Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Method of Characteristics for Hyperbolic PDE

  1. Apr 16, 2013 #1
    I am trying to build a program in Matlab to solve the following hyperbolic PDE by the method of characteristics

    ∂n/∂t + G(t)∂n/∂L = 0

    with the inital and boundary conditions

    n(t,0)=B(t)/G(t) and
    n(0,L)=ns

    Here ns is an intial distribution (bell curve) but I dont have a function to fit it. I just have different values of n at various sizes, L (from 0 to 1000μm).

    I know that the method of characterisitics gives the solution

    ∂n/∂s=0
    ∂t/∂s=1
    ∂L/∂s=G(t)

    which gives the characteristic curve as L-G(t)*t=L0. So the values of n @ any initial size, L0, remain constant along the characteristic curve.

    My problem is how do I incorporate the original n values at time zero and various sizes into the equation. I want to say:

    n@(L0)=n@(L-G(t)*t)

    If the initial condition was given as a function of L0, say n(0,L)= f(L0)=sin(L0), then I could just sub in the characteristic equation and calculate n(t,L) as sin(L-G(t)*t). But I dont have a function that describes the initial condition, just vaues of n at various sizes, L as said above.

    The other problem is the boundary condition, this should make up one section of the solution and the initial condition the other.

    But again how do I actually say n@(t,0)=B(t)/G(t)=n@(L-G(t)*t).

    Say if n(t,0)=g(t), so the boundary condition is a function of time explicitly, not the case for me, B and G are actually a function of concentration which is changing with time, then the solution for the PDE above subject to the boundary would be;

    n(t,L)=g(t-(L/G(t))

    So I get the method overall, Im just confused how I propagate each value when I dont have functions to describe the initial and boundary conditions, like f(L0) and g(t) shown above.

    I posted about this ages ago but if anyone had some feedback on this it would be great.
     
  2. jcsd
  3. Apr 17, 2013 #2
    The characteristic curve is

    [tex]L=L_0+\int_0^t{G(\xi)d\xi}[/tex]
     
Know someone interested in this topic? Share this thread via Reddit, Google+, Twitter, or Facebook