1. Limited time only! Sign up for a free 30min personal tutor trial with Chegg Tutors
    Dismiss Notice
Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Random walk in potential

  1. Apr 6, 2010 #1
    I have a problem concerning a one-dimensional random walk in potentials. Assume a one-dimensional space [0,1] and a probability distribution p(x). At every point x we have a probability p(x) to go left and 1-p(x) to go right. Assume some smooth distribution of p(x) with boundaries p(0) = 0 and p(1) = 1. Now begin a random walk at x=0.5 with some step-size dx (e.g. d = 0.01) and capture the position of the walker at every time-step t. The boundary constraint assures that the walker remains inside [0,1].

    I would assume that after sufficient time steps I get a steady distribution of the position of the walker. This would be equivalent to the probability distribution of finding the walker at some point in the potential.

    However, I have yet no idea how to calculate this distribution from some given p(x). I tried to set up a differential equation using the fact that in the steady case the flow from point x to x + dx and back must be zero. However, I would get a pole at x = 0.5 which is pretty useless. I can post the calculation if someone is interested.

    I would be glad for every hint how one could solve this problem. Thanks in advance!
    Blue2script
     
  2. jcsd
  3. Apr 7, 2010 #2
    I assume a fixed step size ; we can let it tend to zero later.
    Firstly, the boundary conditions don't warrant that the walker won't trip at 0 or 1 ( for instance, take d=1/300). If you want that constraint with initial position =1/2, the step size must be 1/2n for some integer n , making the walk discrete.
    In such a case, this can be modelled as a Markov chain with 2n+1 states ( x=1/2 & other 2n possible positions ;with transition probabilities p(xn)&c. at each state. The limit of the transition probability matrix yields the long run probabilities.
    Finally,letting d->0 gives a continuous time Markov process.

    P.S. : Is this a model of any physicsl phenomenon? I ask out of sheer curiosity.
     
Share this great discussion with others via Reddit, Google+, Twitter, or Facebook