Jim Lambers

MAT 461/561
Spring Semester 2009-10
Lecture 25 Notes
These notes correspond to Sections 11.1 and 11.2 in the text.

The Shooting Method for Two-Point Boundary Value Problems

We now consider the two-point boundary value problem (BVP)
= (, , ),

< < ,

a second-order ODE, with boundary conditions

() = ,

() = .

This problem is guaranteed to have a unique solution if the following conditions hold:
, , and are continuous on the domain
= {(, , ) , < < , < < }.
> 0 on
is bounded on .
There are several approaches to solving this type of problem. The rst method that we will
examine is called the shooting method. It treats the two-point boundary value problem as an initial
value problem (IVP), in which plays the role of the time variable, with being the initial time
and being the nal time. Specically, the shooting method solves the initial value problem
= (, , ),

< < ,

with initial conditions

() = ,

() = ,

where must be chosen so that the solution satises the remaining boundary condition, () = .
Since , being the rst derivative of () at = , is the initial slope of the solution, this
approach requires selecting the proper slope, or trajectory, so that the solution will hit the
target of () = at = . This viewpoint indicates how the shooting method earned its name.
Note that since the ODE associated with the IVP is of second-order, it must normally be rewritten
1

as a system of rst-order equations before it can be solved by standard numerical methods such as
Runge-Kutta or multistep methods.
In the case where = (, , ) is a linear ODE, selecting the slope is relatively simple. Let
1 () be the solution of the IVP
= (, , ),

< < ,

() = ,

() = 0,

< < ,

() = 0,

() = 1.

and let 2 () be the solution of the IVP

= (, , ),

These can be computed using any analytical or numerical method. Then, the solution of the original
BVP has the form
() = 1 () + 2 ()
where is the correct slope, since any linear combination of solutions of the ODE also satises the
ODE, and the initial values are linearly combined in the same manner as the solutions themselves.
To nd the proper value of , we evaluate (), which yields
() = 1 () + 2 () = ,
and therefore = ( 1 ())/2 (). It follows that as long as 2 () = 0, then () is the unique
solution of the original BVP. This condition is guaranteed to be satised due to the previously
stated assumptions about (, , ) that guarantee the existence and uniqueness of the solution.
If the ODE is nonlinear, however, then satises a nonlinear equation of the form
(, ) = 0,
where (, ) is the value of the solution, at = , of the IVP specied by the shooting method, with
initial sope . This nonlinear equation can be solved using an iterative method such as the bisection
method, xed-point iteration, Newtons Method, or the Secant Method. The only dierence is that
each evaluation of the function (, ), at a new value of , is relatively expensive, since it requires
the solution of an IVP over the interval [, ], for which () = . The value of that solution at
= is taken to be the value of (, ).
If Newtons Method is used, then an additional complication arises, because it requires the
derivative of (, ), with respect to , during each iteration. This can be computed using the fact
that (, ) = (, )/ satises the ODE
= + ,

< < ,

(, ) = 0,

(, ) = 1,

which can be obtained by dierentiating the original BVP and its boundary conditions with respect
to . Therefore, each iteration of Newtons Method requires two IVPs to be solved, but this extra
eort can be oset by the rapid convergence of Newtons Method.
2

Suppose that Eulers method,

y+1 = y + f (, y , ),
for the IVP y = f (, y), y(0 ) = y0 , is to be used to solve any IVPs arising from the Shooting
Method in conjunction with Newtons Method. Because each IVP, for (, ) and (, ), is of
second order, we must rewrite each one as a rst-order system. We rst dene
2 = ,

1 = ,

1 = ,

2 = .

We then have the systems

1

= 2,
= (, 1 , 2 ),
= 2,
= (, 1 , 2 ) 1 + (, 1 , 2 ) 2 ,

with initial conditions

1 () = ,

2 () = ,

1 () = 0,

2 () = 1.

The algorithm then proceeds as follows:

Choose (0)
Choose such that = , where is the number of steps
for = 0, 1, 2, . . . until convergence do
= 0, 01 = , 02 = () , 01 = 0, 02 = 1
for = 0, 1, 2, . . . , 1 do
= +
1
+1
= 1 + 2
2
= 2 + ( , 1 , 2 )
+1
1
+1 = 1 + 2
2
+1
= 2 + [ ( , 1 , 2 )1 + ( , 1 , 2 )2 ]
end
1 )/ 1
(+1) = () (

end
Changing the implementation to use a dierent IVP solver, such as a Runge-Kutta method or
multistep method, in place of Eulers method only changes the inner loop.
3