# Numerical Approximation of a 4D System of ODE's

1. Jun 9, 2013

### TylerH

What is the most general method of approximating arbitrary systems of ODEs of 4 variables(x,y,z,t) that fit these conditions? The conditions that are assumed true of the ODEs are:
1) that I require differentials to be explicitly defined (but they can be defined in terms of other differentials).
2) that time is assumed linear (ei, it's a 3 variable system described in terms of 4, but the 4th is known)

That is:
dx=y^2+sin(x)
dy=-dz+y*t
dz=t*z

would be valid, because dx,dy, and dz are explicitly defined and alone on one side.

dx=f(dx)

would be invalid, because I don't want to solve implicit equations.

f(dx,x,y,z,t) = g(x,y,z,t)

would be invalid, because I don't want to have to move all the junk to the other side to get an explicit definition of dx.

2. Jun 10, 2013

### bigfooted

A (4-stage) Runge-Kutta method usually works pretty well for systems of explicit coupled ode's that are not too stiff. If your equation becomes very stiff, you probably need to switch to specialized stiff solvers, which are usually (semi-)implicit.

3. Jun 10, 2013

### TylerH

What are some implicit methods that you would recommend? With generality being the most favored quality. I'm okay with the algorithm being harder and slower, it's all going to be precomputed anyway.

When I said I didn't want to solve implicit equations, I was only really talking about avoiding symbolic math.

Thanks for the response.

4. Jun 11, 2013

### bigfooted

You could start with the backward Euler method, it is unconditionally stable and the easiest of a large class of implicit methods (adams-moulton)