Finite Differences

Finite Difference Approximations

Simple geophysical partial differential equations
Finite differences - definitions
Finite-difference approximations to pdes
Exercises
Acoustic wave equation in 2D
Seismometer equations
Diffusion-reaction equation
Finite differences and Taylor Expansion
Stability -> The Courant Criterion
Numerical dispersion




1
Computational Seismology
dx
x f dx x f
f
x
) ( ) ( +
~ c
+
Finite Differences
) (
2 2 2
2 2
z
y x
t
s p c p
c + c + c = A
+ A = c The acoustic
wave equation
- seismology
- acoustics
- oceanography
- meteorology
Diffusion, advection,
Reaction
- geodynamics
- oceanography
- meteorology
- geochemistry
- sedimentology
- geophysical fluid dynamics
P pressure
c acoustic wave speed
s sources
p RC C C k C
t
+ V - A = v c
C tracer concentration
k diffusivity
v flow velocity
R reactivity
p sources
Partial Differential Equations in Geophysics
Finite Differences
Finite differences
Finite volumes
- time-dependent PDEs
- seismic wave propagation
- geophysical fluid dynamics
- Maxwells equations
- Ground penetrating radar
-> robust, simple concept, easy to
parallelize, regular grids, explicit method
Finite elements
- static and time-dependent PDEs
- seismic wave propagation
- geophysical fluid dynamics
- all problems
-> implicit approach, matrix inversion, well founded,
irregular grids, more complex algorithms,
engineering problems
- time-dependent PDEs
- seismic wave propagation
- mainly fluid dynamics
-> robust, simple concept, irregular grids, explicit
method
Numerical methods: properties
Finite Differences
Particle-based
methods
Pseudospectral
methods
- lattice gas methods
- molecular dynamics
- granular problems
- fluid flow
- earthquake simulations
-> very heterogeneous problems, nonlinear problems
Boundary element
methods
- problems with boundaries (rupture)
- based on analytical solutions
- only discretization of planes
-> good for problems with special boundary conditions
(rupture, cracks, etc)
- orthogonal basis functions, special case of FD
- spectral accuracy of space derivatives
- wave propagation, ground penetrating radar
-> regular grids, explicit method, problems with
strongly heterogeneous media
Other numerical methods
Finite Differences


What is a finite difference?
Common definitions of the derivative of f(x):
dx
x f dx x f
f
dx
x
) ( ) (
lim
0
+
= c

dx
dx x f x f
f
dx
x
) ( ) (
lim
0

= c

dx
dx x f dx x f
f
dx
x
2
) ( ) (
lim
0
+
= c

These are all correct definitions in the limit dx->0.
But we want dx to remain FINITE
Finite Differences


What is a finite difference?
The equivalent approximations of the derivatives are:
dx
x f dx x f
f
x
) ( ) ( +
~ c
+
dx
dx x f x f
f
x
) ( ) (
~ c

dx
dx x f dx x f
f
x
2
) ( ) ( +
~ c
forward difference
backward difference
centered difference
Finite Differences


The big question:
How good are the FD approximations?
~==
This leads us to Taylor series....
Finite Differences


Taylor Series
Taylor series are expansions of a function f(x) for some
finite distance dx to f(x+dx)
What happens, if we use this expression for
dx
x f dx x f
f
x
) ( ) ( +
~ c
+
?
... ) (
! 4
) (
! 3
) (
! 2
) ( dx ) ( ) (
' ' ' '
4
' ' '
3
' '
2
'
+ + = x f
dx
x f
dx
x f
dx
x f x f dx x f
Finite Differences


Taylor Series
... that leads to :
The error of the first derivative using the forward
formulation is of order dx.
Is this the case for other formulations of the derivative?
Lets check!
) ( ) (
... ) (
! 3
) (
! 2
) ( dx
1 ) ( ) (
'
' ' '
3
' '
2
'
dx O x f
x f
dx
x f
dx
x f
dx dx
x f dx x f
+ =
(

+ + + =
+
Finite Differences


... with the centered formulation we get:
The error of the first derivative using the centered
approximation is of order dx
2
.
This is an important results: it DOES matter which formulation
we use. The centered scheme is more accurate!
Taylor Series
) ( ) (
... ) (
! 3
) ( dx
1 ) 2 / ( ) 2 / (
2 '
' ' '
3
'
dx O x f
x f
dx
x f
dx dx
dx x f dx x f
+ =
(

+ + =
+
Finite Differences




xj 1 xj xj + 1 xj + 2 xj + 3
f xj ( )
dx h
desired x location
What is the (approximate) value of the function or its (first,
second ..) derivative at the desired location ?
How can we calculate the weights for the neighboring points?
x
f(x)
Alternative Derivation
Finite Differences




Lets try Taylors Expansion
f x ( )
dx
x
f(x)
dx x f x f dx x f ) ( ' ) ( ) ( + ~ + (1)
(2)
we are looking for something like
f x w f x
i
j
i
index j
j L
( ) ( )
( )
,
( ) ( ) ~

=1
Alternative Derivation
dx x f x f dx x f ) ( ' ) ( ) ( ~
Finite Differences
deriving the second-order scheme
af af af dx
+
~ + '
bf bf bf dx

~ '
+ ~ + +
+
af bf a b f a b f dx ( ) ( ) '
the solution to this equation for a and b leads to
a system of equations which can be cast in matrix form
dx b a
b a
/ 1
0
=
= +
0
1
=
= +
b a
b a
Interpolation Derivative
2nd order weights
Finite Differences
Taylor Operators
... in matrix form ...
|
|
.
|

\
|
=
|
|
.
|

\
|
|
|
.
|

\
|
0
1
1 1
1 1
b
a
Interpolation Derivative
... so that the solution for the weights is ...
|
|
.
|

\
|
=
|
|
.
|

\
|
|
|
.
|

\
|
dx b
a
/ 1
0
1 1
1 1
|
|
.
|

\
|
|
|
.
|

\
|

=
|
|
.
|

\
|

0
1
1 1
1 1
1
b
a
|
|
.
|

\
|
|
|
.
|

\
|

=
|
|
.
|

\
|

dx b
a
/ 1
0
1 1
1 1
1
Finite Differences
... and the result ...
Interpolation Derivative
Can we generalise this idea to longer operators?
|
|
.
|

\
|
=
|
|
.
|

\
|
2 / 1
2 / 1
b
a
|
|
.
|

\
|

=
|
|
.
|

\
|
1
1
2
1
dx b
a
Let us start by extending the Taylor expansion beyond f(xdx):
Interpolation and difference weights
Finite Differences
' ' '
! 3
) 2 (
' '
! 2
) 2 (
' ) 2 ( ) 2 (
3 2
f
dx
f
dx
f dx f dx x f + ~
*a |

*b |

*c |

*d |
... again we are looking for the coefficients a,b,c,d with which
the function values at x(2)dx have to be multiplied in order
to obtain the interpolated value or the first (or second) derivative!

... Let us add up all these equations like in the previous case ...
' ' '
! 3
) (
' '
! 2
) (
' ) ( ) (
3 2
f
dx
f
dx
f dx f dx x f + ~
' ' '
! 3
) (
' '
! 2
) (
' ) ( ) (
3 2
f
dx
f
dx
f dx f dx x f + + + ~ +
' ' '
! 3
) 2 (
' '
! 2
) 2 (
' ) 2 ( ) 2 (
3 2
f
dx
f
dx
f dx f dx x f + + + ~ +
Higher order operators
Finite Differences
~ + + +
+ + +
df cf bf af
+ + + + ) ( d c b a f
+ + + ) 2 2 ( ' d c b a dxf
+ + + + ) 2
2 2
2 ( ' '
2
d
c b
a f dx
)
6
8
6
1
6
1
6
8
( ' ' '
3
d c b a f dx + +
... we can now ask for the coefficients a,b,c,d, so that the
left-hand-side yields either f,f,f,f ...
Higher order operators
Finite Differences
1 = + + + d c b a
0 2 2 = + + d c b a
0 2
2 2
2 = + + + d
c b
a
0
6
8
6
1
6
1
6
8
= + + d c b a
... if you want the interpolated value ...
... you need to solve the matrix system ...
Linear system
Finite Differences
High-order interpolation
|
|
|
|
|
.
|

\
|
=
|
|
|
|
|
.
|

\
|
|
|
|
|
|
.
|

\
|


0
0
0
1
6 / 8 6 / 1 6 / 1 6 / 8
2 2 / 1 2 / 1 2
2 1 1 2
1 1 1 1
d
c
b
a
... with the result after inverting the matrix on the lhs ...
|
|
|
|
|
.
|

\
|

=
|
|
|
|
|
.
|

\
|
6 / 1
3 / 2
3 / 2
6 / 1
d
c
b
a
... Interpolation ...
Finite Differences
|
|
|
|
|
.
|

\
|
=
|
|
|
|
|
.
|

\
|
|
|
|
|
|
.
|

\
|


0
0
/ 1
0
6 / 8 6 / 1 6 / 1 6 / 8
2 2 / 1 2 / 1 2
2 1 1 2
1 1 1 1
dx
d
c
b
a
... with the result ...
|
|
|
|
|
.
|

\
|

=
|
|
|
|
|
.
|

\
|
6 / 1
3 / 4
3 / 4
6 / 1
2
1
dx
d
c
b
a
... first derivative ...
First derivative
Finite Differences
Our first FD algorithm (ac1d.m) !
) (
2 2 2
2 2
z
y x
t
s p c p
c + c + c = A
+ A = c
P pressure
c acoustic wave speed
s sources
Problem: Solve the 1D acoustic wave equation using the finite
Difference method.
Solution:
| |
2
2
2 2
) ( ) ( 2
) ( ) ( 2 ) ( ) (
sdt dt t p t p
dx x p x p dx x p
dx
dt c
dt t p
+ +
+ + = +
Finite Differences
Problems: Stability
| |
2
2
2 2
) ( ) ( 2
) ( ) ( 2 ) ( ) (
sdt dt t p t p
dx x p x p dx x p
dx
dt c
dt t p
+ +
+ + = +
1 ~ s c
dx
dt
c
Stability: Careful analysis using harmonic functions shows that
a stable numerical calculation is subject to special conditions
(conditional stability). This holds for many numerical problems.
(Derivation on the board).
Finite Differences
Problems: Dispersion
| |
2
2
2 2
) ( ) ( 2
) ( ) ( 2 ) ( ) (
sdt dt t p t p
dx x p x p dx x p
dx
dt c
dt t p
+ +
+ + = +
Dispersion: The numerical
approximation has
artificial dispersion,
in other words, the wave
speed becomes frequency
dependent (Derivation in
the board).
You have to find a
frequency bandwidth
where this effect is small.
The solution is to use a
sufficient number of grid
points per wavelength.
True velocity
Finite Differences
Our first FD code!
| |
2
2
2 2
) ( ) ( 2
) ( ) ( 2 ) ( ) (
sdt dt t p t p
dx x p x p dx x p
dx
dt c
dt t p
+ +
+ + = +
% Time stepping

for i=1:nt,

% FD

disp(sprintf(' Time step : %i',i));

for j=2:nx-1
d2p(j)=(p(j+1)-2*p(j)+p(j-1))/dx^2; % space derivative
end
pnew=2*p-pold+d2p*dt^2; % time extrapolation
pnew(nx/2)=pnew(nx/2)+src(i)*dt^2; % add source term
pold=p; % time levels
p=pnew;
p(1)=0; % set boundaries pressure free
p(nx)=0;

% Display
plot(x,p,'b-')
title(' FD ')
drawnow

end
Finite Differences
Snapshot Example
0 1000 2000 3000 4000 5000 6000
0
500
1000
1500
2000
2500
3000
Distance (km)

T
i
m
e

(
s
)

Velocity 5 km/s
Finite Differences
Seismogram Dispersion
Finite Differences
Finite Differences - Summary
Conceptually the most simple of the numerical methods and
can be learned quite quickly
Depending on the physical problem FD methods are
conditionally stable (relation between time and space
increment)
FD methods have difficulties concerning the accurate
implementation of boundary conditions (e.g. free surfaces,
absorbing boundaries)
FD methods are usually explicit and therefore very easy to
implement and efficient on parallel computers
FD methods work best on regular, rectangular grids