Scribd
Upload
18 views

reviewofproblem

Uploaded by

Tayeb Chelirem
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
18 views

reviewofproblem

Uploaded by

Tayeb Chelirem
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

A review of the problem

Andrew P. Ulvestad
September 10, 2009

1 The w-equation
This is just to make sure we’re on the same page. I am going to omit the
subscript x from the v and y from the k to avoid confusion. We have 3 equations
in 3 unknowns, w, v, k
!
∂w ∂ 2βkx0 k(y, t)
w(y, t) + γ0 e−0.1y − D0 e.01y kx0
2
+ k(y, t)2 w(y, t) (1)
 
=− 2
∂t ∂y (kx0 + k(y, t) )
2 2
!
∂v ∂ 2βkx0 k(y, t)
= w(y, t) − νv(y, t) (2)
∂t ∂y (kx0 2 + k(y, t)2 )2

∂k ∂v
= −kx0 (3)
∂t ∂y
2βkx0 k(y, t)
vg = (4)
2 + k(y, t)2 )2
(kx0

A couple of things about the equations. The forcing (γ) and damping (D)
profiles are not fixed. Anything that prevents the solution from reaching and
hitting the far boundary will work for the damping. We also need to make sure
we force enough not to kill off w entirely. I noticed in my ‘Method of Lines’ code
that the solutions seem to be stable until [f orce − damp] < 0 and that seems
to be when it goes unstable. This change in sign happens because k increases
in time. In any event, any profiles, provided the forcing is local relative to the
damping, that give stable solutions would work. I originally wanted to use a
step function (for f orce) to force at a constant value out until, say, y = 50 and
then have nothing but I think this could lead to numerical issues.
The equations are built so when the w equation is added to the v equation
they conserve momentum (up to forcing and damping). This may be an issue
since it yields the flux matrix (when written in matrix from, the 3 × 3 matrix
that ∂y acts on) has a zero eigenvalue. I added an additional term to the v
equation in a test case and didn’t get different results so I’m not sure if this
makes a big difference or not.
I use the following boundary/initial conditions

w(0, t) = w0 ; w(100, t) = 0; w(y, 0) = w0 e−.1y

v(0, t) = v0 ; v(100, t) = 0; v(y, 0) = v0 cos(3y)


k(0, t) = k0 ; k(100, t) = 0; k(y, 0) = k0 e−.05y

1
The initial w and k conditions are not that important provided they are
fairly localized around 0. A delta function would be ideal but again this is
numerically an issue. I have tried a Gaussian and achieved essentially the same
results. The initial v must have some periodicity as the zonal flow is sheared
leading to small w bump propagation. I achieved similar results with cos(y)
and this is fine if it is more stable. I run the following checks on the constants
before solving
2
γ0 > (kx0 + k02 )D0 ; γ0 > ν
γ0 > vg (t = 0)/L; γ0 < β/kx0
Here L is the box length. Provided these relations are satisfied, the values of the
constants are not important. As I mentioned, I was able to get a propagating
bump solution (see movie on website) for the w equation with the following
constants

w0 = 15; v0 = 2; k0 = 5 = kx0 ; γ0 = 7; ν = 0.1 = D0 ; L = 100; β = 80;

I also had to the diffusive term in the k equation, meaning I solved

∂k ∂v ∂2k
= −kx0 + D1 2
∂t ∂y ∂y
where the diffusion coefficient was D1 = 1. As I decrease the diffusion coefficient,
the solution exhibits more and more oscillatory behavior. PDEPE will not
produce a solution when D1 = 0. I should also note that Matlab gives different
solutions depending on the ordering of the equations, which makes me uneasy
about the results. It at least gives a solution when the k equation is solved first,
which is more than my ‘Method of Lines’ code which is unstable to the max. In
any event, I wanted to check the Matlab solution and solve a new system, with
only modifications to the w equation
!  
∂w ∂ 2βkx0 k(y, t)  −0.1y .01y 2 2 ∂ ∂w
− αnl w2 (5)

=− w + γ 0 e − D0 e kx0 + k w + Dnl w
∂t ∂y (kx0 2 + k 2 )2 ∂y ∂y

In this case, α ∼ γ0 and Dnl ∼ D0 . This version of the equations might be more
stable. Matlab gives me a solution, but it is independent of Dnl , which seems
absurd. I tried Dnl = .01 and Dnl = 1000 and it gave the same solution. Matlab
does give different solutions for different coefficients if I am only solving the one
equation, nonlinear diffusion problem. Any help or insight you can provide is
much appreciated. This system has proven quite difficult to handle.

You might also like