I have 2nd order $1+1$ dimensions PDE (a wave equation) that I have rewritten as a coupled system of first order equations and discretized the spatial dimension using a grid of $N$ points labeled by an index $i$ and spacing $h$ such that $r_i = r_0 + i*h$ with $r_0 > 0$ with the intent to solve it numerically using the method of lines.

To reduce the order of the PDE, I've used the following definitions:

$$\Pi(t,r) \equiv \partial_t\phi(t,r) \quad(1) $$ $$\psi(t,r) \equiv r^2f(r)\partial_r\psi \quad(2). $$

At the endpoint the grid, using fourth order finite difference formulas for the spatial derivatives the system takes the form

$$ \partial_t\Pi_N = \frac{f(r_N)}{r_N^2}\left[\frac{25\psi_N - 48\psi_{N-1} + 36\psi_{N-2} - 16\psi_{N-3} + 3\psi_{N-4}}{12h} \right] - \frac{l(l+1)f(r_N)}{r_N^2}\phi_N \quad (3)$$

$$\partial_t\psi_N = r_N^2f(r_N)\left[\frac{25\Pi_N - 48\Pi_{N-1} + 36\Pi_{N-2} - 16\Pi_{N-3} + 3\Pi_{N-4}}{12h} \right] \quad (4)$$

$$\partial_t\phi_N = \Pi_N \quad (5).$$

I would like to apply absorbing boundary conditions at point $r_N$, such that at $r_N$ the following must be true

$$ \partial_t\phi + \partial_r\phi + \frac{\phi}{r} = 0 \quad (6),$$

or equivalently using (1) and (2),

$$\Pi_N + \frac{\psi_N}{r_N^2 f(r_N)} + \frac{\phi_N}{r_n} \quad(7).$$

My question is the following: To apply the abosrbing B. C. to the system, should I:

  1. Substitute $\psi_N$ from (7) into (3) inside the derivative (the bracketed part of the expression) and substitute $\Pi$ from (7) into the derivative part of (4) ?
  2. Substitute the whole derivative (the whole bracketed expression) using (6) and then rewrite $\partial_t\phi$ using (1) ?
  3. Something else entirely?

Anny help would be appreciated.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.