Hamiltonian-Jacobi Solution of the Orbit Equation

Irwin Angelo M. Amago

December 8, 2010

In this problem we are asked to solve for the equation of a particle subjected to a central force eld.
We use, for this, the methods of the Hamiltonian-Jacobi formalism to be able to mathematically acquire an
equation that describes the motion of the particle. The gure below shows the usual motion of a particle
due to some central force eld.

Figure 1: Particle Moving Under the Inuence of a Central force

If we stick with common methodology then the initial step for the problem would be to rst formulate
the Lagrangian of the system, that is the dierence between the Kinetic and Potential energies, and derive
the Hamiltonian of the system from this formulation. For this system, the Hamiltonian is given merely as
the total mechanical energy of the system, which is often the case for systems whose paramters do not have
explicit time dependence. Let us denote this Hamiltonian to be h. Note that in the general case, we actually
want some sort of mapping that transforms our original paramters to new parameters. That is, we want a
new Hamiltonian H with parameters
θ→Θ
 r → R,

p r → Pr

pθ → Pθ .

For convenience, of course, we say that our new Hamiltonian, H must be some constant or simply, must
be zero. We begin solving this problem by noting that our original Hamiltonian, h is given to be
p2r p2θ k
h= + 2
− .
2m 2mr r
As a primary step, we must rst dene the Hamiltonian-Jacobi Equation for this system. Note that to
get there, we must rst remember that
∂F
0=h+ ,
∂t
where F = F (r, θ, t)and is called the generating function. Substituting our Hamiltonian to the afore-
mentioned equation, we get
p2r p2θ k ∂F
0= + 2
− + .
2m 2mr r ∂t
Interestingly, we can use the generating function to get a relationship between our momenta and our
generalized coordinates. Before we invoke these relationship, however, let us assume for simplicity that
our generating function is a composite of three other functions, isolatedly parameterized by the generalized
coordinates and time. Mathematically, this means that

F = F1 (r) + F2 (θ) + F3 (t) .

Now we note that the following relationships must exist:

∂F dF1
pr = =
∂r dr

∂F dF2
pθ = =
∂θ dθ

∂F dF3
= .
∂t dt
Knowing these relations, we nally substitute these to our previous equation and we now have our
Hamiltonian-Jacobi equation for the Keplerian system, which is
 2  2
1 dF1 1 dF2 k dF3
0= + − + .
2m dr 2mr2 dθ r dt

This is our rst landmark in our attempt to solve this problem. The next steps shall be attempts to nd
the composite functions of the generating function that we dened. To do that, we begin by isolating the
time derivative of our generating function and letting it be some constant β3 . Doing that will yield
 2  2
dF3 1 dF1 1 dF2 k
=− − + = β3
dt 2m dr 2mr2 dθ r
 dF3
= β3
dt

F3 (t) = β3 t.

We now have another very important answer indicated. Now that we have this, we then isolate one
composite function, in this case the one involving F2 and we once more equate it to be some constant, which
we shall call β2 . In undertaking this step, we will be left with
 2  2
1 dF1 1 dF2 k
β3 = − − +
2m dr 2mr2 dθ r

 2  2
dF2 dF1
= −r2 + 2mrk − 2mr2 β3 = β22
dθ dr

dF2
= β2
dθ

F2 (θ) = β2 θ.

Finally, we only have one derivative left in our expression. This means that we simply have some non-
linear rst order dierential equation that we have to deal with. We use the same procedure as above to be
able to solve the dierential equation.
 2
2 dF1
−r + 2mrk − 2mr2 β3 = β22
dr

2
β2

dF1 2mk
=− + 2mβ3 + 22
dr r r

ˆ r
2mk β2
F1 (r) = − + 2mβ3 + 22 dr.
r r
Interestingly, we now have expressions for all three composite functions of our generating function. This
means that our generating function is
ˆ r
2mk β2
F = − + 2mβ3 + 22 dr + β2 θ + β3 t.
r r
We now use the assertion that the generating function must be given as F = F (r, θ, Pr , Pθ , t), that is, it
must be a function of the old generalized coordinates, the new momenta and time. Noting this, we notice
that our derived generating function does not indicate any Pr or Pθ thus it must be that β2 and β3 are related
to these new momenta which are both constant. Using this we associate Pr with β3 and Pθ with β2 . Thus
we have
∂F ∂F
R= ∼
∂Pr ∂β3

∂F ∂F
Θ= ∼
∂Pθ ∂β2
 Consider rst the relation involving R and let us introduce the constnat γ1 since we have argued previously
that the generalized coordinates must be mapped to two constant quantities. Doing this we have
ˆ r !
∂F ∂ 2mk β2
Θ= = − + 2mβ3 + 22 dr + β2 θ + β3 t = γ1
∂β2 ∂β2 r r

ˆ r !
∂ 2mk β2
γ1 = − + 2mβ3 + 22 dr + θ
∂β2 r r

ˆ ˆ
β2 1 β2
q dr = √ q dr
r2 − 2mk
2
β2 2m r2
β22
− k
r + 2mβ3 + r2 β3 + 2mr 2 r

Before we succumb to the pleasures of the Integral Table, we rst use the convenient substitution u = 1/r
to make the form of our integrand more appealing.
ˆ ˆ
1 β2 1 β2
√ q dr = − √ q du
2m r2 β3 +
β22
− k 2m β3 +
β22 2
− ku
2mr 2 r 2m u

Now we note the integral of the standard form shown below and apply it to our integrand
ˆ  
1 1 2ax + b
√ dx = − √ sin−1 √
ax2 + bx + c a b2 − 4ac

  
ˆ β22

β2 1 −1 
 mrk − 1 
−√ q du = sin r 
2m β22 2
β3 + 2m u − ku
 
2β3 β22

1 − mk2

  
β22

mrk − 1 
γ1 = sin−1 

 + θ
r 

2β3 β22

1 − mk2

β22
 
mrk −1
sin (θ − γ1 ) = r  .
2β3 β22
1− mk2

Notice that since γ1 is merely a constant, we can let it be − π2 and doing so would give us
β22
 
1
−1
mk r
cos (θ) = r  .
2β3 β22
1 − mk 2

This is exactly the form that we want to relate our result to the desired result. In this case, we must
dene the following variables
β22
α=
mk
 s
2β3 β22
 
ε= 1−
mk 2
Doing so would give us the shorthand form of
α
r (θ) = .
1 + ε cos θ
This is our orbit equation in terms of r. We note the following insights from the problem. First would be
that β3 is merely the negative of the total energy which would concur with our results. This can be shown
explicitly by considering that our old Hamiltonian was argued to be the total energy of the system, let us
call that E thus
∂F
0=h+
∂t
dF3
0=E+ = E + β3
dt

E = −β3 .
Now consider the other condition, that
∂F ∂F
R= ∼ = γ2
∂Pr ∂β3
This tells us that
ˆ r !
∂ 2mk β2
γ2 = − + 2mβ3 + 22 dr + t
∂β3 r r
ˆ
m
γ2 − t = q dr.
β22
− 2mk
r + 2mβ3 + r2

If we let γ2 be equal to zero, then we'll have

ˆ
m
t=− q dr.
β22
− 2mk
r + 2mβ3 + r2

We then use the fact that we already have some value for expression for r previouls in terms of θ and we
merely. Using this transformation, we will have the expression
ˆ
β23 dθ
t= 2.
mk 2 (1 + ε cos (θ))
Now using the fact that
 
θ
1 + cos θ = 2 cos2 ,
2
we can transform our integral to the much more convenient form of
ˆ
β23
 
4θ
t= sec dθ.
4mk 2 2
Notice that this integral form is much easier to evaluate and the result would actually be
β23
 
1
t= tan θ + tan3 θ .
2mk 2 3
We can now settle with the fact that this form actually preserves the form shown in Goldstein for the
inter-dependence of θ and time. Further, we can actually write out an expression that expicitly gives us an
r (t) for which a more involved mathematical treatment is necessary.