Computation and Programming in Physics

Euler & Runge-Kutta Methods

Syeilendra Pramuditya

Physics Program
Institut Teknologi Bandung
1
Grid-based Computation
 Euler-type Methods
 Differential equations
 Temporal domain (variation in time)
 Initial conditions
 Finite Difference Methods
 Differential equations
 Spatial domain (variation in space)
 Boundary conditions: Von Neumann VS
Dirichlet
 3
Differential Equation

dy 2
y( x)   f ( x, y )  x  1
dx
y ( x  0)  0

y ( x)  ..?
 4
Differential Equation
dy
y( x)   f ( x, y )  x 2  1
dx
y ( x  0)  0

Exact Solution
2
   1)dx
dy  ( x

1 3
y  x  xC
3
y ( x  0)  0  C  0
1 3
y ( x )  f ( x, y )  x  x
3
 5
Differential Equation

dy 1
y( x)   f ( x, y ) 
dx x
y ( x  1)  0

y ( x)  ..?
 6
Differential Equation
dy 1
y( x)   f ( x, y ) 
dx x
y ( x  1)  0

Exact Solution
1
 dy   x dx
y  ln( x)  C
y ( x  1)  0  C  0
y ( x)  f ( x, y )  ln( x)
 7
Differential Equation

dy 2 1
y( x)   f ( x, y )  x   1
dx y
y ( x  0)  0

y  ..???
 Euler Methods
 Linear approx. of Taylor series
 “Simple Euler”

y ( x)  y ( x0 )
y( x0 )   f ( x0 , y0 )
x  x0
f ( x)  
y ( x)  y ( x0 )  x  x0 y( x0 )
x  x0  h
y ( x0  h)  y ( x0 )  hf ( x0 , y0 )
f ( x0 ) y ( x0  h)  y ( x0 )  hf ( x0 , y ( x0 ))
y ( x0  h)  y0  hf 0

x0 x
 9
Ordinary Differential Equation (ODE)

dy 1 y ( x0  h)  y ( x0 )  hf ( x0 , y0 )
y( x)   f ( x, y ) 
dx x y ( x0  h)  y0  hf 0
y ( x  1)  0 ynew  yold  hf ( xold , yold )

 Solve the above ODE numerically (find y(x)) using Simple

y ( x0  h)  y ( x0 )  hf ( x0 , y0 )
y ( x0  h)  y0  hf 0
ynew  yold  hf ( xold , yold )
Exact VS Numeric

2.5

1.5

1
Exact
Euler, h = 1.0
0.5
Euler, h = 0.5

0
0 2 4 6 8 10
 Euler Methods
 Linear approx. of Taylor series
 “Modified Euler” Use more correct gradient

 
y ( x)  y ( x0 )  x  x0 y( xmid )
y ( x0  h)  y ( x0 )  hf ( xmid , ymid )
f ( x) xmid  x0 
h
2
h
ymid  y ( x0  )
2
h
f ( x0 ) ymid  y0  f 0
2
y ( x0  h)  y0  hf mid

x0 x
 Euler Methods
 Linear approx. of Taylor series
 “Improved Euler” Use more correct gradient

y( x0 )  f ( x0 , y0 )
y( x)  f ( x, y )  y0  hf 0
f ( x) y ( x0 )  y( x)
 ( x) 
yavg
2
f ( x0 , y0 )  f ( x, y )
f avg 
2
f ( x0 ) f ( x, y )  f ( x0  h, y0  hf ( x0 , y0 ))
y ( x0  h)  y ( x0 )  hf avg

x0 x
4th Order Runge-Kutta (RK4)
Exact VS Numeric

2.5

1.5

1
Exact
Euler, h = 1.0
0.5
Euler, h = 0.5

0
0 2 4 6 8 10
Exact VS Numeric
 Simple Euler
Exact VS Numeric
 Modified Euler
Exact VS Numeric
 Improved Euler
Exact VS Numeric
 RK4
 20
Block-Spring System

F  ma
dv F k
a   x
dt m m
dv k k
 x  v(t )  f ( x, t )  x(t )
dt m m
dx
 v  x(t )  f (v, t )  v(t )
dt
 21
Block-Spring System
y ( x0  h)  y ( x0 )  hf ( x0 , y0 )
y ( x0  h)  y0  hf 0

1 1
v  v0  at  v(t )  v0  v(t )t  v(t )  v(t )  v0  f (v, t )
t t
dv k
   
v t0    v t0  
dt t 0
 v0   x0
m

1 1
x  x0  vt  x(t )  x0  x(t )t  x(t )  x(t )  x0  f ( x, t )
t t
dx
   
x t0    x t0  
dt t 0
 x0   v0

1 2 1 2
E mv  kx
2 2
 22
Block-Spring System (Simplified)
y ( x0  h)  y ( x0 )  hf ( x0 , y0 )
dy
y ( x0  h)  y0  hf 0  y0  h
dx

dv F kx0
v(t0  t )  v0  t  v0  at  v0  t  v0  t
dt m m
k
v(t0  t )  v0  x0 t
m
dx
x(t0  t )  x0  t  x0  v0 t
dt
1 2 1 2
E  mv  kx
2 2
 23
Write a Simple Euler Code
 m=k=1
 0 < t < 20 sec
 dt = 0.2
 x(t=0) = 1
 v(t=0) = 0
 Calculate x(t) and v(t) using Simple Euler
 Output: time,x(t),x_exact,v(t),v_exact,Em

dv F kx0
v(t0  t )  v0  t  v0  at  v0  t  v0  t
dt m m
k
v(t0  t )  v0  x0 t
m
y ( x0  h)  y ( x0 )  hf ( x0 , y0 )
dx
x(t0  t )  x0  t  x0  v0 t dy
dt y ( x0  h)  y0  hf 0  y0  h
1 2 1 2 dx
E  mv  kx
2 2
Simple Euler Code

k
v(t0  t )  v0  x0 t
m
x(t0  t )  x0  v0 t
1 2 1 2
E mv  kx
2 2
 25
Simple Euler: Position
8

x(t) numeric
6
x(t) exact

-2

-4

-6
0 5 10 15 20
 26
Simple Euler: Velocity
8

v(t) numeric
6
v(t) exact

-2

-4

-6

-8
0 5 10 15 20
Simple Euler: Energy
30

E kinetic
25 E potential
E mechanic
20

15

10

0
0 5 10 15 20
 28
Simple Euler: Energy
Energy
1.6
dt = 0.2
1.4
dt = 0.1
1.2

1
Energy [J]

0.8

0.6

0.4

0.2

0
0 1 2 3 4 5 6
Time [s]
 29
Modified Euler: Energy
Energy
1.6
dt = 0.2, Simple Euler
1.4
dt = 0.2, Modified Euler
1.2

1
Energy [J]

0.8

0.6

0.4

0.2

0
0 1 2 3 4 5 6
Time [s]
 30
Damped Block-Spring System

F  ma  kx  cv
dv F kx  cv
a  
dt m m
m  k 1
dv
  x  cv
dt
dx
v
dt
 31
Damped Block-Spring System
1 1
v  v0  at  v(t )  v0  v(t )t  v(t )  v(t )  v0  f (v, t )
t t
dv
   
v t0    v t0  
dt t 0

 v0   x0  cv 

1 1
x  x0  vt  x(t )  x0  x(t )t  x(t )  x(t )  x0  f ( x, t )
t t
dx
   
x t0    x t 0  
dt t 0
 x0   v0

1 2 1 2
E  mv  kx
2 2
 32
Damped Oscillation
 m=k=1
 0 < t < 20 sec
 dt = 0.2
 x(t=0) = 1
 v(t=0) = 0
 C = 0.15
 Calculate x(t) and v(t) using Simple Euler
 Output: time,x(t),x_exact,v(t),v_exact,Em
dv F kx0  cv0
v(t0  t )  v0  t  v0  at  v0  t  v0  t
dt m m
kx0  cv0
v(t0  t )  v0  t
m
y ( x0  h)  y ( x0 )  hf ( x0 , y0 )
dx
x(t0  t )  x0  t  x0  v0 t dy
dt y ( x0  h)  y0  hf 0  y0  h
1 2 1 2 dx
E  mv  kx
2 2
Damped Oscillation: Exact Solution
 Halliday-Resnick, Chapter 15
 34
Damped Oscillation (Mod. Euler)
1.2

0.8

0.6

0.4
Position

0.2

0
0 5 10 15 20
-0.2

-0.4

-0.6

-0.8

-1
TIme