Return Probabilities of a Simple Random Walk on Percolation Clusters

We bound the probability that a continuous time simple random walk on the infinite percolation cluster on $Z^d$ returns to the origin at time $t$. We use this result to show that in dimensions 5 and higher the uniform spanning forest on infinite percolation clusters supported on graphs with infinitely many connected components a.s.


Introduction
Consider Bernoulli bond percolation on Z d with parameter p. Recall that this is the independent process on Z d which retains an edge with probability p and deletes an edge with probability 1 − p.For all d > 1, there exists a critical parameter 0 < p c < 1 that depends on d, such that if p > p c , then a.s.there is a unique connected component with infinitely many edges [1].This component is called the infinite cluster.The phase where p > p c is called supercritical Bernoulli percolation.Supercritical Bernoulli percolation can be viewed as a random perturbation of the original graph.For more on this, see [13].It is then natural to ask what properties persist through this random perturbation.
In particular, we focus on the properties of a simple random walk on the infinite cluster.
Rayleigh's monotonicity principle says that if a simple random walk on a graph G is recurrent then a simple random walk on any subgraph of G is recurrent.This implies that a simple random walk on the infinite percolation cluster in Z 2 is recurrent.
Grimmett, Kesten, and Zhang proved that in Z d , d ≥ 3 a simple random walk on the infinite percolation cluster, C ∞ (Z d , p), is a.s.transient [14].Häggström and Mossel used a method of Benjamini, Pemantle, and Peres to show that a large class of subgraphs of Z d have transient percolation clusters [17] [7].Although on subgraphs transient percolation clusters corresponds well with transience, the methods in these papers tell us nothing about the probability that a simple random walk on the percolation cluster returns to the origin at time t.They also do not answer the question of whether two independent simple random walks on the percolation cluster will intersect infinitely often a.s.
The probability that a simple random walk on Z d returns to the origin at time t decays on the order of Ct −d/2 .The method of Varopoulos and Carne proves that the probability that a simple random walk returns to the origin at time t is greater than or equal to Ct −d/2 /(log t).In this paper we get a bound in the other direction.We show that this probability is bounded above by C t −d/2 (log t) 6d+14 .This result allows us to show that another property of the simple random walk persists after supercritical Bernoulli percolation.Namely, in Z d , d ≥ 5, two simple random walks intersect finitely often a.s.[21].In Section 9 we use the bound on the return probability to show that this property also holds on the infinite percolation cluster in Z d , d ≥ 5.
Together with results of [6], [5] and [19], this in turn shows that for Z d , d ≥ 5, the uniform spanning forest (USF) on percolation clusters is supported on infinitely many components.We define the wired, free and uniform spanning forests in Section 9.This corresponds with a result of Pemantle that the USF in Z d , d ≤ 4, is supported on trees, while in Z d , d ≥ 5 the USF is supported on graphs that have infinitely many connected components [22].The lower bound proves that two independent simple random walks on C ∞ (Z d , p), d ≤ 4, have infinite expected number of intersections.If this could be extended to show that two simple random walks have infinite intersections almost surely then this would prove that the USF on C ∞ (Z d , p), d ≤ 4, is supported on a tree a.s .

Outline
We write [j 1 , k 1 ] × [j 2 , k 2 ] × ... × [j d , k d ] for the subgraph of Z d that includes all vertices (v 1 , ..., v d ) such that j i ≤ v i ≤ k i for all i.When j i = j and k i = k for all i, denote this by [j, k] d .The edges of this graph are those that have both their endpoints in this set of vertices.For any graph G we write nG for the graph that is generated by replacing each edge of G with n edges in parallel.We say a distinguished vertex of nG is a vertex that corresponds with one in G.For any graph G we write V (G) for the vertices of G.
The outline of the return time probability proof is as follows.We want to show that there exist subgraphs of C ∞ (Z d , p) where we can calculate the return times.In particular we will show that for some α > 0 with high probability there exists a embedding of n[−2 αn , 2 αn ] d in C ∞ (Z d , p) centered around the origin.
It is possible to bound the return probabilities of a simple random walk on an embedding of n[−2 αn , 2 αn ] d in C ∞ (Z d , p) centered around the origin.We will then extend the embedding of n[−2 αn , 2 αn ] d to get a graph that spans the percolation cluster intersected with some finite ball.This subgraph will be the embedding with "bushes" attached to various vertices in the grid.The following is a picture of an embedding of 9[−1, 1] 2 with bushes.

an embedding of 9[-1,1]^2 with bushes
The bushes ensure that the subgraph spans the finite section of the percolation cluster.They also do not increase the return probabilities significantly.We then use a result of Benjamini and Schramm (Theorem 7.1) to show that the return probabilities of the percolation cluster are bounded by those on the spanning subgraph.Thus we obtain bounds for the return probabilities on the percolation cluster.
Section 5 shows that the probability that there exists a embedding of n[−2 αn , 2 αn ] d centered near the origin inside C ∞ (Z d , p) is increasing to 1 exponentially in n.The first step is to show that there exists an odd integer ρ such that given that x and y are connected, the conditional probability that there exists a path P that connects x and y such that |P | = ρ|x − y| is approaching 1 as |x − y| approaches ∞.This is done in sections 3 and 4. Then in Section 5 we show how to put these paths together to form a embedding of n[−2 αn , 2 αn ] d .
In Section 6 we show how to extend this grid to a spanning subgraph.We place a measure on such spanning subgraphs.In Section 7 include Benjamini and Schramm's proof of Theorem 7.1.In Section 8 we prove that with high probability the spanning subgraph has low return probabilities.Finally we use Theorem 7.1 to bound the return probabilities on the percolation cluster.In Section 9 we show that the bounds on the return times show that the USF on C ∞ (Z d , p), d ≥ 5 is supported on graphs with infinitely many connected components.

Approximate Pathlengths in the Percolation
In this section we show that if two points x and y are connected in the infinite cluster and far apart, then with high probability there exists a path between them in the cluster of approximately any length in a certain interval.Namely, for any length q ∈ [1.5, 36], there is a path of length qρ|x − y| ± ρ|x − y|.These paths are a first approximation for the sides of the grid we construct in Section 5.In Section 4 we construct paths of length exactly 5ρ|x − y|.These will be the sides of the grid.First we introduce some x and y if x and y are endpoints of edges in P .We also say P is a path from x to y.
We write x ∼ y if there exists an open path from x to y.We write x ∼ ∞ if x is part of the unique infinite cluster.If x ∼ y then we let D(x, y) be the length of the shortest open path from x to y.A path P is in a set of vertices V if the endpoints of every edge in P are in V .We write V (G) for the vertices of a graph.We say that G spans G if Throughout the next two sections we prove that there exists an odd integer ρ, such that if x ∼ y and |x − y| is large, then with high probability there exists a path P which connects x and y such that |P | = 5ρ|x − y|.We show that conditioned on x ∼ y, the probability that there does not exist such a path decreases exponentially in |x − y|.

We use the taxicab metric on
Define L(u, v) to be the elements in Z d which are within d/2 of the line segment joining (y).All constants A i and C i may depend on d and p.There are a number of places in the following sections where in order to be precise in choosing our parameters, it would be necessary to use the greatest integer function.We define our parameters without this as the calculations are already intensive.Similar estimates hold if the parameters are chosen using the greatest integer function when necessary.
We will show that for any q ∈ (1.5ρ|x − y|, 36ρ|x − y|) there exists a path P from x to y, lying in a certain region called the wedge between x and y, such that |P | is approximately q.Then in the next section we will use an inductive argument to show that there exists a path of exactly length 5ρ|x − y|.
There are three main tools that we will use to get a path of approximately the right length.The first is the following result of Antal and Pisztora.The second is the method of Grimmett, Kesten, and Zhang [14].This method was used to show the existence of a tree in C ∞ (Z d , p) with certain branching properties.This tree was used to show that for d ≥ 3, a simple random walk on C ∞ (Z d , p) is transient a.s.
The method works just as well for Z 2 [18].This method requires the following standard facts about the infinite percolation cluster.
Proof: For d ≥ 3 a proof of this can be found in [15].For d = 2 this follows from the work of Seymour and Welsh, and Russo [24] [23].
The third is a combination two theorems, one of Chayes, Chayes and Newman, and the other of Barsky, Grimmett and Newman, which gives the following result.Theorem 3.2 [9], [4] Given p > p c (Z d ) there exists C 3 so that Combining these last two theorems we get the following lemma.
There also exists a C > 0 such that for any w = (w 1 , ...w d ), any i and any j such that j ≤ w i ≤ j + k, Since diam(S) = n there exists i and y, y ∈ S such that y i − y i ≥ n d .Let α = n dk .Then there exists w 1 , ...w α ∈ S such that (w j ) i ∈ [y i + (j − 1)k, y i + jk).Then Now we will use a method similar to the one introduced by Grimmet, Kesten, and Zhang to show that infinite percolation clusters in Z d , d ≥ 3 are a.s.transient.We will use this method to show that for any two points x, y ∈ Z d the probability that there exists a reasonably short and direct path from near x to near y is going to 1 exponentially in |x − y|.
Define E r (z) to be the event that there exists w ∈ B r (z) 1. such that w ∼ ∞ and 2. any point w ∈ (B 2r (z)) C such that w ∼ w satisfies D(w, w ) ≤ ρ|w − w |.
Notice that E r (z) is not an increasing event.This makes some of the future arguments more technical since the FKG inequality will not apply.Lemma 3.3 There exists A 4 , C 5 > 0 such that for all r P(E r (z)) > 1 − A 4 e −C 5 r .Proof: By Lemma 3.1 the probability that line 1 in the definition of E r (z) does not hold is less than e −C 2 r .Theorem 3.1 implies that for an appropriate choice of A 4 and C 5 .
Define F r (z) to be the event that 1. z ∼ ∞ and 2. any point w ∈ B 2r (z) C such that z ∼ w satisfies D(z, w ) ≤ ρ|z − w|.
Let k = 1000.Fix = 1/2k 2 , and ρ > 1 an odd integer from Theorem 3.1.We define sets in Z d , called wedges, around x and y in which the various constructed paths from x to y will lie.The wedges allow enough room to construct paths of many different lengths, while ensuring that when the paths are concatenated, they do not intersect except near x or y.Define ∠xz, xy to be the angle between the line segments xz and xy.Define The following is a picture of the wedge between x and y.
x y . . . . the wedge between x and y detour Definition 3.4 A pair of points x and y in Z d is ( , ρ, ω) good, written x y, if there exists an open path P 0 from x to y such that for all q ∈ [1.5,36], there exists open path 2. P ⊂ W (x, y), Now we show that most pairs of vertices that are connected and far apart are good pairs.Lemma 3.5 There exists A 5 , C 6 > 0 such that Proof: The idea is as follows.By Lemma 3.1, if x ∼ y, then it is likely that there is a path in the percolation cluster whose length is bounded above by ρ|x − y|.We show that it is possible to pick a path, P 0 inside the wedge from x to y with length less than 1.5ρ|x−y|.Moreover we show it is possible to pick a different path so that the length falls within any specified range.To do this, we create P 0 and a disjoint collection of possible detours that are contained in the defined wedges.We lay down a series of evenly spaced points, called markers, on the line segment between x and y.These markers will be labelled z 1 i,0 .We also lay down a series of markers outlining the detours.For a given detour, these markers lie on 3 sides of a rectangle whose fourth side is part of the line segment from x to y.For the jth detour these markers will be labelled z 2 i,j , z 3 i,j and z 4 i,j .The labels 2,3 and 4 indicate on which side of the rectangle the markers lie.Then we show that with high probability there exists a point in a small ball around each marker such that for any two adjacent markers w and w , there is a path between the points corresponding to w and w whose length is bounded by ρ|w − w |.This follows from Lemma 3.3.
We construct P 0 by concatenating the paths between the points in the balls around the markers on the line segment between x and y and removing any loops.We construct the detours in an analogous manner.Each detour has both of its endpoints on P 0 .
Adding a detour means forming a path by replacing a portion of P 0 with the detour that has the same endpoints.We will choose our parameters so that if we add all of the detours to P 0 then the resulting path is long enough (≥ 36ρ|x − y|).On the other hand adding each detour increases the length of the path by a small amount.Hence it is possible to add the right number of detours so that the length of the path falls within the specified range.
Let m = 4ρ be the number of markers on the short side of each rectangle and M = 300ρ 2 be the number of markers on the long side of each rectangle.Define # = 60ρ/ to be the total number of possible detours.Let n = 1000ρ 2 / be the number of markers on the line segment between x and y, η = |x − y|/n be the size of the increments between the markers and v = (y − x)/|y − x|.Finally let v be a unit vector perpendicular to v.
We define the markers equally spaced throughout the region in which the path is permitted.Let ) holds for all i, j and l then x and y are a good pair.First we show that if the events E η/4 (z 1 i,0 ) hold for all i then there is a path P 0 from B η/4 (x) to B η/4 (y) inside L(x, y) + B ρη that has |P 0 | < 1.5ρ|x − y|.To see this, let Furthermore P 0 lies inside L(x, y) + B ρη .This is because the maximum that P 0 can travel away from the line between x and y is bounded by If F η/4 (x) and F η/4 (y) also hold then there is a path Now we show that if in addition the events E η/4 (z l i,j ) hold for all i, j, l then there are paths from near x to near y inside L(x, y) + B (ρ+M )η of many different lengths.Suppose in addition to the events E η/4 (z 1 i,0 ), F η/4 (x), and F η/4 (y) holding that for a fixed j the events E η/4 (z 2 i,j ), E η/4 (z 3 i,j ), and E η/4 (z 4 i,j ) hold for all i.Then there is a path P j such that 2ηM < |P j | < 3ρηM , P j ∩ P 0 = ∅ and the only vertices shared by P j and P 0 are the endpoints of P j .We can form a new path from x to y by cutting out a piece of P 0 and attaching P j .We say this path takes detour j.By this construction, the possible detours occur in the middle half of the line segment between x and y.This is because 2m# ≤ 2(4ρ)(60ρ/ ) < n/2.Furthermore, the maximum height of the detour marker This implies that These containment conditions imply that the P j are all disjoint.
Now suppose E η/4 (z l i,j ) holds for all i, j, l and F η/4 (x) and F η/4 (y) also hold.Then for any i ∈ [0, ..., #] we can form a path that takes exactly i detours.Since 2ηM < |P t | for each t a path that takes i detours has length at least 2ηiM .Thus there is a path that takes # detours which has length at least Adding one extra detour adds at most length to the length of a path.Thus if E η/4 (z l i,j ) holds for all i, j, l and F η/4 (x) and F η/4 (y) also hold then x and y are a good pair.Now we calculate the probability that all of this happens.Notice that n + 3#M is the total number of markers.Also notice that P(x ∼ y) ≥ P(x ∼ ∞) 2 .
Thus the lemma is true for some appropriately chosen A 5 and C 6 .
Corollary 3.6 There exists A 6 and C 7 so that

Exact Pathlengths in the Percolation
This section is devoted to showing that with high probability we can find a path in the percolation cluster from x to y of length exactly 5ρ|x − y|.We do this so that the sides of the constructed grid in Section 3 are all of the same length.This makes the estimates for the random walk easier than if the sides were only approximately the same length.Another approach to solving the problem is to construct a grid whose sides are all approximately the same length.Then include corresponding estimates in Section 8 for the more general graph.This method bypasses section 3.However, we were not able to establish these estimates.Thus we rely on the computationally intensive method of showing exact pathlengths in the cluster.We also feel that this result is of intrinsic interest.On the next page we give an overview of this section.
First we define a pair of points (x, y) to be very good if there exists many paths of specific lengths in the percolation cluster connecting x and y (definition 4.1).The main result of this section is Lemma 4.8 which shows that conditioned on x and y being connected, with high probability, (x, y) is a very good pair.The proof is by induction.
We need to establish the existence of many paths between x and y because the existence of 1 path of a certain length does yield a strong enough induction hypothesis.
The main idea of the induction in Lemma 4.2 is the following.At the nth stage we construct an approximate pathlength path between x and y (we can do this because of the results in Section 3).The path passes through points x and y which are close enough to each other to apply the induction hypothesis.The relative distances of x, y, x and y are important and are laid out in Lemma 4.2.Since (x , y ) is a very good pair, we can modify the path between x and y so that we know the exact length of the entire path.We set the parameters so that if at stage n an approximate pathlength is some length plus or minus , then at stage n − 1 we can construct exact pathlengths of order .The main idea for the induction step in Lemma 4.8 is encompassed in Lemma 4.2, which we outline in the following paragraph.
In Lemma 4.2 we want to construct a path P from x to y which is exactly some linear multiple of |x − y|.We assume that we have points x and y , with paths P 1 between x and x and P 2 between y and y such that |P 1 | + |P 2 | is approximately the desired linear multiple of |x − y|.We further assume that we can find a path P 3 from x to y that is exactly the difference between the linear multiple of |x − y| and |P 1 | + |P 2 |.We want to construct |P | by simple concatenation.The problem is that P may have loops.Hence we replace P 1 , P 2 and P 3 with paths P 1 , P 2 and P 3 which yields a path with no loops when concatenated and such that the new path is the desired length.
After we prove the main part of the induction step in Lemma 4.2 we give a Corollary 4.3 that is more technical but is actually the version we need.It weakens the hypothesis on x and y to ensure that with high probability we can fulfill the hypotheses by possible replacing x and y with nearby points x and y lying on the path between x and y .
The three lemmas after Corollary 4.3 are devoted to proving the base case of the induction.We now give an idea for the structure of this.In Definition 4.4 we define a very good n-ball around x to be one which contains a very good pair (y, z) such that |y − z| = k n .In Lemma 4.5 we show that if there exists very good balls of large diameter around x and x ∼ y, then the probability that (x, y) is a very good pair does not decay exponentially to 0 in |x − y|.Lemma 4.6 shows that with high probability there exists very good balls around a generic point x in the cluster.In Lemma 4.7 we show that the probability that (x, y) is a very good pair, given that the points are connected and a fixed distance apart is bounded away from 0. This lemma is the base case of the induction.
Once we prove Corollary 4.3 and Lemma 4.7, we are ready to prove the main lemma, Lemma 4.8.Notice the first half of the proof in this lemma is verifying that we can fulfill the conditions in Corollary 4.3.The second half of the proof establishes the corresponding estimates on the probabilities that each of these events occurring.Definition 4.1 A pair x and y in Z d is said to be (ρ, ω) very good, written x ∼ = y, if there exists a path P 0 from x to y satisfying the following property.For all q ∈ (3ρ|x − y|, 10ρ|x − y|) such that q + |x − y| = 0 mod 2 there exists P ⊂ ω such that 1. |P | = q, 2. P ⊂ W (x, y), and The proof that most pairs that are connected are very good is inductive.The main idea in the inductive step is contained in this next lemma.The three lemmas after that will allow us to get the induction started.Proof: The basic is idea is that given q we find paths P 1 , P 2 , and P 3 as follows.The definition of a good pair will allow us to choose P 1 so that P 1 connects x and x , P 1 ∈ W (x, x ), and |P 1 | a little less than .5q.It also allows us to choose P 2 so that P 2 connects y and y , P 2 ∈ W (y, y ), and |P 2 | a little less than .5q.Then since x ∼ = y we can choose P 3 so that P 3 connects x and y , P 3 ∈ W (x , y ), and Then form a path P by concatenating P 1 , P 3 and P 2 we get a path from x to y of length q.
The problem is that P may have loops.These can only occur in Let q 1 be the length of the loops near x and q 2 be the length of the loops near y .Notice that q 1 + q 2 < 2ρ|x − y|.Now modify P 1 to find P 1 with This is possible because Modify P 2 to find P 2 with |P 2 | in the same interval.Choose P 3 so that it connects x and y , P 3 ∈ W (x , y ).Finally find Since both ends of P 1 , P 2 , and P 3 are the same as the ends of P 1 , P 2 , and P 3 this modification does not change the concatenation in B 1/10 (x, y) or change the loops.Thus concatenating P 1 , P 3 , and P 2 and then removing the loops near x and y gives a path P from x to y of length q.Finally P ⊂ W (x, x ) ∪ W (x , y ) ∪ W (y, y ) ⊂ W (x, y).
The same argument gives us the following.Since P(w ∼ = w ) is the same for each of these pairs, this implies that P(w ∼ = w ) ≥ P(B 2k n is very good )/(2k nd ) 2 .Since there exist N and C so that By the previous remark, this implies there exists N such that if |x − y| = k n for some Lemma 4.6 There exist A 10 such that if |x − y| = k n for some n then Proof: To prove this we show that there are C 8 and A 7 such that This implies the result by the previous lemma.
The idea of the proof is as follows.It suffices to find a pair of points y, z ∈ B 2k n+1 (x) such that |y − z| = k n+1 and y ∼ = z.Since B 2k n (x) is very good, we first find points y and z in B 2k n (x) such that |y − z | = k n and y ∼ = z .Now we need to find points y and z that are the right distance from y , z and each other, and such that y y, z z.
This would allow us to apply Lemma 4.2.
Technical considerations force us to add another step to the process.We replace y and z with points y and z near y and z respectively, which lie on the path between y and z and such that y y and z z .Once we choose y, we need to choose z satisfying the hypotheses of Lemma 4.2 and such that |y − z| = k n+1 .Thus we choose z lying on ∂B k n+1 (y) and so that it is roughly colinear with y, y and z .In order to do this, we find z 1 and z 2 near this area, but on either side of B k n+1 (y) such that z 1 z 2 .
The path joining them will intersect B k n+1 (y).Let the intersection point be z.Once we find z, we now choose y and z to satisfy the hypotheses in Lemma 4.2.
If B 2k n (x) is very good then there exist y , z ∈ B 2k n (x) such that y , z ∼ ∞, y ∼ = z and |y − z | = k n .Let P be a path connecting y and z .Let v be a unit vector pointing from y to z .If there exist y, y , z, z such that Additionally suppose there exist Thus if all this holds then there exists y, y , z and z .Now we bound the probability that this happens.
The base case of the induction is proved in the next lemma.we can force C 9 to be much smaller than the other constants in the first equation.This implies that the terms with C 9 in them are the dominant terms of the inequality.Since e −.0016C 9 k n+2+α = e −1.6C 9 k n+1+α < e −C 9 k n+1+α the inequality is true for N sufficiently large.
Lemma 4.8 There exist A 8 and C 9 such that Proof: The proof is by induction.We will show that this is true with A 8 = 1 if |x − y| is sufficiently large.This implies the result.The previous lemma establishes the base case, which is n = N .Given that the induction hypothesis is true for k n we will show that it is true for all x, y such that k n+1 ≤ |x − y| ≤ k n+2 .This implies the result for all |x − y| ≥ k N +1 .The idea for the proof is as follows.We start with two points x and y that are connected and |x − y| is approximately k n+1 .We lay down about k disjoint copies of B 2k n in the middle of W (x, y).Using the induction hypothesis, we show that with high probability at least one of these balls contains a pair satisfying the hypotheses of Lemma 4.2.By applying this lemma to the pairs x, x , x , y , and y , y, we obtain that with high probability x ∼ = y.
We claim that if there exist x and y satisfying the following conditions then x and y are a very good pair.
1. x , y ∈ W (x, y) Lines 6 and 7 imply that x, y, x and y are all connected.Combining this with lines 8 and 2 shows that x x .Similarly line 3 and 9 show that y y .Thus we have satisfied the conditions of Lemma 4.2 and x and y are a very good pair.Now we bound the conditional probabilities given x ∼ y that lines 1 -6 are not satisfied, and that line 7, 8, or 9 is not satisfied .
Putting this all together gives Line 8 is by Lemma 4.7 and line 9 is by the choice of α.If there exists y i ∈ B k j (x + k j+1 v i ) \ B k j+1 (x) and z i ∈ P i \ B 2k j /3 (x) such that such that y i , z i ∼ ∞ and y i z i then there exists an appropriate P i .Conditions 1,2, and 4 are satisfied because y i z i .Condition 3 is satisfied because y i z i and

|P
which proves the lemma.
Lemma 5.3 There exists C 11 > 0 such that Proof: The proof is by induction.For the base case we will choose a J and C 11 such that for all n ≤ k J+1 , This is possible by the previous lemma, which states that there exists A 9 and J such that P(x is a k j hub for all j > J) > A 9 .
The ergodic theorem implies that for any j as J → ∞ P(∃x ∈ B k J such that x is a k j hub for all j > J) → 1.
Choose J so that for all j > J P( ∃x ∈ B k J such that x is a k j hub for all j > J) k < .5.
Choose C 11 so that 2. e −C 3 k j /P(h is a k j hub) + 2de −C 2 /2k j + A 6 e −C 7 k j ≤ .5e−C 11 k j+2 for all j > J.
The first condition implies the base case.Suppose the induction hypothesis has been proven for k j .We will now show that it is true for all k j+1 ≤ n ≤ k j+2 .If there is an ∼ ∞ for all i, and there are no y, z ∈ B 2n with y ∼ z, y z, and |y − z| > n/k 2 then h is an n hub.This follows from the argument in the previous lemma.
The events that there exists a k j hub in B k j (x i ) are independent.So By Lemma 3.2 we have that Thus P( ∃h ∈ B n such that h is an n hub) For an α = α(d) which will be fixed later define the graph Then we define the n grid to be nS(α, d, n).Remember a distinguished vertex of nS(α, d, n) is a vertex that corresponds with one in S(α, d, n).In this case, they are the corner points of the n grid.We say there exists a copy of the n grid around 0 in ω if there exists an embedding of the n grid in ω with the distinguished vertex corresponding Proof: By Lemma 5.3 the probability that there are not the desired n hubs is less than ).The probability that they are not all connected to the infinite cluster is less than (2 αn+1 + 1) d e −C 3 n/100ρ /P (x ∼ ∞) 2 .By Lemma 4.8 the probability that the desired pairs are not very good is less than (2n2 αn ) 2d (A 8 e −C 9 n/4ρ ).Thus the constants can be chosen.
This fixes α for the rest of the paper.

Constructing the spanning subgraph
A connected graph Q is a good n graph around 0 if 1. there is a subgraph of Q which is a copy of the n grid around 0 2. each v which is connected to the n grid is connected by a path of length at most 4nd, and 3. for no v does there exist disjoint paths in Q that connect v to the n grid.
Notice that the third condition implies that any walk in Q which leaves the n grid at vertex x must return to the n grid at vertex x.Given a vertex v in the n grid we define the bush attached to v as follows.It contains any vertex v in Q such that all paths from the n grid to v go through v.It also contains all vertices which are connected to v by a single edge.
Any walk in Q generates a unique walk in the n grid.Thus it generates a unique nearest neighbor walk in Z d .This property will be used heavily in the next section.
Define C(x) to be the open cluster containing x. Define then there is a good n graph around 0 that spans C n (0).
Proof: By Lemma 5.4 the last two conditions imply that there is a copy of the n grid around 0. We will inductively show that we can attach each vertex of C n (0) to the n grid in a way that is consistent with the definition of a good graph.Suppose x is a vertex in C n (0) but not in the n grid.Then by the third condition there is a path from x to the n grid of length at most nd.This path can be chosen to intersect the n grid in only one vertex.The union of this path and the n grid satisfies the conditions of a good graph.Now we want to show that if we have an extension E of the n grid that satisfies the conditions of a good graph then we can extend it to include at least one more vertex so that the new extension E still satisfies the conditions of a good graph.Let x be a vertex in C n (0) but which is not in E. By the second property there exists a path, P , connecting x to the n grid with length at most nd.
The problem is P ∪E might not satisfy the conditions of a good graph.We will now pare down P ∪ E to get a new graph E which spans E ∪ P and also satisfies the conditions of a good graph.Starting at the n grid move on P towards x.Stop at the first vertex v in V (E).From v there are two disjoint paths to the n grid.Remove the edge touching v in the longer of the two paths.If they are both the same length then remove either one.Now start at v and move along P toward x and repeat this procedure until you get to x.Call the resulting graph E .Now we claim that E is a good n graph.Suppose there is a vertex y with two disjoint paths to the n grid.Then at least one edge of one of the paths is in P and one edge is in E. Thus there is a vertex in V (E) ∩ V (P ) connected to one edge in E ∩ E and one edge in P ∩ E .But whenever this situation arose we removed one of the two edges.
Thus there is no such vertex.Also by the way we chose to remove edges the distance from any vertex in V (E ∪ P ) to the n grid does not increase.Thus the distance from any point in V(E ∪ P ) is at most nd.Thus the resulting graph is a good graph.Thus we can construct successive extensions until we get a good n graph that spans C n (0).Lemma 6.2 P(∃ a good n graph around 0 that spans Proof: The proof is the same as the proof of Lemma 5.5.
We have shown that good n graphs exist with high probability.We would like to be able to bound the return probabilities on any good n graph.Unfortunately we don't know how to do this.So we will place a measure on good n graphs and then bound the return probabilities on a typical good n graph.This will be done in Section 8.
For a good n graph Q and e = (v to be the portion of Q between the distinguished vertices corresponding to v 1 and v 2 .
Our choice of the measure µ n will be done in such a way that the random variables Lemma 6.3There exists C 13 and measures µ n on subgraphs of ω such that Proof: We will construct a sequence of distributions on subgraphs of ω.At each stage we pick a graph from the previous distribution.Then we extend the graph in a collection of disjoint regions.If at any stage we cannot complete the proscribed extension in a given region then we add nothing to this portion of the graph.We refer to this portion of the graph where we could not extend it as stunted.In future stages, if some portion of the graph in the region that we are trying to extend the subgraph in is stunted then we do not extend the graph in this region.
For each v ∈ S(α, d, n) consider all the which are n/100ρ hubs.We construct a distribution so that all of these have the same probability of being the distinguished vertex corresponding to v. We also do this independently for all v ∈ S(α, d, n).
For a given choice of the distinguished vertex, x v , consider the set of all possible sets of 2d paths which satisfy the definition of x v being an n/100ρ hub.Form a new distribution by assigning each of these sets of paths equal mass.Do this independently for all v. Let H(v) be the n hub and paths corresponding to v. We have defined our measure such that the distribution of H(v) depends only on ω in Fix a choice of H(v) for all v ∈ S(α, d, n).For each pair v 1 , v 2 ∈ S(α, d, n) such that 3. have length n, and 4. agree with Form the next distribution by giving all of these paths equal mass.We form a distribution on n grids by doing this independently for all such pairs in S(α, d, n).Thus for any α, d, n) the distribution of the path connecting x v 1 and x v 2 depends on ω in the region.
Fix a choice, H, of the n grid.For each v ∈ S(α, d, n) consider the set of all subsets S of ω such that: 4. Every vertex y ∈ V (S ) is connected to H by a unique path.
We define the measure µ n by giving all of these graphs equal measure.We do this independently for all hypercubes.For each adjacent pair ) is determined by ω in the region C 13 such that, conditioned on the event that Q is a good n graph around 0 that spans If the appropriate n hubs exist and every pair in the region which are connected are very good then the proof of Lemma 5.5 shows that Q is good n graph around 0 that spans C n (0).Thus Proof: The distributions of W (e 1 ), . . ., W (e k ) are independent as they depend on ω in disjoint regions.The event that none of W (e 1 ), . . ., W (e k ) are stunted means that a certain event did not occur in any of those regions.Thus W (e 1 ), . . ., W (e k ) are conditionally independent.

Monotonicity of average return probability
This section is due to Itai Benjamini and Oded Schramm.Let F be a finite graph where each edge (i, j) has a conductance a i,j ≥ 0. (If the conductances are not specified then they are assumed to be one across every edge of F .)Let Z be the nearest neighbor process on V (F ) where the rate of moving from i to j is a i,j .Define Theorem 7.1 Fix any t > 0. v p F t (v, v) is monotone non-increasing in the conductances.
Proof: Fix some time t > 0, and let z = v p F t (v, v).Observe that z = trace (exp(tA)) where A = (a i,j ) and a i,j is the conductance of the edge {i, j} if i = j and a i,i = − k =i a i,k .Now let tr = trace.It follows from tr(AB) = tr(BA) and the linearity of tr that dz = dtr(exp(tA)) = tr(exp(tA)dA).
We may take w.l.o.g.dA to be of the form a 11 = a The corollary extended to any transitive amenable graph.Note that Barlow and Perkins [3], constructed a subgraph S of Z d , for which for an unbounded sequence of times.
Also define We can bound E because Thus by [12] the covering time of ∪ w ∈N (w) Q((w, w )) is at most (8nd) 3d+3 .This implies Given t choose n to be the smallest integer such that n ≤ log t for all t.This implies that there exists C 14 such that E ≤ C 14 (log t) 3d+3 .

Fix a good graph
We define the set Z = Z Q,v to be the set of all z ∈ {S(α, d, n)} (2 αn/2 ) such that where C 15 will be defined later.Let Ṽ ⊂ V (Q) be the set of all v such that has We say a graph Q is great for t if it is a good n graph and Q(e 1 ), . . ., Q(e k ) are not stunted.
As in line 11, for any distinguished vertices w and w and Q If the site in S(α, d, n) which z visits most often in the first (t/E) steps was visited at most (d + 1)(log(t/E)) 2 times then is the sum of independent random variables which are bounded by Their variance is bounded by C 2 (log t) 6d+10 .Applying Bennett's Inequality (see [25] page 851) we see that if we let Thus there exists constants so that happens with probability 1 − C t −2d 2 .
By [11] the fraction of z for which one vertex is visited too often is less than (t/E) −πd .
Thus the fraction of z and Q such that Q is not good or Q and z don't satisfy line 12 is then for any good graph Q and z the conditional probability that given Q and z is less than t −2d 2 .
Proof: The sequence (indexed by i) )) is at most (8nd) 3d+3 for any Q and z and x > 0 This implies Applying Bernstein's inequality (see [25] page 855) we get Q and z are such that the conditional probability that inf{k : and is at least 1 − t −d 2 .
Proof: We choose C 15 so that C 15 (t/E) Lemma 8.6 There exists C 18 such that the expected value of is less than Proof: As there are at most (8nd) d+1 vertices attached to the n grid between any two distinguished vertices The previous lemmas imply that the expected value of the last sum is at most Thus the expected value of line 18 is at most This implies that P(the n graph centered around 0 is great for t) For any Q and distinguished vertex w, the covering time of ∪ w ∈N (w) Q((w, w )) is at most (8nd) 3d+3 .Thus the conditional probability of line 20 given Q and z is at most 2 −(8nd) 3d+4 /(8nd) 3d+3 ≤ 2 −8nd ≤ t −2d .
To bound the probability of line 21 we condition on any z ∈ V (Q) N which is consistent with z.Let j be such that the ith distinguish vertex of z is z j .The distribution of m given z is the distribution of the jth occurrence of a Poisson process with parameter one.Call that distribution D j (m).Since  9 The Uniform Spanning Forest on C ∞ (Z d , p) In this section we show that two independent simple random walks on C ∞ (Z d , p), d ≥ 5, have finite expected number of intersections.We then define the wired, free and uniform spanning forests on a graph G.We will combine the expectation result with some results about spanning forests to show that the USF on C ∞ (Z d , p), d ≥ 5, is supported on graphs with infinitely many connected components.
Choose ω such that 0 ∈ C ∞ (Z d , p).Let {Z(t)} t≥0 and {Z (t)} t≥0 be two independent copies of simple random walks on C ∞ (Z d , p) with Z(0) = Z (0) = 0. Let Z(t) represent the vertex that the simple random walk occupies at time t.Given a vertex a in the percolation cluster let {X a (t)} t≥0 be a simple random walk in the percolation cluster For each path from 0 to a in the percolation cluster there is a corresponding path from a to 0 (just go in reverse).These two paths have probabilities that differ by a constant.The probability of a path from 0 to a is 1 over the product of the degrees of the vertices at the beginning of each segment.The probability of the corresponding path from a to 0 is the product of the degrees of the vertices at the end of each segment.Since these sets of vertices only differ by one element, these two probabilities differ by a factor of the degree of a / degree of 0. Line 27 is because of the independence of X and Z and the independence of the increments in simple random walk.Line 29 follows from Theorem 8.1 and because the sum of the return times for discrete time simple random walk is bounded by a constant times the sum of the return probabilities for continuous time simple random walk.
We now define spanning forests.Given a finite graph G = (V, E), a spanning tree of that graph is a connected subgraph which contains no cycles and includes all the vertices.
notation.Let ω ∈ {0, 1} Z d be a realization of percolation.An open edge e is one such that ω(e) = 1.An open path P is a connected set of open edges with no cycles such that two vertices have degree 1.These are the endpoints of P .An open path P connects

Lemma 3 . 2
Let S be a connected set with diam(S) = n.Then there exists C 4 such that P(S ∼ ∞) ≤ e −C 4 n .Proof: This argument was given to us by Y. Zhang.For d = 2 the result follows easily from the arguments in [13] on pages 194 and 195.Let d ≥ 3 and p > p c (Z d ).Then there exists k such that for any fixed i, p > p c ({(x 1 , ..

Corollary 4 . 3 1 20Lemma 4 . 5
Suppose x , y ∈ W (x, y) such that 3 8 |x − y| < |x − x |, |y − y | < 5 8 |x − y|, x ∼ = y and P 0 is the base path between them.Also suppose 1 k 2 |x − y| < |x − y | < |x − x | and similarly for |y − y |.If there exists x ∈ V (P 0 ) ∩ B .01|x −y | (x ) and y ∈ V (P 0 ) ∩ B .01|x −y | (y ) such that x x and y y then x ∼ = y.The main result of this section is that P(x ∼ = y | x ∼ y) is converging to 1 exponentially in |x − y|.Before we show this, we show in the next two lemmas that it is not converging to 0 exponentially in |x − y|.This step is necessary to prove the base case of the induction.Definition 4.4 B 2k n (x) is very good if there exists y, z ∈ B 2k n (x) such that y, z ∼ ∞, y ∼ = z and |y − z| = k n .If there exists an C 8 and a A 7 such that P(B 2k n+1 (x) is very good | B 2k n (x) is very good) > 1 − C 8 e −A 7 k n , then there exist an N and A 10 such that if |x − y| = k n for n ≥ N then P(x ∼ = y | x ∼ y) > A 10 |x − y| −2d .Proof: Notice that |w−w |=k n ,w,w ∈B 2k n P(w ∼ = w ) ≥ P(B 2k n is very good ).

Lemma 4 . 7
For any d there exists an N = N (d) and C 9 (N ) such that for all n ≥ N and all 0 ≤ α ≤ 1 k n+1+α and P(x ∼ = y | |x − y| = k N and x ∼ y) < e −C 9 k N .Proof: Fix d.By Lemma 4.6 the last equation is satisfied if e −C 9 k N = 1 − C/k N d .This defines C 9 in terms of N .Notice as N → ∞, C 9 → 0. By choosing N sufficiently large,

2 .P
(line 7 does not hold | x ∼ y) = P(x, y ∼ ∞ | x ∼ y) ≤ e −C 3 k n+1 by Lemma 3.2.Corollary 3.6 tells us that P(lines 8 or 9 do not hold | x ∼ y) ≤ 2P(lines 8 does not hold)/P(x ∼ y) xz ≤ 20}, and 5. V (P i ) ∩ ∂B k j+1 (x) = ∅ then x is a k j+1 hub.Conditions 2 and 5 above and condition 1 from the definition of an n hub show that a subset of P i ∪ P i satisfies condition 1 in the definition of an n + 1 hub.Condition 1 above and condition 2 in the definition of n hub imply condition 2 in the definition of n + 1 hub.Condition 3 above and condition 3 from the definition of an n hub show it satisfies condition 3 in the definition.Condition 4 above shows it satisfies condition 4 in the definition.

Lemma 5. 4 Lemma 5 . 5
If 1. for each v = (v 1 , ..., v d ) ∈ S(α, d, n) there exists an x v ∈ B n/100ρ ((n/5ρ)v) which is an n/100ρ hub and x v ∼ ∞ and 2. each pair z, z ∈ nS(α, d, n) such that |z − z | > n/4ρ and z ∼ z is a very good pair then there is a copy of the n grid around 0 in ω.Proof: Given a v pick the distinguished vertex corresponding to v to be an n/100ρ hub in B n/100ρ ((n/5ρ)v).The fact that distinguished vertices are n/100ρ hubs gives us the start and end of the path between any two neighboring distinguished vertices.The ends of these two paths form a very good pair.Thus we can connect them with a path of the right length so that when we remove loops the distance between any two distinguished vertices is n.It is easy to check that these paths are disjoint.There exist α, C 12 > 0 and A 11 such that P(∃ an n grid around 0 in ω) > 1 − A 11 e −C 12 n .

µCorollary 6 . 4
n (Q is good n graph around 0 that spans C n (0)) > 1 − A 11 e −C 12 n .If e 1 , . . ., e k are separated by at least C 13 then the random variables W (e 1 ), . . ., W (e k ) are independent conditioned on the event that none of W (e 1 ), . . ., W (e k ) are stunted.

Theorem 9 . 1
started at a (X a (0) = a).Let I(Z, Z , C ∞ (Z d , p)) = the number of intersections of Z and Z .For almost every ω the number of intersection of two independent random walks is finite a.s.In other words, with probability oneI(Z, Z , C ∞ (Z d , p)) < ∞.Proof: The following argument was indicated to us by Russell Lyons.We only need to show that the expected value of I(Z, Z , C ∞ (Z d , p)) is finite.E(I(Z, Z , C ∞ )) ≤ Ek j a∈Z d P(Z(k) = a and Z (j) = a) (k) = a and Z(k + j) ≤ because you might double count intersections.Line 24 is because of the independence of Z and Z .Line 25 is true for the following reason.
).There exists C 16 and C 17 such that ify ∈ [−2 αn + 2 αn/2 , 2 αn − 2 αn/2 ] d ,then the fraction of Q and z such that Q is not good or we define an equivalence class by e ∼ f = (w 1 , w 2 ) if ∃k ∈ Z d such thatw 1 = v 1 + C 13 k and w 2 = v 2 + C 13 k,Label the equivalence classes L 1 , L 2 , . . ., L 2dC 13 d .When we condition of the event that Q is a good n graph T Q (e 1 ), . . ., T Q (e k ) are mutually independent By Lemma 6.3 the probability that Q(e 1 ), . . ., Q(e k ) are not stunted is at most A 11 e −C 12 n .The reason for the construction of µ n in Lemma 6.3 is that for any j and By Corollary 6.4 for any set e 1 , . . ., e k ∈ L j , with e i = e j only if i = j, the random variables T Q (e 1 ), . . ., T Q (e k ) are mutually independent conditioned on the event that Ify ∈ [−2 αn + 2 αn/2 , 2 αn − 2 αn/2 ] By Lemma 8.1 the fraction of Q and z for which condition 1 is not satisfied is less thanC 17 (t/E) −πd .By Lemma 8.2 for all good graphs Q and z the conditional probability for which condition 2 is not satisfied is greater than 1 − t −2d 2 .
1/2 (log t) 3d+8 > 3C 16 √ t(log t) 3d+8 .The proof of this lemma is the same as the proof of the previous three except that t/E is replaced by If, for a given Q, y 1 and z lines 15, 16, and 17 are all satisfied then the conditional 1 2 C 15 (t/E) 1/2 (log t) 3d+8 .Let the event A Q,y,z be the event that either line 15, 16, or 17 is not satisfied for y, Q, and z.Lemma 8.5 If 1. v ∈ Q(Q(y 1 ), Q(y 2 )),2.z 1 = y 1 , and3.z / ∈ Z Q,vthen A Q,y 1 ,z occurs.Proof: log t) d+1 .There exists C 19 such that for any t If Q is not great for t then the number of v ∈ V (Q) with v / ∈ Ṽ is at least (t −.51d )|V (Q)|.Thus the sum in line 18 is at least(t −.51d )|V (Q)|(t −.51d )(2d) 2 αn/2 .