Runge-Kutta Methods

for Ordinary Differential Equations

Initial Value Problems

(ODE-IVP )
dyi
= f ( x , yi )
dx
where initial condition is given: yi= ai at x = 0

1 step methods:
yi +1 = yi + φ ( xi , yi , h)h
φ = a1k1 + a2 k 2 + L + an k n (increment function)

1st order: yi +1 = yi + k1h Euler’s method

k1 = f ( xi , yi )
1 1
2nd order: yi +1 = yi + ( k1 + k 2 )h Heun method with a Single corrector
2 2
k1 = f ( xi , yi ) k 2 = ( xi + h, yi + h)

yi +1 = yi + k 2 h Midpoint corrector
1 1
k 2 = ( xi + h, yi + k1h)
k1 = f ( xi , yi ) 2 2
1 2
yi +1 = yi + ( k1 + k 2)h Raltson’s method
3 3 3 3
k1 = f ( xi , yi ) k 2 = f ( xi + h, yi + k 1h)
4 4
Comparison of three 2nd-order RK methods with the true solution

Runge Kutta
1
3rd order: yi +1 = yi + (k1 + 4k 2 + k3 )h
6
k1 = f ( xi , yi )
1 1
k 2 = f ( xi + h, yi + k1h)
2 2
k3 = f ( xi + h, yi − k1h + 2k 2 h)

Classical RK
4th order: 1
yi +1 = yi + (k1 + 2k 2 + 2k3 + k 4 )h
6
k1 = f ( xi , yi )
1 1
k 2 = f ( xi + h, yi + k1h)
2 2
1 1
k3 = f ( xi + h, yi + k2 h)
2 2
k 4 = f ( xi + h, yi + k3h)
 4th order Runge Kutta

Butcher’s
5th order: 1
yi +1 = yi + (7k1 + 32k3 + 12k 4 + 32k5 + 7 k6 )h
90

k1 = f ( xi , yi )
1 1
k 2 = f ( xi + h, yi + k1h)
4 4
1 1 1
k3 = f ( xi + h, yi + k1h + k 2 h)
4 8 8
1 1
k 4 = f ( xi + h, yi − k2 h + k3h)
2 2
3 3 9
k5 = f ( xi + h, yi + k1h + k 4 h)
4 16 16
3 2 12 12 8
k6 = f ( xi + h, yi − k1h + k2 h + k3h − k 4 h + k5 h)
7 7 7 7 7
 f(x,y)=4e0.8x-0.5y

b−a
Effort = n f
h

nf = no. function

System of equations
dy1
= f1 ( x, y1 , y2 , L , yn )
dx
dy2
= f 2 ( x, y1 , y2 , L , yn )
dx

M
dyn
= f n ( x, y1 , y2 , L , yn )
dx

Example: kinetic reactions

 Example 25.10:
Use 4th RK method to solve the ODEs below:
dy1 dy2
= −0.5 y1 = 4 − 0.3 y2 − 0.1y1
dx dx
assuming x = 0, y1=4 and y2 = 6. Integrate to x = 2 with h = 0.5
Solution:
k1 = f ( xi , yi )

1 1
k 2 = f ( xi + h, yi + k1h)
2 2

1 1
k3 = f ( xi + h, yi + k2 h)
2 2

k 4 = f ( xi + h, yi + k3h)
 1
yi +1 = yi + (k1 + 2k 2 + 2k3 + k 4 )h
6

Proceeding in a like manner for the remaining steps:

x y1 y2

0 4 6

0.5 3.115234 6.857670

1.0 2.426171 7.632106

1.5 1.889523 8.326886

2.0 1.471577 8.946865

4th order RK method for system

 ODE with abrupt change:
Adaptive Runge-Kutta Methods

Adapt to the step size

trajectory
– adaptive step-size control

Require local truncation

error at each step.

Basis for lengthening or

decreasing the step
size.

Adaptive RK/step-halving method

Taking each step twice, a full step and 2 half steps.

Then, calculate the difference between 2 results

∆ = y2 − y1
For 4th order RK, the correction is

∆
y2 ← y2 +
15
This estimate is fifth-order accurate.
Example 25.12: (Adaptive RK)

Use adaptive 4th RK method to integrate y’ from x = 0 to x = 2

y ' = 4e 0.8 x − 0.5 yi
step size =2, initial condition is y(0) = 2 (true solutions is y (2) = 14.84392 )
Solution:
- Single prediction with a step of h is computed as
1
y (2) = 2 + (k1 + 2k 2 + 2k3 + k 4 )h = 15.10584
6
- Two half step predictions are:
1
y (1) = 2 + ( k1 + 2k 2 + 2k3 + k 4 )h = 6.20104
6
1
y (2) = 6.20104 + (k1 + 2k 2 + 2k3 + k 4 )h = 14.86249
6
- The approximate error is
14.86249 − 15.10584
Ea = = −0.01622
15
- The true error is Et = 14.84392 − 14.86249 = −0.01857
- Corrected prediction y ( 2) = 14.86249 − 0.01622 = 14.84627
- with true error: Et = 14.84392 − 14.84627 = −0.00235

Runge-Kutta Fehlberg/embedded RK/Cash-Karp RK method

37 250 125 512
4th order est.: yi +1 = yi + ( k1 + k2 + k3 + k 4 )h
378 621 594 1771
2825 18,575 13,525 277 1
5th order est.: yi +1 = yi + ( k1 + k3 + k4 + k5 + k 6 ) h
27,648 48,384 55,296 14,336 4
where k1 = f ( xi , yi )
1 1
k 2 = f ( xi + h, yi + k1h)
5 5
3 3 9
k3 = f ( xi + h, yi + k1h + k2 h)
10 40 40
3 3 9 6
k 4 = f ( xi + h, yi + k1h − k 2 h + k3 h)
5 10 10 5
11 5 70 35
k5 = f ( xi + h, yi − k1h + k 2 h − k3h + k 4 h)
54 2 27 27
7 1631 175 575 44,275 253
k6 = f ( xi + h, yi − k1h + k2h + k3 h − k4 h + k5 h)
8 55,296 512 13,824 110,592 4096
Lower order

- ODE is solved by 5th order estimation.

- The error estimation :

Ea = 5th-order prediction yi - 4th-order prediction yi

Refer to example 25.13.
37 250 125 512
4th order est.: y1 = yi + ( k1 + k2 + k3 + k 4 )h = 14.83192
*

378 621 594 1771

2825 18,575 13,525 277 1
5th order est.: y1 = yi + ( k1 + k3 + k4 + k5 + k6 )h
27,648 48,384 55,296 14,336 4
= 14.83677
Ea = 14.83677 − 14.83192 = 0.004842

Step size control

α
∆ new
hnew = h present
∆ present
h = step size
∆ present = the computed present accuracy
∆ new = the desired accuracy
0.2 when ∆ present ≤ ∆ new (h increase)
α=
0.25 when ∆ present > ∆ new (h decrease)

∆ new = ε yscale ε = an overall tolerance

dy
yscale = y + h
dx