This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
1, is a consequence of Dynkin's proof and Theorem 2.3 of Adler and Lewin (1991).
Adler & Lewin: Intersection local times of superprocesses
9
which we refer to as the time and space coordinates of v. V_ denotes the set of entrances for our graph, and if v E V_ we set s = 0 and yv = x,,. If v is
the exit labelled by j, i < j < n, we set (sv, yv) = (t.i, zj)
Finally, Vo denotes the set of internal vertices, i.e. those vertices that are neither entrances nor exits. We shall, in what follows, be interested only in third and fourth order moments. In the latter case, the set D4 of (2.1) consists of the following six basic graphs, and their various combinatorial rearrangements. The contri-
bution of graph - i.e. the integral appearing in (2.1) - is written after the graph. Since we shall be concerned only with finiteness of moments, we shall not bother with the combinatorial factors associated with each graph.
Graph 1 (O'X 1)
(O,X 2)
(t1,Z1)
(t2'Z2)
4
(O,X 3)
(O,X 4)
(t3,Z3) (t4 ,Z4)
fjP0,ti(Xi,z,),P;(z,)dx;dz;
10
Adler & Lewin: Intersection local times of superprocesses
Graph 2 (O,x 2)
(t 1,Z 1)
%
(S2,Y2)
,YJ)
(t 2,Z 2)
(t 3,Z 3)
(t 4,Z 4)
f dxldx2 f dyidy2 f dslds2ps (xi,yl)p 2(x2,y2) 4
X f pt,-s, (yl, zl )pt2-s, (yl, z2)pt,-s2 (y2, z3)pta-s2 (y2, z4) II y,(z,)dz=
Graph 3 (0, x
(t 3,Z 3)
(t4,Z4)
f dxldx2dx3 f dsl f dyl ps, (x3, yl ) 4
X
f pt,(xi,zi)pt2(x2,z2)pt,-s,(yi,z3)pt,-s,(yi,z4)11cp(z,)dz;
Adler & Lewin: Intersection local times of superprocesses
11
Graph (O,x 1)
(O,x 2)
(s1,y1) (t 4,z4)
J
dxldx2
J
ds1 f dy1 ps (x1, y1)
J
ds2 f dye P32-s, (y1, y2) 47
x J pt,-'1(y1' zl)pt2-s2(y2,z2)pt3-32(y2,z3)pt4(x1,z4) jj(p(zi)dzi icl
Graph 5
J
dxl
J
dsl
J dyl p2 (x1, yl) J
ds2ds3
J
dy2dy3 p 2-31(yl, y2)ps -s, (yl, y3) 4
x f pt,-,,(Y2, zl)pt2-s2(y2, z2)pt3-33(y3, z3)pt4-33(y3, z4) J1 (P(zi)dzi i=1
Adler & Lewin: Intersection local times of superprocesses
12
Graph 6 (O,x 1)
1
021Z 2)
(S1,yi)
(S2,y2)
(S3,y3)
<(t
3'Z3)
(t 4,Z 4)
(t pZ1)
f dxi f dsi f dyi Ps (xi, yi) f ds2 f dye P 2-s, (yi, y2) f ds3 f dy3 Ps -32 (y2, y3) 4
X fPt,-31(Y"ZOK2-32(y2,z2)Pt3-s3(y3, z3)P"-s3(y3, z4) II V(zi)dzi i=1
Similar considerations apply in establishing the graphs and their contributions required for calculating third order moments. Since the pattern should, by now, be clear, we shall merely list the graphs, leaving the calculation of their contributions to the reader, and the discussion of the following section.
The three third order graphs
3 (a)
Proofs Proof of Lemma (1.3) - The Ito Formula. The Ito formula
Adler & Lewin: Intersection local times of superprocesses
13
(1.11) follows in a reasonably straightforward fashion from a standard Ito formula for continuous semi-martingales, and so we shall only sketch the proof, highlighting the technical dificulties. We shall also restrict ourselves to the one-dimensional case (k = 1 in (1.11)) which is all we really need in this paper.
Fix a test function w E W2,r fl W2'1, p > 2, and note that (p, Xt), where Xt is either a super Brownian motion or super stable process, is a continuous semi-martingale (cf., for example Roelly-Coppoletta (1986) for this result for a more restricted class of ep, and Adler and Lewin (1991) for cp E W2,n fl W2'1). The associated finite variation process is given by
Xt), and the martingale part by (p, Zt), with associated increasing process [(o,Z)]t = fo(cp,X,)2ds, (cf.(1.2)).
It thus follows from a standard Ito formula (e.g. Karatzas and Shreve (1988)), that for' E C2(ll + x fit) and cp E W2 P fl W2'1, the formula (1.11)
of Lemma 1.3 holds, with the singular exception that the last term there (remember that k = 1) is replaced by JtP
J0
(s,(p,X,))d(p,Z,).
(3.1)
To complete the proof, we have to show that the above expression is equivalent to t
f
0 J 82d
Wx(s, (cp, X,))cp(x) Z(ds, dx).
Recall that, by assumption, there exists an integral a > 0 such that Wx(s, x) < C(s)IIxIla for all x E Rd. It thus follows from Dynkin's moment
formulae that E(Wx(s,(p,X,)))2 < oo. Hence the integral (3.1) is also in ,C2(P).
We can therefore approximate %Fx(s, (cp, X,)) in £2(P) by a simple function of the form n
Wi` `
L L. CA;IA,(W)I(8,,t;)(s) i=1 A;
where (si, ti] n (si, t7] = 0
if i # j, Ui(si, ti] = R+, and Ai E .F,;, UA; Ai = Q. Thus the integral (3.1) can be approximated in £2(P) by E E CA, I(o,tl(ti) [(gyp, Zt+) - (tP, Z3, )]. i
A;
(3.2)
14
Adler & Lewin: Intersection local times of superprocesses
But, according to Walsh's (1986) formulation of stochastic integration with respect to martingale measures, if we also take an approximation via simple functions to '(x) = Wx(s, (p, X,))cp(x), we immediately have that (3.2) is also an £2(P) approximation to
ff t
0
Wx(s, (4,, Xs)) yo(x) Z(ds, dx).
Std
Going to the £2(P) limit on both sides of this equivalence establishes the required result.
(b) Proof of Lemma 1.4. As in the previous proof, one can apply a regular Ito formula for the continuous semi-martingale (p,Xt) to the non-anticipative functional T (t, x) = x f0 (p, X,) ds, and evaluate it at T(t, (0,Xt)). Noting the various identities for' and its derivatives following (1.12), we thus obtain (1.13) but for the fact that the last term is replaced
f
by
It J ds (V,(x), X3) 0
Zt).
As before, we have to rewrite this as a stochastic integral against the martingale measure Z(dt, dx) to obtain the required form of (1.13). The argument, however, is precisely as before.
(c)
Proof of Lemma 1.5. We commence with functions 'DN(x, y) of the
form 4 N(x, Y)
_
N / E Ok(x)`Vk(y)
(3.3)
k=1
where both cp, i E W2 'P n W2'1. It follows immediately from Lemma 1.4,
with the notation of our lemma, that
f
(I'N(x, y), Xt(dx) XT(dy)) dt
f
T dt
0
t
J0
ds (L Sf
+j
y), X,(dx) Xt(dy)) dt t
+j T Jtd J
N(x, Y), X,(dx) Xt(dy))
ds
y), X,(dx)) Z(dt, dy).
(3.4)
This, of course, is the required (1.14) but for the restriction of the form (3.3). We need (3.4) for ' (x, y) E C2(Rd X Rd) of compact support.
Note, however, that every such function can be represented as the limit of functions 'PN of the form (3.3), with convergence in the supremum norm
Adler & Lewin: Intersection local times of superprocesses
15
be as required, and let {41iN}N<1 be an jjcpjj. = sup,, Icp(x)I. Thus, let appropriate approximating sequence. We shall proceed by showing that each
of the four terms in (3.4) is Cauchy in G2(P), and that the limit variable is the corresponding expression with 4N replaced by 4). We shall give details only for the first and last terms. The first three are all similar, but the last, being a stochastic integral, is somewhat different. (A)
Because of the form of
The First Term:
as a sum, it is clear
that £2(P) Cauchy convergence of fo ("DN(x, y), Xt(dx) XT(dy)) dt will result from (3.5) y), Xt(dx) XT(dy)) I2 G C C. II'1)NIIq EI
for some q > 0, where C is a constant that may depend on T and
In we expect
particular, if D C Rd X Rd is the union of the supports of the C to depend on D. (D, of course, can be assumed compact.) The left-hand side of (3.5) is equivalent to l2
(N
EI 1: (
NN /'
"/,
= E E E {(ak, Xt) (tOj, Xt)(?k, XT)(Wj, XT)} .
(3.6)
j=1 k=1
Each expectation here can be treated via the graph formulae of Section 2. Since these are fourth order moments, there are six different graphs to consider. While the requisite calculations are not difficult, they are tedious, so we shall look, in detail, at only two cases. Consider the case arising from Graph 1, the simplest of the six. By the formula following the graph, the corresponding contribution to (3.6) is
NN
Ef j=1 k=1
(
d
/'
/'
Vk(z1)Pj(z2)4'k(z3)'l'j(Z4) 4
x pi(xl, zl)pt (x2, z2)pT(x3, z3)pT(x4, z4) jj dxidzi. i=1
Note that the p" are symmetric, and do the xi integrals. The resulting expression is
NN
E If j=1 k=1 f3Z4d
z4) dzldz2dz3dz4.
16
Adler & Lewin: Intersection local times of superprocesses
Since the ''N all have support within a common compact set D, this last integral is bounded above by IDl4ll4kNlI'.0, where IDS denotes the Lebesgue measure of D.
As a second example, consider the cases corresponding to the evaluation of the summands of (3.6) via Graph 2. There are really two different subcases here, depending on whether the two disjoint subgraphs there end with {(t,cpk),(t,Wj)} and {(T,1/'k),(T,'j)} or with {(t,cpk),(T,ij)} and {(t,Vj),(T,Ok)}. Consider the first of these. Their contribution to (3.6) is of the form rN N LS E I
dxldx2 J dy1dy2 J dslds2 ps (xl, yl)p 2(x2, y2)
x
IPi .,, (y1,z1)Pt'-.,, (y1,z2)PT-12(y2,z3)pT-12(y2,z4)
j=1 k=1
74
x Wj(zl)Vk(z2)4'j(z3)4'k(z4) 11 dzi. i-1
As before, note the symmetry pt (u, v) = p (v, u), and thus integrate out x1 and x2. Perform the yl, y2 integrals using the Chapman-Kolmogorov formula to obtain that (3.8) is equivalent to
E f dslds2 J P2(t-s,)(zl, z2)P2(T-sp)(z3, z4) j=1 k=1 x
pj(zl)Wk(z2)0j(z3)0k(z4)dzldz2dz3dz4
= f dslds2 f P2(t-s,)(Zl, z2)P2(T_32)(z3, z4) X
-I)N(zl, z3)'I'N(z2, z4)dzldz2dz3dz4.
Applying the same argument as before, and noting that $1, s2 < T, we obtain that (3.9) is bounded above by
The remaining graphs are treated similarly, as are the other three nonstochastic integrals in (3.4). It remains to show, however, that the Cauchy, G2(P), convergence is to the correct limit. In the current case, however, this is easy. Since the convergence of the 4 N to 1 is in the supremum norm, convergence is also uniform. For each
s, t > 0, Xt x X, is a.s. a finite measure on D, and so (IN, X, X Xt) 4 (4), X, x Xt). We have just shown that (IN, X, X Xt) is G2(P) convergent. Since the £2(P) and a.s. limit must agree, we are done. (B) The Stochastic Integral Term: We must now apply a similar argument to the last term of (3.4). We start with Cauchy convergence, for which we
Adler & Lewin: Intersection local times of superprocesses
17
take the 4)N as above. We shall show that (fT
E
0
ff 82d
2
t
y), Xs(dx)) Z(dt, dy)< ) C. II N00
ds
(3.10)
II2
0
where C = C(T, D). Since Z is a martingale measure with associated increasing process [(cp, Z.)]T = fo (p2, Xt)dt, (cf. Walsh (1986)), the left-hand side of (3.10) is equivalent to
E I f (ft(N(x, y), Xs(dx))ds)
X(dy) dt.
(3.11)
0 SNote
the form of 4)N, and expand the square to obtain that this is equivalent to
f
N N It ds' 1: 1: E {
T
dt
It
0
ds
0
0
Xs)
Xsi)
Yak, Xt) }
(3.12)
.
i=1 j=1
Evaluation of each summand here hinges on the three graphs at the end of Section 2. Consider, for example, the third and most complex of these. The corresponding contributions to (3.12) are of the form N N T t3 0t3 IT dt3 dtl f dt2 E > fed dx dsl fed dy1 ps, (x, y1)
f
T
x
f ds2
dye p32-s1 (y1, y2) f823d pt, -31 (y1) zl )pt2-s2 (y2, z2)pt3-s2 (y2, z3)
(3.13)
dzldz2dz3.
(There are, of course, a number of contributions of this kind with the roles of the various pairs of times and test functions interchanged.) As in the previous case, use the symmetry of pt to integrate out x, and ChapmanKolmogorov to integrate over yl. After rearrangement, (3.13) becomes
IVd N(zl, x
z3)
IT T dt3
t3
f dtl It' dt2
IT
T dsl f T ds2
JRd dy2 pt,+s2-2s, (y2, zi )
x pt2-s2(y2, z2)pt3-s2(y2, z3)dzidz2dz3.
(3.14)
Bound 'DN(z1, z3)IN(z2i z3) by II(DNII'00, and then perform the z3 integral over all of Jd rather than just D. Then integrate out Y2 via ChapmanKolmogorov, to bound (3.14) by T
t3
T
t3
1I4NII00 f dt3 f dt1 f dt2 f ds1 f 0
0
0
0
DxDp(t,+t2-2s,)(zl, z2) dzldz2.
(3.15)
18
Adler & Lewin: Intersection local times of superprocesses
Integrate z2 over Rd, and then z1 over D to finally find a bound of the form IDI as required. All other terms can be handled similarly.
As before, to complete the proof, we must show that the £2(P) limit, assured by Cauchy convergence is, in fact, the stochastic integral IJOT
JBtd JOt
ds (4)(x, y), Xt(dx)) Z(dt, dz).
(3.16)
This, however, is relatively straightforward by a subsequencing argument.
Implicit in the above calculations is the fact that, for every y E Dy = {y y), Xt(dx)) is Cauchy in £2(P). Thus (x)y) E D} and t E [0,T], there is an a.s. convergent subsequence ('Nk(x, y), Xt(dx)), which, since Xt(Dx) < oo, where Dx = {x : (x, y) E D}, and the convergence of'kN to is uniform, implies (-I)Nk(x, y), Xt(dx)) $ y), Xt(dx)) as k - oo. Furthermore, both ('kN(x, y), Xt(dx)) and (4k(x, y), Xt(dx)) are uniformly :
continuous in y E Dy and t E [0, T]. Hence, along the subsequence {Nk}k>1 we must have JOT
(x, y), Xt(dx)) Z(dt, dy)
,/d lot ds
T0d 0
y), Xt(dx)) Z(dt, dy).
ds
Thus the £2(P) convergence along the full sequence must be to the appropriate limit.
Proof of Theorem 1.2 We can now, finally, turn to the proof of Theorem 1.2. We shall concentrate on establishing the evolution equation (d)
(1.10). Since the proof is of an G2 nature, the convergence of -y' (T, gyp), which
forms the first part of the theorem, will be proven en passant. To start, let {GE},>o be a sequence of Cc functions of bounded support such that G, + Ga as a -> 0. Take the evolution equation (1.14), and replace the function -1)(x, y) appearing there by G,(x - y)cp(x). Subtract l dt Jot ds (G,(x - y)cp(x), Xt(dx) Xt(dy))
from both sides of the resultant equation. Note the definition (1.9) of y, (T, cp), as well as (1.8), to obtain -tE (T, tP)
IrT r 0 I'd + AfT
IT
/'t
ds (GE(x - y)cp(x), Xt(dx)) Z(dt, dy)
0 dtIt
dS
(G,(x - y)cp(x), X3(dx)Xt(dy))
(G,(x - y)cp(x), Xt(dx) XT(dy)) dt.
(3.17)
Adler & Lewin: Intersection local times of superprocesses
19
If we now show that each of the three terms on the right-hand side of this equation converges in G2, as e --> 0, to a limit that is independent of the sequence {GE}, then the same must be true of the left-hand side, and so the central Theorem 1.2 will be established.
The proof hinges on the moment formulae given in Section 2, and very often closely parallels similar calculations in Rosen (1990). Because of the fact that we have three terms to consider, each one involving a number of graphs, and because the examples given in Rosen's paper show how these calculations must be carried out, we shall now carry out only one of the moment calculations in detail.
The term that we choose is the stochastic integral with respect to Z. There are two reasons for this. Firstly, Rosen's calculations do not involve an expression of this kind, and secondly, it will become clear during the calculations where the main conditions of the theorem (i.e. d = 4,5 in the Brownian case and d/3 < cti < d/2 in the stable case) come from. Thus, we now turn to proving that for any sequence {GE}, of the form described above, converging in G' to Ga,
{jT Eli o E
lods ((GE - GE)(x - y)(x), X3(dx)) Z(dt,
J
d)}= 0 . (3.18)
To save on notation, fix e, e' > 0 and set D(x) = G,(x) - G,, (x). From Section 2, three different graphs arise in computing (3.18), described there as (a)-(c). We claim that (a) is easy, and so leave it to the reader. A typical term arising from (b) is J d dxl
4
J T dsl 1
d
dyi ps (xi, yi) f T dti fed dzi pig-s, (yi,
dt2 f d dz2 pts-,, (yi, z2)D(zl - z2) XJ
/T
d
dx2 f dt3 fed dz3 pt3(x2, z3)D(zl - z3) .
(3.19)
(If this expression is not "obvious", then write D(x, y) as E;_1 dj (x)d'j (y),
Adler & Lewin: Intersection local times of superprocesses
20
and consider a term in this expansion corresponding to the graph 2
dj (z1) dk (z1)
xto']
,
which immediately yields the upper estimates in iii) and iv). As for the lower bound part one uses (1.15), which in the one dimensional situation boils down to 2
li mt-OOt-, log(pt(tax, oo)) > -vx - 2p
,
a> d+2
with the constraint p < 1 when a = 1. This immediately provides the lower bound part of iii) and iv). The next observation is that i), (a similar remark could in fact be applied in the context of Theorem 1.1), is a consequence of ii). So there now remains to prove ii). Let us begin with the lower bound. Pick /3 E (0,1/3), and consider the interval J = (-t'3,0 +(xVfo)t1/3). One has: (1.16)
log(µt(t1/3x, co)) > -v(2a + 20 + (x V £o)t1/3) + log(PJ [TJ > t , Zt > xt1/3])
Using an eigenfunction expansion for the last term, it is standard to argue that PJ [Tj > t , Zt > xt1 /3] is equivalent to the principal term (with obvious notations): Oi (0) exp{-
IT2
2I
(xVto)t" +to
JI2 t}
Oi (u) du Ix1/3
From the explicit behavior of Oi it is now easy to deduce: 7r22
limt-.t-1/3
log(PJ [Tj > t , Zt > xt1/3]) > -
2(x V to)2
Sznitman: Brownian motion among Poissonian obstacles
361
This and (1.16) now proves the lower bound part of ii). For the upper bound
part, it is enough to assume that x > to in ii). Considering the points of the cloud falling immediately at the left and at the right of 0, we see that: (1.17)
P ®Po [Zt > xt1/3 , T > t] e-v(u+v)tl/3 Po [T(_u,v) > t1/3
< J >0 , v>x du dv v212/3 Now
Po [T(-u,,,) > t1/3] < sup P., [T(o,I) > (0,1)
dx Pz [T(o,l) >
(27r)-1
(u + v)2
]
11/3
JoI
<
tl/3
.2
(u - v)2
1]
t1/3
- 1) } , ((u + v)2 where the last inequality follows from the spectral theorem. If we use this estimate in (1.17), Laplace method now yields: <
(2ir)-1
exp { -
log
2
µt(t1/3x,
CO)
2 <-
2
inf(vt + 2Q2) = vx + 2x2
This finishes the proof of ii) and of Theorem 1.3. 0 Remark 1.4. It should be mentioned that the proof above works wi th trivial modifications if we replace the quantity pt(t"x, oo) in the four cases of Theorem 1.3 by P x Po[sup,
t], with rate function I(x) = vy+
property under P ®Po
22
-c(1, v),
with y = (2x) A (to V x), whereas t-1/31Ztl under the same2yconditional measure satisfies a large deviation property with rate function I'(x) = v(x V
to) +
2(x V X0)2
- 41, v).
o
H. Upper bound: a preliminary reduction step. We suppose now that d > 2. We want to derive a preliminary estimate which will help us in proving upper bounds on the probability of existence prior to time t of an excursion of surviving Brownian motion at a distance larger than const. This estimate will show that with no loss of generality we can assume that Brownian motion does not leave a suitable cube T of size const. t(d+1)/(d+2), that in this cube the obstacles form finitely td/(d+2).
many "clearings" of size - t1/(d+2). The estimate will also allow us to
Sznitman: Brownian motion among Poissonian obstacles
362
assume that the number of excursions of the process "in the forest", that is outside a neighborhood of size tl/(d+2) of the clearings, compared to td/(d+2),
as well as the fraction of time spent during these excursions are arbitrarily small. We will now make these notions precise. First, it will be convenient to adopt respectively tl/(d+2) and t2/(d+2) as new space and time units. With this new choice of units, we now study a standard Brownian motion until time s Lef td/(d+2), evolving among a Poisson cloud with intensity vs of points xi of Rd surrounded by closed balls of radius at-1/(d+2) = as-1/d
We now introduce a number L > 0, and an integer N > 1, and for s > 1 we define the cube T of Rd:
T = (- N[s]L , N[s]L)d
(2.1)
,
([s] : integer part of s) .
For each multiindex m E [-N[s], N[s] - 1]d, Cm will be the open subcube of T:
Cm = {zERd, miL
(2.2)
With no loss of generality we will assume that no point of the Poisson cloud has one of its coordinates which is an integer multiple of L.
We now introduce two numbers b > a, and b > 0. Following [5], we define a point xi E Cm of the Poisson cloud to be a good point if for all closed balls C = B(xi , where e 1!1':f s-1/d, 0 <.£ and 10t+1eb < L/2, (2.3)
CmnCn( U B(x;,bE))I>3 ICmnCI. x j EC,,,
We will denote by G(m) the set of good points of Cm. Notice that for r > 0, and z E Cm, using a homothety of ratio 1/3 at z, one has: ICm n B(z, r/3)I > 3-dICm n B(z, r) I It now follows from the covering Lemma 1.3 of [5], that for each Cm: (2.4)
1c,. n ( U B(xi, be) )I
5ICmI = SLd
xi¢G(m)
xjECm
In other words, the union of balls of radius be at bad points of Cm cover a small fraction of the volume of Cm.. We then chop identically each segment [kL, (k+1)L] in at most
[iS1/d} +1 intervals of length
s-1/d, except
Sznitman: Brownian motion among Poissonian obstacles
363
maybe for the "last one". This yields closed boxes of diameter less than
\d
be = bs/d, in number less than ([V'dsh/c] + 1 I , whose union is C,,,.. We then introduce a number r > 0, and denote/ / by Clr the event that there is a "clearing of size r in the cube C,,,", that is: (2.5)
Clm = (w, I Um(w)I >_ 2-d I B(O, r)I = 2-dwdrd }
,
if U,n(w) is the random open set of C,,,, obtained by taking the complement in C,,, of the closed boxes where a good point of the cloud falls. We also
C T, where there is a
denote by B(w) the union of all closed cubes clearing of size r: (2.6)
1B(w)(z) _
1Cm(z) -
We now pick a number 2 > 0. Let us mention that we will not need to let £ vary and we could very well pick f = 1, once and for all. We now define the successive excursions of the process Z.(w), at distance f from B(w):
Di=inf{v>0,
,
Rn+1 = R1 0 8R + R. ,
here Be denotes the open neighborhood of point at distance less than f of B(w) (which is empty if B is empty). We set N. to be the number of excursions completed by time s: (2.7)
{N,=k}={Rk<s
with the convention Ro = 0. Finally, we define L, to be the fraction of time spent up to time s in the excursions: (2.8)
L,= 1E(RiAs-Dins). s i>1
We can now state the main object of this section:
364
Sznitman: Brownian motion among Poissonian obstacles
Theorem 2.1. For anyL>0,r1>0,2>0: (2.9)
limN.00 limr-o limno-00 limb.oo, 6-0 lim,_,. llog(PxPo[{T>s} n[{TT<s}U{JBI>noLd}U{N,>[,s]}U{L,>7/}]])=-oo.
Here TT denotes the entrance time in T'. Proof: Our claim (2.9) will follow if we show that: (2.10)
(2.11)
limN
moo
lim,-.
s
log (Po [TT < s])
limno-.°° 1imb_oo,6-.0 lims-
,
s
= -00 ,
logP[IBI > noLd] = -oo
and with the notation T A TT =!h (2.12)
limr.o limb-.oo, 6-.0 lima-oo 1 log sup Po If T > s l S
W
n{N,>[iis]}]=-oo, (2.13)
limr.o limb-.., 6-.o lim,.,,. -log sup Po [{T > s} s
W
n{N,<[77s]}n{L,>i7}]_ -oo Let us start with the proof of (2.10). Using scaling, we immediately find:
Po[TT<s]
Vis-
where Po is the one dimensional Wiener measure. It follows that: 2
li m8-00
s log Po [TT <s] < - 2 L2
,
which yields (2.10). Let us now show (2.11). From the definition (2.3), we see that for any
point xi E Cm, the property that xi is a good point of C,n just depends on the restriction of the Poisson measure to C,n. From this observation, it follows that the events Cl,,, ("presence of a clearing of size r" in Cm), for m E [-N[s], N[s] -1]d are i.i.d. Let us now give an upper bound on P[Cl,n].
Sznitman: Brownian motion among Poissonian obstacles
365
If U,,,(w) denotes the complement of the boxes in Cm. where some point falls (recall that for Um(w), it is a good point instead), we have since each box has diameter less than be:
Um C Um U ( U B(xi, bs-1Id ) n Cm) . .i G(m) si ECm
Thanks to (2.4), this yields: (2.14)
IUmI
This shows that cam = {IUmI >
2-dIB(O,r)I} c {IUmI > 2-d9B(O,r)I
- 6L d}
and it follows that: P[Clm] < P[IUmI > 2-dIB(0, r)I - 6Ld] dsi/d + 1)d < 2( b
exp{-vs(2-dIB(0, r)I
- 6Ld)}
Now, there are at most (2N s)dno possible subsets of no elements in the set of subcubes Cm of T. As a consequence, P[IBI ? noLd]
< (2Ns)dno2((LI6)vde'id+1)dn,, exp{-novs(2-dIB(0,r)I
- 6Ld)}
so that lime.
s log P[IBI > noLd] < no (log 2(b V d + v6Ld - v2 -d I B(0, r)I)
From this we obtain (2.11) immediately. Let us now prove (2.12). We first introduce the constant cl defined as: (2.15)
cl= 12 CEC of
Po[Hc0,
where H3 = inf{v > 0, IZ,, -Zo I > 3}, He is the entrance time in C, and C is the class of compact subsets of W(0,2), such that ICI > 2-d(1-2 -d )I B(0, 2)[. cl is easily seen to be positive, for instance by observing that the transition density ql (0, ) at time 1 of Brownian motion kill ed outside B(0, 3) has a
366
Sznitman: Brownian motion among Poissonian obstacles
positive infimum on B(0, 2). The notion of clearing we introduced will be of special help to us thanks to
Lemma 2.2. For any b > a, 1 > S > 0, 0 < r < L/4, N > 1, there exists so > 0 such that for s > so: inf
(2.16)
w, zEB(w)°f1T
Pz[T < H4r] > C1
.
Proof: First define the positive number: (2.17) a(S, b, a) =
inf
Izl<1, CEC'
PP[HC < Hlo] x inf Pz[HB(o,a) < HB(0,3b)-] IzI
where C' is the class of compact subsets of B(O,1) with relative volume no less than S/6d. Then set m(S, b, a) to be the smallest integer such that:
m log(1 - a) < - log 2 .
(2.18)
Consider now z E Cm a point at distance smaller or equal to be = bs-1/d of a good point x= E Cm. Suppose 10'+1 be + be < L/2, we have: (2.19)
Pz[H1om+4e+b,,
with Hp = inf {u > 0
,
I Zu - x; I > p}. Since x; is a good point of Cm, we
know that: ICm n c n
(U
B(x;
,
bs-11d))
I > 3 ICm n cl ? s Icl
x, EC-
for all balls C = B(x;,10e+1eb), with 0 < 2 and 10l+leb < L/2. Now from the definition (2.17) of the constant a, using scaling and Markov property, we see that 1 Pz[H1m+lb, < T] < (1 - a)n, <
As a consequence, we see that for any z in T within distance be of some good point in some cube Cm, we have (2.20)
Pz[Hr > T] > Pz[H1Om+be+be > T] > 1 - (1 - a)f1 > 2
provided s is large enough so that (2.21)
10'+1 be + be < r < L/4
.
Sznitman: Brownian motion among Poissonian obstacles
367
n B(w)', we know that lCim(,,,) = 0, and IUmI < Now, when z E 2-djB(0, r)j. As a consequence, the intersection of B(z, 2r) with the union of boxes in C,,, containing some good point has a volume bigger than or equal to:
I - 2-d1 B(0,r)I > (1 - 2-d)2-dIB(0,2r)I , since r < L/4
JB(z,2r) fl
Consequently, using scaling and the definition (2.15) of cl, (2.22)
Pz[Hcm\um < H3r] > 2c1
,
for z ECfl B(w)`
.
Combining (2.22) and (2.20), we see that when s is large enough so that (2.21) holds, for z E T fl B(w)c, Pz[T < H4r] > 2 x 2c1 = c1. This yields our claim.
Lemma 2.2 has the following consequence: for b > a, 1 > S > 0, 0 < r < L/4, N > 1, s > so (so appearing in Lemma 2.2), and p > 0: sup
(2.23)
PP[HB < T] < (1 - c1)[P/4r]
w, zE(BP)'nT
Indeed, if z E (BP)c fl T, the process has to exit successively at least 12-1 4r balls of radius 4r before entering B, so that (2.23) follows from (2.16) and a repeated use of strong the Markov property. Let us now continue the proof of (2.12). We have: sup Po[{T> s} fl {N8 > [is]}] < supPo[R[,,8]
W
Observe now that for k > 0: Po [Rk+l < T] < Po [Rk < T, Dk+1 < T, HB 0 ODk}1 < T o ODk+1 ]
< Po[Rk < T] x
sup Pz[HB < T]
(B<)cnT
< ( sup Pz [HB < T]) k+1 (BL)cnT
using induction. It now follows from (2.23), that when s is large enough:
supPo[{T > s} fl {N8 > [is]}] < (1 U;
from which we obtain (2.12) immediately.
Sznitman: Brownian motion among Poissonian obstacles
368
Let us now prove (2.13). Pick .X0 small enough so that P0[exp{A0H1}] < 1 + 2
(2.24)
Define H° = 0, and H'+1 = H` + H4r o 9H:, the successive times when Z. travels at distance 4r. Then choose A < Ao/16r2 and s > s0 given by Lemma 2.2, for z E (Bt)c nT, we have: (2.25) E,z[exp{ARi AT}]
= PZ[T = 0] +
Ez[Hk < HB A T < Hk+1 , ea(HBnT)] k>O
< 1 + 1: EZ[H' < HB AT , aHk] EO[eaH9, ] k>O
From scaling, E0[exp{AH4r}] = E0[exp{A16r2H1}] < 1 + cl/2. Now for
k>1: EZ[Hk < HB A T , eXH'] < Ez[Hk-1 < HB A T , eAHk-1 (EZHk-1 [H4r < HB A T] + EZHk_1 [e- \H--] - 1)] Cl)
< Ez[Hk_1 < HB A T, eAHk-1] (1
-
where we use the fact that ZHk-1 E B(w)c n T, on the event Hk-1 < HB A T, together with (2.16). It now follows that for A < A°/16r2, s > s0, z E (Bt)° n T, and any w: Ez[exp{ARi A T}] < 1 + (1 +
(2.26)
2)
(1 k>O
2
2
Picking A = A0/16r2, we find (2.27)
)k = 2 +
supPO[{T>s}n{N, <[is]}n{L,>ij}] W
< supexp{-Arks} Po [exp {A[(R1 A T) o 0D1nT W
+(R1 AT)o0D2AT+...+(Ri AT)o9D(ne]nT]}1 Using now the strong Markov property, and the fact that (2.26) obviously holds when z T, we see that the expression in (2.27) for s > so, is smaller than: exp{-A s}(2+2/c1)'?9+1 from which it follows that:
lifls-c
s
log sup P0[{T>s}n{N,<[r)s]}n{L,>rl}] W
6r2
+ log (2 + 2/cl)
This immediately yields (2.13) and finishes the proof of Theorem 2.1.
Sznitman: Brownian motion among Poissonian obstacles
369
III. Upper bound We suppose here, as in Section 2, that d > 2. Our main purpose is to derive an upper bound for the probability that surviving Brownian motion This will provide a companion in time t travels at distance of order estimate to the lower bound of Theorem 1.1, ii). td/(d+2).
Theorem 3.1. (3.1) FMt_""t-d/(d+2) log p ®Po[T > t , sup IZu I > x td/(d+2)] u
< -(c(d, v) + uwd_1ad-lx)
Proof: Using the same scaling argument and same notations as in Section 2, we want to show that (3.2)
limsy00s-1 log P ®Po[T
xs(d-1)1d]
> s , sup IZu I > u<9
< -(c(d, v) +
vwd-lad-1x)
,
where we recall that s = td/(d+2) and P now denotes a Poisson point process of intensity vs, the points of the random cloud being surrounded by closed
balls of radius as-l/d, constituting the obstacles. Thanks to Theorem 2.1, we see that for a given 0 < i < 1, and L > 0, we can pick N(i1, L), r(77, L), r7o(r7, L), b(r7, L), b(i, L) so that the probability of the event which appears in
(2.9), decays at a faster exponential rate in s than -(c(d, v) + vwd_1ad-1x) So, (3.2) will follow if we show that (3.3)
limL-. limn-o 11m,-.0 log P ®Po [T > s I BI < noLd
,
N. < [11s]
,
L. < t sup IZ.I >
, x8(d-1)/d]
v<s
< -(c(d, v) + vwd-lad-lx) Observe that the condition L, < 77 < 1, forces B(w) not to be empty. The number of nonempty subsets M of [-N[s], N[s] - 1]d, with no more than no elements is smaller than (2Ns)dno x no. It is consequently enough to show that: (3.4)
limL_,oo lim,7_,o limn-oc
N. < [17s]
,
L, < 17
,
sup IZuI > U<3
sup
1
logP ® Po[T > s ,
C -(c(d, v) + vwd-lad-1x)
370
Sznitman: Brownian motion among Poissonian obstacles
where now N. denotes the number of excursions between B = UM C. and (BL)c, as in (2.7) and L9 is the fraction of time spent up to time s, coming back from these excursions as in (2.8). Observe that the restrictions of the Poisson cloud to B2t and (B2e)c, are independent. We denote by PB2t and P(B2<). their respective laws. Conditioning on the restriction to Bet of the cloud, we see that the probability under study in (3.4) equals: (3.5) PB2t ® Po[T > s , N. < [ris]
, L. < q, sup IZuI > X3(d-l)/d u<s
exp{-vsJWW
3_l
n (B21)cI}]
where T = TT A T, and T denotes now the entrance time in the obstacles coming from the cloud with law PB2t. It will now be convenient to derive:
Lemma 3.2. Let 0 E Q[0, TI, Rd), ¢(0) = 0. For p > 0,
(3.6) IW(q)n(B2c)cI >
wd_lPd-l
sup ICI-no(L+4e+2p)J) -nowdpd . [0,T]
Proof: With no loss of generality, possibly reducing the interval [0, T], we assume that Ic(T)I = max[o,T] 101. Define for m E M, Cm 3 to be the closed cube with same center and parallel to Cm of side length L + V + 2p, and set U rim 3 B2c+a
.
M
We then define sl = inf{u E [0, T] , 0(u) E B} (Si = oo, if the previous set is empty). Then there is one or maybe several m E M such that q5(sl) E Cm. Define tl to be the supremum of the v E [sl,T] where r¢(u) belongs to the union o f these Cm. Continue then by defining s2 = inf{u > t1, 4(u) E B}, and construct a finite sequence, 0 < sl < tl < s2 t2 < < sk < tk < T, with 0 < k < no, such that for u V Ul
IWp(t')I -no(
(L+4Q+2p)Vdwd-lpd-l+wdPd)
>Wd-1Pd-l(IO(T)I -no(L+4Q+2p)Vrd-)-nowdpd which proves our claim (3.6) since IO(T)I = Iq(T)I = max[o,T[ 101.
11
,
Sznitman: Brownian motion among Poissonian obstacles
371
If we now apply Lemma 3.2 to (3.5), we find that the expression in (3.5) is smaller than: PB2c ®Po [T > s , N, < [77s] , L, < r7]
(x - 2nov(L + 4t +
exp{-vswd_lad-1
2as-'Id) 8-(d-1)/d) - 2nowdad}
So our claim (3.4) will follow if we show that: (3.7) limL-,oo lim,l_,o lime-0
sup
1109 PB24 (D Po [T > s,
1<J.MJ<no(,),L) S
N, < [r7s], L, < r7] < -c(d, v) .
Let us introduce the torus of size L, TL = (R/LZ)d, and denote by proj'r, the canonical projection on TL. We have:
IW;'-'/d n B2`I >
IprojT.L(W;'-'/d n B21)1,
where we use I I to denote the usual volume on TL as well. As soon as as-1ld < Q, the last quantity is bigger than ¢e-1/d
(ta-1/d
0.8-1/d
IProJTi(W[o,D1) U W[Ri,Dz) ... U W[RN,6ADN,))I
For as-11d < .e, we find
PBz<®Po[T>s, N.<[r7s], L,
e-11d
n B211} , N. < [77s]
,
L. < q
,
TT > s]
Po[exp{-vlprojT.L(GV[0 D0d U W[R1,Dz) U ... U W[RN.,enDN.))I }
e
N,<[17s], L,
(3.8)
Let us now denote by PL the law of the random point process obtained by picking a Poisson configuration of points with intensity vs in (0, L)' and extending it to the whole of Rd by periodicity. As before let us define the obstacles as closed balls of radius as-11d centered at each point of the periodic configuration. If T denotes as before the entrance time in the obstacles, we see that (3.8) equals PL
®Po[TT>s, T>D1, ToOR1>D1oOR1, ..., T o 9RN, > (s - RN,) A D1 o 0RN, , N. < [17s], L. < rl]
Denote by AL(w) the principal Dirichlet eigenvalue for the generator of the self adjoint trace class semigroup on L2(TL), corresponding to Brownian
Sznitman: Brownian motion among Poissonian obstacles
372
motion on TL killed on the obstacles on TL, projection of the periodic obstacles on Rd. Observe that on the event whic h appears in (3.9), T A D1 + T A D1 0 ORI +
+ T A D1 0 9RN, > (1-77)s
It follows that for M > 0, and p > 0, (3.9) is smaller than: PL®Po [TT > s, exp{(AL A M - p)+(T A D1 + T A D1 0 ORS + .. .
+T AD1 0ORN. - (1 -q)s)} , N, < [17s]] < pL [exp{-(JCL A M - p)+(1 - 71)s} Po [exp{(AL A M - p)+
T A D1 0 9R; 1(R; < oo)}] ] 0
< PL [exp{-(JCL A M - p)+(1 - 77)s}
(sup PZ[exp{(AL A M - p)+T}])ns+1 ] ZERd
If now PL denotes the law of Brownian motion on TL starting from z, using obvious notations, we have sup Px [exp{(AL A M - p)+T}] = sup Pz [exp{(AL A M - p)+TL}] Rd
TL
This last quantity, by inequality (1.22) of [5] is smaller than eM(1 + CL + cL v ), where CL stands for the supremum of the transition density at time one, with respect to the normalized volume measure of Brownian motion on TL. It is may be helpful to mention that the notations of [5] are somewhat different and TL here corresponds to Tb in [5], and AL to .fib. We now see that (3.9) is smaller than f (L,q, M, p, s) dew-[ exp{(2M71
)ns+1 pL [exp{-(AL A M)s}]
+ p)s + M}(l + CL + CL
P
It now follows that the left member of (3.7) is smaller than: limL-oo limM-.oo, p--.o limn-o lim,_,oo s log(f (L, r/, M, p, s) )
= lifL.oo FIMM.oo lim,.oo 1 log PL[exp{-(AL A M)s}] = -c(d, v)
,
S
The last equality comes from the upperbound we derive for PL[exp{-(AL A M)s}] in [5], see what follows there (1.39), Theorem 2.2 and Lemma 3.3.
Sznitman: Brownian motion among Poissonian obstacles
373
We should also mention that in the notations of [5], AL plays the role of \KT his proves our claim (3.7) and finishes the proof of Theorem 3.1. As a counterpart to the lower bounds of Theorem 1.1, we have
Corollary 3.3. Suppose d > 2, and x > 0, (3.10) limt_oot-d/(d+2) log(µt(td/(d+2)x, oo) ) vwd_Iad_1x)
< -(c(d, v) +
,
log(/1t(t°'x, oo)) < -vwd-iad-1 x
(3.11)
,
d
d
2
2
(3.12) limt_oot-1 log(pt(tx, oo)) < -(2 + vwd-lad-1 x)
Proof: (3.10) is an immediate consequence of Theorem 3.1. The proof of (3.11) and (3.12) is straightforward, it relies on the already observed fact that: (3.13)
IWt (Z)I > vwd-lad-'IZt - ZoI , from which we deduce pt(t' x,oo) < exp{-vwd-lad_1xt'} P0[Zf > xta]. (3.11) and (3.12) now follow easily.
Remark 3.4. 1) It should be observed that (3.10) does not hold in the one dimensional case (see Theorem 1.3). 2) From Theorem 1.1 ii) and Theorem 3.1, we see that, when d > 2:
-k(d, v, a)x < lim t-d/(d+2) log p ® Po[sup IZSI > xtd/(d+2)/T > t] s
< -vwd_1ad-lx . This strongly suggests the possibility of a large deviation property for the law of t-d/(d+2) sup,
IV. An application We are now going to apply our results to the study of the large t behavior of Eo[exp{h Zt - vlWf I }]
,
when
IhI <
vwd-lad-1
This quantity was investigated by Eisele and Lang, in their study of survival probability for Brownian motion with a constant drift evolving among Poissonian obstacles. They showed that (4.1)
lim 1 logEo[exp{h Zt - vlWW I }] = 0 , t-.oo t
when
IhI < a(d, v, a)
>0,
when
I hl > a(d, v, a)
,
Sznitman: Brownian motion among Poissonian obstacles
374
for a critical value a(d, v, a) in the interval [vwd_lad-1, k(d, v, a)], where
k(d, v, a) is defined in (0.6). We are now going to apply the results of section III, in the d > 2 case, and section I in the d = 1 case, to refine (4.1) when IhI < vwd_lad-1.
Theorem 4.1. When d = 1, and IhI < v: (4.2)
lim t-113 logEo[exp{hZt - vlWW I }] = -c(1, v - IhI) _ -3(ir(v
t-+00
When d > 2, and IhI <
-
Ih1))213
.
vwd_lad-1:
lim t-d/(d+2) logEo[exp{h Zt - viWf I fl _ -c(d, v) t-00
(4.3)
Proof: It is enough to study Eo[exp{hZ' -vlWW l}] for h > 0. This quantity is bigger than (4.4) f 00 exp{hx} ditt(x) = Eo[exp{-viWW I }] + h 0
J0
ehxpt(x, oo) dx
,
and anyway smaller than: /'oo
2Eo[exp{-vjWW I}]+h J
ehxµt(x,oo)dx
.
0
It follows that we simply have to study the second term to the right of (4.4).
It is the sum of A a1 deJ- htd/(d+2)
exp{td/(d+2)hx} µt(td/(d+2)x, 00) dx
J0
and
oo
exp{hx} µt(x, oo) dx
a2d e-L h J
,
with A a positive constant.
Atd/(d+2)
Using (3.13), we see that a2 is smaller than 00
J
dx hexp{x(h - vwd_lad-1)}, td/(d+2)
so that (4.5)
limt__o, _d/(d+2) log a2 < -A(vwd_1 ad-1 - h)
.
We see our claims (4.2), (4.3) will be proved if we show that for large enough A:
(4.6)
log a1 < -c(d, v)
,
when d > 2
,
Sznitman: Brownian motion among Poissonian obstacles
lim t1/3 log al = -c(1, v - ]h`) ,
(4.7)
t-00
375
when d = I
Observe that Ftt(td/(d+2)x, oo) is a decreasing function of x. Using Riemann sums, and Laplace method, we see that when d = 2, thanks to (3.10),
l1mt_",)t-d/(d+2) log al < sup {hx - c(d, v) - vwd_lad-lx} = -c(d, v) xE[O,A]
since h <
vwd_jad-1
This proves (4.6). On the other hand when d = 1,
using Theorem 13, we find 2
lim t-1/31oga1 = sup {hx - v(x V 4o) t---oo xE[O,A]
2(4 V x)2
= c(1, v -
provided A is bigger than [r2/(v - h)]1/3. This yields (4.7) and finishes the proof of our claim. O
Bibliography BOLTHAUSEN, E.: "Localization of a two dimensional random walk with an attractive path interaction", preprint. DONSKER, M. D., VARADHAN, S.R.S.: "Asymptotics for the Wiener sausage", Comm. Pure Appl. Math., 28, 525-565, (1975). EISELE, T., LANG, R.: "Asymptotics for the Wiener sausage with drift", Prob. Th. Rel. Fields, 74, 1, 125-140, (1987). GRASSBERGER, P., PROCACCIA, I.: "Diffusion and drift in a
medium with randomly distributed traps", Phys. Rev., A 26, 36863688, (1982).
SZNITMAN, A.S.: "Lifschitz tail and Wiener sausage I", in J. Funct. Anal., 94, 223-246 (1990). "Long time asymptotics for the shrinking Wiener sausage", Comm. Pure Appl. Math., 43, 809-820, (1990). "On the confinement property of two dimensional Brownian
motion among Poissonian obstacles", to appear in Comm. Pure Appl. Math.