Potential Analysis 21: 263288, 2004.

2004 Kluwer Academic Publishers. Printed in the Netherlands.

263

Symmetric Stable Processes in Cones

RODRIGO BAUELOS1,
and KRZYSZTOF BOGDAN2,


1 Department of Mathematics, Purdue University, West Lafayette, IN 47907-1395, U.S.A.

(e-mail: banuelos@math.purdue.edu)
2 Institute of Mathematics, Wrocaw University of Technology, 50-370 Wrocaw, Poland
(e-mail: bogdan@im.pwr.wroc.pl)
(Received: 9 October 2002; accepted: 5 September 2003)
Abstract. We study exponents of integrability of the first exit time from generalized cones for
conditioned rotation invariant stable Lvy processes. Along the way, we introduce the spherical
fractional Laplacian and derive some of its spectral properties.
Mathematics Subject Classifications (2000): Primary: 31B05, 60J45.
Key words: symmetric stable process, -harmonic function, cone, conditional process, exit time.

1. Introduction
For x Rd \{0}, we denote by (x) the angle between x and the point (0, . . . , 0, 1).
The right circular cone of angle  (0, ) is the domain  = {x Rd \{0} :
(x) < }. The exact moments of integrability of the first exit time,  , for
both Brownian motion and conditioned Brownian motion from  have been extensively studied in the literature. These investigations began with the work of
Burkholder [14] who showed that there exists a constant p(, d) such that for
p
any x  , Ex ( ) < if and only if p < p(, d). This critical exponent, as it
was shown by Burkholder, can be expressed in terms of zeros of confluent hypergeometric function. Extensions of the result were given by DeBlassie [20], Davies
and Zang [19], and Bauelos and Smits [2]. In particular, the results in [19] and [2]
provide the analogue of Burkholders result for conditioned Brownian motion. In
[21], DeBlassie obtained a counterpart of Burkholders result for symmetric stable
processes in R2 . DeBlassies result was extended to all dimensions by MndezHernndez [30]. In [26], Kulczycki gave results on the asymptotics of the critical
exponent in right circular cones of decreasing aperture.
The purpose of the present paper is to obtain an analogue of the results in [19]
and [2] for conditioned symmetric stable processes in generalized cones and to
extend the results in [21] and [30] for the unconditioned processes to more general

The first author was supported in part by NSF grant # 9700585-DMS.


The second author was supported in part by KBN and by RTN contract HPRN-CT-2001-00273-

HARP.

264

RODRIGO BAUELOS AND KRZYSZTOF BOGDAN

cones. Our method is motivated by, but different then, the method in [19]. The key
step in our proof is to identify, for generalized cones, the Martin kernel with pole at
infinity as a homogeneous function of degree [0, ). Once this is done we use
the scaling of the stable process and homogeneity of the kernel, together with tools
related to the boundary Harnack principle of Song and Wu [33], to estimate the
distribution of the exit time of the process from the cone. The exponent depends
on the geometry of the cone and relates the considered problem to the asymptotics
of harmonic functions at the vertex of the cone, see [7] for boundary points of
Lipschitz domains. We use to identify the critical moments of integrability of the
exit time of the stable processes and the conditioned stable process from the cone as
/ and (d + 2)/, respectively, see Theorem 4.1 below. Our method should
apply to other processes with scaling such as the censored processes studied in
[11].
The paper is organized as follows. In Section 2 we recall the basic properties
of the symmetric -stable processes and the definition of -harmonic functions.
In Section 3 we recall the boundary Harnack principle of [33] and state Theorem 3.2 and Theorem 3.3 which identify the Martin kernels at infinity and at
0 as homogeneous functions of degree and d , respectively. We also
give some explicit examples of cones where we identify these exponents in terms
of the parameter . Theorems 3.2 and 3.3 are proved in Section 6. In Section 4
we obtain the above-mentioned critical moments of integrability of the lifetime of
the stable processes in the generalized cones. In Section 5 we study a spherical
fractional Laplacian, which is related in a natural way to the symmetric stable
process. In the classical case of the Brownian motion, the spherical Laplacian plays
a crucial role in understanding the moments of integrability of the corresponding
process in cones. Indeed, the exponent of integration is obtained from the Dirichlet
eigenvalues of the spherical Laplacian on the set of the sphere which generates the
cones, see for example [2]. While in the current case we were not able to obtain this
precise relation to the exponent of integrability, nevertheless the spherical operator
does provide some additional information. We also believe this operator may be
useful in other settings and a more detailed study of its spectral properties will be
of interest. In a sense, we give a polar coordinate decomposition of the fractional
Laplacian, which as far as we know has not been given before.

2. Preliminaries
We begin by reviewing the notation used in this paper. By | | we denote the
Euclidean norm in Rd . For x Rd , r > 0 and a set A Rd we let B(x, r) =
{y Rd : |x y| < r} and dist(A, x) = inf{|x y| : y A}. A is the closure,
and Ac is the complement of A. We always assume Borel measurability of the
considered sets and functions. The notation c = c(, , . . . , ) means that c is a
constant depending only on , , . . . , . Constants will always be (strictly) positive
and finite. Throughout the paper we use the convention that 0 = 0.

265

EXIT TIMES FROM CONES

For the rest of the paper, unless stated otherwise, is a number in (0, 2) and
d = 1, 2, . . . . By (Xt , Px ) we denote the standard [5] rotation invariant -stable
Rd -valued Lvy process (that is, homogeneous, with independent increments), with
index of stability and characteristic function
Ex ei(Xt x) = et | | ,

x Rd , Rd , t
0.

As usual, Ex denotes the expectation with respect to the distribution Px of the

process starting from x Rd . The sample paths of Xt are right-continuous with
left limits almost surely. (Xt , Px ) is a Markov process with transition probabilities
given by


p(t; x, y) dy
Pt (x, A) = Px (Xt A) =
A

and it is a strong Markov processes with respect to the standard filtration. For this
and other basic properties, we refer the reader to [5].
/ U }, the first exit time
For an open set U Rd , we put U = inf{t
0; Xt
of U . Given x Rd , the Px distribution of XU , which is given by
Px {XU A, U < +},
for every Borel set A, is a subprobability measure on U c (probability measure if U
is bounded) called the -harmonic measure.
The scaling property of Xt plays a key role in this paper. Namely, for r > 0
we have that for every x Rd the Px distribution of Xt is the same as the Prx
distribution of r 1 Xr t . In particular, the Px distribution of U is the same as the Prx
distribution of r rU . In short, rU = r U in distribution.
When r > 0, |x| < r and B = B(0, r) Rd , the corresponding -harmonic
measure has the density function Pr (x, ) (the Poisson kernel) given by
 2
2 /2
d r |x|
|y x|d if |y| > r,
(1)
Pr (x, y) = C
|y|2 r 2
with Cd = (d/2) d/21 sin( /2), and 0 otherwise. The formula for the Poisson kernel for the exterior of the ball {y Rd : |y x| > r} is similar. Namely,
for  d and |x| > r we have
 2
2 /2
d |x| r

|y x|d if |y| < r,

(2)
Pr (x, y) = C 2
r |y|2
and Pr (x, y) = 0 if |y|
r. (1) and (2) can be found in [6], see Theorem B
on p. 241 in [6] (see also Theorem C in [6] for a counterpart of (2) in the case
1 = d < ).
DEFINITION 2.1. We say that f defined on Rd is -harmonic in an open set
D Rd if it satisfies the mean value property
f (x) = Ex f (XU ),

x U,

(3)

266

RODRIGO BAUELOS AND KRZYSZTOF BOGDAN

for every bounded open set U with closure contained in D. It is called regular
-harmonic in D if (3) holds for U = D.
In (3) we always assume that the expectation is absolutely convergent. If D is
unbounded then by the usual convention Ex u(XD ) = Ex [u(XD ); D < ]. By
the strong Markov property a regular -harmonic function is -harmonic. The
converse is not generally true as shown in [8, 16]. An alternative description of
-harmonic functions as those annihilating the fractional Laplacian


f (y) f (x) f (x) (y x)1{|yx|<1}
/2
dy
 f (x) = Ad,
|y x|d+
Rd


f (y) f (x)
dy,
= Ad, lim+
d+
0
B(x,)c |y x|
is given in [10]. Here and below, Ad, = 21 d/2 [(d + )/2]/ (1 /2)
is an appropriate normalizing constant ([5, 27]). It follows from (1) and (3) that an
-harmonic function f in D satisfies


Pr (x , y )f (y) dy, x B(, r),
(4)
f (x) =
|y|>r

for every ball B(, r) of closure contained in D. In fact, the condition characterizes
functions -harmonic in D. The integral in (4) is absolutely convergent and, by (1),
f is smooth in D and


|f (y)|
dy < .
(5)
d+
Rd (1 + |y|)
If, in addition, f is nonnegative on Rd and nonzero in D, then it is positive in D,
regardless of the connectedness of D. This is a consequence of Harnack inequality,
see, e.g., Lemma 1, p. 138 in [9].
3. Kernel Functions
The cones  described in the introduction are called right circular cones. By a
generalized cone in Rd we shall mean in this paper an open set  Rd with the
property that if x  and r > 0 then rx . If 0  then  = Rd . Otherwise,
 is characterized by its intersection with the unit sphere Sd1 Rd . Namely, let
= be a relatively open subset of Sd1. Without loosing generality in what
follows, we assume that 1 = (0, 0, . . . , 0, 1) . The generalized cone spanned
by is then
 =  = {x Rd : x = 0 and x/|x| }.
Note that we do not impose any regularity properties on , in particular  may
be disconnected.
For n = 0, 1, . . . , we let Bn = {|x| < 2n } and n =  Bn .
The following result follows from [33].

267

EXIT TIMES FROM CONES

LEMMA 3.1 (Boundary Harnack principle). There is a constant C1 = C1 (, )

such that for all functions u, v
0 on Rd which vanish on  c B1 and satisfy:
u(x) = v(x) for some x 0 ,
u(x) = Ex u(X1 ),

for all x Rd ,

(6)

v(x) = Ex v(X1 ),

for all x Rd ,

(7)

and

we have
C11 v(x)  u(x)  C1 v(x),

x B0 .

(8)

We note that by (6) and (7), u and v are regular -harmonic on 1 . By considering the regular -harmonic function v(x) = Px [X1 B1c ]  1 defined by the
boundary condition: v = 1 on B1c and v = 0 on B1  c , we see that (8) implies
that every function u satisfying the assumptions of Lemma 3.1 is bounded on B0
and in fact,
u(x)  C2 u(1),

|x| < 1,

(9)

where C2 = C12 /P1 [X1 B1c ] < depends only on and .

Recall that a function h : Rd \ {0} R is homogeneous of degree if
h(rx) = r h(x),

r > 0, x = 0,

or, equivalently, h(x) = |x| h(x/|x|), x = 0.

THEOREM 3.2. There exists a unique nonnegative function M on Rd such that
M(1) = 1, M = 0 on  c and for every open bounded set B 
M(x) = Ex M(XB ),

x Rd .

(10)

The function is locally bounded on Rd and homogeneous of degree = (, ).

That is,
M(x) = |x| M(x/|x|),

x = 0.

(11)

Here, = 0 if  c is a polar set for Xt and 0 < < , otherwise.

The function M will be called the Martin kernel with pole at infinity for . The
proof of Theorem 3.2 is given in Section 6 below.
Note that if  is a right circular cone then it is a Lipschitz domain. In this
case the uniqueness of M follows from the uniqueness of the Martin representation
of nonnegative -harmonic functions in Lipschitz domains [8, 16] (see also the
beginning of p. 240 and Example 2 in [8] in this connection). The homogeneity of
M then follows by a simple argument given at the end of the proof of Theorem 3.2.
This observation was stated without proof in Example 4.1 in [12].

268

RODRIGO BAUELOS AND KRZYSZTOF BOGDAN

By (10) the function M is regular -harmonic on every open bounded subset of

 and by Lemma 3.1 the decay rate of M, namely |x| as x 0 along radii, is
universal among all functions satisfying the assumptions of Lemma 3.1. By (9)
M(x)  C2 |x| ,

x Rd .

(12)

We shall now describe some examples of generalized cones where the exponent
can be explicitly identified.
EXAMPLE 3.1. Consider  = Rd . The only nonnegative -harmonic functions
on the whole of Rd are constants, see, e.g., Lemma 3.2, p. 545, in [12]. Thus M 1
and = 0 for this cone.
EXAMPLE 3.2. We put
 /2
if xd > 0,
M(x) = xd
0
if xd  0,
where x = (x1 , x2 , . . . , xd ) Rd . Then M is -harmonic on Rd + = Rd {xd > 0}
by checking that /2 M(x) = 0, x Rd + (see the calculations pp. 2729 in [11]
and notice that (, /2) = 1/, with the notation therein). By Theorem 3.2, it
is the Martin kernel at infinity for  = Rd + . In this case, = /2. (This rate of
decay is characteristic for smooth domains [15, 24].)
EXAMPLE 3.3. Let > 1 and consider the function M(x) = |xd |1 , x Rd .
Using the calculations in [11], pp. 2729 and noticing that


1
t 1 (1 + t)1 dt =

0
for every (0, 2), one obtains /2 M(x) = 0, x Rd , xd = 0. Again by
Theorem 3.2, M is the Martin kernel at infinity for cone  = Rd \ {xd = 0} and
= 1 in this case. For the same cone but for  1, M is the indicator function
of Rd \ {xd = 0} since {xd = 0} is a polar set if  1 [27]. This time we have
= 0.
Let ,  be generalized cones in Rd , and let m, M be their respective Martin kernels
with pole at infinity. If  then
Px {X1 B1c }  Px {X1 B1c },

x Rd .

(13)

Here 1 = B1 . Hence, by BHP there is a constant c such that

m(x)  cM(x),

x B0 .

(14)

We conclude that the respective homogeneity exponents satisfy ( , )
(, ),

0 < < 2. In fact, we have the following result.

269

EXIT TIMES FROM CONES

LEMMA 3.3. If  then ( , )
(, ). Furthermore, ( , ) > (, )

if and only if  \ is a non-polar set.
Proof. We adopt the notation above. We only need to prove the second statement
of the lemma. Let  \ be nonpolar. We (falsely) assume that ( , ) = (, ).
There exists a compact K 1 \ such that for some (hence every) starting point
x 1 , TK = 1 < 1 with positive probability, where we set TK = inf{t

0; Xt K}. We clearly have
M(x) > Ex {M(X1 )1 (X1 )},

x 1 .

(15)

Let
a = inf
x

M(x)
.
m(x)

By (14) and our assumption ( ) = () we have a > 0. We define H = M1

am, and h(x) = Ex H (X1 ), x Rd . By (15) we have
h(x) < M(x) am(x) = H (x),

x 1 .

We note that H is homogeneous of degree ( ) = (), nonnegative and continuous, so if H (y) > 0 for some y (hence on a set of positive Lebesgue measure)
then h(x) > 0 for all x 1 by (4). By BHP we have that h
m on B(0, 1)
with some > 0, thus H
m everywhere. In particular M am
m or
a = infx M(x)/m(x)
a +, which is a contradiction. We conclude that H 0
or M1 = am. By the mean value property of m
M(x) = Ex {M(X1 )1 (X1 )},

x 1 .

This contradicts (15) and thus ( ) = ().

On the other hand, if  \ is a polar set then there is equality in (13) for x ;
in particular ( ) = ().
2
The above lemma gives a positive answer to the question of Ewa Damek whether
the asymptotics of harmonic functions in obtuse cones is different than in the
half-space, where = /2. On the other hand the proof of the lemma does not
give quantitative information on ( ) () when  \ is non-polar. We expect
that spectral analysis of the spherical fractional Laplacian defined below may give
such quantitative results.
By an application of Kelvin transform (see Lemma 8 in [8]) the function
K(x) = |x|d M(x/|x|2 ) = |x|d K(x/|x|),

x = 0,

(16)

is -harmonic in T  = {x/|x|2 ; x } = , if  = Rd . For completeness we put

K(0) = 0 so that K = 0 on  c . We call K the Martin kernel at 0 for , which is
justified by the following theorem.

270

RODRIGO BAUELOS AND KRZYSZTOF BOGDAN

THEOREM 3.4. K given by (16) is the unique nonnegative function on Rd such

that K(1) = 1, K = 0 on  c and for every open set B  such that dist(0, B) > 0,
K(x) = Ex {K(XB ); B < },

x Rd .

(17)

The proof of Theorem 3.4 is given in Section 6 below.

EXAMPLE 3.4. In the context of Example 3.2 we obtain
 /2
d
if xd > 0,
K(x) = xd |x|
0
if xd  0.
On the other hand the Martin kernel at 0 for  in Example 3.3 is K(x) =
|xd |1 |x|2d provided > 1 (in particular K(x) = 1 for x = 0, if d = 1)
and

|x|d , xd = 0,
K(x) =
0,
xd = 0,
provided  1. For  1, this is essentially Riesz kernel (see [27], p. 43).

4. Integrability of Exit Times

We will write P0x and E0x for the probability and expectation associated with our
stable process killed off the cone  and conditioned by the Martin kernel K of
 with the pole at 0, as defined above. The process is a special case of the Doob
h-process, in particular for any bounded or nonnegative function f on  we have
E0x {f (Xt );  > t} =

1
Ex {K(Xt )f (Xt );  > t}.
K(x)

THEOREM 4.1. Let be the homogeneity degree of the Martin kernel M of the
cone . For p > 0 and x  we have
p

if and only if

p < /,

(18)

if and only if

p < (d + 2)/.

(19)

Ex  <
and
E0x  <

The proof of Theorem 4.1 follows immediately from the formula

p
t p1 P( > t) dt, p > 0,
E = p
0

valid for any positive random variable on any probability space, and the following
two lemmas.

271

EXIT TIMES FROM CONES

LEMMA 4.2. There is C3 = C3 (, ) such that for all t > 0 and x Rd
satisfying |x| < t 1/ we have
C31 M(x)t /  Px { > t}  C3 M(x)t / .

(20)

Proof. We first prove that there is c1 = c1 (, ) such that

c11 M(x)  Px { > 1}  c1 M(x),

|x| < 1.

(21)

This is a consequence of the boundary Harnack principle. Indeed, we let, as usual,

n =  Bn ,

Bn = {|x| < 2n }, n = 0, 1, . . . ,

and we have
Px { > 1}  Px {1 > 1} + Px {1 <  },

x Rd .

By the boundary Harnack principle

Px {1 <  }  C1 P1 {1 <  }M(x),

|x| < 1,

where C1 is the constant of Lemma 3.1. We let

Ad,
dy.
c2 = inf
v1 \ |y v|d+
1
Clearly, c2 > 0. We denote by G1 the Green function of 1 for our process {Xt }.
By IkedaWatanabe formula ([23]) and the boundary Harnack principle


G1 (x, v) dv
Px {1 > 1}  Ex 1 =



1
Ad,
G1 (x, v)
dv dy
 c21
|y v|d+
\1 1
= c21 Px {X1 }  c21 C1 P1 {X1 }M(x),

|x| < 1.

This verifies the upper bound in (21). We then have

Px { > 1}
Px {3 > 1}

Ex [X1 B(41, dist(1,  c )); PX {3 > 1}].
1

Here, B(41, dist(1,  )) is the ball centered at 41 = (0, 0, . . . , 0, 4)  and of

radius dist(1,  c ).
It is easy to verify that there is c3 = c3 (, ) such that Pz {3 > 1} > c3 for all
z B(41, dist(1,  c )). Thus, by the boundary Harnack principle, for |x| < 1 we
have
c

Px {3 > 1}
c3 Px [X1 B(41, dist(1,  c ))]

c3 C11 P1 [X1 B(41, dist(1,  c ))]M(x).

272

RODRIGO BAUELOS AND KRZYSZTOF BOGDAN

The proof of (21) is complete.

To prove (20), we use the scaling of {Xt }, (21) and the homogeneity of M. For
t > |x| the upper bound in (21) gives
Px { > t} = Pt 1/ x { > 1}  c1 M(t 1/ x)
= c1 M(x)t / .
Similarly, the lower bound in (21) gives Px { > t}
c11 M(x)t / , completing
the proof of Lemma 4.2.
2
LEMMA 4.3. There is C4 = C4 (, ) such that for all t > 0 and x  satisfying
|x| < t 1/ we have
C41 t (d2)/ |x|d+2  P0x { > t}  C4 t (d2)/ |x|d+2 .

(22)

Proof. For clarity we note that M(x)/K(x) = |x|d+2 , x , which is one

factor in (22). As in the proof of Lemma 4.2, we first consider t = 1 in (22). We
have
P0x { > 1} = K(x)1 Ex {K(X1 );  > 1}.
To prove (22) for t = 1 we only need to verify that there is c1 = c1 (, ) such that
c11 M(x)  Ex {K(X1 );  > 1}  c1 M(x),

x 0 .

(23)

We have
Ex {K(X1 );  > 1} = Ex {K(X1 ); 3  1,  > 1} + Ex {K(X1 ); 3 > 1}.
By the -harmonicity of K and Fatous lemma
K(x)
Ex K(X3 ),

x Rd .

Thus,
Ex {K(X1 ); 3 > 1}

Ex {EX1 K(X3 ); 3 > 1} = Ex {K(X3 ); 3 > 1}

Ex {EX {K(X3 ); 3 > 1}; X1 B(41, dist(1,  c ))}.
1

There is c2 = c2 (, ) such that Ez {K(X3 ); 3 > 1}
c2 for z B(41,

dist(1,  c )). By the boundary Harnack principle and the above
Ex {K(X1 ); 3 > 1}
c2 Ex {X1 B(41, dist(1,  c ))}

c2 C11 M(x)E1 {X1 B(41, dist(1,  c ))},
which gives the lower bound in (23). To prove the upper bound we note that the
transition density of the stable process killed off 3 satisfies
p13 (x, y)  c3 M(x)M(y),

x, y Rd ,

(24)

273

EXIT TIMES FROM CONES

where c3 = c3 (, ), as we shall see in Lemmas 4.5 and 4.6 below. It follows that


Ex {K(X1 ); 3 > 1} =
p 3 (1, x, y)K(y) dy
3


M(x)M(y)K(y) dy  c4 M(x),
 c3
3

with c4 = c4 (, ). Note that for every x Rd and s
0, Ex {K(Xs );  > s} 

K(x) by -harmonicity of K and Fatous lemma. By strong Markov property, this
inequality and Lemma 3.1 we have
Ex {K(X1 ); 3  1,  > 1} = Ex {E{K(X1 );  > 1|F3 }; 3  1}
= Ex {EX {K(XS );  > S}S=13 ; 3  1}
3

 Ex K(X3 )  C1 E1 K(X3 )M(x)

 C1 K(1)M(x) = C1 M(x).
The proof of (23) is complete.
To prove (22) we use (23), the scaling of {Xt }, and the homogeneity of M and
K. That is, for t > 0 we have
1
Ex {K(Xt );  > t}
K(x)
1
Et 1/ x {K(t 1/ X1 );  > 1}
=
K(x)
 c1 (t 1/ )d M(t 1/ x)/K(x) = c1 t (d2)/ |x|d+2 .

P0x { > t} =

The lower bound

Px0 { > t}
c11 t (d2)/ |x|d+2
is proved similarly. This completes the proof of Lemma 3.3 under the assumption
that (24) holds.
2
The estimate (24) is a direct consequence of the intrinsic ultracontractivity of the
semigroup of the killed process Xt which is valid for any bounded domain in Rd
[25], see also [17]. This estimate, however, can be proved by other more elementary
means, comp. [32]. For a domain D Rd let s(x) = sD (x) = Ex D , x Rd . If
sup{s(x) : x Rd } < then D is called Green bounded [10, 18]. If the volume,
|D|, of D is finite and D denotes the ball of same volume as D centered at the
origin, then (see [1]),
sup sD (x)  sD (0) = cd, |D|/d < .
xRd

Thus domains of finite volume are Green bounded. It is also well known that there
exist Green bounded domains of infinite volume.

274

RODRIGO BAUELOS AND KRZYSZTOF BOGDAN

LEMMA 4.4. Suppose D Rd is such that there exits a constant C such that
s(x)  C0 for all x Rd . Then for n = 1, 2, . . . ,
Ex Dn  n!C0n1 s(x),

x Rd ,

Ex exp(D ) 1  s(x)/(1 C0 ),

x Rd , 0 < < 1/C0 ,

(25)

and
Px {D > t} 

s(x)
,
1 C0 exp(t) 1

x Rd , t > 0, 0 < < 1/C0 .

Proof. By the strong Markov property of Xt we have for any r > 0,

Px {D > t} dt =

Px {D > t + r}

0

Ex {PXr {D > t}; D > r} dt





PXr {D > t} dt
= Ex 1{D >r}
=

 C0 Px {D > r}.
For n
2 we multiply both sides of the above inequality by (n 1)r n2 and
integrate, applying Fubinis theorem on the left-hand side, to obtain
Ex Dn  nC0 Ex Dn1 .
The first asserted inequality follows by induction. The other two inequalities are
even easier and are left to the reader.
2
Let ptD (x, y) be the transition density of the process Xt killed off D:


ptD (x, y) dy.
Px {D > t, Xt A} =
A

Clearly
ptD (x, y)  ptR (x, y) = t d/ p1R (xt 1/ , yt 1/ )

d/
d
(2 )
e| | d = ct d/ , x, y D, t > 0,
 t
d

Rd

where c = (2 )d d (d/)/ and d is the surface measure of the unit sphere in

Rd . Recall that


ptD (x, y) dy.
Px {D > t} =
D

275

EXIT TIMES FROM CONES

From these two observations and the semigroup property of ptD (x, y) we have


D
D
p3t (x, y) =
p2t
(x, z)ptD (z, y) dz
D




D
D
pt (x, w)pt (w, z) dw ptD (z, y) dz
=
D

c
t d/

Px {D > t}Py {D > t}.

(26)

This together with Lemma 4.4 gives the following result.

LEMMA 4.5. Under the assumptions of Lemma 4.4 there is C = C(, D, t0 , )
such that
ptD (x, y)  Ce2t /3s(x)s(y),

x, y Rd , t > t0 , 0 < < C0 .

Finally, the next lemma yields the estimate (24).

LEMMA 4.6. Let s(x) = Ex k , k = 1, 2, . . . . There is c = c(, d, , k) such
that
s(x)  cM(x),

x Rd .

Also,
s(x)
c1 M(x),

x k1 .

The lemma can be proved in a very similar way as Lemma 4.2, by using the
boundary Harnack principle and the IkedaWatanabe formula. We skip the details.
From (20) and (22) we immediately obtain the following result.
COROLLARY 4.7. There are constants C3 and C4 depending only on , and d,
such that for all x ,
C31 |x| M(x/|x|)  lim inf t / Px { > t}
t

 lim sup t / Px { > t}



t
C3 |x| M(x/|x|)

(27)

and
C41 |x|d+2  lim inf t (2+d)/ P0x { > t}
t

 lim sup t (2+d)/ P0x { > t}

t

 C4 |x|d+2 .

(28)

We also have the following heat-kernel version of this corollary. Recall that
pt (x, y) are the transition densities (heat kernel) of Xt killed off .

276

RODRIGO BAUELOS AND KRZYSZTOF BOGDAN

COROLLARY 4.8. There is a constants C5 depending only on , and d, such

that for all x, y ,
C51 |x| |y| M(x/|x|)M(y/|y|)
 lim inf t (2+d)/ pt (x, y)
t

 lim sup t (2+d)/ pt (x, y)



t
C5 |x| |y| M(x/|x|)M(y/|y|).

(29)

Proof. By (26) we have


(x, y) 
p3t

c
t d/

Px { > t}Py { > t}

with c = (2 )d d (d/)/. The upper bound follows from this and the previous
result.
For the lower bound, suppose first that |x|, |y| < 1. By domain monotonicity,
we have p1 (x, y)
p13 (x, y). However, since 3 is a bounded domain, it is
intrinsically ultracontractive by [25, 17]. Therefore
p13 (x, y)
c1 Px {3 > 1}Py {3 > 1}.
From our proof of (21) it follows that for |x| < 1 we have Px {3 > 1}
c2 M(x).
From this we conclude that for |x|, |y| < 1,
p1 (x, y)
c3 M(x)M(y).
This inequality and scaling gives that for all t > max(|x| , |y| ),
pt (x, y) = t d/ p1 (t 1/ x, t 1/ y)

c3 t d/ M(t 1/ x)M(t 1/ y)
= c3 t (2d)/ M(x)M(y).
The left-hand side of (29) follows from this.

Corollaries 4.7 and 4.8 should be compared with the corresponding results for the
Brownian motion in generalized cones ([2] (1.5), (1.10) and (2.2)). However, the
exact limits are computed in [2]. It would be interesting to have the exact limits in
the current setting as well. In addition to generalized cones, the above limits have
been studied in [3, 34, 28, 29] in the case of Brownian motion in parabolas and in
other parabolic regions of the form D = {(x, y) R Rd1 : x > 0, |y| < Ax p },
0 < p < 1. The asymptotic behavior of the distribution of the exit time and of the
heat kernel for such regions is shown to be subexponential. It would be interesting
to determine the behavior of the distribution of the exit time and the transition
densities for stable processes in these regions. The scaling techniques we use here
do not seem to apply.

277

EXIT TIMES FROM CONES

5. Spherical Fractional Laplacian

In this section we study the action of /2 on homogeneous functions. We also
introduce the corresponding spherical operator and give some of its properties
which may be of importance in studying the exponent (, ).
We first consider an arbitrary function on Rd such that /2 (1) is well
defined. This is satisfied if


|(y)|
dy < ,
(30)
d+
Rd (1 + |y|)
and, say, |(1 + x) (1) (1) x|  c|x|2 for |x| < 1, e.g., is C 2 at 1.
LEMMA 5.1. We have


(y 1)1{|y1|<1}
dy = 0,
lim
0+ Rd \
|y 1|d+
and

/2

(1) = Ad, lim+

0

Rd \

(31)

(y) (1)
dy.
|y 1|d+

(32)

The proof of (31) is somewhat tedious (because  is not symmetric about 1) but
elementary and will be omitted. The formula (32) follows from (31) immediately.
Let be also homogeneous of degree ; (x) = |x| (x/|x|), x = 0. In view
of (30) we will only consider d < < .
By (32) and polar coordinates


(y) (1)
/2
dy
 (1) = Ad, lim+
d+
0
Rd \ |y 1|




()r (1)
(d)
r d1
dr
= Ad, lim+
0
|r 1|d+
0
Sd1 \


[()u (d ) (1)u0 (d )] (d).
(33)
= Ad, lim+
0

Here

u (t) =

Sd1 \

r d+ 1 (r 2 2rt + 1)(d+)/2 dr,

1  t  1

(34)

(we drop d and from the notation); and (33) holds because




d+ 1
d
r
|r 1|
dr =
r d+ 1 (r 2 2rd + 1)(d+)/2 dr.
0

LEMMA 5.2. For every 1  t  1

u (t) = ud (t),

d < < ,

(35)

and the function  u (t) is increasing on [( d)/2, ) with u (t) = .

278

RODRIGO BAUELOS AND KRZYSZTOF BOGDAN

Proof. Let 1  t  1. By a change of variable

1
u (t) =
r d+ 1 (r 2 2rt + 1)(d+)/2 dr +
0


s d 1 (s 2 2s 1 t + 1)(d+)/2 ds
+
0

1
1 d+
(r
+ r )(r 2 2rt + 1)(d+)/2 dr,
=
0 r
which proves (35). Let ( d)/2  1 < 2 < . We have
u2 (t) u1 (t)

1
1 d+2
+ r 2 r d+1 r 1 )(r 2 2rt + 1)(d+)/2 dr
(r
=
r
0

1
1 1
=
(r r 2 )(r 2 1 r d )(r 2 2rt + 1)(d+)/2 dr > 0,
(36)
r
0
because 2 1 < 21  ( d) = d. To verify that u (t) =
we note that for


r d+ 1 (r 2 2rt + 1)(d+)/2 dr
u (t)


1
r d+1 (r 2 2rt + 1)(d+)/2 dr

1


r d+1 (r + 1)d dr = .


1

The proof is complete.

By (33) for homogeneous of degree we have


/2
[() (1)]u0 (d ) (d) +
 (1) = Ad, P.V.
Sd1


()[u (d ) u0 (d )] (d).
+ Ad,

(37)

Sd1

The principal value integral (P.V.) above is understood as in (33). We note that the
second integral in (37) vanishes if is homogeneous of degree 0 and the first one
vanishes whenever is constant on the unit sphere. The observation can be used
to show that the integrals converge and the second one converges absolutely (conf.
Lemma 4.5), a fact that can be also verified by a detailed inspection of u at t = 1.
A formula similar to (37) clearly holds for every vector Sd1:


[() ()]u0 ( ) (d) +
/2 () = Ad, P.V.
Sd1


()[u ( ) u0 ( )] (d),
(38)
+ Ad,
Sd1

279

EXIT TIMES FROM CONES

where denotes the usual scalar product of and . The operator

/2
[() ()]u0 ( ) (d)
Sd1 () = Ad, P.V.
Sd1


= Ad, lim+
[() ()]u0 ( ) (d)
0

(39)

Sd1 {1>}

will be called the spherical fractional Laplacian. The second integral in (38) will
be called the radial part and denoted R below.
REMARK 1. By the proof of Lemma 5.2 we see that (x) = |x| is -harmonic
on Rd \ {0} if and only if = d.
We will be concerned with nonnegative definiteness of the kernel u ( ).
We first consider dimension d = 2 and we will identify R2 with the complex
plane C. By a rotation, Ox2 axis in R2 will be identified with z axis in C so that
1 R2 is now represented by 1 C. S1 will be replaced by ei , T
[0, 2 ) in formulas. With this identification in mind we define


u(2)
()
=
r 2+ 1 |rei 1|2 dr.

For r [0, 1) and > 0 let f () = |rei 1| =

in
,
cn e

[0, 2 ).

LEMMA 5.3. For every > 0 and r (0, 1) we have cn > 0,

n Z.
n
Proof. Let z = rei . We have |z| < 1 and log(1 z) =
n=1 z /n, thus
log |1 z| =

n=1

n in

r e

/n =

r |n| ein /2|n|.

n=,n=0

Since |1 z| = exp( log |1 z|), the sequence {cn = cn (r)} is the convolution exponent of the sequence {1{n=0} r |n| /(2|n|)}
. The latter sequence is
2
nonnegative and so {cn } is positive (c0
1).
Let be a test function and let , [0, 2 ). We consider the operator

2
(2)
(2)
()[u(2)
R () =
( ) u0 ( )] d.
0

For
0 we have that R(2)
0 in the sense that

2
R(2)()() d
0
0

for all such . In fact we have the following result.

R(2)

0.
LEMMA 5.4. If ( d)/2  1  2 < then R(2)
2
1

280

RODRIGO BAUELOS AND KRZYSZTOF BOGDAN

Proof. Let () = n= an ein L2 (0, 2 ). By (36) and Lemma 5.3 with

= d + we have

2
[R(2)
R(2)
]()() d
2
1
0

1
1 1
(r r 2 )(r 2 1 r d )
=
r
0

2
2

cn ein() ()() d d dr
0

= 4

|an |

2
0

1 1
(r r 2 )(r 2 1 r d )cn (r) dr > 0,
r

unless = 0 a.e. (comp. the proof of Lemma 5.2).

THEOREM 5.5. {R , [0, )} is an increasing family of nonnegative definite

operators on L2 (Sd1 , ), d
2.
Proof. Let ( d)/2  1  2 < . We only need to prove that K = R2 R1
is nonnegative definite. For dimension d = 2 this is proved above under a slightly
different notation. We now let d
3. For a test function on Sd1


K() = Ad,
()[u2 ( ) u1 ( )] (d),
(40)
Sd1

compare (38). K has the form of spherical convolution hence it is diagonalized by

spherical harmonics;
KYm = Km Ym ,

m = 0, 1, . . . ,

(41)

where Ym is any spherical harmonics of degree m, and, by FunkHecke formula

([22], (11.4.24), p. 248 or [31], (2.15)(2.19), p. 10),

1
[u2 (t) u1 (t)]Cmd/21 (t)[1 t 2 ](d3)/2 dt,
(42)
Km = cm
1

where
cm =

2 (d1)/2Ad,
d/21

((d 1)/2))Cm

d/21

(1)

(43)

are the Gegenbauer (ultraspherical)

and Cm

d+m3 polynomials ([22], (10.9), p. 174).

d/21
(1) =
.
Notice that cm > 0 since Cm
m
We only need to verify that Km
0, m = 0, 1, . . . . By (36) it is enough to
prove that for every r [0, 1) the spherical convolution


()[r 2 2r + 1](d+)/2 (d)
k() =
Sd1

EXIT TIMES FROM CONES

281

has nonnegative eigenvalues

1
[r 2 2rt + 1](d+)/2Cmd/21 (t)[1 t 2 ](d3)/2 dt.
km = cm

(44)

By Rodrigues formula ([22], (10.9.11), page 175),

Cmd/21 (t) = (1)m dm (1 t 2 )(3d)/2

dm
[(1 t 2 )m+(d3)/2],
dt m

where
dm =

(d 2)m
,
2m m!(d/2 1/2)m

and we used the notation ()k = ( + 1) . . . ( + k 1). Integrating by parts we

obtain

1
dm
m
km = (1) cm dm
[r 2 2rt + 1](d+)/2 m [(1 t 2 )m+(d3)/2] dt
dt
1

1 m
d
[r 2 2rt + 1](d+)/2 (1 t 2 )m+(d3)/2 dt.
= cm dm
m
dt
1
By induction we easily check that


dm 2
(d+)/2
m d +
[r 2rt + 1]
=2
r m [r 2 2rt + 1](d+)/2m .
dt m
2
m
Thus


km = 2m

d +
2
1


cm dm r m
m

[r 2 2rt + 1](d+)/2m (1 t 2 )m+(d3)/2 dt > 0.

2

The proof is complete.

/2

For clarity we note that the operator Sd1 is negative semi-definite on

L2 (Sd1 , ). Namely, for test functions we have


/2
Sd1 ()() (d)
Sd1


1
[() ()]2 u0 ( ) (d) (d),
= Ad,
2
Sd1 Sd1
which is negative unless is constant on Sd1 . The proof of the above equality
using the approximation (39) is standard and will be omitted.

282

RODRIGO BAUELOS AND KRZYSZTOF BOGDAN

/2

EXAMPLE 5.1. Let d = 2 and = 1. The kernel of Sd1 is

1
(r + 1)(r 2 2rt + 1)3/2 dr, |t|  1,
u0 (t) =
0

see the proof of Lemma 5.2. We use Euler change of variable:

x r, to get

1+2(1t ) 2
x + 2x 2t 1
dx.
u0 (t) = 2
(x 2 2tx + 1)2
1

r 2 2rt + 1 =

The primitive function of the last integrand is (x + 1)/(x 2 2tx + 1), which
yields
u0 (t) =

1
.
1t

Thus, for a test function on S1 we have

1
() ()
1/2
P.V.
(d).
S1 () =
1
2
1
S
In particular, if m () = (1 + i2 )m , where = (1 , 2 ), 12 + 22 = 1, m =
0, 1, 2, . . . (exponential basis on the circle), then, after a calculation, we obtain
1/2
/2
S1 m = |m|m . Similarly, trigonometric functions diagonalize S1 .
/2
We conclude this section with a variant of FunkHecke formula for Sd1 when
d
3. For every spherical harmonics Ym of order m = 0, 1, . . . we have
/2

Sd1 Ym = m Ym ,
where

m = cm

[Cmd/21(t) Cmd/21 (1)]u0 (t)[1 t 2 ](d3)/2 dt,

(45)

and cm is given in (43). The proof of (45) follows from the usual FunkHecke
formula via the approximation (39).
d/21
1. For m > 0 we recall a formula for the
Note that 0 = 0 because C0
Gegenbauer polynomials ([22], (10.9.17), p. 175):
Cm (cos ) =

m

()n ()mn
n=0

d/21

n!(m n)!

cos(m 2n).

attains its maximum at t = 1, and so m < 0.

Clearly, Cm
/2
A further study of spectral properties of Sd1 in relation to the increasing operator family {R , 0  < } discussed above may help to quantitatively describe
(, ) in terms of the trace of  on the unit sphere. Apart from such a program

283

EXIT TIMES FROM CONES

there is also an interesting problem to understand to what extent our spherical

fractional Laplacian is related to the spherical operator introduced in Ch. 8 of [31].
/2
With the notation of Ch. 8 in [31], it turns out that Sd1 = (Id1 )1 when d = 2
and = 1 but this may be more the exception than the rule.
6. Proofs of Theorems 3.2 and 3.4
Below we construct and prove the uniqueness of the Martin kernel M at infinity
for generalized cone . The intersection of  with the unit sphere may be highly
irregular however the scaling property of the cone: k =  (k > 0) allows for
application of arguments similar to those used in Lemma 16 in [7] and in [8] for
Lipschitz domains. Note that the uniqueness of the Martin kernel with the pole at a
boundary point of a general domain is an open problem for our stable process. For
more on this, we also refer the reader to [33], where results on the so called fat
sets are given.
Proof of Theorem 3.2. For s > 0 we write Ts = {|x|<s} and we define
hs (x) = Px {XTs },

x Rd .

By scaling we have that for all s, t > 0 and x Rd ,

 
t
hs (sx) = ht (tx) or hs (x) = ht x .
s
We claim there is
= (, ) < ,

(46)

and c1 = c1 (, ) such that

h1 (r1)
c1 r ,

0 < r  1.

(47)

The argument verifying (47) is given in the proof of Lemma 5 in [7] and we refer
the reader to that paper for details (take Q = 0, As (Q) = r1 and Ar (Q) = 1 in the
notation of [7]). The fact that is strictly less that is important and distinguishes
the present situation from that of the potential theory of Brownian motion. By
scaling we have hs (1)
c1 s , s
1. We define
us (x) =

hs (x)
,
hs (1)

s > 0, x Rd ,

so that us (1) = 1, s > 0. Note that

us (x) 

1
= c11 s ,
c1 s

s
1, x Rd .

(48)

We claim that
ut (x)  2 C1 c11 [|x| 1] ,

t
2, x Rd .

(49)

284

RODRIGO BAUELOS AND KRZYSZTOF BOGDAN

To verify (49), assume that t
2. If |x|
t/2 then we put s = 2(1 |x|) and by
Lemma 3.1 we have
ut (x)  C1 us (x)  C1 c11 2 [|x| 1] .
By Harnack inequality and our normalization ut (1) = 1, the functions ut are
uniformly bounded on F for any compact F  and equicontinuous on F for
all large t. The last assertion follows from the Poisson formula for the ball or the
gradient estimates of [12]. Therefore there is a sequence tn and a function
M such that
M(x) = lim utn (x),
n

x Rd

and we take M such that M = 0 on  c . Notice that, by (49),

M(x)  2 C1 c11 [|x| 1] ,

x Rd .

(50)

Let x  and set B = B(0, r), where r > |x|. We have for tn > r,


x
utn (y)B
(dy),
utn (x) =
(B)c

x
x
the -harmonic measure of  B. Since B
 Bx on
where we denote by B
c
B we have by (2)
x
(dy)  2d+ 3/2 Cd r |y|d dy,
B

|y| > 2r.

(51)

Since utn (y) M(y) for all y, by (49), (51) and dominated convergence we have




x
x
utn (y)B (dy) =
M(y)B
(dy),
M(x) = lim
n

(B)c

(B)c

proving that M is a regular -harmonic function on  B. We note that we used

here the integrability of |y|d+ at infinity, which is a consequence of (46).
By the strong Markov property, M is regular -harmonic on every open bounded
subset of . This proves the existence part of the theorem.
To prove the uniqueness of M, we assume that there is another function m
0
on Rd which vanishes on  c , satisfies m(1) = 1 and for which
m(x) = Ex m(XU ),

x Rd ,

for every open bounded U . By Lemma 3.1, for every x B0 we have

C11 m(x)  M(x)  C1 m(x).
Note that Lemma 3.1 is invariant upon dilations of the domain, so the above inequality holds with the same constant in arbitrarily large ball Bn , hence for all
x Rd . Let a = infx m(x)/M(x). For clarity, we observe that C11  a  1.
Let H (x) = m(x) aM(x), so that H
0 on Rd .

285

EXIT TIMES FROM CONES

Assume that H (x) > 0 for some, and therefore for every, x . Once again by
Lemma 3.1 and scaling
H (x)
M(x),

x Rd ,

for some > 0. This gives

m(x)
aM(x) + H (x)
= inf

a + ,
x M(x)
x
M(x)

a = inf

which is a contradiction.
Thus H 0 and hence m = aM. The normalizing condition m(1) = M(1) = 1
yields a = 1 and the uniqueness of M is verified.
It remains to prove the homogeneity property of M. By the scaling of Xt , for
every k > 0 the function M(kx)/M(k1) satisfies the hypotheses used to construct
M. By uniqueness this function is equal to M, that is, M(kx) = M(x)M(k1) for
x Rd . In particular, M(kl1) = M(l1)M(k1) for every positive k, l. By continuity
there exists such that M(k1) = k M(1) = k and
M(kx) = k M(x),

x Rd .

By (50), M is locally bounded, thus
0 and  < .

We claim that  c is non-polar if and only if 0 is a regular point of  c , that is,
P0 { = 0} = 1, where  = inf{t > 0; Xt  c } is the first hitting time of
 c . Indeed, it is enough to verify that if  c is non-polar, then P0 { = 0} = 1.
But in this case P0 { < 1/}
for some > 0. By scaling of Xt and ,
P0 { = 0}
. By the 01 law, P0 { = 0} = 1.
Consider u(x) = Px {X1 }, x Rd . If 0 is regular for  c , then u(x) 0
as x 0. In consequence, by Lemma 3.1, M(k1) 0 as k 0+ . Therefore
> 0. If  c is polar, the indicator function of  satisfies the hypotheses defining
M and so = 0 in this case. This completes the proof of Theorem 3.2.
2
We now consider K, the Martin kernel with the pole at 0 for the generalized cone
 Rd . Before we prove Theorem 3.4 we note that the function K as defined
by (16) is bounded in the complement of every neighborhood of 0. For dimension
d = 1 this follows by inspection of Example 3.4. For d
2 we even have that
K(x) 0 as |x| .
Proof of Theorem 3.4. Consider K defined by (16). Let > > 0 and U =
 { > |y| > }. Consider an increasing sequence
 {Un } of open sets such that
each closure Un is a compact subset of U and U = Un . Let x Rd . Since K is
-harmonic in  we have K(x) = Ex K(XUn ), n = 1, 2, . . . . Let
O = {Un = U for some n},

(52)

P = {Un < U for every n}.

(53)

286

RODRIGO BAUELOS AND KRZYSZTOF BOGDAN

For every n,
K(x) = Ex {K(XUn ); O} + Ex {K(XUn ); P }
= Ex {K(XU ); O, Un = U } + Ex {K(XUn ); O, Un < U } +
+ Ex {K(XUn ); P }.

(54)

We have that K is bounded on U and continuous on Rd , except for a polar subset.

By monotone convergence, dominated convergence and the quasi-left-continuity
of the paths of Xt ,
K(x) = Ex {K(XU ); O} + Ex {K(XU ); P } = Ex K(XU ).
We now let U =  {|y| > }, 0 < < 1 and Un =  {n > |y| > },
n = 1, 2 . . . . For these new sets Un we define O and P by (52) and (53), and we
obtain (54). If < d or  c is non-polar, then K(x) 0 as |x| . Hence, the
second and the third terms in (54) tend to 0 as n . We thus obtain
K(x) = Ex {K(XU ); O} = Ex {K(XU ); U < }.
If d = 1  and  = R \ {0}, then K is given by Example 3.4. Since the process
Xt is recurrent in this case, we obtain
K(x) = 1 = Ex K(XU ), if x = 0.
We used here the observation that XU = 0 Px -a.s., for |x| > . In fact, the Px
distribution of XU is absolutely continuous with respect to the Lebesgue measure
on U c (see Theorems B and C, p. 541, in [6]).
The case of general U in (17) now follows by strong Markov property.
We now sketch a proof of uniqueness of K. Assume that K is a nonnegative

= 1, K = 0 on  c and for every open set U 

function on Rd such that K(1)
such that dist(0, U ) > 0
U ); U < },

K(x)
= Ex {K(X

x Rd .

(55)

By [33], K is locally bounded in Rd \ {0}. By (5) it is integrable in any bounded

neighborhood of 0. Given a cone  Rd , r > 0, U = Rd {|y| > r} and
A Rd {|y|  r}, we have


Pr (x, y)dy,
Px {U  < , XU A}  Px {U < , XU A} =
A

for every x U , where Pr (x, y) is given by (2) for  d (see Theorem C in [6]
for the case d = 1 < ). If  d then


Pr (x, y) dy  C|x|d , x U,
{|y|<r}

EXIT TIMES FROM CONES

287

as can be shown from [4], Lemma 2.2. From this it follows that K(x)
 C|x|d
for all sufficiently large x. Hence,
2

)C
T K(x)
= |x|d K(x/|x|

in every neighborhood of 0. Since {0} is polar if  d, it follows that T K is

regular -harmonic in every bounded subset of . Thus T K = M by Theorem 3.2
and so K = K.
If d = 1 < and  = R1+ , then a similar argument works (see [8] for the
Poisson kernel for the half-line).
If d = 1 < and  = R \ {0}, then, since the process is recurrent, by
(55) and Harnack inequality, K is bounded in a neighborhood of 0, hence in R.

Using the Kelvin transform we see that T K(x)

 c|x|1 and this goes to 0 as x

goes to 0. Thus T K is regular -harmonic in every open bounded subset of . By

Theorem 3.2 T K = M or K = K, as before.
2

Acknowledgement
We are very grateful to an anonymous referee for the careful reading of this paper
and the many useful comments which have made the paper easier to read.
References
1.

2.
3.
4.
5.
6.
7.
8.
9.
10.
11.

Bauelos, R., Lataa, R. and Mndez-Hernandez, P.J.: A BrascampLiebLuttinger-type inequality and applications to symmetric stable processes, Proc. Amer. Math. Soc. 129(10)
(2001), 29973008 (electronic).
Bauelos, R. and Smits, R.: Brownian motion in cones, Probab. Theory Related Fields 108
(1997), 299319.
Bauelos, R., DeBlassie, R.D. and Smits, R.: The first exit time of planar Brownian motion
from the interior of a parabola, Ann. Probab. 29 (2001), 882901.
Bass, R.F. and Cranston, M.: Exit times for symmetric stable processes in Rn , Ann. Probab.
11(3) (1983), 578588.
Blumenthal, R.M. and Getoor, R.K.: Markov Processes and Potential Theory, Springer-Verlag,
New York, 1968.
Blumenthal, R.M., Getoor, R.K. and Ray, D.B.: On the distribution of first hits for the
symmetric stable process, Trans. Amer. Math. Soc. 99 (1961), 540554.
Bogdan, K.: The boundary Harnack principle for the fractional Laplacian, Studia Math.
123(1) (1997), 4380.
Bogdan, K.: Representation of -harmonic functions in Lipschitz domains, Hiroshima Math.
J. 29(2) (1999), 227243.
Bogdan, K. and Byczkowski, T.: Probabilistic proof of boundary Harnack principle for
-harmonic functions, Potential Anal. 11 (1999), 135156.
Bogdan, K. and Byczkowski, T.: Potential theory for the -stable Schrdinger operator on
bounded Lipschitz domains, Studia Math. 133(1) (1999), 5392.
Bogdan, K., Burdzy, K. and Chen, Z.-Q.: Censored stable processes, to appear in Probab.
Theory Related Fields (2001).

288
12.
13.
14.
15.
16.
17.

18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.

RODRIGO BAUELOS AND KRZYSZTOF BOGDAN

Bogdan, K., Kulczycki, T. and Nowak, A.: Gradient estimates for harmonic and q-harmonic
functions of symmetric stable processes, Illinois J. Math. 46(2) (2002), 541556.
Burdzy, K.: Multidimensional Brownian Excursions and Potential Theory, Pitman Research
Notes in Mathematics 164, Longman Scientific & Technical, 1987.
Burkholder, D.L.: Exit times of Brownian motion, harmonic majorization, and Hardy spaces,
Adv. Math. 26(2) (1977), 182205.
Chen, Z.Q. and Song, R.: Estimates on Green functions and Poisson kernels for symmetric
stable process, Math. Ann. 312(3) (1998), 465501.
Chen, Z.-Q. and Song, R.: Martin boundary and integral representation for harmonic functions
of symmetric stable processes, J. Funct. Anal. 159(1) (1998), 267294.
Chen, Z.Q. and Song, R.: Intrinsic ultracontractivity, conditional lifetimes and conditional
gauge for symmetric stable processes on rough domains, Illinois J. Math. 44(1) (2000), 138
160.
Chung, K.L. and Zhao, Z.: From Brownian Motion to Schrdingers Equation, Springer-Verlag,
New York, 1995.
Davis, B. and Zhang, B.: Moments of the lifetime of conditioned Brownian motion in cones,
Proc. Amer. Math. Soc. 121(3) (1994), 925929.
DeBlassie, R.: Exit times from cones in R n of Brownian motion, Probab. Theory Related
Fields 74 (1987), 129.
DeBlassie, R.: The first exit time of a two-dimensional symmetric stable process from a
wedge, Ann. Probab. 18 (1990), 10341070.
Erdlyi, A. (ed.): Higher Transcendental Functions, Vols. I, II, McGraw-Hill, New York, 1953.
Ikeda, N. and Watanabe, S.: On some relations between the harmonic measure and the Lvy
measure for a certain class of Markov processes, J. Math. Kyoto Univ. 2(1) (1962), 7995.
Kulczycki, T.: Properties of Green function of symmetric stable process, Probab. Math.
Statist. 17(2) (1997), 339364.
Kulczycki, T.: Intrinsic ultracontractivity for symmetric stable processes, Bull. Polish Acad.
Sci. Math. 46(3) (1998), 325334.
Kulczycki, T.: Exit time and Green function of cone for symmetric stable processes, Probab.
Math. Statist. 19(2) (1999), 337374.
Landkof, N.S.: Foundations of Modern Potential Theory, Springer-Verlag, New York, 1972.
Li, W.: The first exit time of Brownian motion from unbounded domain, Preprint.
M. Lifshitz and Z. Shi, The first exit time of Brownian motion from parabolic domain,
Bernoulli 8(6) (2002), 745767.
Mndez-Hernndez, P.J.: Exit times from cones in Rn of symmetric stable processes, Preprint,
2000.
Rubin, B., Fractional Integrals and Potentials, Pitman Monographs 82, Addison-Wesley, 1996.
Ryznar, M.: Estimates of Green function for relativistic -stable process, Potential Anal. 17
(2002), 123.
Song, R. and Wu, J.-M.: Boundary Harnack principle for symmetric stable processes, J.
Funct. Anal. 168(2) (1999), 403427.
van den Berg, M.: Subexponential behavior of the Dirichlet heat kernel, Preprint.