Picard’s Method - Project

1 Derivation of the Method

If we start with the general, first order nonlinear ordinary differential equation

dy


f x y y x0  y0
dx
we can integrate both sides of the equation for x0  x  x1 to get

x1 x1


y x1  y x0   dy
 f x y x dx
x0 x0

Using t as the variable for integration, instead of x, and setting x x1 to be the

arbitrary final point

x


y x y x0    f t  y t  dt (1)
x0

Although a correct representation of the solution, this is not completely satisfatory.

We
proceed to try to solve this by iteration. Substituting an initial guess of y x
φ0 x into the right hand side of (1), we get

x


φ1 x  y x0    f t  φ0 t  dt (2)
x0



In the unlikely event that φ1 x φ0 x , we are finished, we have found a solution.
Otherwise, we keep on iterating:

x


φn  1 x y x0   f t  φn t  dt
x0


If no further information on the solution is available, we normally choose φ0 x
y0 .

1
1.1 Linear Example



To start off with a simple example let f x y y, and y 0 1 so we are trying to
solve:

dy


y x y 0 1
dx

In this case, we actually know the solution:

y x ex


Now let’s see what Picard’s method is doing, given the initial guess φ0 x 1.

x
x
φ1 x  1  φ0 t  dt 1  1dt 1 x
0 0

x
x 1 2
φ2 x  1  φ1 t  dt 1  1  tdt 1 x x
0 0 2

x
x 1 2 1 2 1 3
φ3 x  1  φ2 t  dt 1  1 t  t dt 1 x x  x
0 0 2 2 6

x
x 1 2 1 3 1 2 1 3 1 4
φ4 x  1  φ3 t  dt 1  1 t  t  t dt 1 x x  x  x
0 0 2 6 2 6 24

These are precisely the first four terms in the Maclaurin (Taylor) series expansion
for ex about x 0.

2
1.2 Non linear Example with Unique Solution

For the next case, consider

dy


x  y2 x y 0 0
dx
Although this can be solved (see attached Maple worksheet), it is in terms of very
complicated functions! A
simpler method for getting an approximate solution is
Picard’s method. With φ0 x 0

x

x 1 2
φ1 x  t  φ20 t  dt  tdt x
0 0 2


x


x

1 2 2 1 2 1 5
φ2 x  t  φ21 t  dt t t   dt x  x
0 0 2 2 2


x


x

1 2 1 5 2 1 2 1 5 1 8 1 11
φ3 x  t  φ22 t  dt t t  t   dt x  x  x  x
0 0 2 20 2 20 160 4400

The iterations rapidly converge, as seen below:

0.5

0.4

0.3

0.2

0.1

0 0.2 0.4 0.6 0.8 1

3
1.3 Non linear Example with Non Unique Solution

dy


3  y x 3  2  y 2 0
dx


If we start with φ0 x 0 then the iterations satisfy

x
x
φ1 x  3  φ0 t  3  2dt  0dt 0
2 2

and

x
x
φ2 x   3  φ1 t  2  3dt  0dt 0
2 2


Each iterate is identically 0. This means that y x 0 is a solution of the initial
value problem. Is it the only one, however?


If we start with an alternative initial guess, φ0 x x  2, then we have

x
x 9
φ1 x   3  φ0 t  2  3dt  3  t  2 2  3 dt  x  2 
5 3
2 2 5

x
x 9
x 9

φ2 x  3  φ1 t  2  3dt  3 t  2 5  3  2  3 dt
 3   2  3  t  2 10  9
 dt
2 2 5 2 5

Although this looks complicated (and it is!) we can gain some insight by setting

φn x  cn  x  2 rn

which means that

x
φn  1 x
 3  cn  t  2 rn  2  3 dt
2

4
 x
2 3
3cn  t  2 2rn  3 dt
2
2 3
cn
3  x  2 2rn  3 1
2rn  3  1
cn 1 x  2 rn 1

2 3
if rn 1 1  2rn  3 and cn 1 3 2rncn3 1

If rn  r exists, then r must satisfy r 1  2r  3, which implies r 3. If the limit

of cn exists, then c must satisfy c c2  3 , which means that c 0  1

c 0 gives

the trivial solution, so we take c 1. The limit of the iterations is
therefore x  2 3 .