Professional Documents
Culture Documents
Stochastic Differential Equations With Multi-Marko
Stochastic Differential Equations With Multi-Marko
Research Article
Stochastic Differential Equations with
Multi-Markovian Switching
Copyright © 2013 M. Liu and K. Wang. This is an open access article distributed under the Creative Commons Attribution License,
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
This paper is concerned with stochastic differential equations (SDEs) with multi-Markovian switching. The existence and
uniqueness of solution are investigated, and the pth moment of the solution is estimated. The classical theory of SDEs with single
Markovian switching is extended.
For the sake of convenience, we take the following two- properties of the system (8) and to find out whether the
dimensional competitive model as an example: presence of two Markovian switchings affects some known
results. The first step and the foundation of those studies are
𝑑𝑥1 = 𝑥1 [𝑟10 (𝛾 (𝑡)) − 𝑎11 (𝛾 (𝑡)) 𝑥1 − 𝑎12 (𝛾 (𝑡)) 𝑥2 ] 𝑑𝑡 to establish the theorems for the existence and uniqueness of
the solution to system (8). So in this paper, we will give some
+ 𝛼1 (𝛾 (𝑡)) 𝑥1 𝑑𝐵1 (𝑡) , theorems for the existence and uniqueness of the solution to
(6)
𝑑𝑥2 = 𝑥2 [𝑟20 (𝛾 (𝑡)) − 𝑎21 (𝛾 (𝑡)) 𝑥1 − 𝑎22 (𝛾 (𝑡)) 𝑥2 ] 𝑑𝑡 system (8) and study some properties of this solution. The
theory developed in this paper is the foundation for further
+ 𝛼2 (𝛾 (𝑡)) 𝑥2 𝑑𝐵2 (𝑡) , study and can be applied in many different and complicated
situations, and hence the importance of the results in this
where 𝑥𝑖 is the size of 𝑖th species at time 𝑡, 𝑟𝑖0 (𝑗) represents paper is clear.
the growth rate of 𝑖th species in regime 𝑗 for 𝑖 = 1, 2, It should be pointed out that the theory developed in this
𝑗 ∈ S, and 𝐵1 and 𝐵2 are independent standard Brownian paper can be generalized to cope with the more general SDEs
motions. However, there are many stochastic factors that with more Markovian chains
affect some coefficients intensely but have little impact on
𝑑𝑥 (𝑡) = 𝑓 (𝑥 (𝑡) , 𝑡, 𝛾1 (𝑡) , . . . , 𝛾𝑛 (𝑡)) 𝑑𝑡
other coefficients in (6). For example, suppose that the (11)
stochastic factor is rain falls and 𝑥1 is able to endure a damp + 𝑔 (𝑥 (𝑡) , 𝑡, 𝛾𝑛+1 (𝑡) , . . . , 𝛾𝑛+𝑚 (𝑡)) 𝑑𝐵 (𝑡) .
weather while 𝑥2 is fond of a dry environment, then the rain
falls will affect 𝑥2 intensely but have little impact on 𝑥1 . Thus, The reason we concentrate on (8) rather than (11) is to avoid
a more appropriate model is governed by the notations becoming too complicated. Once the theory
developed in this paper is established, the reader should be
𝑑𝑥1 = 𝑥1 [𝑟10 (𝛾10 (𝑡)) − 𝑎11 (𝛾11 (𝑡)) 𝑥1 − 𝑎12 (𝛾12 (𝑡)) 𝑥2 ] 𝑑𝑡 able to cope with the more general (11) without any difficulty.
The remaining part of this paper is as follows. In Section 2,
+ 𝛼1 (𝛾1 (𝑡)) 𝑥1 𝑑𝐵1 (𝑡) , the sufficient criteria for existence and uniqueness of solution,
𝑑𝑥2 = 𝑥2 [𝑟20 (𝛾20 (𝑡)) − 𝑎21 (𝛾21 (𝑡)) 𝑥1 − 𝑎22 (𝛾22 (𝑡)) 𝑥2 ] 𝑑𝑡 local solution, and maximal local solution will be established,
respectively. In Section 3, the 𝐿𝑃 -estimates of the solution
+ 𝛼2 (𝛾2 (𝑡)) 𝑥2 𝑑𝐵2 (𝑡) , will be given. In Section 4, we will introduce an example to
(7) illustrate our main result. Finally, we will close the paper with
conclusions in Section 5.
where 𝛾𝑖𝑗 (𝑡) and 𝛾𝑘 (𝑡) are right-continuous homogenous
Markovian chains taking values in finite state spaces S𝑖𝑗 for 2. SDEs with Markovian Chains
𝑖 = 1, 2, 𝑗 = 0, 1, 2, and S𝑘 for 𝑘 = 1, 2, respectively.
Thus the above examples show that the study of the Throughout this paper, let (Ω, F, {F𝑡 }𝑡∈𝑅+ , P) be a complete
following SDEs with multi-Markovian switchings is essential probability space. Let 𝐵(𝑡) = (𝐵1 (𝑡), . . . , 𝐵𝑚 (𝑡))𝑇 be an 𝑚-
and is of great importance from both theoretical and practical dimensional Brownian motion defined on the probability
points: space.
In this section, we will consider (8). Let 𝛾(𝑡) = (𝛾1 (𝑡),
𝑑𝑥 (𝑡) = 𝑓 (𝑥 (𝑡) , 𝑡, 𝛾1 (𝑡)) 𝑑𝑡
𝛾2 (𝑡)). We impose a hypothesis.
(8)
+ 𝑔 (𝑥 (𝑡) , 𝑡, 𝛾2 (𝑡)) 𝑑𝐵 (𝑡) , 𝑡0 ≤ 𝑡 ≤ 𝑇 (H1): 𝛾1 (𝑡) is independent of 𝛾2 (𝑡).
Then 𝛾(𝑡) is a homogenous vector Markovian chain with
with initial conditions 𝑥(𝑡0 ) = 𝑥0 ∈ 𝐿2F0 and 𝛾𝑖 (𝑡0 ), where transition probabilities
𝛾𝑖 (𝑡) is a right-continuous homogenous Markovian chain 𝑃 {𝛾 (𝑠 + 𝑡) = (𝑗1 , 𝑗2 ) | 𝛾 (𝑠) = (𝑖1 , 𝑖2 )}
on the probality space taking values in a finite state space
S𝑖 = {1, 2, . . . , 𝑁𝑖 } and is F𝑡 -adapted but independent of the = 𝑃 {𝛾 (𝑡) = (𝑗1 , 𝑗2 ) | 𝛾 (0) = (𝑖1 , 𝑖2 )} (12)
Brownian motion 𝐵𝑖 (𝑡), 𝑖 = 1, 2, and
=: 𝑃𝑖1 𝑖2 ;𝑗1 𝑗2 (𝑡) ,
𝑓 : R𝑛 × R+ × S1 → R𝑛 ,
(9) where (𝑖1 , 𝑖2 ), (𝑗1 , 𝑗2 ) ∈ S = {(1, 1), (1, 2), . . . , (1, 𝑁2 ), (2, 1),
𝑔 : R𝑛 × R+ × S2 → R𝑛+𝑚 . . . . , (𝑁1 , 𝑁2 )}.
Now, we will prepare some lemmas which are important
Equation (8) can be regarded as the result of the 𝑁1 × 𝑁2 for further study.
equations
Lemma 1. 𝑃𝑖1 𝑖2 ;𝑗1 𝑗2 (𝑡) has the following properties:
𝑑𝑥 (𝑡) = 𝑓 (𝑥 (𝑡) , 𝑡, 𝑖) 𝑑𝑡
(10) (i) 𝑃𝑖1 𝑖2 ;𝑗1 𝑗2 (𝑡) ≥ 0 for (𝑖1 , 𝑖2 ), (𝑗1 , 𝑗2 ) ∈ S;
+ 𝑔 (𝑥 (𝑡) , 𝑡, 𝑗) 𝑑𝐵 (𝑡) , 𝑖 ∈ S 1 , 𝑗 ∈ S2 (ii) ∑(𝑗1 ,𝑗2 )∈S 𝑃𝑖1 𝑖2 ;𝑗1 𝑗2 (𝑡) = 1 for (𝑖1 , 𝑖2 ) ∈ S;
switching among each other according to the movement of (iii) 𝑃𝑖1 𝑖2 ;𝑗1 𝑗2 (0) = 𝛿𝑖1 𝑗1 ⋅ 𝛿𝑖2 𝑗2 , where 𝛿𝑖𝑘 𝑗𝑘 = 1 if 𝑖𝑘 = 𝑗𝑘 ,
the Markovian chains. It is important for us to discover the otherwise 𝛿𝑖𝑘 𝑗𝑘 = 0, 𝑘 = 1, 2;
Journal of Applied Mathematics 3
(iv) (Chapman-Kolmogorov equation) For 𝑠, 𝑡 ≥ 0 and for sufficient large 𝑛. Then making use of Chapman-Kolm-
(𝑖1 , 𝑖2 ), (𝑗1 , 𝑗2 ) ∈ S, ogorov equation
𝑃𝑖1 𝑖2 ;𝑗1 𝑗2 (𝑠 + 𝑡) 𝑃𝑖1 𝑖2 ;𝑖1 𝑖2 (𝑠 + 𝑡) = ∑ 𝑃𝑖1 𝑖2 ;𝑘1 1𝑘2 (𝑡) 𝑃𝑘1 𝑘2 ;𝑖1 𝑖2 (𝑠)
(13) (𝑘1 ,𝑘2 )∈S (16)
= ∑ 𝑃𝑖1 𝑖2 ;𝑘1 𝑘2 (𝑡) 𝑃𝑘1 𝑘2 ;𝑗1 𝑗2 (𝑠) .
(𝑘1 ,𝑘2 )∈S ≥ 𝑃𝑖1 𝑖2 ;𝑖1 𝑖2 (𝑡) 𝑃𝑖1 𝑖2 ;𝑖1 𝑖2 (𝑠)
gives
Proof. The proofs of (i), (ii), and (iii) are obvious. Now, let us
prove (iv): 𝑡 𝑛
𝑃𝑖1 𝑖2 ;𝑖1 𝑖2 (𝑡) ≥ (𝑃𝑖1 𝑖2 ;𝑖1 𝑖2 ( )) > 0, (17)
𝑛
𝑃𝑖1 𝑖2 ;𝑗1 𝑗2 (𝑠 + 𝑡)
which is the desired assertion.
= 𝑃 {𝛾 (𝑠 + 𝑡) = (𝑗1 , 𝑗2 ) | 𝛾 (0) = (𝑖1 , 𝑖2 )}
Lemma 3. Under Assumption (H2), for all (𝑖1 , 𝑖2 ) ∈ S,
= ∑ 𝑃 {𝛾 (𝑠 + 𝑡) = (𝑗1 , 𝑗2 ) ,
(𝑘1 ,𝑘2 )∈S 1 − 𝑃𝑖1 𝑖2 ;𝑖1 𝑖2 (𝑡)
−𝑞𝑖1 𝑖2 ;𝑖1 𝑖2 := lim (18)
𝑡→0 𝑡
𝛾 (𝑡) = (𝑘1 , 𝑘2 ) | 𝛾 (0) = (𝑖1 , 𝑖2 )}
exists (but may be ∞).
= ∑ (𝑃 {𝛾 (𝑠 + 𝑡) = (𝑗1 , 𝑗2 ) ,
(𝑘1 ,𝑘2 )∈S Proof. Define 𝜙(𝑡) = − ln 𝑃𝑖1 𝑖2 ;𝑖1 𝑖2 (𝑡) ≥ 0. Then making use of
(16) gives
𝛾 (𝑡) = (𝑘1 , 𝑘2 ) , 𝛾 (0) = (𝑖1 , 𝑖2 )}
−1 𝜙 (𝑠 + 𝑡) ≤ 𝜙 (𝑠) + 𝜙 (𝑡) . (19)
×(𝑃 {𝛾 (𝑡) = (𝑘1 , 𝑘2 ) , 𝛾 (0) = (𝑖1 , 𝑖2 )}) )
Set −𝑞𝑖1 𝑖2 ;𝑖1 𝑖2 = sup𝑡>0 (𝜙(𝑡)/𝑡). It is easy to see that
𝑃 {𝛾 (𝑡) = (𝑘1 , 𝑘2 ) , 𝛾 (0) = (𝑖1 , 𝑖2 )}
×
𝑃 {𝛾 (0) = (𝑖1 , 𝑖2 )} 𝜙 (𝑡)
0 ≤ −𝑞𝑖1 𝑖2 ;𝑖1 𝑖2 ≤ ∞, lim sup ≤ −𝑞𝑖1 𝑖2 ;𝑖1 𝑖2 . (20)
𝑡→0 𝑡
= ∑ 𝑃 {𝛾 (𝑠 + 𝑡) = (𝑗1 , 𝑗2 ) | 𝛾 (𝑡) = (𝑘1 , 𝑘2 ) ,
(𝑘1 ,𝑘2 )∈S
Now we will assert
𝛾 (0) = (𝑖1 , 𝑖2 )} 𝜙 (𝑡)
lim inf ≥ −𝑞𝑖1 𝑖2 ;𝑖1 𝑖2 . (21)
× 𝑃 {𝛾 (𝑡) = (𝑘1 , 𝑘2 ) | 𝛾 (0) = (𝑖1 , 𝑖2 )}
𝑡→0 𝑡
In fact, for 0 < ℎ < 𝑡, ∃𝑛, 0 ≤ 𝜀 ≤ ℎ such that 𝑡 = 𝑛ℎ + 𝜀.
= ∑ 𝑃 {𝛾 (𝑠 + 𝑡) = (𝑗1 , 𝑗2 ) | 𝛾 (𝑡) = (𝑘1 , 𝑘2 )} Applying (19) yields
(𝑘1 ,𝑘2 )∈S
𝜙 (𝑡) 𝑛ℎ 𝜙 (ℎ) 𝜙 (𝜀)
× 𝑃 {𝛾 (𝑡) = (𝑘1 , 𝑘2 ) | 𝛾 (0) = (𝑖1 , 𝑖2 )} ≥ + . (22)
𝑡 𝜀 ℎ 𝑡
= ∑ 𝑃𝑘1 𝑘2 ;𝑗1 𝑗2 (𝑠) 𝑃𝑖1 𝑖2 ;𝑘1 𝑘2 (𝑡) . Note that 𝜀 → 0, 𝑛ℎ/𝜀 → 1, 𝜙(𝜀) → 0 whenever ℎ → 0+ ,
(𝑘1 ,𝑘2 )∈S then for all 𝑡 > 0 we have
(14)
𝜙 (𝑡) 𝜙 (ℎ)
≤ lim inf . (23)
This completes the proof. 𝑡 ℎ→0 ℎ
Now, we impose another hypothesis, which is called This implies −𝑞𝑖1 𝑖2 ;𝑖1 𝑖2 ≤ lim inf 𝑡 → 0 (𝜙(𝑡)/𝑡). Thus
standard condition.
(H2): lim𝑡 → 0 𝑃𝑖1 𝑖2 ;𝑗1 𝑗2 (𝑡) = 𝛿𝑖1 𝑖2 ;𝑗1 𝑗2 . 𝜙 (𝑡)
−𝑞𝑖1 𝑖2 ;𝑖1 𝑖2 = lim . (24)
𝑡→0 𝑡
Lemma 2. Under Assumption (H2), for all 𝑡 ≥ 0, (𝑖1 , 𝑖2 ) ∈ S,
one has 𝑃𝑖1 𝑖2 ;𝑖1 𝑖2 (𝑡) > 0. Using the definition of 𝜙(𝑡) gives
𝜀
Proof. By (H2), we note that for all 0 < 𝜀 < 1/3, ∃0 < 𝛿 < 1, ≥ 𝑛 (1 − 𝜀 − )𝑃 (ℎ) (1 − 𝜀)
1 − 𝜀 𝑖1 𝑖2 ;𝑗1 𝑗2
such that
≥ 𝑛 (1 − 3𝜀) 𝑃𝑖1 𝑖2 ;𝑗1 𝑗2 (ℎ) .
𝑃𝑖1 𝑖2 ;𝑖1 𝑖2 (𝑡) > 1 − 𝜀, 𝑃𝑗1 𝑗2 ;𝑗1 𝑗2 (𝑡) > 1 − 𝜀,
(33)
(27)
𝑃𝑗1 𝑗2 ;𝑖1 𝑖2 (𝑡) < 𝜀
Dividing both sides of the above inequality by ℎ and noting
provided 0 < 𝑡 ≤ 𝛿. 𝑛ℎ → 𝑡 whenever ℎ → 0 yield
For ∀0 ≤ ℎ < 𝑡, set 𝑛 = ⟨𝑡/ℎ⟩, where ⟨𝑎⟩ = max𝑛≤𝑎 {𝑛 ∈
𝑃𝑖1 𝑖2 ;𝑗1 𝑗2 (ℎ) 1 𝑃𝑖1 𝑖2 ;𝑗1 𝑗2 (𝑡)
Z}. Let lim sup ≤ < ∞. (34)
ℎ→0 ℎ 1 − 3𝜀 𝑡
(𝑗 𝑗 )
𝑃𝑖 𝑖1 ;𝑘2 𝑘 (ℎ) = 𝑃𝑖1 𝑖2 ;𝑘1 𝑘2 (ℎ) ,
1 2 1 2
Then letting 𝑡 → 0 gives
(𝑗 𝑗 )
𝑃𝑖 𝑖1 ;𝑘2 𝑘 (𝑚ℎ)
1 2 1 2 (28) 𝑃𝑖1 𝑖2 ;𝑗1 𝑗2 (ℎ) 1 𝑃𝑖 𝑖 ;𝑗 𝑗 (𝑡)
lim sup ≤ lim inf 1 2 1 2 , (35)
= ∑
(𝑗 𝑗 )
𝑃𝑖1 𝑖12 ;𝑟21 𝑟2 ((𝑚 − 1) ℎ) 𝑃𝑟1 𝑟2 ;𝑘1 𝑘2 (ℎ) , ℎ→0 ℎ 1 − 3𝜀 𝑡 → 0 𝑡
(𝑟1 ,𝑟2 ) ≠ (𝑗1 ,𝑗2 )
and the required assertion follows immediately by letting
(𝑗 𝑗 ) 𝜀 → 0. This completes the proof.
where 𝑃𝑖 𝑖1 ;𝑘2 𝑘 (𝑚ℎ) means that the probability of the 𝛾(𝑡) will
1 2 1 2
not reach to (𝑗1 , 𝑗2 ) at times ℎ, 2ℎ, . . . , (𝑚 − 1)ℎ but will reach Set 𝛾(𝑡) = (𝛾1 (𝑡), 𝛾2 (𝑡)), then it is easy to see that
to (𝑘1 , 𝑘2 ) at time 𝑚ℎ. Note that if ℎ ≤ 𝑡 ≤ 𝛿, then almost every sample path of 𝛾(𝑡) is a right continuous step
function. Now letting P(𝑡) = (𝑃𝑖1 𝑖2 ;𝑗1 𝑗2 (𝑡))𝑁1 𝑁2 ×𝑁1 𝑁2 , Q =
𝜀 > 1 − 𝑃𝑖1 𝑖2 ;𝑖1 𝑖2 (𝑡) (𝑞𝑖1 𝑖2 ;𝑗1 𝑗2 )𝑁1 𝑁2 ×𝑁1 𝑁2 = P (0). Then by Chapman-Kolmogorov
equation
= ∑ 𝑃𝑖1 𝑖2 ;𝑘1 𝑘2 (𝑡) ≥ 𝑃𝑖1 𝑖2 ;𝑗1 𝑗2 (𝑡)
(𝑘1 ,𝑘2 ) ≠ (𝑖1 ,𝑖2 ) P (𝑡 + ℎ) = P (𝑡) P (ℎ) = P (ℎ) P (𝑡) , (36)
𝑛 (29)
(𝑗 𝑗 ) we have
≥ ∑ 𝑃𝑖1 𝑖12 ;𝑗21 𝑗2 (𝑚ℎ) 𝑃𝑗1 𝑗2 ;𝑗1 𝑗2 (𝑡 − 𝑚ℎ)
𝑚=1
P (𝑡 + ℎ) − P (𝑡) P (ℎ) − I P (ℎ) − I
𝑛 = P (𝑡) [ ]=[ ] P (𝑡) .
(𝑗 𝑗 ) ℎ ℎ ℎ
≥ (1 − 𝜀) ∑ 𝑃𝑖1 𝑖12 ;𝑗21 𝑗2 (𝑚ℎ) , (37)
𝑚=1
Definition 6. An R𝑛 -valued stochastic process {𝑥(𝑡)}𝑡0 ≤𝑡≤𝑇 is with initial conditions 𝑥(𝑡0 ) = 𝑥0 , 𝛾(𝑡0 ) = (𝛾1 (𝑡0 ), 𝛾1 (𝑡0 )).
called a solution of (8) if it has the following properties: Then by the theory of SDEs, we obtain that (46) has a unique
(i) {𝑥(𝑡)}𝑡0 ≤𝑡≤𝑇 is continuous and F𝑡 -adapted; solution which obeys 𝑥(𝜏1 ) ∈ 𝐿2F𝜏 (Ω; R𝑛 ). We next consider
1
(8) on 𝑡 ∈ [[𝜏1 , 𝜏2 [[ which becomes
(ii) {𝑓(𝑥(𝑡), 𝑡, 𝛾1 (𝑡))}𝑡0 ≤𝑡≤𝑇 ∈ L1 ([𝑡0 , 𝑇]; R𝑛 ) while {𝑔(𝑥(𝑡),
𝑡, 𝛾2 (𝑡))}𝑡0 ≤𝑡≤𝑇 ∈ L2 ([𝑡0 , 𝑇]; R𝑛×𝑚 ); 𝑑𝑥 (𝑡) = 𝑓 (𝑥 (𝑡) , 𝑡, 𝛾1 (𝜏1 )) 𝑑𝑡
(47)
(iii) for any 𝑡 ∈ [𝑡0 , 𝑇], equation + 𝑔 (𝑥 (𝑡) , 𝑡, 𝛾2 (𝜏1 )) 𝑑𝐵 (𝑡) .
𝑡
𝑥 (𝑡) = 𝑥 (𝑡0 ) + ∫ 𝑓 (𝑥 (𝑠) , 𝑠, 𝛾1 (𝑠)) 𝑑𝑠 Again by the theory of SDEs, (47) has a unique solution which
𝑡0 obeys 𝑥(𝜏1 ) ∈ 𝐿2F𝜏 (Ω; R𝑛 ). Repeating this procedure, we
(41) 2
𝑡 conclude that (8) has a unique solution 𝑥(𝑡) on [𝑡0 , 𝑇].
+ ∫ 𝑔 (𝑥 (𝑠) , 𝑠, 𝛾2 (𝑠)) 𝑑𝐵 (𝑠) Now, let us prove (44). For every 𝑘 ≥ 1, define the
𝑡0
stopping time
holds with probability 1.
𝜏𝑘 = 𝑇 ∧ inf {𝑡 ∈ [𝑡0 , 𝑇] : |𝑥 (𝑡)| ≥ 𝑘} . (48)
A solution {𝑥(𝑡)}𝑡0 ≤𝑡≤𝑇 is said to be unique if any other
It is obvious that 𝜏𝑘 ↑ 𝑇 a.s. Set 𝑥𝑘 (𝑡) = 𝑥(𝑡∧𝜏𝑘 ) for 𝑡 ∈ [𝑡0 , 𝑇].
𝑥(𝑡)}𝑡0 ≤𝑡≤𝑇 is indistinguishable from {𝑥(𝑡)}𝑡0 ≤𝑡≤𝑇 .
solution {̃
Then 𝑥𝑘 (𝑡) obeys the equation
Now we can give our main results in this section.
𝑡
Theorem 7. Assume that there exist two positive constants 𝐾 𝑥𝑘 (𝑡) = 𝑥0 + ∫ 𝑓 (𝑥𝑘 (𝑠) , 𝑠, 𝛾1 (𝑠)) 𝐼[[𝑡0 ,𝜏𝑘 ]] (𝑠) 𝑑𝑠
𝑡0
and 𝐾 such that. (49)
(Lipschitz condition) for all 𝑥, 𝑦 ∈ R𝑛 , 𝑡 ∈ [𝑡0 , 𝑇] and 𝑡
(𝑖1 , 𝑖2 ) ∈ S + ∫ 𝑔 (𝑥𝑘 (𝑠) , 𝑠, 𝛾2 (𝑠)) 𝐼[[𝑡0 ,𝜏𝑘 ]] (𝑠) 𝑑𝐵 (𝑠) .
𝑡0
2
𝑓 (𝑥, 𝑡, 𝑖1 ) − 𝑓 (𝑦, 𝑡, 𝑖1 ) Making use of the elementary inequality |𝑎 + 𝑏 + 𝑐|2 ≤ 3(|𝑎|2 +
(42)
2 2
∨ 𝑔 (𝑥, 𝑡, 𝑖2 ) − 𝑔 (𝑦, 𝑡, 𝑖2 ) ≤ 𝐾𝑥 − 𝑦 . |𝑏|2 + |𝑐|2 ), the Hölder inequality, and (43), we can see that
𝑡
(Linear growth condition) for all (𝑥, 𝑡, (𝑖1 , 𝑖2 )) ∈ R𝑛 × 2 2 2
𝑥𝑘 (𝑡) = 3𝑥0 + 3𝐾 (𝑇 − 𝑡0 ) ∫ (1 + 𝑥𝑘 (𝑠) ) 𝑑𝑠
[𝑡0 , 𝑇] × S 𝑡 0
2 2 𝑡 2
𝑓 (𝑥, 𝑡, 𝑖1 ) ∨ 𝑔 (𝑥, 𝑡, 𝑖2 ) ≤ 𝐾(1 + |𝑥|) .
2
+ 3∫ 𝑔 (𝑥𝑘 (𝑠) , 𝑠, 𝛾2 (𝑠)) 𝐼[[𝑡0 ,𝜏𝑘 ]] (𝑠) 𝑑𝐵 (𝑠) .
(43)
𝑡0
Then there exists a unique solution 𝑥(𝑡) to (8) and, moreover, (50)
the solution obeys
Thus, applying the Doob martingale inequality and (43), we
2
𝐸 ( sup |𝑥 (𝑡)| ) can further show that
𝑡0 ≤𝑡≤𝑇 (44)
2 2
≤ (1 + 3𝐸𝑥0 ) exp {3𝐾 (𝑇 − 𝑡0 ) (𝑇 − 𝑡0 + 4)} . 𝐸 ( sup 𝑥𝑘 (𝑠) )
𝑡0 ≤𝑠≤𝑡
Proof. Recall that almost every sample path of 𝛾(𝑡) = 𝑡
2 2
(𝛾1 (𝑡), 𝛾2 (𝑡)) is a right continuous step function with a finite = 3𝐸𝑥0 + 3𝐾 (𝑇 − 𝑡0 ) ∫ (1 + 𝐸𝑥𝑘 (𝑠) ) 𝑑𝑠
number of jumps on [𝑡0 , 𝑇]. Thus there exists a sequence of 𝑡0
stopping times {𝜏𝑘 }𝑘≥0 such that 𝑡
2
+ 12𝐸 ∫ 𝑔 (𝑥𝑘 (𝑠) , 𝑠, 𝛾2 (𝑠)) 𝐼[[𝑡0 ,𝜏𝑘 ]] (𝑠) 𝑑𝑠
(i) for almost every 𝜔 ∈ Ω there is a finite 𝑘𝜔 for 𝑡0 = 𝑡 0
Using the Gronwall inequality leads to Theorem 10. Under condition (57), there exists a unique
maximal local solution of (8).
2
1 + 𝐸 ( sup 𝑥𝑘 (𝑠) )
𝑡0 ≤𝑠≤𝑡 Proof. Define functions
(53)
2
≤ (1 + 3𝐸𝑥0 ) exp {3𝐾 (𝑇 − 𝑡0 ) (𝑇 − 𝑡0 + 4)} . {𝑓 (𝑥, 𝑡, 𝑖1 ) , if |𝑥| ≤ 𝑘,
𝑓𝑘 (𝑥, 𝑡, 𝑖1 ) = { 𝑘𝑥
Consequently, 𝑓 ( , 𝑡, 𝑖1 ) , if |𝑥| > 𝑘,
{ |𝑥|
(58)
1 + 𝐸 ( sup |𝑥 (𝑡)|2 ) {𝑔 (𝑥, 𝑡, 𝑖2 ) , if |𝑥| ≤ 𝑘,
𝑡0 ≤𝑡≤𝜏𝑘
(54) 𝑔𝑘 (𝑥, 𝑡, 𝑖2 ) = { 𝑘𝑥
𝑔 ( , 𝑡, 𝑖2 ) , if |𝑥| > 𝑘.
2 { |𝑥|
≤ (1 + 3𝐸𝑥0 ) exp {3𝐾 (𝑇 − 𝑡0 ) (𝑇 − 𝑡0 + 4)} .
Then 𝑓𝑘 and 𝑔𝑘 satisfy the Lipschitz condition and the linear
Then the required inequality (44) follows immediately by growth condition. Thus by Theorem 7, there is a unique
letting 𝑘 → ∞. solution 𝑥𝑘 (𝑡) of the equation
Condition (42) indicates that the coefficients 𝑓(𝑥, 𝑡, 𝑖1 )
and 𝑔(𝑥, 𝑡, 𝑖2 ) do not change faster than a linear function of 𝑑𝑥𝑘 (𝑡) = 𝑓𝑘 (𝑥𝑘 (𝑡) , 𝑡, 𝛾1 (𝑡)) 𝑑𝑡
𝑥 as change in 𝑥. This means in particular the continuity of (59)
𝑓(𝑥, 𝑡, 𝑖1 ) and 𝑔(𝑥, 𝑡, 𝑖2 ) in 𝑥 for all 𝑡 ∈ [𝑡0 , 𝑇]. Then functions + 𝑔𝑘 (𝑥𝑘 (𝑡) , 𝑡, 𝛾2 (𝑡)) 𝑑𝐵 (𝑡)
that are discontinuous with respect to 𝑥 are excluded as the with the initial conditions 𝑥𝑘 (𝑡0 ) = 𝑥0 and 𝛾(𝑡0 ) = (𝛾1 (𝑡0 ),
coefficients. Besides, there are many functions that do not 𝛾2 (𝑡0 )). Define the stopping times
satisfy the Lipschitz condition. These imply that the Lipschitz
condition is too restrictive. To improve this Lipschitz condi- 𝜎𝑘 = 𝑇 ∧ inf {𝑡 ∈ [𝑡0 , 𝑇] : 𝑥𝑘 (𝑡) ≥ 𝑘} . (60)
tion let us introduce the concept of local solution.
Clearly, if 𝑡0 ≤ 𝑡 ≤ 𝜎𝑘 ,
Definition 8. Let 𝜎∞ be a stopping time such that 𝑡0 ≤ 𝜎∞ ≤ 𝑇
a.s. An R𝑛 -valued F𝑡 -adapted continuous stochastic process 𝑥𝑘 (𝑡) = 𝑥𝑘+1 (𝑡) , (61)
{𝑥(𝑡)}𝑡0 ≤𝑡<𝜎∞ is called a local solution of (8) if 𝑥(𝑡0 ) = 𝑥0 and,
which indicates that 𝜎𝑘 is increasing so 𝜎𝑘 has its limit 𝜎∞ =
moreover, there is a nondecreasing sequence {𝜎𝑘 }𝑘≥1 such that
lim𝑘 → +∞ 𝜎𝑘 . Define {𝑥(𝑡) : 𝑡0 ≤ 𝑡 < 𝜎∞ } by
𝑡0 ≤ {𝜎𝑘 } ↑ 𝜎∞ a.s. and
𝑡∧𝜎𝑘 𝑥 (𝑡) = 𝑥𝑘 (𝑡) , 𝑡 ∈ [[𝜎𝑘−1 , 𝜎𝑘 [[ , 𝑘 ≥ 1, (62)
𝑥 (𝑡) = 𝑥 (𝑡0 ) + ∫ 𝑓 (𝑥 (𝑠) , 𝑠, 𝛾1 (𝑠)) 𝑑𝑠
𝑡0 where 𝜎0 = 𝑡0 . Applying (61), one can show that 𝑥(𝑡 ∧ 𝜎𝑘 ) =
(55) 𝑥𝑘 (𝑡 ∧ 𝜎𝑘 ). It then follows from (59) that
𝑡∧𝜎𝑘
+∫ 𝑔 (𝑥 (𝑠) , 𝑠, 𝛾2 (𝑠)) 𝑑𝐵 (𝑠) 𝑡∧𝜎𝑘
𝑡0
𝑥 (𝑡 ∧ 𝜎𝑘 ) = 𝑥0 + ∫ 𝑓𝑘 (𝑥 (𝑠) , 𝑠, 𝛾1 (𝑠)) 𝑑𝑠
𝑡0
holds for any 𝑡 ∈ [𝑡0 , 𝑇) and 𝑘 ≥ 1 with probability one. If,
furthermore, 𝑡∧𝜎𝑘
+∫ 𝑔𝑘 (𝑥 (𝑠) , 𝑠, 𝛾2 (𝑠)) 𝑑𝐵 (𝑠)
lim sup |𝑥 (𝑡)| = ∞ whenever 𝜎∞ < 𝑇, (56) 𝑡0
𝑡 → +𝜎∞ (63)
𝑡∧𝜎𝑘
then it is called a maximal local solution and 𝜎∞ is called the = 𝑥0 + ∫ 𝑓 (𝑥 (𝑠) , 𝑠, 𝛾1 (𝑠)) 𝑑𝑠
𝑡0
explosion time. A maximal local solution {𝑥(𝑡) : 𝑡0 ≤ 𝑡 < 𝜎∞ }
is said to be unique if any other maximal local solution {𝑥(𝑡) : 𝑡∧𝜎𝑘
𝑡0 ≤ 𝑡 < 𝜎∞ } is indistinguishable from it, namely, 𝜎∞ = 𝜎∞ +∫ 𝑔 (𝑥 (𝑠) , 𝑠, 𝛾2 (𝑠)) 𝑑𝐵 (𝑠) ,
𝑡0
and 𝑥(𝑡) = 𝑥(𝑡) for 𝑡0 ≤ 𝑡 < 𝜎∞ with probability one.
for any 𝑡 ∈ [𝑡0 , 𝑇) and 𝑘 ≥ 1. It is easy to see that if 𝜎∞ ≤ 𝑇,
Definition 9 (local Lipschitz condition). For every integer then
𝑘 ≥ 1, there exists a positive constant ℎ𝑘 such that, for all
𝑡 ∈ [𝑡0 , 𝑇], 𝑖 = (𝑖1 , 𝑖2 ) ∈ S and those 𝑥, 𝑦 ∈ R𝑛 with lim sup |𝑥 (𝑡)|
𝑡 → 𝜎∞
|𝑥| ∨ |𝑦| ≤ 𝑘 (64)
2 ≥ lim sup 𝑥 (𝜎𝑘 ) = lim sup 𝑥𝑘 (𝜎𝑘 ) = ∞.
𝑓 (𝑥, 𝑡, 𝑖1 ) − 𝑓 (𝑦, 𝑡, 𝑖1 ) 𝑘 → +∞ 𝑘 → +∞
(57)
2 2 Therefore {𝑥(𝑡) : 𝑡0 ≤ 𝑡 < 𝜎∞ } is a maximal local solution.
∨ 𝑔 (𝑥, 𝑡, 𝑖2 ) − 𝑔 (𝑦, 𝑡, 𝑖2 ) ≤ ℎ𝑘 𝑥 − 𝑦 .
Now, we will prove the uniqueness. Let {𝑥(𝑡) : 𝑡0 ≤ 𝑡 <
The following theorem shows the existence of unique 𝜎∞ } be another maximal local solution. Define
maximal local solution under the local Lipschitz condition
without the linear growth condition. 𝜎𝑘 = 𝜎∞ ∧ inf {𝑡 ∈ [[𝑡0 , 𝜎∞ [[ : 𝑥 (𝑡) ≥ 𝑘} . (65)
Journal of Applied Mathematics 7
Then 𝜎𝑘 → 𝜎∞ a.s. and Theorem 11. Assume that the local Lipschitz condition (57)
holds. Assume also that there is a function 𝑉(𝑥, 𝑡, (𝑖1 , 𝑖2 )) ∈
𝑃 {𝑥 (𝑡) = 𝑥 (𝑡) , 𝑡 ∈ [[𝑡0 , 𝜎𝑘 ∧ 𝜎𝑘 ]]} = 1 ∀𝑘 ≥ 1. (66) 𝐶2,1 (R𝑛 × R+ × S; R+ ) and a constant 𝜃 > 0 such that
Letting 𝑘 → ∞ gives
lim ( inf 𝑉 (𝑥, 𝑡, (𝑖1 , 𝑖2 ))) = ∞, (72)
|𝑥| → ∞ (𝑡,(𝑖1 ,𝑖2 ))∈R+ ×S
𝑃 {𝑥 (𝑡) = 𝑥 (𝑡) , 𝑡 ∈ [[𝑡0 , 𝜎∞ ∧ 𝜎∞ [[} = 1. (67)
𝐿𝑉 (𝑥, 𝑡, (𝑖1 , 𝑖2 )) ≤ 𝜃 (1 + 𝑉 (𝑥, 𝑡, (𝑖1 , 𝑖2 ))) ,
In order to complete the proof, we need only to show that (73)
𝜎∞ = 𝜎∞ a.s. In fact, for almost every 𝜔 ∈ {𝜎∞ < 𝜎∞ }, we ∀ (𝑥, 𝑡, (𝑖1 , 𝑖2 )) ∈ R𝑛 × R+ × S.
have
Then there exists a unique global solution 𝑥(𝑡) to (69).
𝑥 (𝜎∞ , 𝜔)
(68) Proof. We need only to prove the theorem for any initial
= lim 𝑥 (𝜎𝑘 , 𝜔) = lim 𝑥 (𝜎𝑘 , 𝜔) = ∞, condition 𝑥0 ∈ R𝑛 and (𝛾1 (𝑡0 ), 𝛾2 (𝑡0 )) ∈ S. From Theorem 10,
𝑘 → +∞ 𝑘 → +∞
we know that the local Lipschitz condition guarantees the
which contradicts the fact that 𝑥(𝑡, 𝜔) is continuous on 𝑡 ∈ existence of the unique maximal solution 𝑥(𝑡) on [𝑡0 , 𝜎∞ ),
[𝑡0 , 𝜎∞ (𝜔)). This implies 𝜎∞ ≥ 𝜎∞ a.s. In the same way, one where 𝜎∞ is the explosion time. We need only to show 𝜎∞ =
can show 𝜎∞ ≤ 𝜎∞ a.s. Thus we must have 𝜎∞ = 𝜎∞ a.s. This ∞ a.s. If this is not true, then we can find a pair of positive
completes the proof. constants 𝜀 and 𝑇 such that
In many situations, we often consider an SDE on [𝑡0 , ∞) 𝑃 {𝜎∞ ≤ 𝑇} > 2𝜀. (74)
𝑑𝑥 (𝑡) = 𝑓 (𝑥 (𝑡) , 𝑡, 𝛾1 (𝑡)) 𝑑𝑡 For each integer 𝑘 ≥ 1, define the stopping time
(69)
+ 𝑔 (𝑥 (𝑡) , 𝑡, 𝛾2 (𝑡)) 𝑑𝐵 (𝑡) , 𝑡0 ≤ 𝑡 < ∞ 𝜎𝑘 = inf {𝑡 ≥ 𝑡0 : |𝑥 (𝑡)| ≥ 𝑘} . (75)
with initial data 𝑥(𝑡0 ) = 𝑥0 and 𝛾(𝑡0 ). If the assumption of Since 𝜎𝑘 → 𝜎∞ almost surely, we can find a sufficiently large
the existence-and-uniqueness theorem holds on every finite integer 𝑘0 for
subinterval [𝑡0 , 𝑇] of [𝑡0 , ∞), then (69) has a unique solution
𝑃 {𝜎𝑘 ≤ 𝑇} > 𝜀, ∀𝑘 ≥ 𝑘0 . (76)
𝑥(𝑡) on the entire interval [𝑡0 , ∞). Such a solution is called
a global solution. To establish a more general result about Fix any 𝑘 ≥ 𝑘0 , then for any 𝑡0 ≤ 𝑡 ≤ 𝑇, by virtue of the
global solution, we need more notations. To this end, we generalized Itô formula (see, e.g., [1])
introduce an operator 𝐿𝑉 from R𝑛 × R+ × S to R which is
given by 𝐸𝑉 (𝑥 (𝑡 ∧ 𝜎𝑘 ) , 𝑡 ∧ 𝜎𝑘 , 𝛾 (𝑡 ∧ 𝜎𝑘 ))
𝐿𝑉 (𝑥, 𝑡, (𝑖1 , 𝑖2 )) 𝑡∧𝜎𝑘
= 𝑉 (𝑥0 , 𝑡0 , 𝛾0 ) + 𝐸 ∫ 𝐿𝑉 (𝑥 (𝑠) , 𝑠, 𝛾 (𝑠)) 𝑑𝑠
𝑡0
= 𝑉𝑡 (𝑥, 𝑡, (𝑖1 , 𝑖2 )) + 𝑉𝑥 (𝑥, 𝑡, (𝑖1 , 𝑖2 )) 𝑓 (𝑥, 𝑡, 𝑖1 ) (77)
≤ 𝑉 (𝑥0 , 𝑡0 , 𝛾0 ) + 𝜃 (𝑇 − 𝑡0 )
+ 0.5 trace [𝑔𝑇 (𝑥, 𝑡, 𝑖2 ) 𝑉𝑥𝑥 (𝑥, 𝑡, (𝑖1 , 𝑖2 )) 𝑔 (𝑥, 𝑡, 𝑖2 )]
𝑡
+ ∑ 𝑞𝑖1 𝑖2 ;𝑗1 𝑗2 𝑉 (𝑥, 𝑡, (𝑗1 , 𝑗2 )) , + 𝜃 ∫ 𝐸𝑉 (𝑥 (𝑠 ∧ 𝜎𝑘 ) , 𝑠 ∧ 𝜎𝑘 , 𝛾 (𝑠 ∧ 𝜎𝑘 )) 𝑑𝑠.
𝑡0
( 1 2 )∈S
𝑗 ,𝑗
(70) Making use of the Gronwall inequality gives
Then (72) means 𝑔𝑘 → ∞. It follows from (76) and (79) that Then one has
[𝑉 (𝑥0 , 𝑡0 , 𝛾0 ) + 𝜃 (𝑇 − 𝑡0 )] exp {𝜃 (𝑇 − 𝑡0 )}
(81) 𝐸|𝑥 (𝑡)|𝑝
≥ 𝑔𝑘 𝑃 {𝜎𝑘 ≤ 𝑇} ≥ 𝜀𝑔𝑘 . 𝑝
≤ 2(𝑝−2)/2 (1 + 𝐸𝑥0 ) exp {𝑝𝜃 (𝑡 − 𝑡0 )} ∀𝑡 ∈ [𝑡0 , 𝑇] .
Letting 𝑘 → ∞ yields a contradiction, that is to say, 𝜎∞ = ∞. (89)
The proof is complete.
Proof. Applying the elementary inequality |𝑎 + 𝑏|𝑝 ≤ Proof. Making use of the generalized Itô’s formula and
2𝑝−1 (|𝑎|𝑝 + |𝑏|𝑝 ), the Hölder inequality, and the linear growth condition (96), we have
condition, we can derive that
|𝑥 (𝑡)|𝑝
𝑝
≤ 𝑥0 + 𝑀 (𝑡)
𝐸|𝑥 (𝑡) − 𝑥 (𝑠)|𝑝
𝑡
𝑡 𝑝 + ∫ 𝑝|𝑥 (𝑠)|𝑝−2 [𝑥𝑇 (𝑠) 𝑓 (𝑥 (𝑠) , 𝑠, 𝛾1 (𝑠))
≤ 2 𝐸∫ 𝑓 (𝑥 (𝑢) , 𝑢, 𝛾1 (𝑢)) 𝑑𝑢
𝑝−1
𝑡0
𝑠
𝑡 𝑝 𝑝 − 1 2
+ 𝑔 (𝑥 (𝑠) , 𝑠, 𝛾2 (𝑠)) ] 𝑑𝑠
+ 2𝑝−1 𝐸∫ 𝑔 (𝑥 (𝑢) , 𝑢, 𝛾2 (𝑢)) 𝑑𝐵 (𝑢) 2 (98)
𝑠
𝑝
𝑡
≤ 𝑥0 + 𝑀 (𝑡) + 0.5𝑝 (𝑝 + 1) 𝐾
𝑝
≤ [2 (𝑡 − 𝑠)]𝑝−1 𝐸 ∫ 𝑓 (𝑥 (𝑢) , 𝑢, 𝛾1 (𝑢)) 𝑑𝑢 (94) 𝑡
𝑠
× ∫ |𝑥 (𝑠)|𝑝−2 (1 + |𝑥 (𝑠)|2 ) 𝑑𝑠
𝑝/2 𝑡0
+ 0.5[2𝑝 (𝑝 − 1)] (𝑡 − 𝑠)(𝑝−2)/2 𝐸
𝑡
𝑝
𝑡
𝑝 ≤ 𝑥0 + 𝑀 (𝑡) + 𝑝 (𝑝 + 1) 𝐾 ∫ (1 + |𝑥 (𝑠)|𝑝 ) 𝑑𝑠,
× ∫ 𝑔 (𝑥 (𝑢) , 𝑢, 𝛾2 (𝑢)) 𝑑𝑢 𝑡0
𝑠
𝑡 𝑝/2
where
≤ 𝐶1 (𝑡 − 𝑠)(𝑝−2)/2 ∫ 𝐸(1 + |𝑥 (𝑢)|2 ) 𝑑𝑢, 𝑡
𝑠
𝑀 (𝑡) = ∫ 𝑝|𝑥 (𝑠)|𝑝−2 𝑥𝑇 (𝑠) 𝑔 (𝑥 (𝑠) , 𝑠, 𝛾2 (𝑠)) 𝑑𝐵 (𝑠) . (99)
𝑡0
where 𝐶1 = 2(𝑝−2)/2 𝐾𝑝/2 ([2(𝑇 − 𝑡0 )]𝑝/2 + [𝑝(𝑝 − 1)]𝑝/2 ). Using Therefore, for any 𝑡1 ∈ [𝑡0 , 𝑇],
(91) yields
𝑝
𝐸 ( sup |𝑥 (𝑡)|𝑝 ) ≤ 𝐸𝑥0 + 𝐸 ( sup |𝑀 (𝑡)|)
𝑡0 ≤𝑡≤𝑡1 𝑡0 ≤𝑡≤𝑡1
𝐸|𝑥 (𝑡) − 𝑥 (𝑠)|𝑝 𝑡1
(𝑝−2)/2
+ 𝑝 (𝑝 + 1) 𝐾𝐸 ∫ (1 + |𝑥 (𝑠)|𝑝 ) 𝑑𝑠.
≤ 𝐶1 (𝑡 − 𝑠) 𝑡0
(100)
𝑡
𝑝
×∫ 2 (𝑝−2)/2
(1 + 𝐸𝑥0 ) exp {𝑝𝜃 (𝑢 − 𝑡0 )} 𝑑𝑢 (95)
𝑠 At the same time, applying the well-known Burkholder-
Davis-Gundy inequality (see, e.g., [2]) gives
𝑝
≤ 𝐶1 2(𝑝−2)/2 (1 + 𝐸𝑥0 )
𝑡 0.5
1
2
which is the desired inequality. ≤ 3𝐸(∫ 𝑝2 |𝑥 (𝑠)|2𝑝−2 𝑔 (𝑥 (𝑠) , 𝑠, 𝛾2 (𝑠)) 𝑑𝑠)
𝑡0
𝑝
Theorem 15. Let 𝑝 ≥ 2 and 𝑥0 ∈ 𝐿 F𝑡 (Ω; R𝑛 ).
Assume that
0
𝑛
≤ 3𝐸 ( sup |𝑥 (𝑡)|𝑝
there is a 𝐾 > 0 such that for all (𝑥, 𝑡, (𝑖1 , 𝑖2 )) ∈ R × [𝑡0 , 𝑇] × S 𝑡0 ≤𝑡≤𝑡1
𝑡 0.5
1
2
× ∫ 𝑝2 |𝑥 (𝑠)|𝑝−2 𝑔 (𝑥 (𝑠) , 𝑠, 𝛾2 (𝑠)) 𝑑𝑠)
2
𝑥𝑇 𝑓 (𝑥, 𝑡, 𝑖1 ) ∨ 𝑔 (𝑥, 𝑡, 𝑖2 ) ≤ 𝐾(1 + |𝑥|)2 . (96) 𝑡0
Substituting the above inequality into (100) gives (H1) : both 𝛾1 (𝑡) and 𝛾2 (𝑡) are right-continuous homoge-
nous Markovian chains such that 𝛾(𝑡) = (𝛾1 (𝑡), 𝛾2 (𝑡)) is a
𝑝 homogenous vector chain.
𝐸 ( sup |𝑥 (𝑡)|𝑝 ) ≤ 2𝐸𝑥0 Under hypothesis (H1) , the results given in this paper
𝑡0 ≤𝑡≤𝑡1
can be established similarly. It is easy to see that if 𝛾1 (𝑡) ≡
𝑡1 𝛾2 (𝑡) and 𝛾1 (𝑡) is a right-continuous homogenous Markovian
+ 2𝑝 (10𝑝 + 1) 𝐾 ∫ (1 + |𝑥 (𝑠)|𝑝 ) 𝑑𝑠. chain, then (H1) is fulfilled immediately. At the same time,
𝑡0
(102) if 𝛾1 (𝑡) ≡ 𝛾2 (𝑡), (8) will reduce to the classical SDEs with
single Markovian chain; that is to say, the classical theory
Thus about SDEs with single Markovian chain is a special case
of our theory. On the other hand, many theorems in this
paper will play important roles in further study. For example,
1 + 𝐸 ( sup |𝑥 (𝑡)|𝑝 ) Theorem 15 will be useful when one studies the approximate
𝑡0 ≤𝑡≤𝑡1
solutions.
𝑝
≤ 1 + 2𝐸𝑥0 + 2𝑝 (10𝑝 + 1) 𝐾 (103) Some important and interesting questions can be further
investigated using the results in this paper. For example,
𝑡1 approximate solutions, boundedness and stability, stochas-
× ∫ (1 + sup |𝑥 (𝑡)|𝑝 ) 𝑑𝑠. tic functional differential equations with vector Markovian
𝑡0 𝑡0 ≤𝑡≤𝑠
switching and their applications. In particular, the stability of
Then the required assertion follows from the Gronwall (8) is one of the most important and interesting topics, and
inequality. those investigations are in progress.
Up to now, we have discussed the 𝐿𝑝 -estimates for the
solution in the case when 𝑝 ≥ 2. As for 0 < 𝑝 < 2, the similar Acknowledgments
results can be given without any difficulty as long as we note
that the Hölder inequality implies The authors thank the editor and referees for their very
important and helpful comments and suggestions. The
0.5𝑝
𝐸|𝑥 (𝑡)|𝑝 ≤ [𝐸|𝑥 (𝑡)|2 ] . (104) authors also thank Dr. C. Zhang for helping them to improve
the English exposition. This research is supported by the
NSFC of China (nos. 11171081 and 11171056).
4. Example
References
Consider the following Black-Scholes model:
[1] A. V. Skorokhod, Asymptotic Methods in the Theory of Stochastic
Differential Equations, vol. 78 of Translations of Mathematical
𝑑𝑋 (𝑡) = 𝜇 (𝛾1 (𝑡)) 𝑋 (𝑡) 𝑑𝑡 + ] (𝛾2 (𝑡)) 𝑋 (𝑡) 𝑑𝐵 (𝑡) , (105)
Monographs, American Mathematical Society, Providence, RI,
USA, 1989.
where 𝛾1 (𝑡) is a right-continuous homogenous Markovian
chain taking values in finite state spaces S1 = {1, 2} and 𝛾2 (𝑡) [2] X. Mao and C. Yuan, Stochastic Differential Euations with
Markovian Switching, Imperial College Press, London, UK,
is a right-continuous homogenous Markovian chain taking
2006.
values in finite state spaces S2 = {1, 2, 3}, 𝜇(𝑖) = 𝑖, 𝑖 = 1, 2,
[3] G. Yin and X. Y. Zhou, “Markowitz’s mean-variance portfolio
](𝑗) = 𝑗 + 1, 𝑗 = 1, 2, 3. Taking 𝐾 = 16, 𝐾 = 9, then (42)
selection with regime switching: from discrete-time models to
and (43) hold. Therefore, by Theorem 7, (105) has a unique their continuous-time limits,” IEEE Transactions on Automatic
solution. Control, vol. 49, no. 3, pp. 349–360, 2004.
[4] J. Buffington and R. J. Elliott, “American options with regime
5. Conclusions and Further Research switching,” International Journal of Theoretical and Applied
Finance, vol. 5, no. 5, pp. 497–514, 2002.
This paper is devoted to studying the existence and unique- [5] C. Zhu and G. Yin, “On competitive Lotka-Volterra model in
ness of solution of SDEs with multi-Markovian switchings random environments,” Journal of Mathematical Analysis and
and estimating the 𝑝th moment of the solution. We have used Applications, vol. 357, no. 1, pp. 154–170, 2009.
two continuous-time Markovian chains to model the SDEs. [6] Q. Luo and X. Mao, “Stochastic population dynamics under
This area is becoming increasingly useful in engineering, regime switching,” Journal of Mathematical Analysis and Appli-
economics, communication theory, active networking, and cations, vol. 334, no. 1, pp. 69–84, 2007.
so forth. The sufficient criteria for existence and unique- [7] X. Li, D. Jiang, and X. Mao, “Population dynamical behavior
ness of solution, local solution, and maximal local solution of Lotka-Volterra system under regime switching,” Journal of
were established. Those results indicate that (8) keeps many Computational and Applied Mathematics, vol. 232, no. 2, pp.
properties that (89) owns. At the same time, although the 427–448, 2009.
hypothesis (H1) is used in this paper, we want to point out that [8] X. Li, A. Gray, D. Jiang, and X. Mao, “Sufficient and neces-
this hypothesis is not essential. In fact, (H1) can be replaced sary conditions of stochastic permanence and extinction for
by the following generalized hypothesis. stochastic logistic populations under regime switching,” Journal
Journal of Applied Mathematics 11
International
Journal of Journal of
Mathematics and
Mathematical
Discrete Mathematics
Sciences