Take the 2-minute tour ×
MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

Let $G$ be the Green function of the simple random walk on $\mathbb{Z}^d,\:d\geq 3$; i.e. $$G(x) = E \sum_{i=0}^{+\infty} 1_{X_i=x},$$ where $X$ is the simple random walk starting from $0$. The behaviour of $G(x)$ is well know when $x$ is large. But do we know how the function behaves when $x$ is small?

I tried to look in book of Spitzer (Principles of RW) and Lawler, Limic (Random Walk: Modern Introduction). I only found "inverse Fourier transform" formula which I do not have idea how to use (besides some numeric calculations).

May be some of you know other references?

My question stems from esimation of $$S_d:=\sum_{x\in \mathbb{Z}^d\setminus \lbrace 0 \rbrace} a_x \left(\sqrt{1+\frac{G(x)^2}{G(0)}}-1\right),$$ where $a_x$ are some coefficients. In the simplest case $a_x =1$ (or $a_x$ = 1 if $\sum_{i=1}^d x_i$ is odd and -1 otherwise; "chessboard of -1, 1").

In the case of $a_x=1$ one can easily prove that $S=\infty$ if $d=3,4$ and $S<\infty$ if $d\geq5$ (just using the fact that $G(x) \sim |x|^{2-d}$ when $x$ is large.)

But is it possible to say anything more about $S_d$ when $d>5$? E.g. I expect that $S_d \rightarrow 0$ as $d\rightarrow +\infty$.

I made a mistake in the first version the sum was over $\mathbb{Z}^d$ and it should be over $\mathbb{Z}^d\setminus \lbrace0\rbrace$.

Update 24.05.2011 Thanks to observations made in Didier Piau proof I was able to prove that indeed $S_d \rightarrow 0$ (in the case when $a_x=1$). I put the proof here.

share|improve this question
    
See section 2 of google.com/… –  Steve Huntsman May 22 '11 at 2:01
add comment

1 Answer

In fact $S_d$ does not converge to zero when $d\to\infty$, at least if $a_x=1$ for every $x$. Here is a proof.

For every $x$, $G_d(x)=P_0^{(d)}(\mathtt{hit}\ x)G_d(0)\le G_d(0)$ hence $G_d(x)^2/G_d(0)\le G_d(0)$. For every $z$ in $[0,G_d(0)]$, $\sqrt{1+z}-1\ge c_dz$ with $$ c_d=\frac1{G_d(0)}[\sqrt{1+G_d(0)}-1]=\frac1{\sqrt{1+G_d(0)}+1}. $$ Hence, for every $(a_x)$, $$ S_d\ge c_dT_d,\quad\mbox{with}\ T_d=\sum_xa_xG_d(x)^2/G_d(0). $$ In the special case where $a_x=1$ for every $x$, by reversibility, $$ \sum_xG_d(x)^2=\sum_{i,j}\sum_xp_i^{(d)}(0,x)p_j^{(d)}(x,0)=\sum_{i,j}p_{i+j}^{(d)}(0,0)=\sum_{i}(i+1)p_i^{(d)}(0,0), $$ hence $$ \sum_xG_d(x)^2>\sum_{i}p_i^{(d)}(0,0)=G_d(0). $$ In particular, $T_d>1$. Now $G_d(0)$ is a nonincreasing function of $d$ hence $c_d\ge c_5$ for every $d\ge5$ and $S_d>c_5$ for every $d\ge5$.

Edit Much simpler: if $a_x\ge0$ for every $x$, then $S_d>a_0(\sqrt{1+G_d(0)}-1)$ and $G_d(0)>1$ hence $S_d>a_0(\sqrt2-1)$.

share|improve this answer
    
What is the error? –  Did May 22 '11 at 19:35
    
Sorry, no error. I corrected my previous comment. –  Piotr Miłoś May 24 '11 at 9:16
    
As you observed the question in the first version was somehow trivial. What I meant was a sum over Zd∖{0} :), sorry for that. I am not sure at the momement if your proof can be improved to handle this situations as well –  Piotr Miłoś May 24 '11 at 9:17
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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