S. Pang et al. / Journal of Computational and Applied Mathematics 213 (2008) 127 – 141 129
where > 0. Here
ij
is the transition rate from i to j and
ij
> 0ifi = j while
ii
=−
j=i
ij
. We note that
almost every sample path of r(·) is a right continuous step function with a finite number of sample jumps in any finite
subinterval of R
+
:= [0, ∞).
We will use |·|to denote the Euclidean norm of a vector and the trace norm of a matrix. We will denote the indicator
function of a set G by I
G
.Forx ∈ R, int(x) denotes the integer part of x.
We begin our study with the special but important case of scalar linear hybrid SDEs of the form
dx(t) = (r(t))x(t) dt + (r(t))x(t) dB(t), t 0 (2.1)
with initial data x(0) = x
0
∈ R and r(0) = r
0
∈ S. Here, to avoid complicated notations, we let B(t) be a scalar
Brownian motion while and are mappings from S → R. The SDE (2.1) is known as the hybrid Brownian motion
or the volatility-switching geometric Brownian motion. One motivation for studying this class is that sharp stability
results can be derived, allowing us to test the efficiency of a numerical method. One more motivation is that it is a
realistic model in mathematical finance [10] and hence the qualitative behaviour of numerical methods on this model
is of independent interest.
As a standing hypothesis, we assume moreover in this paper that the Markov chain is irreducible. This is equivalent
to the condition that for any i, j ∈ S, we can find i
1
,i
2
,...,i
k
∈ S such that
i,i
1
i
1
,i
2
···
i
k
,j
> 0.
Note that always has an eigenvalue 0. The algebraic interpretation of irreducibility is rank () = N − 1. Under this
condition, the Markov chain has a unique stationary (probability) distribution = (
1
,
2
,...,
N
) ∈ R
1×N
which
can be determined by solving
⎧
⎨
⎩
= 0
subject to
N
j=1
j
= 1 and
j
> 0 for all j ∈ S.
It is known that the linear hybrid SDE (2.1) has the explicit solution
x(t) =x
0
exp
t
0
(r(s)) −
1
2
2
(r(s))
ds +
t
0
(r(s)) dB(s)
. (2.2)
Making use of this explicit form we are able to discuss almost sure and moment exponential stability precisely. The
following theorem gives a necessary and sufficient condition for the SDE (2.1) to be almost surely exponentially stable
(see [20]).
Theorem 2.1. The sample Lyapunov exponent of the SDE (2.1) is
lim
t→∞
1
t
log(|x(t)|) =
N
j=1
j
j
−
1
2
2
j
a.s. (2.3)
(for x
0
= 0 of course). Hence the SDE (2.1) is almost surely exponentially stable if and only if
N
j=1
j
j
−
1
2
2
j
< 0. (2.4)
The following theorem gives a sufficient and necessary condition for the SDE (2.1) to be pth moment exponentially
stable. It should be pointed out that the proof for the pth moment exponential stability of a linear scalar (non-hybrid)
SDE is rather simple (see e.g. [17]) while the proof below for the hybrid SDE becomes much more complicated.
Theorem 2.2. The pth moment Lyapunov exponent of the hybrid SDE (2.1) is
lim
t→∞
1
t
log(E|x(t)|
p
) =
N
j=1
j
p
j
+
1
2
(p − 1)
2
j
(2.5)