Scribd
Upload
52 views2 pages

MAT 461/561: 12.2 One-Step Methods, Cont'd

This document discusses different numerical methods for solving initial value problems (IVPs) where y' = f(t,y). It introduces Euler's method and higher-order Runge-Kutta methods, including derivations of the midpoint method and explicit trapezoidal method. It also covers implicit methods like backward Euler and the trapezoidal method. MATLAB's suite of ODE solvers is mentioned as using mostly high-order methods.

Uploaded by

Debisa
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
52 views2 pages

MAT 461/561: 12.2 One-Step Methods, Cont'd

This document discusses different numerical methods for solving initial value problems (IVPs) where y' = f(t,y). It introduces Euler's method and higher-order Runge-Kutta methods, including derivations of the midpoint method and explicit trapezoidal method. It also covers implicit methods like backward Euler and the trapezoidal method. MATLAB's suite of ODE solvers is mentioned as using mostly high-order methods.

Uploaded by

Debisa
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

MAT 461/561: 12.

2 One-Step Methods, cont’d

James V. Lambers

April 1, 2020

Announcements

One-Step Methods
For initial value problem y 0 = f (t, y):
Euler’s method: error is O(h), where h is the time step size–convergence is slow. Can we obtain
higher-order accurate methods, error O(hp ), p > 1?
Can use more terms in Taylor expansion, except must approximate derivatives of f (t, y)

Solving IVPs in MATLAB


MATLAB has a suite of ODE solvers, using mostly high-order methods. See MATLAB diary.

Runge-Kutta Methods
Runge-Kutta methods are high-order time-stepping methods that use several evaluations of f (t, y)
at various points, for tn ≤ t ≤ tn+1 .
Sample derivation: assume a method of the form

yn+1 = yn + ahf (tn , yn ) + bhf (tn + α, yn + β)

To obtain a, b, α, β, match Taylor expansion of exact solution:


hd
f (t, y(t)) + [f (t, y(t))] + · · ·
2 dt
with
af (t, y(t)) + bf (t + α, y(t) + β)
Chain Rule:
h
f (t, y(t)) + [ft (t, y(t)) + f (t, y(t))fy (t, y(t))] + · · ·
2
Multivariable Taylor expansion:

(a + b)f (t, y(t)) + b [ft (t, y(t))α + fy (t, y(t))β] + · · ·

Matching:
h h
a + b = 1, bα = , bβ = f (t, y(t)).
2 2

1
Option 1: a = 0, b = 1, α = h/2, β = hf (t, y(t))/2 yields
 
h h
yn+1 = yn + hf tn + , yn + f (tn , yn )
2 2

which is the Midpoint Method. Note the second argument to f is an approximation of y(tn +h/2)
by Euler’s Method!
Option 2: a = b = 1/2, α = h,

h
yn+1 = yn + [f (tn , yn ) + f (tn+1 , yn + hf (tn , yn ))]
2
This is the Explicit Trapezoidal Method. Note 2nd argument to f in second term is the Euler’s
approximation of y(tn+1 ). Note integral viewpoint:
Z tn+1
y 0 = f (t, y) =⇒ y(tn+1 ) = y(tn ) + f (s, y(s)) ds
tn

Both methods are second-order accurate because first non-matched term in Taylor expansion of
y(t) is O(h3 ).

Implicit Methods
Backward Euler:
yn+1 = yn + hf (tn+1 , yn+1 )
This is the simplest implicit method. Because yn+1 appears on both sides, we must solve for it. If
f is nonlinear in y, must use an iterative method like Newton’s Method, unlike explicit methods
like the ones we have seen, that use “plug-and-chug” to get yn+1
Implicit methods require more computational effort per time step, but can often take fewer steps.
Backward Euler is also first-order accurate. Second-order implicit method: Trapezoidal Method
h
yn+1 = yn + [f (tn , yn ) + f (tn+1 , yn+1 )]
2
When using iterative methods, for initial guess can use yn , or yn + hf (tn , yn ) ≈ yn+1 by Euler’s
Method
Systems of linear ODEs: y0 = Ay, A is n × n matrix. Backward Euler:

yn+1 = yn + hAyn+1

Rearrange:
(I − hA)yn+1 = yn
So we have the same system solved many times, different right-hand sides

You might also like