Numerical methods for solving ordinary

differential equations
Some methods will be discussed in this course:

For first-order differential equations given the initial condi-

tion, for example:
y ′ (x) = x3 − x y (x) + y 2 (x) , x ∈ (0, 1) ,
{
y (0) = 5.3,
we discuss here three methods to obtain approximate solutions:
• Euler’s method.
• Improved Euler’s method.
• Runge-Kutta method.

For the linear second-order boundary-value problem, for ex-

ample:
x3 y” (x) − x y ′ (x) + 2 y (x) = x2, x ∈ (0, 2) ,
{
y (0) = 0.3, y (2) = 1.2,
we consider:
• The finite-difference method.

All methods here are based on the same technique:

using Taylor expansion.

1
 Example. Let y (x) be a function satisfying
2
y ′ (x) = x3 − x y (x) + (y (x)) , x ∈ (0, 1) ,
(⋆) {
y (0) = 2.2.
Find an approximation for y (0.1).
There are three methods discussed here:
• Euler’s method.
• Improved Euler’s method.
• Runge-Kutta method.

For simplicity, the equation (⋆) can be written as

y ′ (x) = f (x, y (x)) , x ∈ (0, 1) ,
{
y (0) = 2.2.
Here
f (s, t) ∶= s3 − st + t2.

First, let’s divide the interval [0, 1] into 10 equal intervals, so

the step size is h = 0.1.

x0 x1 x2 x3 x10

0.1

2
Euler’s method.

y (0.1) = y (x1) = y (x0 + h)

(First-order Taylor expansion) ≈ y (x0) + h y ′ (x0)

≈ 2.2 + 0.1 f (0, 2.2)

≈ 2.6840 .

We can continue this process to find approximate values for

y (0.2), y (0.3) , . . .:

y (0.2) = y (x2)

≈ y (x1) + h f (x1, y (x1))

≈ 2.6840 + 0.1 f (0.1, 2.6840)

≈ 3.3777.

3
Euler’s method.

Given the equation:

y ′ (x) = f (x, y (x)) , x ∈ (a, b) ,
{
y (a) = α.

Consider the Euler’s method with step size h, that is

x0 ∶= a, x1 = x0 + h, . . . , xk = x0 + hk, . . .

Define the sequence {yk } by:

y0 = α,
{
yk+1 = yk + h f (xk , yk ) , for k ≥ 0.

Then we have the approximation:

y (xk ) ≈ yk

Example. Let y (x) be a function satisfying

2
y ′ (x) = x3 − x y (x) + (y (x)) , x ∈ (0, 1) ,
{
y (0) = 2.2.
Use the Euler’s method with step size h = 0.1 to find an
approximation for y (0.5).

4
Answer.
y (0.5) ≈ 10.0294

Implementation:

Y = Y + 0.1 (X 3 − XY + Y 2) ∶ X = X + 0.1

5
 Improved Euler’s method
The idea of the improved Euler’s method: instead of using
first-order Taylor expansion, we consider second-order Taylor
expansion:
Example. Let y (x) be a function satisfying
2
y ′ (x) = x3 − x y (x) + (y (x)) , x ∈ (0, 1) ,
(⋆) {
y (0) = 2.2.
Find an approximation for y (0.1).

Improved Euler’s method:

Assume that the step size is h = 0.1, then

y (0.1) = y (x1) = y (x0 + h)

′ 1 2
≈ y (x0) + hy (x0) + h y” (x0)
2

f (x0, y (x0)) + f (x1, y0 + hf (x0, y (x0)))

≈ y (x0) + h
2

f (0, 2.2) + f (0.1, 2.2 + 0.1 f (0, 2.2))

≈ 2.2 + 0.1
2

≈ 2.7888.
6
 Improved Euler’s method.

Given the equation:

y ′ (x) = f (x, y (x)) , x ∈ (a, b) ,
{
y (a) = α.

Consider the improved Euler’s method with step size h,

that is
x0 ∶= a, x1 = x0 + h, . . . , xk = x0 + hk, . . .

Define the sequence {yk } by:

⎧
⎪y0
⎪
⎪ = α,
⎨ f (xk , yk ) + f (xk+1, yk + h f (xk , yk ))
⎪
⎪
⎪ yk+1 = yk + h , for k ≥ 0.
⎩ 2

Then we have the approximation:

y (xk ) ≈ yk

Example. Let y (x) be a function satisfying

2
y ′ (x) = x3 − x y (x) + (y (x)) , x ∈ (0, 1) ,
{
y (0) = 2.2.
Use the improved Euler’s method with step size h = 0.1 to
find an approximation for y (0.5).
7
Answer.
y (0.5) ≈ 39.5783

Implementation:

A = X + 0.1 ∶ B = Y + 0.1 (X 3 − XY + Y 2) ∶
0.1
Y =Y + (X 3 − XY + Y 2 + A3 − AB + B 2) ∶ X = A
2

8
Runge-Kutta method.

Given the equation:

y ′ (x) = f (x, y (x)) , x ∈ (a, b) ,
{
y (a) = α.

Consider the Runge-Kutta method with step size h,

that is
x0 ∶= a, x1 = x0 + h, . . . , xk = x0 + hk, . . .

Define the sequence {yk } by:

⎧
⎪y0
⎪ = α,
⎨ 1
⎪
⎪ y k +1 = y k + (Ak + 2Bk + 2Ck + Dk ) , for k ≥ 0,
⎩ 6
where
Ak ∶= h f (xk , yk ) ,

h Ak
Bk ∶= h f (xk + , yk + ) ,
2 2

h Bk
Ck ∶= h f (xk + , yk + ) ,
2 2

Dk ∶= h f (xk + h, yk + Ck ) .

Then we have the approximation:

y (xk ) ≈ yk
9
 Idea behind the Runge-Kutta formula: similar to the Euler’s
method or the improved Euler’s method, we also apply the Tay-
lor expansion. However, for the Runge-Kutta method, we use
Taylor expansion of order 4.

Example. Let y (x) be a function satisfying

y ′ (x) = sin (x) + x2 y (x) , x ∈ (1, 3) ,
{
y (1) = 0.3.
Use the Runge-Kutta method with step size h = 0.1 to
find an approximation for y (1.1).

10
 Answer.

h = 0.1, x0 = 1, y0 = 0.3, f (x, y ) = sin (x) + x2y.

Since y (1.1) = y (x1) ≈ y1, therefore our objective is to
find y1.
We have
1
y1 = y0 + (A0 + 2B0 + 2C0 + D0)
6

A0 = h f (x0, y0) = 0.1 f (1, 0.3) → STO A,

h A0 0.1 A
B0 = h f (x0 + , y0 + ) = 0.1 f (1 + , 0.3 + ) → STO B,
2 2 2 2
h B0 0.1 B
C0 = h f (x0 + , y0 + ) = 0.1 f (1 + , 0.3 + ) → STO C,
2 2 2 2
D0 = h f (x0 + h, y0 + C0) = 0.1 f (1 + 0.1, 0.3 + C ) → STO D.

Thus
1
y1 = 0.3 + (A + 2B + 2C + D) = 0.4268.
6

11
 Linear second-order boundary-value problem:

p (x) y” (x) + q (x) y ′ (x) + r (x) y (x) = f (x) , a < x < b,

{
y (0) = α, y (b) = β.

• The method: the finite-difference method (the behind idea

is still Taylor expansion).
• We use Taylor expansion to approximate the derivatives: y ′,
y”. In the end, we transfer the differential equation into a
system of linear equations.

Assume that the step size is h.

x0 x1 x2 x3 ⋯ xn

First, let’s approximate y ′ (xk ) and y” (xk ):

y (xk+1) = y (xk + h) ≈ y (xk ) + hy ′ (xk ) (Taylor expansion) ,
y (xk−1) = y (xk − h) ≈ y (xk ) − hy ′ (xk ) (Taylor expansion) .

12
 ⇓

y (xk+1) − y (xk−1)
y ′ (xk ) ≈ (that’s why we call this method:
2h
finite-difference method).

Similarly, let’s use Taylor expansion of order 2 to approxi-

mate y” (xk ):
1
y (xk+1) = y (xk + h) ≈ y (xk ) + hy ′ (xk ) + h2y” (xk ) ,
2
′ 1 2
y (xk−1) = y (xk − h) ≈ y (xk ) − hy (xk ) + h y” (xk ) .
2
⇓

y (xk+1) − 2y (xk ) + y (xk−1)

y” (xk ) ≈ 2
.
h

For simplicity, let’s denote yk for y (xk ).

Also, pk for p (xk ), qk for q (xk ), rk for r (xk ), and fk for f (xk ).
It’s time to insert our approximations into the equation, we
get that:

13
 yk+1 − 2yk + yk−1 yk+1 − yk−1
pk 2
+ q k + rk yk ≈ fk .
h 2h

This is equivalent to:

p k qk 2pk p k qk
( 2 − ) yk−1+(rk − 2 ) yk +( 2 + ) yk+1 ≈ fk , 1 ≤ k ≤ n−1.
h 2h h h 2h

Remember that we also have the boundary conditions:

y0 = α, yn = β.
From here we arrive that a system of n − 1 equations with n − 1
unknowns: y1, y2, . . . , yn−1:

⎧
⎪ pk qk 2pk p k qk
⎪
⎪
⎪ ( 2 − ) yk−1 + (rk − 2 ) yk + ( 2 + ) yk+1 ≈ fk ,
⎪ h 2h
⎪
⎪ h h 2h
1 ≤ k ≤ n − 1,
⎨
⎪
⎪
⎪
⎪
⎪
⎪
⎩y0 = α, yn = β.
⎪

Remark. Even though this is not a real system (since we

have the approximate signs here, not the equal signs), however,
solving this system would give you approximation values for
y (xk ), that is
y (xk ) ≈ yk .

14
Example. Consider the linear second-order boundary-value
problem:
(x + 0.23) y” + x2y ′ − 1.23 y = 0.29x (x + 2) , 1 < x < 1.8,
{
y (1) = 0.27, y (1.8) = 1.67.
Use the finite-difference method with step size h = 0.2 to
approximate the values of y (1.2) , y (1.4) , y (1.6).
Answer. We have

p (x) = x + 0.23, q (x) = x2, r (x) = −1.23,

f (x) = 0.29x (x + 2) .

α = 0.27, β = 1.67, h = 0.2,

x0 = 1, x1 = 1.2, x2 = 1.4, x3 = 1.6 , x4 = 1.8.

Objective Find y1, y2, y3 by solving the system:

⎡ r − 2p21 p12 + q1 0 RRR f − α ( p1 − q1 ) ⎤
⎢ 1 h h 2h R 1 h2 2h ⎥
⎢ p2 q2 2p2 p2 q2 RRR ⎥
⎢ h2 − 2h r2 − h2 h2 + 2h RR f2 ⎥.
⎢ p3 q3 2p3 R R p 3 q 3
⎥
⎢ 0
⎣ h2
− 2h r3 − h2 RRR f3 − β ( h2 + 2h ) ⎥⎦
Here
pk = p (xk ) , rk = r (xk ) , qk = q (xk ) , fk = f (xk ) .
We get that
y1 = 0.6504, y2 = 1.0100, y3 = 1.3498.
15
Therefore
y (1.2) ≈ 0.6504, y (1.4) ≈ 1.0100, y (1.6) ≈ 1.3498.

16
 Exercise.
1. Consider the Cauchy equation:
y ′ (x) = x2 + x y 3, x > 1,
{
y (1) = 0.7.

Use the Euler’s method with step size h = 0.2 to find an ap-
proximation for y (1.8).
2. Consider the Cauchy equation:
y ′ (x) = sin (x) + x2 y, x > 1,
{
y (1) = 0.3.

Use the improved Euler’s method with step size h = 0.1 to

find an approximation for y (2.0).
3. Consider the Cauchy equation:
y ′ (x) = y 2 + x3 y, x > 1,
{
y (1) = 0.7.

Use the Runge-Kutta method with step size h = 0.2 to find an

approximation for y (1.2).
4. Consider the linear second-order boundary-value problem:
(x2 + 1) y” + 5x y ′ − 10 y = −8x2, 1.4 < x < 1.8,
{
y (1.4) = 0, y (1.8) = 0.8.

Use the finite-difference method with step size h = 0.1 to ap-

proximate the values of y (1.5) , y (1.6) , y (1.7).

17
 Answer.
1. y (1.8) ≈ 10.0387.
2. y (2.0) ≈ 7.6538.
3. y (1.2) ≈ 1.0879.
4. y (1.5) ≈ 0.3416, y (1.6) ≈ 0.5722, y (1.7) ≈ 0.7190.

18