I'm curious if there is a known asymptotic scaling for the return-to-origin (i.e. recurrence) probability for a random on $Z^d$ as a function of $d$?

Mathworld gives the recurrence probability: http://mathworld.wolfram.com/PolyasRandomWalkConstants.html

$p(d) = 1-\frac{1}{u(d)} = 1- \left(\int_{t=0}^{\infty}I_0(\frac{t}{d})^de^{-t} dt\right)^{-1}$

Where $I_0$ is a modified Bessel function of the first kind: http://mathworld.wolfram.com/ModifiedBesselFunctionoftheFirstKind.html

Taking a wild guess, I found that a quartic fit seems to work well for values of $p(d)$ for $d = 1$ to $8$: http://www.wolframalpha.com/input/?i=quartic+fit+%7B3%2C+0.340537%7D%2C%7B4%2C+0.193206%7D%2C%7B5%2C+0.135178%7D%2C%7B6%2C+0.104715%7D%2C%7B7%2C+0.0858449%7D%2C%7B8%2C+0.0729126%7D

Also, what's the best way to approximate the integral expression necessary to calculate $p(d)$ if we only need $k$ digits of precision?

Sorry folks: Pólya's Random Walk Constants at infinity

I should have done a more thorough job checking for previous questions about this topic. =(

  • $\begingroup$ corrected parentheses in formula $\endgroup$ – Gerald Edgar Dec 26 '12 at 15:32
  • $\begingroup$ Montroll's paper referred in your link to wolframalpha has asymptotics when d tends to infinity. $\endgroup$ – Alexandre Eremenko Dec 26 '12 at 16:32

Asymptotic in the sense $d \to \infty$? From Maple I get: $$ p(d) = \frac{1}{2 d} + \frac{1}{2 d^{2}} + \frac{7}{8 d^{3}} + \frac{35}{16 d^{4}} + O \Bigl(d^{-5}\Bigr) $$ as $d \to \infty$.


Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.