1

The non linear advection equation

(inviscid Burgers equation)
equation (one dimensional): 0 0
2
=

=
x
u
t
u
x
u
u
t
u
dt
du
Solution: ) ( ) , ( ut x f t x u =
with any function f
Like linear advection, but f depends on u itself (implicit equation)
Solvable (for practical use) in few special cases only.
A formal (geometrical) solution can be constructed by using the method of
characteristics
) sin( ) 0 , ( x t x u = = Initial condition:

=
=
1
) sin( ) , (
~
) , (
k
s
kx t k u t x u Solution:
kt
kt I
t k u
k
s
) (
2 ) , (
~
= with
I
k
(kt) = first kind Bessel function
e.g. Platzmann 1964, Tellus:
The non linear advection equation
0 =

=
x
u
u
t
u
dt
du
The method of characteristics:
0 0
) 0 , ( x t t x u x + = =
Lagrangian perspective: discussing the paths of individual parcels (points)
(paths = characteristics)
0 =
dt
du
=> velocity of individual parcel is constant in time
=> parcels are moving on a straight lines:
=> characteristics are straight lines with slope u(x
0
,t=0)
if u(x
0
,t=0) varies in space: characteristics cross at a certain time t
c
=> b) after t
c
one parcel can have more than one location (unphysical!)
=> physical solutions only possible up to t=t
c
from
=> a) discontinuities (shock waves) are formed
Example: Characteristics of Platzmanns Solution
Analytic solution characteristics
t=0
t=1
t=2
t=3
After: Platzmann, G.W., 1964: An exact integral of complete spectral equations for unsteady one-
dimensional flow. Tellus, 16, 422-431.
u
x
2
The non linear advection equation
Non linear term after inserting sine-series:

+ + =
=
=

1 2
2 1
1 2
2 1
2
2
1
1
]) ) sin[( ] ) )(sin[( ( ) (
) cos( ) sin( ) ( ) (
) cos( ) ( ) sin( ) (
2 1 2 1 2
2 1 2
2 2 1
k k
k k
k k
k k
k
k
k
k
x k k x k k t u t u k
x k x k t u t u k
x k t u k x k t u
x
u
u
The non linear wave-wave interaction of the waves k
1
and k
2
forces waves with
wave number k
1
+k
2
and k
1
-k
2
.
=> Energy/momentum is redistributed within the wave spectrum
0 =

x
u
u
t
u
L
n
k
2
= Example:
Sine-series (waves) with time dependent amplitudes

=
k
k
kx t u t x u ) sin( ) ( ) , (
The nonlinear term:
The non linear advection equation
grid point method: aliasing
1 2 3 4 0 j =
Problem: With a finite number of grid points (e.g. 0-M) waves with wave
numbers k > M/2 get erroneously interpreted (as M-k)
Example: M=4; k=3 => false wave with k=1
The non linear advection equation: Aliasing
consequence for energy
2 0 cos ) sin( ) 0 , (
0 0
= =

= = x (kx) ku
x
u
kx u t x u let
=> kin. Energy:

= =

2
0
2
0 2
2
0
0
2
) ( sin
2
u
dx kx
u
E
Energy change:

|

\
|

2
0
2
0
2
0
2
2
1
dx
x
u
u u dx
t
u
u dx
t
u
t
E
Example:

= =

= = |

\
|

2
0
3
0 2
0
2
0
0 ) 2 sin( ) sin(
2
) 2 sin(
2
) cos( ) sin( with dx kx kx
ku
t
E
kx u
k
kx kx ku
x
u
u
But
With resolution M=3k

= =

2
0
3
0
3
0
sin
2
) sin( ) sin(
2
) sin( ) 2 sin(
ku
dx kx kx
ku
t
E
kx kx
g Alia
=> Energy increase (due to aliasing from limited resolution)
=> analytically the total energy is conserved
3
The non linear advection equation:
Non linear instability
At each time level short waves (k
1
+k
2
) may be forced which (potentially)
are not resolved and, therefore, erroneously interpreted.
This leads to unrealistic increase of wave amplitudes (i.e. energy), in
particular in the short wave part of the spectrum, and finally to a blow up of
the numerical solution.
This mechanism is, in principal, independent of the time step length or the
grid size.
Non linear instability
The non linear advection equation:
Non linear instability
Solutions
Artificial elimination of short waves
Diffusive schemes (e.g. upstream)
Explicit diffusion
Appropriate discretization of non linear terms
The non linear advection equation:
Discretization of the non linear term
Analytically:
i.e. total energy (~u
2
) is conserved

x
dx
t
u
x
u
t
u
x
u
u
t
u
u
x
u
u
t
u
0
2
1
3
1
2
1
2 3 2
2
If this would also be the case numerically => no non linear instability
Solution: Discretization of the non linear term
x
u u
u
x
u
u
i i
i
i

\
|

+
2
1 1
1
st
try:
0
2
1
2
1
2
1
2 1 1 2
=

\
|

+
+
j
j j j j
j
j j
j
u u u u
u u
u dx
t
u
u u dx
t
u
u
t
E
( ) 0
2
1
2
2
3 1
2
3 1
2
2 3
2
2 3
2
1 2
2
1
+ + = u u u u u u u u u u u u example: 3 Points, cyclic boundary conditions
Not appropriate
4
The non linear advection equation:
Discretization of the non linear term
( )

\
|

+ +
+ + + + +
j
j j j j j
j j j j j j j j j j j j j
j
u u u u u
x t
E
x
u u u
x
u u u
x
u u u
x
u u u u
x
u
u
0
6
1
2
) (
2
) (
2
) (
3
1
6
2 1 2 1
2 1 2 1 1 1 2 1 2 1 2
nd
try:
But: Still accumulation of energy in wrong waves (unphysical)
( ) 0
6
1
1 2 3 2 1 3 3 1 2 1 3 2 2 3 1 3 2 1
= + + = u u u u u u u u u u u u u u u u u u example: 3 points, cyclic boundary
conditions
appropriate
[ ]
[ ]

= +

=
|
|

\
|

|
|

\
| + +
|

\
|

+ +
+ +
+ +
j
j j j j j j j
j j j j j j
j j j j j
j
u u u u u u u
x t
E
u u u u u u
x x
u u u u u
x
u
u
0
6
1
6
1
2 3
2
1 1 1
2
1
2
1 1 1
2
1
1 1 1 1 3
rd
try:
appropriate
The non linear advection equation:
finite differences
x
u u u u
t
u u
x
u
u
t
u
n n n n
n
i
n
i i i i i

+ +
+
6
2 1 2 1
1
options
Two level Euler:
x
u u u u
t
u u
x
u
u
t
u
n n n n
n
i
n
i i i i i

+ +
+
6 2
2 1 2 1
1 1
Three level Leap frog:
Caution: a) slightly unstable due to t-discretization
b) linear instability may occur => mind the CFL criterion
The non linear advection equation
spectral method
=> right hand side:

=
=
+ =
=

N
N k
k
k k
k k
k
k
k
k
ikx F
x k k i u u ik
x ik u ik x ik u
x
u
u
2
2
2 1 2
2 2 1
) exp(
) ) ( exp(
) exp( ) exp(
1 2
2 1
2
2
1
1
0 =

x
u
u
t
u

=
=
N
N k
k
ikx t u t x u ) exp( ) ( ) , (
Ansatz: transformation of T into new basis functions which are differentiable orthogonal
functions of x, e.g. Fourier-series:

N
N k
k
ikx
t
u
t
u
) exp( => left hand side:
Since contributions to
wavenumbers -2N to 2N are
obtained
with k=k
1
+k
2
and

=
2
1
) (
L
L l
l k l k
u u l k i F (interaction coefficients)
L
1
=max(-N,k-N); L
2
=min(N,k+N), i.e. L
1
=k-N, L
2
=N for k 0, and L
1
=-N, L
2
=k+N for k < 0
5
The non linear advection equation
spectral method
with

= =
=

N
N k
k
N
N k
k
ikx F ikx
t
u
x
u
u
t
u
2
2
) exp( ) exp( with

=
2
1
) (
L
L l
l k l k
u u l k i F
L
1
=max(-N,k-N); L
2
=min(N,k+N)

< = =
+ =
N k N
k
N
N k
k
N
N k
k
ikx F ikx F ikx F
2
2
2
) exp( ) exp( ) exp(
resolved
waves
unresolved
waves
Neglecting the unresolved waves leads to 2N coupled ordinary differential equations
(ODEs):

= =

2
1
) (
L
L l
l k l k
k
u u l k i F
t
u
L
1
=max(-N,k-N); L
2
=min(N,k+N)
-N k N
Cut off: no aliasing but not conserving moments higher than u
2
To be solved using common methods
The non linear advection equation
Numerics: The spectral transform method
spectral method versus finite differences:
spectral: exact spatial derivatives but numerical effort goes with N
2
finite differences: numerical effort goes with N but spatial derivatives not exact
Improvement by combination: The spectral transform method
Idea: compute derivatives (and linear terms) in spectral space and non linear term on
corresponding grid ( 3N+1 grid points to avoid aliasing). The effort goes with N ln(N)
(using a fast fourier transform (fft)).
Non linear advection:
Step 1: transform u into grid point domain (inverse Fourier transformation (via fft))
Step 2: compute u
2
on grid points
Step 3: transform u
2
into spectral space (via fft)
Step 4: compute the x-derivative of u
2
in spectral space, and do the time stepping
Step 1: .
0
2
1
0
2
=

x
u
t
u
x
u
u
t
u
Extension: The (viscid) Burgers-equation
Advection & Diffusion
equation (one dimension):
2
2
x
u
K
x
u
u
t
u

Various applications to describe processes in science and technology in a simple way

(e.g. road traffic) and important test bed for numerical schemes.
Numerical Solution: Spectral or finite differences with the known schemes. In most
cases by using time splitting (separating advection and diffusion)
(formal) analytic solution using Cole-Hopf transformation:
x
z
K
x
z
z
K
u

=
ln
2
2
2
2
2
2
2
2
2
2
2 2
ln
2
ln
2
ln
2
2
1
x
z
K
t
z
x
z
x
K
x
z
x
K
t
z
x
K
x
u
K
x
u
t
u

= |

\
|

d
K
t x G
d
K
t x G
t
x
u
)
2
) , ; (
exp(
)
2
) , ; (
exp(
=>
t
x
d u t x G
2
) (
) ( ) , ; (
2
0
0

+ =

Linear heat equation

6
Example: Numerical solution of Burgers equation
using different K
The non linear advection/transport equation
(Burgers equation)
Summary
Viscid and inviscid Burgers-equation
Aliasing
Non linear instability
Energy conserving discretization
Spectral transform method