8 26

Stiff Differential Equations

Differential eeuations
for which some numerical

methods are
going to
be unstable
In our meth J for
the error
solving Irp
bounded the
was

value of
by maximum

a
higher derivative
of the solution

If that gets larger then

the error could increase
and this is called Equation
When the solution
to the differential
equation changes very
rapidly
The test eduction
161 2
4 Xy X L 0

If X is
very large
then the solution
rapidly
and nt derivative
decays
Ly e't can become
large
e't
Ylt
consider Euler's method

Witi Wit hXW

 ft thx wi

By induction
il wi i L Its

Then
used
1
yet wit ai

ti Idith
I L e w
I 12 I
1411 e't Leith 51
Kille Kathy Y
 Athx will needs
to be close to
how
Error depends on

weil thx
approximates
exh
e 10 so we

require Ithx 10

as
j 22
So we
requite
11th I E L
fine
it

h for which the

method will be

sable
I
he

stiffness means

that we need to take

very small values of h in

order to ensure
stability
2 h C
explicit
for the Ehler method

11 12 1 LL
r
 1Inct

fi

In general a one
step
to the test
applied
etuation gives
Witi QCXh Wi
 how
accuracy depends on

well achx approximates em

so in this case

IQlah 1cL
Eater's method
Example Implicit
to the test Equation
applied
wit Witch Witi
Wit Wit
74

QCxhli.tn
let E ha then I I 1
The for
stability
Eater
region
implicit

4th

A mil
Definition
The of absolute stability
region
12 of a method is the set

12 xhe Wi 70 as its

Where Wi is
generated by
the numerical method
applying
to the test IVP
 t

Haation
storm

order
Higher systems
and educations

A n m order
system of
first order initial value

problem has the form

1 f ft Ui Uzi Hum

Um
dYf f 2 t U U2 a
 firm
tuff t Ui Ua Um

for a E 6 with initial

conditions

Ui Li fist M

to find
Our
goal is U.lt UmCly

that satisfy the differential

ejections with the initial
conditions

Extend tools from single

variable the
systems
Definition The function
f t U Um defined on

the set

D Let.mu am
f If
 is said to satisfy a Lipschitz
condition on D in the variables
U Um if a constant L
exists with

1 fct.ie am fCt.Z Zm

E L IU Zit
for all f U Um AZ tn iuD

We can use the mean

Value theorem to show

that if f and its

first partial derivatives

are continuous on D
and
cu 1
I EL
for each i 1,2 h
and all ct.ua Un inD
then f satisfies Lipschitz
a

condition with constant L

on D

Example Let h 0.5

the Euler method
apply
to solve
Ivp system
Ui's U2 Ui U
pete
U.co I Udo O

Yeti Lift
F HI I 1.44
w I
WE WothFCt w
 d to.SE EI.sJWz

w.thF t.w
I Is to 5
g
5

5
I
Higher order differential
etuations
A general mth order

Ivp
ycntctl
aEt
fH.y.yiyi.it
Eb
t

with initial conditions

Li fat Lz Hm
flat y
 u.HlLy
u.lt

tf

Letting u.lt
ylt a

Ctl
U.lt
Un
y'ttl.u.ltl
Ct Ct
y
y
It can be converted
into first order system
JU
Ua
Tt

II
 Y sun

SUIT y fft.y.yi.in
f t U u Um with

initial conditions
U.la yCa L

U at y'ial
da

u a
y fats dm

Example Transform the

second order
Ivp
2 e sinft
Y y 124 0.4
oft El 1 101
Cole 0.6
y
Sol
U.lt
yCtlUzCtl
y'It
from which we obtain

µ
Itt U.lt

Uz e2tsinH 2U.H
t2Ualt
Uilo 0.4 U2 of 0.6

At home write a

short code that

solves with
initial conditions
 of ODE
System
2 ox IS
y
Is X Zo
y y
X lol 2
so
yeol
I A I

A t.is
I

s
ii l l
A
Xi A l I XIA 35

I
 39T t
Xlt e e
t
I
yett e e

Two for
components St t
the solution e e

The 2
eigenvalues
provide two
decay rates
of the Solution

Explicit Vs
Implicit
Ewer

Using Explicit Euler

Xun Xu 1hL Lox

syn

yn.ir nthf csxn Loyn

 X 2 i 1 0
0

We can use induction

to obtain n

Xue I 354 74 41

I 35h1 a h
Yu
Xn lo ya 10

for n 18

l fl 3941cL
2 It the
ht 21
At home
try
similar analysis
with implicit