0 r

and then
3 3
t
3 rg L3  rg   dr 
t r
rg µ
∫t Γ dt = ∫t 2 r 2 r 6  r   dϕ 
 1 −    dt − ∫r r 3 dr . (22)
0 0 0

The first term on the right side of eq. (22) is worked so:
2
t
3 rg  dr  L3  3rg   dr 
∫t 2 r 2   6 1 −
 dϕ  r 
  dt =
r   dϕ 
0

2 2 2
3 rg  dr  L3  3rg  3 rg L  dr 
r r
r2
∫ 2   6 1 − r
r0 2 r  d ϕ  r 

 L 1 − rg 
dr = ∫
r0 2 r
6   dr.
 dϕ 
 
 r 
Finally the eq. (22) becomes
2
t r
3 rg L
2
 dr  r
rg µ
∫ Γdt = ∫
t0 r0
2 r6
  dr − ∫ 3 dr = R .
 dϕ  r0 r
(23)

Putting eq. (23) into (21) we have

µ µ
1 2
2
( 2
v − v0 = − + R .
r r0
) (24)

The quantity R, by means of eq. (23), is splitted into two terms

2
r 2
3 rg L  dr 
r
rg µ
R = R1 + R2 = ∫ 6   dr − ∫ dr . (25)
 dϕ 
3
r0 2 r r0 r

We note that R2 is readily calculated, while R1 requires a little attention in order to gain
some insight into integration problem. At first, we see that the integrand contains the factor

Page 138 APEIRON Vol. 7 Nr. 3-4, July-October, 2000

 2
 dr 
  . As it is well known the orbit, in the simplest Schwarzschild solution of the problem of
 dϕ 
motion, is the so-called semielliptic having the equation (Berry, 1989)
p
r= (26)
3rg
1 + e cos (1 − ϕ)
2r
and corresponding to a special plane curve whose name is polygasteroid (Loria, 1930).
The motion of the particle exhibits a trajectory which fill up everywhere densely the ring
region bounded by two circles, respectively, with the maximum and the minimum r as the radii.
dr
As second step, we see from eq. (26) that (following GRT paradigm) differs from
dϕ
 dr 
  (following classical celestial mechanics) on account of O(rg) terms.
 d ϕ  newt
2
 dr 
Since O(rgn)
terms, with n≥2, are negligible, we can substitute   into R1, in order to
 d ϕ  newt
perform calculation.
By means of simple calculations we deduce
2
r4µ 2  2 p  
2
 dr 
  = e −  − 1 
 d ϕ  newt L4  r  
and putting it into the expression for R1 we will have
3 rg µ  2  p  
r 2 2

R1 = ∫ 2 2 
e −  − 1  dr =
r0 2 r L   r  
(27)
3 rg µ 1  rg µ p  1 1  3 rg µ 2 p  1 1 
2 2 2
2 1

2 L2
(1 − e )  −  +
2 L2
 3 − 3  + 2  2 − 2 
 r r0  r r0  2 L  r0 r 
whereas R2 is provided by the formula
rg µ  1 1 
R2 = −  2 − 2  .
2  r0 r 
Remembering eq. (15) and p=a(1-e2), finally we obtain
3 rg µ  1 1   1 1  1 1 1 
R = R1 + R2 =  −  + rg µ  2 − 2  + rg µ a (1 − e )  3 − 3  .
2

2 a  r r0   r0 r  2 r r0 
As in newtonian celestial mechanics, also in relativistic celestial mechanics are available
two prime integrals. In fact during the particle motion remains constant the quantity
 rg 
r 2ϕ& − L 1 −  = first prime integral,
 r 
particularly of value zero, but also the quantity
1 2 µ
v − − E = second prime integral;
2 r
APEIRON Vol. 7 Nr. 3-4, July-October, 2000 Page 139
this last expression is the total energy of the particles. The correction term E is provided by the
following formula
3 rg µ rg µ 1 rg µ
E= − 2 + a (1 − e 2 )
2 ar r 2 r3
and it depends on µ, r, rg and also on a, e (size and shape of newtonian ellipse).

3. Concluding Remarks
The conclusions are derived here in the first order approximation of quantity rg. The terms
O(rgn), with n≥2, have been dropped out. Similar approximations are used in textbooks using
“parametrized post-newtonian (PPN) formalism” (Brumberg, 1991). Possible application of
foregoing results to practical spacecraft astrodynamics (orbital maneuvers problems and so on)
seems to us not reasonable. In fact the corrections would be negligible ones. Perhaps these
results shall be taken into account in the far future in the case of mission towards neutron stars,
black holes and similar massive heavenly bodies.
References
Finlay-Freundlich (1958), Celestial Mechanics, Pergamon Press, New York.
A. L. Eddington (1952), The Mathematical Theory of Relativity, Cambridge University Press (p.26,
ch.III).
M. Berry (1989), Principles of Cosmology and Gravitation, Adam Hilger, Bristol.
G. Loria (1930), Curve piane speciali, Hoepli, Milano.
V. A. Brumberg (1991), Essential Relativistic Celestial Mechanics, Adam Hilger, Bristol.

Appendix
The aim of this appendix is to obtain eqs. (18) starting from eqs. (3), (4) and (5). It is substantially
followed the Eddington’s method. Let us write down eqs. (3) and (4).
rg rg
2 2 2
d 2r 2  dr   rg   dϕ  2  dt 
−   − r 1 −   + c 2   = 0. (A.1)
ds 2  r   ds   r   ds   r   ds 
r 2 1 − g  r 2 1 − g 
 r  r

d 2ϕ 2 dr dϕ
+ = 0. (A.2)
ds 2 r ds ds
d2 d
It is covenient to transform the operator 2
in the following way. Let us write the operator so
ds ds
d d dt dt d
= =
ds dt ds ds dt
and then we have
2
d 2 dt d  dt d   dt  d 2 dt d  dt  d
= =  +   . (A.3)
ds 2 ds dt  ds dt   ds  dt 2 ds dt  ds  dt
In eq. (A.3) the right hand second term is rewritten so
dt d  dt  dr d
  . (A.4)
ds dr  ds  dt dt

Page 140 APEIRON Vol. 7 Nr. 3-4, July-October, 2000

 d  dt 
Now we calculate the   term using eq. (7):
dr  ds 

  r 
k  g2 
d  dt  d  k  r
 =   = −  2 .
dr  ds  dr  rg   rg 
1−  1 − 
 r   r

Eq. (A.3) becomes

r 
2 2 k  g2 
d 2  dt  d 2 dt dr d  dt  d  dt  d 2  r  dr dt d .
2
=  2 +   =  2 − 2
ds  ds  dt ds dt dr  ds  dt  ds  dt  rg  dt ds dt
1 − 
 r
dt k
Again owing eq. (7), i.e. = , the foregoing equation becomes
ds r
1− g
r
  rg  
2  2
2  2 
d  dt   d  r  dr d  .
==    −  (A.5)
ds 2  ds   dt
2
 rg  dt dt 
 1 − 
  r 

Let us come back to (A.1); applying eq. (A.5) we deduce

 rg   rg 
2 2 2  2  2 2   2
 dt  d r  dt   r   dr   dt  2  dr 
  −    −    −
 
ds dt 2
 
ds  r      2
dt ds  r   dt 
1 −  r 1 − g
g

 r  r 
 g
r
2 2   2
 dt   dϕ   rg   2  c 2  dt  = 0,
−    r 1 −  +  
 ds   dt   r  r   ds 
r 2 1 − g 
 r

and also
2 2
d 2 r 3 rg 1  dr   rg   dϕ  µ 1
−   − r 1 −    + 2 =0. (A.6)
dt 2 2 r 2  rg   dt   r  dt  r  rg 
1 −  1 − 
 r   r

The same procedure applied to (A.2) gives

  rg  
2 2  2  2
  
dt d ϕ r
  dr d ϕ  2 dr dϕ  dt 
   −  +   =0
 ds   dt
2
 rg  dt dt  r dt dt  ds 
 1 − 
  r 

and then
d 2ϕ rg dr dϕ 2 dr dϕ
− + = 0.
dt 2 r 2 dt dt r dt dt
Multiplying the foregoing equation by r

APEIRON Vol. 7 Nr. 3-4, July-October, 2000 Page 141

 d 2ϕ dr dϕ rg dr dϕ rg
r +2 = = r&ϕ& = transverse component of acceleration.
dt 2 dt dt r dt dt r
From (A.6) we have
2
d 2r  dϕ  µ 3 rg 2 rg µ
2
− r  =− 2 + r& − rgϕ& 2 − 3 = radial component of acceleration.
dt  dt  r 2 r2 r
We conclude that
3 rg 2 rg µ
Fpr = 2
r& − rgϕ& 2 − 3
2r r
rg
Fpϕ = r&ϕ& .
r

Page 142 APEIRON Vol. 7 Nr. 3-4, July-October, 2000