In this paper, we are concerned with Reflected Geometric Brownian Motion (RGBM) with two barriers. And the stationary distribution of RGBM is derived by Markovian infinitesimal generator method. Consequently the first passage time of RGBM is also discussed.
1. Introduction
We consider a finite-capacity fluid queue, the level of which at time
is denoted by
. And
satisfies the following differential equation:
(1)
This model shows fluid arrives into this queue at rate
and leaves the queue at rate
. This fluid level can be also varied by a local variance function
and a standard Brownian motion
.
and
are nondecreasing processes, interfering only when
hits a or d and make
vary between a and d.
In particular, when
and
,
and
disappear. Then the process
becomes Geometric Brownian Motion. So we call
determined by (1) Reflected Geometric Brownian Motion(RGBM).
Speak precisely, we are concerned with RGBM
with two barriers a and d (d > a > 0), which is defined by
(2)
where
is a standard Brownian Motion,
and
are constants and satisfy
.
Moreover, the processes L and U are uniquely determined by the following property [1,2]:
1) Both L and U are continuous nondecreasing processes with
;
2) L and U increase only when
and
; respectively, i.e.,

According to the theory of stochastic differential equation, (2) is equivalent to
(3)
Such a process is a regenerative Markov process with state space [a,d] compact. Then it has a unique stationary distribution [1,3,4]. In the coming section, our objective is to derive the stationary distribution and give an expression for the Laplace Transform of the first passage time of RGBM
by the method in [5-7].
2. Main Results on RGBM
2.1. On the Stationary Distribution of RGBM
In this section, we firstly give a Lemma on the stationary distribution of the reflected process
with two-sided barriers and omit its proof.
Lemma 2.1 Let Z be the RGBM defined by (2) (or(3)). Then, as a Markov process, the stationary distribution
of the process must satisfy the following equation
(4)
where
and
which denotes the space of all bounded continuous functions having twicely continuous derivatives on [a,d].
Proof. See similar argument in [1].
Suppose
be a probability distribution on [a,d] and satisfies that
(5)
for
.
Define
then by (4) and (5) it is equivalent to the following equation (Note that
)
(6)
where
and 
On one hand,
and
can be computed by the same method in [5].
Proposition 2.1 Choose
and
, then they respectively satisfy the following equations,

Then we have

Proof. A straightforward calculation.
On the other hand, since
satisfies that for all
,

By twice integral changes, the above equation becomes that


i.e.,
(7)
Assume that
, satisfying that
,
, and
satisfies
and
, then it follows from (7) that
(8)
Summarizing the discussion, we get the following theorem.
Theorem 2.1
is the solution of

Then for all
satisfying
,
, (5) holds, i.e.,
.
Furthermore (5) holds for all
. This implies that
is a stationary distribution of the corresponding Markov process
.
Remark 2.1 This theorem is a standard application of renewal theorems, so we sketch its proof.
Thus
is the density of the stationary distribution of RGBM. Finally we will give an expression for the Laplace transform of the first passage time of RGBM.
2.2. On the First Passage Time of RGBM
In this section, we consider equation (2). Let
, define the first passage time by
with the usual convention
. On the other hand, suppose
, for
, define a operator

Finally we are going to give the expression of the Laplace transform of
.
Theorem 2.2. For
and
, then
(10)
(11)
where

and

Proof. Let
for
. Then applying
formula for
, we have




(12)
The last equation holds, for
and
increase only when
and
. Let
be a stopping time and
. It follows from martingale optional theorem, that
(13)
In particular, take
for
, and note that

and

Then
(14)
and
(15)
Replace
by
in (14) and by
in (15), we immediately get (10) and (11) by
,
and
. Thus the Proof of the theorem is completed.
3. Conclusions
This paper studies Reflected Geometric Brownian Motion (RGBM) with two barriers. Both the stationary distribution and Laplace transform of the first passage time of RGBM are derived. The studies for RGBM not only have practical significance, but also give an important result in theory of stochastic process.
4. Acknowledgements
This research is supported by the National Natural Science foundation of China (Grant No.70671074) and the Research Foundation of Tianjin university of Science and technology (Grant No.20080207). The authors would like to thank an anonymous referee for his constructive comments and suggestions on the first version of the manuscript.
[1] J. M. Harrison, “Brownian motion and stochastic flow systems,” John Wiley & Sons Ltd., New York, 1986.
[2] X. Y. Xing, W. Zhang and Y. J. Wang, “The Stationary Distributions of Two Classes of Reflected OrnsteinUhlenbeck Processes,” Journal of Applied Probability, Vol. 46, No. 3, 2009, pp. 709-720.
[3] S. Asmussen and O. Kella, “A Multi-Dimensional Martingale for Markov Additive Process and its Applications,” Advances in Applied Probability, Vol. 32, No. 2, 2000, pp. 376-393.
[4] S. Either and T. Kurtz, “Markov Processes: Characterization and Convergences,” John Wiley and Sons, New York, 1986.
[5] L. J. Bo, L. D. Zhang and Y. J. Wang, “On the First Passage Times of Reflected O-U Processes with Two-Sided Barriers,” Queueing Systems, Vol. 54, No. 4, December 2006, pp. 313-316.
[6] V. Linetsky, “On the Transition Densities for Reflected Diffusions,” Advances in Applied Probability, Vol. 37, No. 2, 2005, pp. 435-460.
[7] L. Rabehasaina and B. Sericola, “A Second-Order Markov-Modulated Fluid Queue with Linear Service Rate,” Journal of Applied Probability, Vol. 41, No. 3, 2004, pp. 758-777.