Numerical Methods for Partial

Differential Equations
CAAM 452
Spring 2005
ecture !
"nstructor# $im %ar&urton
CAAM 452
Spring 2005
Summar' of Small $heta Anal'sis
(
$he dominant remainder term in this anal'sis
relates to a commonl' used) ph'sicall' moti*ated
description of the shortfall of the method#
( )
( )
{ }
( )
( )
{ }
( )
( )
{ }
( )
3
3
3
3
2
5
2
5
1
2
2
2
2
0
1
4
2
5
4
3!
!
!
0
0
0
0
m
m
m
m
m
m
m
t
d
t
m
O
i ct
dx
L
m m
t
m
c O
i ct
dx
L
m m
it
m
c O
i ct
dx
L
m m
i i
m
c t
i c
i
t
t
dx
t
dx
m
x
x
L
d
m
du
c u u u e e
dt
du
c u u u e e
dt
du
c u u u e e
dt
du
c u u u e e
dt

+
+
+

+
+

r
r
r
r
r
r
r
r
( )
{ }
7
m
t
O
dx

Dissipati*e
+nsta&le
Dispersi*e
Dispersi*e
CAAM 452
Spring 2005
Dissipation in Action
(
Consider the right difference *ersion#
(
%e are going to e,perimentall' determine ho-
much dissipation the solution e,periences.
( )
( )
{ }
2
3
3
1
2
0
m m
t
m
O
i ct
dx
L
m m
t
dx
du
c u u u e e
dt

+
+

r
r
CAAM 452
Spring 2005
$esting Methodolog'
/0 1educe the time2stepping error to a secondar' effect &' choosing a 4
th

order 1unge23utta 45S$0 time2integrator and a small dt.
20 6i, the method 4choose one of the difference operators for the spatial
deri*ati*e0
70 Do a parameter stud' in M) i.e. -e as8 the questions# ho- does
increasing the num&er of data points change the error.
40 %e need to understand -hat questions -e are as8ing#
/0 "s the computer code sta&le 4as predicted &' the theor'09
20 Does the computer code con*erge 4as predicted &' the theor'09
70 %hat order of accurac' in M do -e achie*e9
40 %e h'pothesi:e that if the theor' holds then -e should achie*e#
50 %hat is the actual appro,imate p achie*ed9
( ) ( )
, ,
p
m m
u T x u T x Cdx %
CAAM 452
Spring 2005
"nitial Condition
(
;ecause -e do not -ish to introduce uncertaint'
o*er the source of errors in the computation -e use
an initial condition -hich is infinitel' smooth.
( )
( )
2
4cos
, 0 with [0, 2)
x
u x e x


CAAM 452
Spring 2005
Computing Appro,imate
<rder of Accurac'
(
%e compute the error &'#
(
%e -ill compute the appro,imate order of accurac' &'
assuming#
(
-here C is independent of M
(
"f -e measure the error for t-o different M then -e can
eliminate the constant#
( "n the case -hen M
2
=2M
/
p
p
L
Cdx C
M

_


,
1
1
1
2 1 2
2 1 2
2
1
2
log
log
p
p
p
L
C
M
M
p
M M
L
C
M
M

_
_



_
,
,

)

_
,
_


,
,
( )
1
2
log
log 2
p

_

,

( ) ( ) : max , 8 , 8
m m
m
u x t u x t %
CAAM 452
Spring 2005
After 4 Periods
(
$he numerical pulses are in the right place &ut
ha*e se*erel' diminished amplitude.

du
c u
dt

+

r
r
CAAM 452
Spring 2005
After 4 Periods
M Ma,imum error at t=> ?order
20 0.@A/4A/@A4A5/!A
40 0.@A/7!5747@744! 0
>0 0.@>72A40>/24>5@ 0.0/
/@0 0.@7/@7A@04A2/>5 0.//
720 0.52!425@0@5040> 0.2@
After 4 periods the solution is totall' flattened in all &ut the last 2 results. "f -e &othered
to 8eep increasing M -e -ould e*entuall' see the error decline as /BM

du
c u
dt

+

r
r
CAAM 452
Spring 2005
$he +nsta&le eft Stencil
(
" repeated this -ith e*er'thing the same) e,cept the choice
of instead of

du
c u
dt

r
r

+
( )
( )
2
3
1 2
2!
2
0
m
m
t
d
t
m
c O
i ct
dx
L
x M
m m
du
c u u u e e
dt

+


_

,
+
' )

r
r
%e clearl' see that there is initial gro-th
near the pulses) &ut e*entuall' the
dominant feature is the highl' oscillator'
and e,plosi*e gro-th
4large m in the a&o*e red term0.
CAAM 452
Spring 2005
Cont (snapshot)
( )
( )
2
3
1 2
2!
2
0
m
m
t
d
t
m
c O
i ct
dx
L
x M
m m
du
c u u u e e
dt

+


_

,
+
' )

r
r
CAAM 452
Spring 2005
Dispersion in Action
(
Consider the central difference *ersion#
(
%ith the same time integrator &efore) M=/00)C
(
%e note that the remainder terms are dispersi*e corrections
D i.e. the' indicate that modes of different frequenc' -ill
tra*el -ith different speed.
(
6urthermore) to leading order accurac' the higher order
modes -ill tra*el more and more slo-l' as m increases.
( )
( )
{ }
3 5 2
3!
0
0
m m
it
m
i
t c O
i ct
dx
L
m m
dx
du
c u u u e e
dt

r
r
0

du
c u
dt

r
r
CAAM 452
Spring 2005
(
%e notice that the humps start to shed rear oscillations as
the higher frequenc' 6ourier components lag &ehind the
lo-er frequenc' 6ourier components.
2
nd
<rder Central Difference
After 4 Periods
0

du
c u
dt

r
r
CAAM 452
Spring 2005
Con*ergence Stud' 4t=>0
%hat should -e use as an error
"ndicator 99
0

du
c u
dt

r
r
CAAM 452
Spring 2005
2
nd
<rder Central Difference 0
du
c u
dt

r
r
M Ma,imum error at t=> ?order
20 0.A4240>0/4!450@
40 0.7502A>AA>50!77 /.42
>0 0.7>2A>5/2!!@>5A 0./7
/@0 0./A//>/2!4@@>!> /.00
720 0.054!/!705244>2 /.>0
( %e do not see a con*incing 2
nd
order accurac'
( " computed this &' log24errorEMBerrorFM2/G0
CAAM 452
Spring 2005
4
th
<rder Central Differencing
( )
( )
{ }
5 7
4
2
5!
4
0
m m
i it
m
c t O
i ct
dx dx
L
m m
du
c u u u e e
dt

+

r
r
4
du
c u
dt

r
r
CAAM 452
Spring 2005
4
th
<rder Central Difference
( %e see prett' con*incing 4
th
order accurac'
( " computed this &' log24errorEMBerrorFM2/G0
M Ma,imum error at t=> ?order
20 0.44757>!>5422!!
40 0./2>52A52A!2//@ /.!A
>0 0.02040/A0>>@!@! 2.@@
/@0 0.00/740!@4!>25A 7.A7
720 0.0000>7/A47>4/4 4.0/
4
du
c u
dt

r
r
CAAM 452
Spring 2005
@
th
<rder Central Difference 6
du
c u
dt

r
r
M Ma,imum error at t=> ?order
20 0./A40A5!5@/5520
40 0.04A57@5252@A>7 /.A!
>0 0.00/@A2AA/55@07 4.>!
/@0 0.00002>A/!24/05 5.>!
720 0.0000004@04/4/5 5.A!
( %e see prett' con*incing @
th
order accurac'
( " computed this &' log24errorEMBerrorFM2/G0
CAAM 452
Spring 2005
Summar' of $esting Procedure
/0 +nderstand -hat 'ou -ant to test
20 3eep as man' parameters fi,ed as possi&le
70 "f possi&le) perform an anal'sis &efore hand
40 1un parameter s-eeps to determine if code agrees -ith anal'sis.
50 Estimate error scaling -ith a single parameter if possi&le.
@0 "t should &e apparent here that the errors computed in the simulations
onl' appro,imate the tid' po-er la- 4d,Hp0 in the limit of small d,
4large M0.
!0 "t is quite common to refer to the range of M &efore the error scaling
applied as the Ipreasymptotic convergence rangeJ.
>0 $he preas'mptotic range is due to the other factors in the error
estimate -hich are much larger than the d,Hp term 4i.e. the pKth
deri*ati*es ma' &e *er' large) or the constant ma' &e largeC0
A0 "n this case -e heuristicall' set dt small) using a high2order 1unge2
3utta time integrator) and then performed a parameter s-eep on M.
%e could ha*e &een unluc8' and the time2stepping errors ma' ha*e
&een dominant -hich -ould mas8 the d, &eha*ior.
/00 $o &e careful -e -ould tr' this e,periment -ith e*en smaller dt and
chec8 if the scalings still hold.
CAAM 452
Spring 2005
Summar' of our Approach to Designing
6inite Difference Methods
(
%e ha*e s'stematicall' created finite difference methods &'
separating the treatment of space and time deri*ati*es.
(
$hen designing a sol*er consists of choosingBtesting#
D
a time integrator 4so far -e co*ered Euler26or-ard) eap6rog)
Adams2;ashford42)7)40) 1unge23utta0
D
a discreti:ation for spatial deri*ati*es
D a discrete differential operator -hich has all eigen*alues in the
left half of the comple, plane 4assuming the PDE onl' admits
non2gro-ing solutions0.
D
Possi&l' using LerschgorinKs theorem to locali:e the
eigen*alues of the discrete differentiation operator
D
dt so that dtMlargest eigen*alue 4&' magnitude0 of the
deri*ati*e operator sits inside the sta&ilit' region sta&ilit'.
D
4small d,0 spatial truncation anal'sis consistenc' and
accurac'.
D 6ourier anal'sis for classification of the differential operator.
D
%riting code and testing.
CAAM 452
Spring 2005
Some Classical 6inite Difference Methods
(
No-e*er) there are a num&er of classical methods
-hich -e ha*e not discussed and do not quite fit
into this frame-or8.
(
%e include these for completeness..
CAAM 452
Spring 2005
eap 6rog 4space and time0
(
%e use centered differencing for &oth space and time.
(
%e 8no- that the leap frog time stepping method is onl'
sta&le for operators -ith eigen*alues in the range#
(
No-e*er) -e also 8no- that the centered difference
deri*ati*e matri, is a s8e- s'mmetric matri, -ith
eigen*alues#
(
So -e are left -ith a condition# i.e.
1 1
1 1
2 2
n n n n
m m m m
u u u u
c
dt dx
+
+
_

,
[ ]
1,1 dt i
2
sin
ic m
dx M

_

,
1
cdt
dx

dx
dt
c

CAAM 452
Spring 2005
cont
(
%e can perform a full truncation anal'sis 4in space
and time0#
(
%e 8no- that dt O= d, so
( ) ( ) ( ) ( )
( ) ( )
1 1 1 1
3 3
, , , ,
2 2
n
m n m n m n m n
m
u x t u x t u x t u x t
T cdt
dx
cdt
O dt O dx
dx
+ +

_


,

( ) ( )
3 3 n
m
T O dt cO dx
CAAM 452
Spring 2005
a,26riedrichs Method
(
%e immediatel' conclude that the follo-ing method is not
sta&le#
(
;ecause the Euler26or-ard time integrator onl' admits one
sta&le point 4the origin0 on the imaginar' a,is) &ut the
central differencing matri, has all eigen*alues on the
imaginar' a,is.
(
No-e*er) -e can sta&ili:e this formula &' replacing the
second term in the time2stepping formula#
1
0
n n
n
m m
m
u u
c u
dt

+

( )
1
1 1
0
0.5
n n n
m m m
n
m
u u u
c u
dt

+
+
+

CAAM 452
Spring 2005
cont
(
$his formula does not quite fit into our constructi*e
process 4method of lines approach0.
(
%e ha*e admitted spatial a*eraging into the
discreti:ation of the time deri*ati*e.
(
%e can rearrange this#
( )
1
1 1
0
0.5
n n n
m m m
n
m
u u u
c u
dt

+
+
+

( ) ( )
1
1 1 1 1
1 1
1
2 2
1 1
1 1
2 2
n n n n n
m m m m m
n n
m m
dt
u u u c u u
dx
dt dt
c u c u
dx dx
+
+ +
+
+ +
_ _
+ +

, ,
CAAM 452
Spring 2005
cont
(
%e can immediatel' determine that this is a sta&le method as
long as cMdtBd, O=/
(
Li*en) this condition -e o&ser*e that the time updated
solution is al-a's &ounded &et-een the *alues of the left and
right neigh&or at the pre*ious time &ecause this is an
interpolation formula.
1
1 1
1 1
1 1
2 2
n n n
m m m
dt dt
u c u c u
dx dx
+
+
_ _
+ +

, ,
CAAM 452
Spring 2005
cont
(
A formal sta&ilit' anal'sis -ould in*ol*e#
(
%hich gi*es sta&ilit' for each mode if#
( ) ( )
{ }
1
1 1
1
1
1 1
1 1
2 2
1 1
1 1
2 2
1 1
1 1
2 2
cos sin
m m
n n n
m m m
n n n
i i n n n
m m m
n
m m m
dt dt
u c u c u
dx dx
dt dt
u c u c u
dx dx
dt dt
u c e u c e u
dx dx
dt
c i u
dx


+
+
+
+
_ _
+ +

, ,
_ _
+ +

, ,
_ _
+ +

, ,
+
R L
r r r
% % %
%
( ) ( )
cos sin 1
m m
dt
c i
dx
+
CAAM 452
Spring 2005
cont
(
$hus considering all the possi&le modes
(
%e note that the middle mode requires#
(
$his condition gi*es a sufficient condition for all modes to &e
&ounded.
(
;' the in*erti&ilit' and &oundedness of the 6ourier transform
-e conclude that the original equation is sta&le.
cos sin 1
2 2
1
dt
c i
dx
dt
c
dx

_ _
+

, ,

2
m
m
M

CAAM 452
Spring 2005
cont
(
%e can recast the a,26riedrichs method again
(
$his method consists of Euler 6or-ard) central
differencing for the space deri*ati*e and an e,tra
dissipati*e term 4i.e. a grid dependent ad*ection
diffusion equation# 0
( )
1
1 1
1 2 2
0
1 1
1 1
2 2
1
1
2
n n n
m m m
n n n
m m m
dt dt
u c u c u
dx dx
u cdt u dx u
+
+
+
_ _
+ +

, ,

+ +
2 2
2
2
u u dx u
c
t x x

+

CAAM 452
Spring 2005
C6 Num&er
(
$he ratio appears repeatedl') in particular in
the estimates for the ma,imum possi&le time step.
(
%e refer to this quantit' as the C6 4Courant2
6riedrichs2e-'0 num&er.
(
;ounds of the form# -hich result from a
sta&ilit' anal'sis are frequentl' referred to as C6
conditions.
dt
dx

dt
C
dx

CAAM 452
Spring 2005
a,2%endroff Method
(
%e are fairl' free to choose the parameter in the sta&ili:ing
term#
(
$he artificial viscosity term acts to shift the eigen*alues of
the spatial operator into the left half2plane.
(
1ecall D the Euler26or-ard sta&ilit' region is the unit circle
centered at 2/ in the comple, plane. So pushing the
eigen*alues off the imaginar' a,is allo-s us to choose a dt
small enough to push the eigen*alues of the discrete spatial
operator into the sta&ilit' regionC
( )
1 2 2
0
1
1
2
n n n
m m m
u cdt u cdt u
+
+ +
CAAM 452
Spring 2005
Class E,ercise
(
Please pro*ide a C6 condition for#
(
"n terms of#
(
%hat is the truncation error9
( ) ( )
1 1
1 1 2 2
4 1
3 6
n n n n n n
m m m m m m
u u c u u c u u
+
+ +
+
dt
dx

CAAM 452
Spring 2005
Pon Neumann Anal'sis
(
;' stealth " ha*e introduced the classical Pon
Neumann anal'sis.
(
$he first step is to 6ourier transform the finite
difference equation in space.
(
A short cut is to ma8e the follo-ing su&stitutions#
1
1
2
2
2
2
......
m
m
m
m
n n
m m
i n n
m m
i n n
m m
i n n
m m
i n n
m m
u u
u e u
u e u
u e u
u e u

%
%
%
%
%
CAAM 452
Spring 2005
Anal'sis cont
(
%e can also ma8e the su&stitutions for the
difference operators#
( )
( )
{ }
( )
2
2
1
1
2
0
2
2
2
2
1 1
1 2 sin
2
1 1
1 2 sin
2
1 1
sin
2
......
m
m
m
m
m
m
m m
m
m
i
i
m
n n
m m
i
i
m
i n n
m m
i n n
m m
i i
i n n
m
m m
i n n
m m
e ie
dx dx
u u
e ie
u e u
dx dx
u e u
e e i
u e u
dx dx
u e u

_
+
' )

,

' )

%
%
%
%
%
( )
{ }
2
2
2
1
4sin
2
2
1 cos
m
m
dx
dx

+

_

' )

,

CAAM 452
Spring 2005
Example
(
$he a,2%endroff scheme
(
&ecomes
( )
1 2 2
0
1
1
2
n n n
m m m
u cdt u cdt u
+
+ +
2 2
1 2
2
1 2
1
2 2
m m m m
i i i i
n n n
m m m
e e e e
u cdt u cdt u
dx dx

+
_ _ _ +
+ +

, , ,
% % %
CAAM 452
Spring 2005
Example cont
(
So the a,2%endroff scheme &ecomes a neat one
step method for each 6ourier coefficient#
(
%e ha*e neatl' decoupled the 6ourier coefficients
so -e are &ac8 to sol*ing a recurrence relation in n
for each m#
(
Nence -e need to e,amine the roots of the sta&ilit'
pol'nomial for the a&o*e
2 2
1 2
2
1 2
1
2 2
m m m m
i i i i
n n n
m m m
e e e e
u cdt u cdt u
dx dx

+
_ _ _ +
+ +

, , ,
% % %
n
m
u%
2 2
2
2
1 2
1
2 2
m m m m
i i i i
e e e e
z cdt cdt
dx dx

_ _ _ +
+ +
' )

, , ,
CAAM 452
Spring 2005
E,ample cont
(
$his case is tri*ial as -e Qust need to &ound the
single root &' /
(
Plot it this as a function of thetaC
( ) ( ) ( )
{ }
2 2
2
2
2
1 2
1
2 2
1 sin cos 2 1
m m m m
i i i i
m
e e e e
z cdt cdt
dx dx
c i c



_ _ _ +
+ +
' )

, , ,
+ +
CAAM 452
Spring 2005
6inal Pon Neuman Shortcut
(
%e can s8ip lots of steps &' ma8ing the direct
su&stitution#
(
Nere g is referred to as the amplification factor and
iMmMtheta is our pre*ious thetaEm
n n im
m
u g e

CAAM 452
Spring 2005
Ne,t $ime
(
Definition of consistenc'
(
a,21ichtm'er equi*alence theor'
(
$reating higher order deri*ati*es
CAAM 452
Spring 2005
Nome-or8 7
R/0 ;uild a finite2difference sol*er for
R/a0 use 'our Cash23arp 1unge23utta time integrator from N%2 for time stepping
R/&0 use the 4
th
order central difference in space 4periodic domain0
R/c0 perform a sta&ilit' anal'sis for the time2stepping 4&ased on *isual inspection of
the C3 123 sta&ilit' region containing the imaginar' a,is0
R/d0 &ound the spectral radius of the spatial operator
R/e0 choose a dt -ell in the sta&ilit' region
R/f0 perform four runs -ith initial condition
4use M=20)40)>0)/@00 and compute ma,imum error at t=>
R/g0 estimate the accurac' order of the solution.
R/h0 e,tra credit# perform adapti*e time2stepping to 8eep the local truncation error
from time stepping &ounded &' a tolerance.
u u
c
t x

( )
( )
[ ]
2
4cos
, 0 , 0, 2
x
u x e x