# Runge Kutta to solve higher order ODE

1. Feb 20, 2017

### MechEngrStdnt

1. The problem statement, all variables and given/known data

Edit* should say F'(0) = F(0) = 0

2. Relevant equations
I know that I typically need 3 equations for a 3rd order ODE, does this apply if the is no F'? In the picture above are the equations I came up with, am I on the right trail? Lastly I am familiar with RK4, however I have never used it to solve a system... how does one go about this? can it be done in excel? I took Diff Eq years ago and they really glossed over the numerical methods.

3. The attempt at a solution
The picture above is the question statement (at the top), and my attempt at breaking it into a system of equations.

2. Feb 20, 2017

### BvU

Yes.
Yes. Looks good.
Yes. I have a proposal: start with a simple integrator: forward Euler
You have two initial conditions and one boundary condition. The easiest approach is the shooting method : guess a value for F''(0) and integrate. Adjust until for $x\rightarrow \infty$ F'(x) is close enough to one. $x\rightarrow \infty$ is a bit bothersome, so if F'(x) $\approx$ 1 and doesn't change too much any more you should be satisfied.

So what you want to solve is F''' = $-$F F'' / 2

x, F, F', F'', F''' and if your first guess for F''(0) is, e.g., 1, the first row looks like: 0, 0, 0, 1, '= $-F * F'/ 2$ ' (if you know what I mean).

Choose a step size d, e.g. 0.01 then your next row looks like (supposing x(0) is in cell A4 )
Code (Text):

=A4+dx $\quad$ =B4+dx*C4 $\quad$ =C4+dx*D4 $\quad$ =D4+dx*E4 $\quad$ =(-1)*(B5*D5*2)
which you can copy/paste from A6 all the way down to infinity (for me 400).

Take some time to understand these formulas: Euler is simply F(x+dx) = F(x) + F' dx $\quad$.

Then fumble with F''(0) until the boundary condition appears to be met reasonably well.

It's handy to use a defined name for dx so you can vary it to check if your step size is reasonable. (dx 0.01 and then Formulas | Create from selection)

3. Feb 20, 2017

### MechEngrStdnt

Thanks for the reply! I am still a little hung up on accounting for the boundary condition of F'(Inf) = 1, how do I account for this? I've only dealt with initial conditions before, which are much more obvious. Thanks!

4. Feb 21, 2017

### BvU

So deal with guessed initial conditions and keep guessing until the boundary condition is satisfied. Do you understand the explanation in the shooting method link ?