stochastic differential equations with multi-markovian ... › journals › hoa › jam ›...
TRANSCRIPT
Hindawi Publishing CorporationJournal of Applied MathematicsVolume 2013 Article ID 357869 11 pageshttpdxdoiorg1011552013357869
Research ArticleStochastic Differential Equations withMulti-Markovian Switching
Meng Liu1 and Ke Wang2
1 School of Mathematical Science Huaiyin Normal University Huairsquoan 223300 China2Department of Mathematics Harbin Institute of Technology Weihai 264209 China
Correspondence should be addressed to Meng Liu liumeng0557sinacom
Received 31 August 2012 Revised 2 March 2013 Accepted 5 March 2013
Academic Editor Jose L Gracia
Copyright copy 2013 M Liu and K WangThis is an open access article distributed under the Creative Commons Attribution Licensewhich 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 anduniqueness of solution are investigated and the pth moment of the solution is estimated The classical theory of SDEs with singleMarkovian switching is extended
1 Introduction
Stochastic modeling has played an important role in manybranches of industry and science SDEs with single continu-ous-time Markovian chain have been used to model manypractical systems where they may experience abrupt changesin their parameters and structure caused by phenomenasuch as abrupt environment disturbances SDEs with singleMarkovian switching can be denoted by
119889119909 (119905) = 119891 (119909 (119905) 119905 120574 (119905)) 119889119905
+ 119892 (119909 (119905) 119905 120574 (119905)) 119889119861 (119905) 1199050le 119905 le 119879
(1)
with initial conditions 119909(1199050) = 119909
0isin 1198712
F1199050
and 120574(1199050) = 120574
0
where 120574(119905) is a right-continuous homogenous Markovianchain on the probability space taking values in a finite statespace S = 1 2 119873 and is F
119905-adapted but independent
of the Brownian motion 119861(119905) and
119891 R119899times R+times S 997888rarr R
119899
119892 R119899times R+times S 997888rarr R
119899+119898
(2)
Owing to their theoretical and practical significance (1)has received great attention and has been recently studiedextensively andwe heremention Skorokhod [1] andMao andYuan [2] among many others
However in the real world the condition that coefficients119891 and 119892 in (1) are perturbed by the same Markovian chainis too restrictive For example in the classical Black-Scholesmodel the asset price is given by a geometric Brownianmotion
119889119883 (119905) = 120583119883 (119905) 119889119905 + ]119883(119905) 119889119861 (119905) (3)
where 120583 is the rate of the return of the underlying assert ]is the volatility and 119861(119905) is a scalar Brownian motion Sincethere is strong evidence to indicate that 120583 is not a constant butis a Markovian jump process (see eg [3 4]) many authorsproposed the following model
119889119883 (119905) = 120583 (120574 (119905))119883 (119905) 119889119905 + ] (120574 (119905))119883 (119905) 119889119861 (119905) (4)
However many stochastic factors that affect 120583 are differentfrom those that affect ] Then the following model is moreappropriate than model (105) to describe this problem
119889119883 (119905) = 120583 (1205741(119905))119883 (119905) 119889119905 + ] (120574
2(119905))119883 (119905) 119889119861 (119905) (5)
where 120574119894(119905) is a right-continuous homogenous Markovian
chain taking values in a finite state space 119894 = 1 2 Anotherexample is the stochastic Lotka-Volterra model with singleMarkovian switching which has received great attentionand has been studied extensively recently (see eg [5ndash12])
2 Journal of Applied Mathematics
For the sake of convenience we take the following two-dimensional competitive model as an example
1198891199091= 1199091[11990310
(120574 (119905)) minus 11988611
(120574 (119905)) 1199091 minus 11988612
(120574 (119905)) 1199092] 119889119905
+ 1205721(120574 (119905)) 119909
11198891198611(119905)
1198891199092= 1199092[11990320
(120574 (119905)) minus 11988621
(120574 (119905)) 1199091minus 11988622
(120574 (119905)) 1199092] 119889119905
+ 1205722(120574 (119905)) 11990921198891198612 (119905)
(6)
where 119909119894is the size of 119894th species at time 119905 119903
1198940(119895) represents
the growth rate of 119894th species in regime 119895 for 119894 = 1 2119895 isin S and 119861
1and 119861
2are independent standard Brownian
motions However there are many stochastic factors thataffect some coefficients intensely but have little impact onother coefficients in (6) For example suppose that thestochastic factor is rain falls and 119909
1is able to endure a damp
weather while 1199092is fond of a dry environment then the rain
falls will affect 1199092intensely but have little impact on 119909
1 Thus
a more appropriate model is governed by
1198891199091= 1199091[11990310
(12057410
(119905)) minus 11988611
(12057411
(119905)) 1199091minus 11988612
(12057412
(119905)) 1199092] 119889119905
+ 1205721(1205741 (119905)) 11990911198891198611 (119905)
1198891199092= 1199092[11990320
(12057420
(119905)) minus 11988621
(12057421
(119905)) 1199091minus 11988622
(12057422
(119905)) 1199092] 119889119905
+ 1205722(1205742 (119905)) 11990921198891198612 (119905)
(7)
where 120574119894119895(119905) and 120574
119896(119905) are right-continuous homogenous
Markovian chains taking values in finite state spaces S119894119895for
119894 = 1 2 119895 = 0 1 2 and S119896for 119896 = 1 2 respectively
Thus the above examples show that the study of thefollowing SDEs with multi-Markovian switchings is essentialand is of great importance fromboth theoretical and practicalpoints
119889119909 (119905) = 119891 (119909 (119905) 119905 1205741(119905)) 119889119905
+ 119892 (119909 (119905) 119905 1205742 (119905)) 119889119861 (119905) 1199050le 119905 le 119879
(8)
with initial conditions 119909(1199050) = 119909
0isin 1198712
F0
and 120574119894(1199050) where
120574119894(119905) is a right-continuous homogenous Markovian chain
on the probality space taking values in a finite state spaceS119894= 1 2 119873
119894 and isF
119905-adapted but independent of the
Brownian motion 119861119894(119905) 119894 = 1 2 and
119891 R119899times R+times S1997888rarr R
119899
119892 R119899times R+times S2997888rarr R
119899+119898
(9)
Equation (8) can be regarded as the result of the 1198731times 1198732
equations
119889119909 (119905) = 119891 (119909 (119905) 119905 119894) 119889119905
+ 119892 (119909 (119905) 119905 119895) 119889119861 (119905) 119894 isin S1 119895 isin S
2
(10)
switching among each other according to the movement ofthe Markovian chains It is important for us to discover the
properties of the system (8) and to find out whether thepresence of two Markovian switchings affects some knownresults The first step and the foundation of those studies areto establish the theorems for the existence and uniqueness ofthe solution to system (8) So in this paper we will give sometheorems for the existence and uniqueness of the solution tosystem (8) and study some properties of this solution Thetheory developed in this paper is the foundation for furtherstudy and can be applied in many different and complicatedsituations and hence the importance of the results in thispaper is clear
It should be pointed out that the theory developed in thispaper can be generalized to cope with the more general SDEswith more Markovian chains
119889119909 (119905) = 119891 (119909 (119905) 119905 1205741 (119905) 120574119899 (119905)) 119889119905
+ 119892 (119909 (119905) 119905 120574119899+1
(119905) 120574119899+119898
(119905)) 119889119861 (119905)
(11)
The reason we concentrate on (8) rather than (11) is to avoidthe notations becoming too complicated Once the theorydeveloped in this paper is established the reader should beable to cope with the more general (11) without any difficulty
The remaining part of this paper is as follows In Section 2the sufficient criteria for existence anduniqueness of solutionlocal solution andmaximal local solution will be establishedrespectively In Section 3 the 119871
119875-estimates of the solutionwill be given In Section 4 we will introduce an example toillustrate our main result Finally we will close the paper withconclusions in Section 5
2 SDEs with Markovian Chains
Throughout this paper let (ΩF F119905119905isin119877+
P) be a completeprobability space Let 119861(119905) = (119861
1(119905) 119861
119898(119905))119879 be an 119898-
dimensional Brownian motion defined on the probabilityspace
In this section we will consider (8) Let 120574(119905) = (1205741(119905)
1205742(119905)) We impose a hypothesis(H1) 120574
1(119905) is independent of 120574
2(119905)
Then 120574(119905) is a homogenous vector Markovian chain withtransition probabilities
119875 120574 (119904 + 119905) = (1198951 1198952) | 120574 (119904) = (119894
1 1198942)
= 119875 120574 (119905) = (1198951 1198952) | 120574 (0) = (119894
1 1198942)
= 1198751198941119894211989511198952
(119905)
(12)
where (1198941 1198942) (1198951 1198952) isin S = (1 1) (1 2) (1119873
2) (2 1)
(1198731 1198732)
Now we will prepare some lemmas which are importantfor further study
Lemma 1 1198751198941119894211989511198952
(119905) has the following properties
(i) 1198751198941119894211989511198952
(119905) ge 0 for (1198941 1198942) (1198951 1198952) isin S
(ii) sum(11989511198952)isinS 1198751198941119894211989511198952
(119905) = 1 for (1198941 1198942) isin S
(iii) 1198751198941119894211989511198952
(0) = 12057511989411198951
sdot 12057511989421198952
where 120575119894119896119895119896
= 1 if 119894119896
= 119895119896
otherwise 120575119894119896119895119896
= 0 119896 = 1 2
Journal of Applied Mathematics 3
(iv) (Chapman-Kolmogorov equation) For 119904 119905 ge 0 and(1198941 1198942) (1198951 1198952) isin S
1198751198941119894211989511198952
(119904 + 119905)
= sum
(1198961 1198962)isinS
1198751198941119894211989611198962
(119905) 1198751198961119896211989511198952
(119904) (13)
Proof The proofs of (i) (ii) and (iii) are obvious Now let usprove (iv)
1198751198941119894211989511198952
(119904 + 119905)
= 119875 120574 (119904 + 119905) = (1198951 1198952) | 120574 (0) = (119894
1 1198942)
= sum
(11989611198962)isinS
119875 120574 (119904 + 119905) = (1198951 1198952)
120574 (119905) = (1198961 1198962) | 120574 (0) = (119894
1 1198942)
= sum
(11989611198962)isinS
(119875 120574 (119904 + 119905) = (1198951 1198952)
120574 (119905) = (1198961 1198962) 120574 (0) = (119894
1 1198942)
times(119875 120574 (119905) = (1198961 1198962) 120574 (0) = (119894
1 1198942))minus1
)
times119875 120574 (119905) = (119896
1 1198962) 120574 (0) = (119894
1 1198942)
119875 120574 (0) = (1198941 1198942)
= sum
(11989611198962)isinS
119875 120574 (119904 + 119905) = (1198951 1198952) | 120574 (119905) = (119896
1 1198962)
120574 (0) = (1198941 1198942)
times 119875 120574 (119905) = (1198961 1198962) | 120574 (0) = (119894
1 1198942)
= sum
(1198961 1198962)isinS
119875 120574 (119904 + 119905) = (1198951 1198952) | 120574 (119905) = (119896
1 1198962)
times 119875 120574 (119905) = (1198961 1198962) | 120574 (0) = (119894
1 1198942)
= sum
(1198961 1198962)isinS
1198751198961119896211989511198952
(119904) 1198751198941119894211989611198962
(119905)
(14)
This completes the proof
Now we impose another hypothesis which is calledstandard condition
(H2) lim119905rarr0
1198751198941119894211989511198952
(119905) = 1205751198941119894211989511198952
Lemma 2 Under Assumption (H2) for all 119905 ge 0 (1198941 1198942) isin S
one has 1198751198941119894211989411198942
(119905) gt 0
Proof From 1198751198941119894211989411198942
(0) gt 0 and (H2) we know that forarbitrary fixed 119905 gt 0 we have
1198751198941119894211989411198942
(119905
119899) gt 0 (15)
for sufficient large 119899 Then making use of Chapman-Kolm-ogorov equation
1198751198941119894211989411198942
(119904 + 119905) = sum
(1198961 1198962)isinS
11987511989411198942119896111198962
(119905) 1198751198961119896211989411198942
(119904)
ge 1198751198941119894211989411198942
(119905) 1198751198941119894211989411198942
(119904)
(16)
gives
1198751198941119894211989411198942
(119905) ge (1198751198941119894211989411198942
(119905
119899))
119899
gt 0 (17)
which is the desired assertion
Lemma 3 Under Assumption (H2) for all (1198941 1198942) isin S
minus1199021198941119894211989411198942
= lim119905rarr0
1 minus 1198751198941119894211989411198942
(119905)
119905(18)
exists (but may be infin)
Proof Define 120601(119905) = minus ln1198751198941119894211989411198942
(119905) ge 0 Then making use of(16) gives
120601 (119904 + 119905) le 120601 (119904) + 120601 (119905) (19)
Set minus1199021198941119894211989411198942
= sup119905gt0
(120601(119905)119905) It is easy to see that
0 le minus1199021198941119894211989411198942
le infin lim sup119905rarr0
120601 (119905)
119905le minus1199021198941119894211989411198942
(20)
Now we will assert
lim inf119905rarr0
120601 (119905)
119905ge minus1199021198941119894211989411198942
(21)
In fact for 0 lt ℎ lt 119905 exist119899 0 le 120576 le ℎ such that 119905 = 119899ℎ + 120576Applying (19) yields
120601 (119905)
119905ge
119899ℎ
120576
120601 (ℎ)
ℎ+
120601 (120576)
119905 (22)
Note that 120576 rarr 0 119899ℎ120576 rarr 1 120601(120576) rarr 0 whenever ℎ rarr 0+
then for all 119905 gt 0 we have
120601 (119905)
119905le lim infℎrarr0
120601 (ℎ)
ℎ (23)
This implies minus1199021198941119894211989411198942
le lim inf119905rarr0
(120601(119905)119905) Thus
minus1199021198941119894211989411198942
= lim119905rarr0
120601 (119905)
119905 (24)
Using the definition of 120601(119905) gives
lim119905rarr0
1 minus 1198751198941119894211989411198942
(119905)
119905= lim119905rarr0
1 minus exp 120601 (119905)
120601 (119905)
120601 (119905)
119905= minus1199021198941119894211989411198942
(25)
which is the required assertion
4 Journal of Applied Mathematics
Lemma 4 Under Assumption (H2) for (1198941 1198942) (1198951 1198952) isin
S (1198941 1198942) = (1198951 1198952)
1199021198941119894211989511198952
= 1198751015840
1198941119894211989511198952
(0) = lim119905rarr0
1198751198941119894211989511198952
(119905)
119905(26)
exists and is finite
Proof By (H2) we note that for all 0 lt 120576 lt 13 exist0 lt 120575 lt 1such that
1198751198941119894211989411198942
(119905) gt 1 minus 120576 1198751198951119895211989511198952
(119905) gt 1 minus 120576
1198751198951119895211989411198942
(119905) lt 120576
(27)
provided 0 lt 119905 le 120575For forall0 le ℎ lt 119905 set 119899 = ⟨119905ℎ⟩ where ⟨119886⟩ = max
119899le119886119899 isin
Z Let
119875(11989511198952)
1198941119894211989611198962
(ℎ) = 1198751198941119894211989611198962
(ℎ)
119875(11989511198952)
1198941119894211989611198962
(119898ℎ)
= sum
(11990311199032) = (1198951 1198952)
119875(11989511198952)
1198941119894211990311199032
((119898 minus 1) ℎ) 1198751199031119903211989611198962
(ℎ)
(28)
where 119875(11989511198952)
1198941119894211989611198962
(119898ℎ)means that the probability of the 120574(119905)willnot reach to (119895
1 1198952) at times ℎ 2ℎ (119898 minus 1)ℎ but will reach
to (1198961 1198962) at time 119898ℎ Note that if ℎ le 119905 le 120575 then
120576 gt 1 minus 1198751198941119894211989411198942
(119905)
= sum
(1198961 1198962) = (1198941 1198942)
1198751198941119894211989611198962
(119905) ge 1198751198941119894211989511198952
(119905)
ge
119899
sum
119898=1
119875(11989511198952)
1198941119894211989511198952
(119898ℎ) 1198751198951119895211989511198952
(119905 minus 119898ℎ)
ge (1 minus 120576)
119899
sum
119898=1
119875(11989511198952)
1198941119894211989511198952
(119898ℎ)
(29)
which indicates119899
sum
119898=1
119875(11989511198952)
1198941119894211989511198952
(119898ℎ) le120576
1 minus 120576 (30)
Then making use of
1198751198941119894211989411198942
(119898ℎ) = 119875(11989511198952)
1198941119894211989411198942
(119898ℎ)
+
119898minus1
sum
119897=1
119875(11989511198952)
1198941119894211989511198952
(119897ℎ) 1198751198951119895211989411198942
((119898 minus 119897) ℎ)
(31)
we obtain
119875(11989511198952)
1198941119894211989411198942
(119898ℎ) ge 1198751198941119894211989411198942
(119898ℎ) minus
119898minus1
sum
119897=1
119875(11989511198952)
1198941119894211989511198952
(119897ℎ)
ge 1 minus 120576 minus120576
1 minus 120576
(32)
Consequently
1198751198941119894211989511198952
(119905)
gt
119899
sum
119898=1
119875(11989511198952)
1198941119894211989411198942
((119898 minus 1) ℎ) 1198751198941119894211989511198952
(ℎ) 1198751198951119895211989511198952
(119905 minus 119898ℎ)
ge 119899 (1 minus 120576 minus120576
1 minus 120576)1198751198941119894211989511198952
(ℎ) (1 minus 120576)
ge 119899 (1 minus 3120576) 1198751198941119894211989511198952
(ℎ)
(33)
Dividing both sides of the above inequality by ℎ and noting119899ℎ rarr 119905 whenever ℎ rarr 0 yield
lim supℎrarr0
1198751198941119894211989511198952
(ℎ)
ℎle
1
1 minus 3120576
1198751198941119894211989511198952
(119905)
119905lt infin (34)
Then letting 119905 rarr 0 gives
lim supℎrarr0
1198751198941119894211989511198952
(ℎ)
ℎle
1
1 minus 3120576lim inf119905rarr0
1198751198941119894211989511198952
(119905)
119905 (35)
and the required assertion follows immediately by letting120576 rarr 0 This completes the proof
Set 120574(119905) = (1205741(119905) 1205742(119905)) then it is easy to see that
almost every sample path of 120574(119905) is a right continuous stepfunction Now letting P(119905) = (119875
1198941119894211989511198952
(119905))11987311198732times11987311198732
Q =
(1199021198941119894211989511198952
)11987311198732times11987311198732
= P1015840(0) Then by Chapman-Kolmogorovequation
P (119905 + ℎ) = P (119905)P (ℎ) = P (ℎ)P (119905) (36)
we have
P (119905 + ℎ) minus P (119905)
ℎ= P (119905) [
P (ℎ) minus Iℎ
] = [P (ℎ) minus I
ℎ]P (119905)
(37)
Letting ℎ rarr 0 and taking limits give
P1015840(119905) = P (119905)Q
P1015840(119905) = QP (119905)
(38)
Note that
P (0) = I (39)
Then by solving the ordinary differential equations (38) and(39) we obtain the following lemma
Lemma 5 For P(119905) and Q one has
P (119905) = exp Q119905 (40)
We are now in the position to give the sufficient condi-tions for the existence and uniqueness of the solution of (8)For this end let us first give the definition of the solution
Journal of Applied Mathematics 5
Definition 6 An R119899-valued stochastic process 119909(119905)1199050le119905le119879
iscalled a solution of (8) if it has the following properties
(i) 119909(119905)1199050le119905le119879
is continuous andF119905-adapted
(ii) 119891(119909(119905) 119905 1205741(119905))1199050le119905le119879
isinL1([1199050 119879]R119899)while 119892(119909(119905)
119905 1205742(119905))1199050le119905le119879
isin L2([1199050 119879]R119899times119898)
(iii) for any 119905 isin [1199050 119879] equation
119909 (119905) = 119909 (1199050) + int
119905
1199050
119891 (119909 (119904) 119904 1205741(119904)) 119889119904
+ int
119905
1199050
119892 (119909 (119904) 119904 1205742(119904)) 119889119861 (119904)
(41)
holds with probability 1
A solution 119909(119905)1199050le119905le119879
is said to be unique if any othersolution (119905)
1199050le119905le119879
is indistinguishable from 119909(119905)1199050le119905le119879
Now we can give our main results in this section
Theorem 7 Assume that there exist two positive constants 119870
and 119870 such that(Lipschitz condition) for all 119909 119910 isin R119899 119905 isin [119905
0 119879] and
(1198941 1198942) isin S
1003816100381610038161003816119891 (119909 119905 1198941) minus 119891 (119910 119905 119894
1)10038161003816100381610038162
or1003816100381610038161003816119892 (119909 119905 119894
2) minus 119892 (119910 119905 119894
2)10038161003816100381610038162le 119870
1003816100381610038161003816119909 minus 11991010038161003816100381610038162
(42)
(Linear growth condition) for all (119909 119905 (1198941 1198942)) isin R119899 times
[1199050 119879] times S
1003816100381610038161003816119891 (119909 119905 1198941)10038161003816100381610038162or
1003816100381610038161003816119892 (119909 119905 1198942)10038161003816100381610038162le 119870(1 + |119909|)
2 (43)
Then there exists a unique solution 119909(119905) to (8) and moreoverthe solution obeys
119864( sup1199050le119905le119879
|119909 (119905)|2)
le (1 + 311986410038161003816100381610038161199090
10038161003816100381610038162) exp 3119870 (119879 minus 119905
0) (119879 minus 119905
0+ 4)
(44)
Proof Recall that almost every sample path of 120574(119905) =
(1205741(119905) 1205742(119905)) is a right continuous step function with a finite
number of jumps on [1199050 119879] Thus there exists a sequence of
stopping times 120591119896119896ge0
such that
(i) for almost every 120596 isin Ω there is a finite 119896120596for 1199050
=
1205910lt 1205911lt sdot sdot sdot lt 120591
119896= 119879 and 120591
119896= 119879 if 119896 gt 119896
120596
(ii) both 1205741(sdot) and 120574
2(sdot) in 120574(sdot) are constants on interval
[[120591119896 120591119896+1
[[ namely
1205741 (119905) = 120574
1(120591119896) on 120591
119896le 119905 le 120591
119896+1forall119896 ge 0
1205742(119905) = 120574
2(120591119896) on 120591
119896le 119905 le 120591
119896+1forall119896 ge 0
(45)
First of all let us consider (8) on 119905 isin [[1205910 1205911[[ then (8)
becomes119889119909 (119905) = 119891 (119909 (119905) 119905 120574
1(1199050)) 119889119905
+ 119892 (119909 (119905) 119905 1205742(1199050)) 119889119861 (119905)
(46)
with initial conditions 119909(1199050) = 119909
0 120574(1199050) = (120574
1(1199050) 1205741(1199050))
Then by the theory of SDEs we obtain that (46) has a uniquesolution which obeys 119909(120591
1) isin 1198712
F1205911
(ΩR119899) We next consider(8) on 119905 isin [[120591
1 1205912[[ which becomes
119889119909 (119905) = 119891 (119909 (119905) 119905 1205741 (1205911)) 119889119905
+ 119892 (119909 (119905) 119905 1205742(1205911)) 119889119861 (119905)
(47)
Again by the theory of SDEs (47) has a unique solutionwhichobeys 119909(120591
1) isin 119871
2
F1205912
(ΩR119899) Repeating this procedure weconclude that (8) has a unique solution 119909(119905) on [119905
0 119879]
Now let us prove (44) For every 119896 ge 1 define thestopping time
120591119896= 119879 and inf 119905 isin [119905
0 119879] |119909 (119905)| ge 119896 (48)
It is obvious that 120591119896uarr 119879 as Set 119909
119896(119905) = 119909(119905and120591
119896) for 119905 isin [119905
0 119879]
Then 119909119896(119905) obeys the equation
119909119896(119905) = 119909
0+ int
119905
1199050
119891 (119909119896(119904) 119904 120574
1(119904)) 119868[[1199050120591119896]](119904) 119889119904
+ int
119905
1199050
119892 (119909119896(119904) 119904 120574
2(119904)) 119868[[1199050120591119896]](119904) 119889119861 (119904)
(49)
Making use of the elementary inequality |119886+119887+119888|2le 3(|119886|
2+
|119887|2+ |119888|2) the Holder inequality and (43) we can see that
1003816100381610038161003816119909119896 (119905)10038161003816100381610038162= 3
1003816100381610038161003816119909010038161003816100381610038162+ 3119870 (119879 minus 119905
0) int
119905
1199050
(1 +1003816100381610038161003816119909119896 (119904)
10038161003816100381610038162) 119889119904
+ 3
100381610038161003816100381610038161003816100381610038161003816
int
119905
1199050
119892 (119909119896(119904) 119904 120574
2(119904)) 119868[[1199050120591119896]](119904) 119889119861 (119904)
100381610038161003816100381610038161003816100381610038161003816
2
(50)
Thus applying the Doob martingale inequality and (43) wecan further show that
119864( sup1199050le119904le119905
1003816100381610038161003816119909119896 (119904)10038161003816100381610038162)
= 311986410038161003816100381610038161199090
10038161003816100381610038162+ 3119870 (119879 minus 119905
0) int
119905
1199050
(1 + 1198641003816100381610038161003816119909119896 (119904)
10038161003816100381610038162) 119889119904
+ 12119864int
119905
1199050
1003816100381610038161003816119892 (119909119896(119904) 119904 120574
2(119904))
10038161003816100381610038162119868[[1199050120591119896]](119904) 119889119904
le 311986410038161003816100381610038161199090
10038161003816100381610038162+ 3119870 (119879 minus 119905
0+ 4)int
119905
1199050
(1 + 1198641003816100381610038161003816119909119896 (119904)
10038161003816100381610038162) 119889119904
(51)
That is to say
1 + 119864( sup1199050le119904le119905
1003816100381610038161003816119909119896 (119904)10038161003816100381610038162) le 1 + 3119864
1003816100381610038161003816119909010038161003816100381610038162+ 3119870 (119879 minus 119905
0+ 4)
times int
119905
1199050
[1 + 119864 sup1199050le119903le119904
1003816100381610038161003816119909119896 (119903)10038161003816100381610038162]119889119904
(52)
6 Journal of Applied Mathematics
Using the Gronwall inequality leads to
1 + 119864( sup1199050le119904le119905
1003816100381610038161003816119909119896 (119904)10038161003816100381610038162)
le (1 + 311986410038161003816100381610038161199090
10038161003816100381610038162) exp 3119870 (119879 minus 119905
0) (119879 minus 119905
0+ 4)
(53)
Consequently
1 + 119864( sup1199050le119905le120591119896
|119909 (119905)|2)
le (1 + 311986410038161003816100381610038161199090
10038161003816100381610038162) exp 3119870 (119879 minus 119905
0) (119879 minus 119905
0+ 4)
(54)
Then the required inequality (44) follows immediately byletting 119896 rarr infin
Condition (42) indicates that the coefficients 119891(119909 119905 1198941)
and 119892(119909 119905 1198942) do not change faster than a linear function of
119909 as change in 119909 This means in particular the continuity of119891(119909 119905 119894
1) and 119892(119909 119905 119894
2) in 119909 for all 119905 isin [119905
0 119879] Then functions
that are discontinuous with respect to 119909 are excluded as thecoefficients Besides there are many functions that do notsatisfy the Lipschitz conditionThese imply that the Lipschitzcondition is too restrictive To improve this Lipschitz condi-tion let us introduce the concept of local solution
Definition 8 Let120590infinbe a stopping time such that 119905
0le 120590infin
le 119879
as An R119899-valuedF119905-adapted continuous stochastic process
119909(119905)1199050le119905lt120590infin
is called a local solution of (8) if 119909(1199050) = 1199090and
moreover there is a nondecreasing sequence 120590119896119896ge1
such that1199050le 120590119896 uarr 120590infin
as and
119909 (119905) = 119909 (1199050) + int
119905and120590119896
1199050
119891 (119909 (119904) 119904 1205741(119904)) 119889119904
+ int
119905and120590119896
1199050
119892 (119909 (119904) 119904 1205742(119904)) 119889119861 (119904)
(55)
holds for any 119905 isin [1199050 119879) and 119896 ge 1 with probability one If
furthermore
lim sup119905rarr+120590
infin
|119909 (119905)| = infin whenever 120590infin
lt 119879 (56)
then it is called a maximal local solution and 120590infinis called the
explosion time Amaximal local solution 119909(119905) 1199050le 119905 lt 120590
infin
is said to be unique if any other maximal local solution 119909(119905)
1199050le 119905 lt 120590
infin is indistinguishable from it namely 120590
infin= 120590infin
and 119909(119905) = 119909(119905) for 1199050le 119905 lt 120590
infinwith probability one
Definition 9 (local Lipschitz condition) For every integer119896 ge 1 there exists a positive constant ℎ
119896such that for all
119905 isin [1199050 119879] 119894 = (119894
1 1198942) isin S and those 119909 119910 isin R119899 with
|119909| or |119910| le 119896
1003816100381610038161003816119891 (119909 119905 1198941) minus 119891 (119910 119905 119894
1)10038161003816100381610038162
or1003816100381610038161003816119892 (119909 119905 119894
2) minus 119892 (119910 119905 119894
2)10038161003816100381610038162le ℎ119896
1003816100381610038161003816119909 minus 11991010038161003816100381610038162
(57)
The following theorem shows the existence of uniquemaximal local solution under the local Lipschitz conditionwithout the linear growth condition
Theorem 10 Under condition (57) there exists a uniquemaximal local solution of (8)
Proof Define functions
119891119896(119909 119905 119894
1) =
119891 (119909 119905 1198941) if |119909| le 119896
119891(119896119909
|119909| 119905 1198941) if |119909| gt 119896
119892119896(119909 119905 119894
2) =
119892 (119909 119905 1198942) if |119909| le 119896
119892 (119896119909
|119909| 119905 1198942) if |119909| gt 119896
(58)
Then 119891119896and 119892
119896satisfy the Lipschitz condition and the linear
growth condition Thus by Theorem 7 there is a uniquesolution 119909
119896(119905) of the equation
119889119909119896(119905) = 119891
119896(119909119896(119905) 119905 120574
1(119905)) 119889119905
+ 119892119896(119909119896(119905) 119905 120574
2(119905)) 119889119861 (119905)
(59)
with the initial conditions 119909119896(1199050) = 119909
0and 120574(119905
0) = (120574
1(1199050)
1205742(1199050)) Define the stopping times
120590119896= 119879 and inf 119905 isin [119905
0 119879]
1003816100381610038161003816119909119896 (119905)1003816100381610038161003816 ge 119896 (60)
Clearly if 1199050le 119905 le 120590
119896
119909119896(119905) = 119909
119896+1(119905) (61)
which indicates that 120590119896is increasing so 120590
119896has its limit 120590
infin=
lim119896rarr+infin
120590119896 Define 119909(119905) 119905
0le 119905 lt 120590
infin by
119909 (119905) = 119909119896(119905) 119905 isin [[120590
119896minus1 120590119896[[ 119896 ge 1 (62)
where 1205900= 1199050 Applying (61) one can show that 119909(119905 and 120590
119896) =
119909119896(119905 and 120590119896) It then follows from (59) that
119909 (119905 and 120590119896) = 1199090+ int
119905and120590119896
1199050
119891119896(119909 (119904) 119904 120574
1(119904)) 119889119904
+ int
119905and120590119896
1199050
119892119896(119909 (119904) 119904 1205742 (119904)) 119889119861 (119904)
= 1199090+ int
119905and120590119896
1199050
119891 (119909 (119904) 119904 1205741(119904)) 119889119904
+ int
119905and120590119896
1199050
119892 (119909 (119904) 119904 1205742 (119904)) 119889119861 (119904)
(63)
for any 119905 isin [1199050 119879) and 119896 ge 1 It is easy to see that if 120590
infinle 119879
then
lim sup119905rarr120590infin
|119909 (119905)|
ge lim sup119896rarr+infin
1003816100381610038161003816119909 (120590119896)1003816100381610038161003816 = lim sup119896rarr+infin
1003816100381610038161003816119909119896 (120590119896)1003816100381610038161003816 = infin
(64)
Therefore 119909(119905) 1199050le 119905 lt 120590
infin is a maximal local solution
Now we will prove the uniqueness Let 119909(119905) 1199050
le 119905 lt
120590infin
be another maximal local solution Define
120590119896= 120590infin
and inf 119905 isin [[1199050 120590infin
[[ 119909 (119905) ge 119896 (65)
Journal of Applied Mathematics 7
Then 120590119896
rarr 120590infin
as and
119875 119909 (119905) = 119909 (119905) 119905 isin [[1199050 120590119896and 120590119896]] = 1 forall119896 ge 1 (66)
Letting 119896 rarr infin gives
119875 119909 (119905) = 119909 (119905) 119905 isin [[1199050 120590infin
and 120590infin
[[ = 1 (67)
In order to complete the proof we need only to show that120590infin
= 120590infin
as In fact for almost every 120596 isin 120590infin
lt 120590infin
wehave
1003816100381610038161003816119909 (120590infin
120596)1003816100381610038161003816
= lim119896rarr+infin
1003816100381610038161003816119909 (120590119896 120596)
1003816100381610038161003816 = lim119896rarr+infin
1003816100381610038161003816119909 (120590119896 120596)
1003816100381610038161003816 = infin(68)
which contradicts the fact that 119909(119905 120596) is continuous on 119905 isin
[1199050 120590infin
(120596)) This implies 120590infin
ge 120590infin
as In the same way onecan show 120590
infinle 120590infinasThus wemust have 120590
infin= 120590infinasThis
completes the proof
In many situations we often consider an SDE on [1199050infin)
119889119909 (119905) = 119891 (119909 (119905) 119905 1205741(119905)) 119889119905
+ 119892 (119909 (119905) 119905 1205742(119905)) 119889119861 (119905) 119905
0le 119905 lt infin
(69)
with initial data 119909(1199050) = 119909
0and 120574(119905
0) If the assumption of
the existence-and-uniqueness theorem holds on every finitesubinterval [119905
0 119879] of [119905
0infin) then (69) has a unique solution
119909(119905) on the entire interval [1199050infin) Such a solution is called
a global solution To establish a more general result aboutglobal solution we need more notations To this end weintroduce an operator 119871119881 from R119899 times R
+times S to R which is
given by
119871119881 (119909 119905 (1198941 1198942))
= 119881119905(119909 119905 (119894
1 1198942)) + 119881
119909(119909 119905 (119894
1 1198942)) 119891 (119909 119905 119894
1)
+ 05 trace [119892119879 (119909 119905 1198942) 119881119909119909
(119909 119905 (1198941 1198942)) 119892 (119909 119905 119894
2)]
+ sum
(11989511198952)isinS
1199021198941119894211989511198952
119881 (119909 119905 (1198951 1198952))
(70)
where 119881(119909 119905 (1198941 1198942)) isin 119862
21(R119899 times R
+times SR
+) and
119881119905(119909 119905 (119894
1 1198942)) =
120597119881 (119909 119905 (1198941 1198942))
120597119905
119881119909(119909 119905 (119894
1 1198942))
= (120597119881 (119909 119905 (119894
1 1198942))
1205971199091
120597119881 (119909 119905 (119894
1 1198942))
120597119909119899
)
119881119909119909
(119909 119905 (1198941 1198942)) = (
1205972119881 (119909 119905 (119894
1 1198942))
120597119909119894120597119909119895
)
119899times119899
(71)
Theorem 11 Assume that the local Lipschitz condition (57)holds Assume also that there is a function 119881(119909 119905 (119894
1 1198942)) isin
11986221
(R119899 times R+times SR
+) and a constant 120579 gt 0 such that
lim|119909|rarrinfin
( inf(119905(1198941 1198942))isinR+timesS
119881 (119909 119905 (1198941 1198942))) = infin (72)
119871119881 (119909 119905 (1198941 1198942)) le 120579 (1 + 119881 (119909 119905 (119894
1 1198942)))
forall (119909 119905 (1198941 1198942)) isin R
119899times R+times S
(73)
Then there exists a unique global solution 119909(119905) to (69)
Proof We need only to prove the theorem for any initialcondition 119909
0isin R119899 and (120574
1(1199050) 1205742(1199050)) isin S FromTheorem 10
we know that the local Lipschitz condition guarantees theexistence of the unique maximal solution 119909(119905) on [119905
0 120590infin
)where 120590
infinis the explosion time We need only to show 120590
infin=
infin as If this is not true then we can find a pair of positiveconstants 120576 and 119879 such that
119875 120590infin
le 119879 gt 2120576 (74)
For each integer 119896 ge 1 define the stopping time
120590119896= inf 119905 ge 119905
0 |119909 (119905)| ge 119896 (75)
Since 120590119896
rarr 120590infinalmost surely we can find a sufficiently large
integer 1198960for
119875 120590119896le 119879 gt 120576 forall119896 ge 119896
0 (76)
Fix any 119896 ge 1198960 then for any 119905
0le 119905 le 119879 by virtue of the
generalized Ito formula (see eg [1])
119864119881 (119909 (119905 and 120590119896) 119905 and 120590
119896 120574 (119905 and 120590
119896))
= 119881 (1199090 1199050 1205740) + 119864int
119905and120590119896
1199050
119871119881 (119909 (119904) 119904 120574 (119904)) 119889119904
le 119881 (1199090 1199050 1205740) + 120579 (119879 minus 119905
0)
+ 120579int
119905
1199050
119864119881 (119909 (119904 and 120590119896) 119904 and 120590
119896 120574 (119904 and 120590
119896)) 119889119904
(77)
Making use of the Gronwall inequality gives
119864119881 (119909 (119905 and 120590119896) 119905 and 120590
119896 120574 (119905 and 120590
119896))
le [119881 (1199090 1199050 1205740) + 120579 (119879 minus 119905
0)] exp 120579 (119879 minus 119905
0)
(78)
Therefore
119864 (119868120590119896le119879
119881 (119909 (120590119896) 120590119896 120574 (120590119896)))
le [119881 (1199090 1199050 1205740) + 120579 (119879 minus 119905
0)] exp 120579 (119879 minus 119905
0)
(79)
At the same time set
119892119896= inf 119881 (119909 119905 (119894
1 1198942)) |119909| ge 119896 119905 isin [119905
0 119879] (119894
1 1198942) isin S
(80)
8 Journal of Applied Mathematics
Then (72) means 119892119896
rarr infin It follows from (76) and (79) that
[119881 (1199090 1199050 1205740) + 120579 (119879 minus 119905
0)] exp 120579 (119879 minus 119905
0)
ge 119892119896119875 120590119896le 119879 ge 120576119892
119896
(81)
Letting 119896 rarr infin yields a contradiction that is to say120590infin
= infinThe proof is complete
3 119871119875-Estimates
In the previous section we have investigated the existenceand uniqueness of the solution to (8) In this section as abovelet 119909(119905) 119905
0le 119905 le 119879 be the unique solution of (8) with initial
conditions 119909(1199050) = 1199090and 120574(119905
0) and we will estimate the 119901th
moment of the solution
Theorem 12 Assume that there is a function 119881(119909 119905 (1198941 1198942)) isin
11986221
(R119899 times R+times SR
+) and positive constants 119901 120578 120579 such that
for all (119909 119905 (1198941 1198942)) isin R119899 times R
+times S
120578|119909|119901le 119881 (119909 119905 (119894
1 1198942)) (82)
119871119881 (119909 119905 (1198941 1198942)) le 120579119881 (119909 119905 (119894
1 1198942)) (83)
Assume also the initial condition 119909(1199050) = 119909
0and 120574(119905
0) obeys
119864119881(1199090 1199050 120574(1199050)) lt infin then one has
119864|119909 (119905)|119901
le119864119881 (119909
0 1199050 120574 (1199050)) exp 120579 (119905 minus 119905
0)
120578 forall119905 isin [119905
0 119879]
(84)
Proof For each integer 119896 ge 1 define the stopping time
120590119896= 119879 and inf 119905 ge 119905
0 |119909 (119905)| ge 119896 (85)
Thus 120590119896
rarr 119879 as Using the generalized Itorsquos formula and(83) we obtain that for 119905 isin [119905
0 119879]
119864119881 (119909 (119905 and 120590119896) 119905 and 120590
119896 120574 (119905 and 120590
119896))
le 119864119881 (1199090 1199050 120574 (1199050))
+ 120579int
119905
1199050
119864119881 (119909 (119904 and 120590119896) 119904 and 120590
119896 120574 (119904 and 120590
119896)) 119889119904
(86)
Then the Gronwall inequality indicates
119864119881 (119909 (119905 and 120590119896) 119905 and 120590
119896 120574 (119905 and 120590
119896))
le 119864119881 (1199090 1199050 120574 (1199050)) exp 120579 (119905 minus 119905
0)
(87)
for all 119905 isin [1199050 119879] By virtue of condition (83) we obtain the
required assertion (84)
Corollary 13 Assume 119901 ge 2 and 1199090isin 119871119901
F1199050
(ΩR119899) Assumealso that there exists a constant 120579 gt 0 such that for all(119909 119905 (119894
1 1198942)) isin R119899 times R
+times S
119909119879119891 (119909 119905 119894
1) +
119901 minus 1
2
1003816100381610038161003816119892 (119909 119905 1198942)10038161003816100381610038162le 120579 (1 + |119909|
2) (88)
Then one has
119864|119909 (119905)|119901
le 2(119901minus2)2
(1 + 11986410038161003816100381610038161199090
1003816100381610038161003816119901) exp 119901120579 (119905 minus 119905
0) forall119905 isin [119905
0 119879]
(89)
Proof Define 119881(119909 119905 (1198941 1198942)) = (1 + |119909|
2)1199012 Making use of
(88) yields
119871119881 (119909 119905 (1198941 1198942))
= 119901(1 + |119909|2)(119901minus2)2
119909119879119891 (119909 119905 119894
1)
+119901
2(1 + |119909|
2)(119901minus2)21003816100381610038161003816119892 (119909 119905 119894
2)10038161003816100381610038162
+119901 (119901 minus 2)
2(1 + |119909|
2)(119901minus4)210038161003816100381610038161003816
119909119879119892 (119909 119905 119894
2)10038161003816100381610038161003816
2
le 119901(1 + |119909|2)(119901minus2)2
[119909119879119891 (119909 119905 119894
1) +
119901 minus 1
2
1003816100381610038161003816119892 (119909 119905 1198942)10038161003816100381610038162]
le 119901120579(1 + |119909|2)1199012
(90)
Then byTheorem 12 we get
119864(1 + |119909 (119905)|2)1199012
le 119864(1 +10038161003816100381610038161199090
10038161003816100381610038162)1199012
exp 119901120579 (119905 minus 1199050)
(91)
and the required assertion (89) followsIt is useful to point out that if the linear growth condition
(43) is satisfied then (88) is fulfilled with 120579 = radic119870 + 119870(119901 minus
1)2 Now we will show the other important properties of thesolution
Theorem 14 Let 119901 ge 2 and 1199090
isin 119871119901
F1199050
(ΩR119899) Assume alsothat the linear growth condition (43) holds Then one has
119864|119909 (119905) minus 119909 (119904)|119901le 119862(119905 minus 119904)
1199012 (92)
where
119862 = 2119901minus2
(1 + 11986410038161003816100381610038161199090
1003816100381610038161003816119901) exp 119901120579 (119879 minus 119905
0)
times ([2 (119879 minus 1199050)]1199012
+ [119901 (119901 minus 1)]1199012
)
(93)
and 120579 = radic119870+119870(119901minus1)2 Particularly the 119901th moment of thesolution is continuous on [119905
0 119879]
Journal of Applied Mathematics 9
Proof Applying the elementary inequality |119886 + 119887|119901
le
2119901minus1
(|119886|119901+ |119887|119901) the Holder inequality and the linear growth
condition we can derive that
119864|119909 (119905) minus 119909 (119904)|119901
le 2119901minus1
119864
10038161003816100381610038161003816100381610038161003816int
119905
119904
119891 (119909 (119906) 119906 1205741(119906)) 119889119906
10038161003816100381610038161003816100381610038161003816
119901
+ 2119901minus1
119864
10038161003816100381610038161003816100381610038161003816int
119905
119904
119892 (119909 (119906) 119906 1205742(119906)) 119889119861 (119906)
10038161003816100381610038161003816100381610038161003816
119901
le [2 (119905 minus 119904)]119901minus1
119864int
119905
119904
1003816100381610038161003816119891 (119909 (119906) 119906 1205741(119906))
1003816100381610038161003816119901119889119906
+ 05[2119901 (119901 minus 1)]1199012
(119905 minus 119904)(119901minus2)2
119864
times int
119905
119904
1003816100381610038161003816119892 (119909 (119906) 119906 1205742 (119906))1003816100381610038161003816119901119889119906
le 1198621(119905 minus 119904)
(119901minus2)2int
119905
119904
119864(1 + |119909 (119906)|2)1199012
119889119906
(94)
where1198621= 2(119901minus2)2
1198701199012
([2(119879minus 1199050)]1199012
+ [119901(119901minus1)]1199012
) Using(91) yields
119864|119909 (119905) minus 119909 (119904)|119901
le 1198621(119905 minus 119904)
(119901minus2)2
times int
119905
119904
2(119901minus2)2
(1 + 11986410038161003816100381610038161199090
1003816100381610038161003816119901) exp 119901120579 (119906 minus 119905
0) 119889119906
le 11986212(119901minus2)2
(1 + 11986410038161003816100381610038161199090
1003816100381610038161003816119901)
times exp 119901120579 (119879 minus 1199050) (119905 minus 119904)
(1199012)
(95)
which is the desired inequality
Theorem 15 Let 119901 ge 2 and 1199090
isin 119871119901
F1199050
(ΩR119899) Assume thatthere is a119870 gt 0 such that for all (119909 119905 (119894
1 1198942)) isin R119899times[119905
0 119879]timesS
119909119879 1003816100381610038161003816119891 (119909 119905 119894
1)1003816100381610038161003816 or
1003816100381610038161003816119892 (119909 119905 1198942)10038161003816100381610038162le 119870(1 + |119909|)
2 (96)
Then
119864( sup1199050le119905le119879
|119909 (119905)|119901)
le (1 + 211986410038161003816100381610038161199090
1003816100381610038161003816119901) exp 2119901 (10119901 + 1)119870 (119879 minus 119905
0)
(97)
Proof Making use of the generalized Itorsquos formula andcondition (96) we have
|119909 (119905)|119901
le10038161003816100381610038161199090
1003816100381610038161003816119901+ 119872(119905)
+ int
119905
1199050
119901|119909 (119904)|119901minus2
[119909119879(119904) 119891 (119909 (119904) 119904 120574
1(119904))
+119901 minus 1
2
1003816100381610038161003816119892 (119909 (119904) 119904 1205742 (119904))10038161003816100381610038162] 119889119904
le10038161003816100381610038161199090
1003816100381610038161003816119901+ 119872(119905) + 05119901 (119901 + 1)119870
times int
119905
1199050
|119909 (119904)|119901minus2
(1 + |119909 (119904)|2) 119889119904
le10038161003816100381610038161199090
1003816100381610038161003816119901+ 119872(119905) + 119901 (119901 + 1)119870int
119905
1199050
(1 + |119909 (119904)|119901) 119889119904
(98)
where
119872(119905) = int
119905
1199050
119901|119909 (119904)|119901minus2
119909119879(119904) 119892 (119909 (119904) 119904 1205742 (119904)) 119889119861 (119904) (99)
Therefore for any 1199051isin [1199050 119879]
119864( sup1199050le119905le1199051
|119909 (119905)|119901) le 119864
100381610038161003816100381611990901003816100381610038161003816119901+ 119864( sup
1199050le119905le1199051
|119872 (119905)|)
+ 119901 (119901 + 1)119870119864int
1199051
1199050
(1 + |119909 (119904)|119901) 119889119904
(100)
At the same time applying the well-known Burkholder-Davis-Gundy inequality (see eg [2]) gives
119864( sup1199050le119905le1199051
|119872 (119905)|)
le 3119864(int
1199051
1199050
1199012|119909 (119904)|
2119901minus21003816100381610038161003816119892 (119909 (119904) 119904 1205742 (119904))10038161003816100381610038162119889119904)
05
le 3119864( sup1199050le119905le1199051
|119909 (119905)|119901
timesint
1199051
1199050
1199012|119909 (119904)|
119901minus21003816100381610038161003816119892 (119909 (119904) 119904 1205742(119904))
10038161003816100381610038162119889119904)
05
le 05119864( sup1199050le119905le1199051
|119909 (119905)|119901) + 45119864
times int
1199051
1199050
1199012119870|119909 (119904)|
119901minus2(1 + |119909 (119904)|
2) 119889119904
le 05119864( sup1199050le119905le1199051
|119909 (119905)|119901) + 9119901
2119870119864int
1199051
1199050
(1 + |119909 (119904)|119901) 119889119904
(101)
10 Journal of Applied Mathematics
Substituting the above inequality into (100) gives
119864( sup1199050le119905le1199051
|119909 (119905)|119901) le 2119864
100381610038161003816100381611990901003816100381610038161003816119901
+ 2119901 (10119901 + 1)119870int
1199051
1199050
(1 + |119909 (119904)|119901) 119889119904
(102)
Thus
1 + 119864( sup1199050le119905le1199051
|119909 (119905)|119901)
le 1 + 211986410038161003816100381610038161199090
1003816100381610038161003816119901+ 2119901 (10119901 + 1)119870
times int
1199051
1199050
(1 + sup1199050le119905le119904
|119909 (119905)|119901)119889119904
(103)
Then the required assertion follows from the Gronwallinequality
Up to now we have discussed the 119871119901-estimates for the
solution in the case when 119901 ge 2 As for 0 lt 119901 lt 2 the similarresults can be given without any difficulty as long as we notethat the Holder inequality implies
119864|119909 (119905)|119901le [119864|119909 (119905)|
2]05119901
(104)
4 Example
Consider the following Black-Scholes model
119889119883 (119905) = 120583 (1205741(119905))119883 (119905) 119889119905 + ] (120574
2(119905))119883 (119905) 119889119861 (119905) (105)
where 1205741(119905) is a right-continuous homogenous Markovian
chain taking values in finite state spaces S1= 1 2 and 120574
2(119905)
is a right-continuous homogenous Markovian chain takingvalues in finite state spaces S
2= 1 2 3 120583(119894) = 119894 119894 = 1 2
](119895) = 119895 + 1 119895 = 1 2 3 Taking 119870 = 16 119870 = 9 then (42)and (43) hold Therefore by Theorem 7 (105) has a uniquesolution
5 Conclusions and Further Research
This paper is devoted to studying the existence and unique-ness of solution of SDEs with multi-Markovian switchingsand estimating the119901thmoment of the solutionWe have usedtwo continuous-time Markovian chains to model the SDEsThis area is becoming increasingly useful in engineeringeconomics communication theory active networking andso forth The sufficient criteria for existence and unique-ness of solution local solution and maximal local solutionwere established Those results indicate that (8) keeps manyproperties that (89) owns At the same time although thehypothesis (H1) is used in this paper wewant to point out thatthis hypothesis is not essential In fact (H1) can be replacedby the following generalized hypothesis
(H1)1015840 both 1205741(119905) and 120574
2(119905) are right-continuous homoge-
nous Markovian chains such that 120574(119905) = (1205741(119905) 1205742(119905)) is a
homogenous vector chainUnder hypothesis (H1)1015840 the results given in this paper
can be established similarly It is easy to see that if 1205741(119905) equiv
1205742(119905) and 120574
1(119905) is a right-continuous homogenous Markovian
chain then (H1)1015840 is fulfilled immediately At the same timeif 1205741(119905) equiv 120574
2(119905) (8) will reduce to the classical SDEs with
single Markovian chain that is to say the classical theoryabout SDEs with single Markovian chain is a special caseof our theory On the other hand many theorems in thispaper will play important roles in further study For exampleTheorem 15 will be useful when one studies the approximatesolutions
Some important and interesting questions can be furtherinvestigated using the results in this paper For exampleapproximate solutions boundedness and stability stochas-tic functional differential equations with vector Markovianswitching and their applications In particular the stability of(8) is one of the most important and interesting topics andthose investigations are in progress
Acknowledgments
The authors thank the editor and referees for their veryimportant and helpful comments and suggestions Theauthors also thank Dr C Zhang for helping them to improvethe English exposition This research is supported by theNSFC of China (nos 11171081 and 11171056)
References
[1] A V SkorokhodAsymptoticMethods in theTheory of StochasticDifferential Equations vol 78 of Translations of MathematicalMonographs American Mathematical Society Providence RIUSA 1989
[2] X Mao and C Yuan Stochastic Differential Euations withMarkovian Switching Imperial College Press London UK2006
[3] G Yin and X Y Zhou ldquoMarkowitzrsquos mean-variance portfolioselection with regime switching from discrete-time models totheir continuous-time limitsrdquo IEEE Transactions on AutomaticControl vol 49 no 3 pp 349ndash360 2004
[4] J Buffington and R J Elliott ldquoAmerican options with regimeswitchingrdquo International Journal of Theoretical and AppliedFinance vol 5 no 5 pp 497ndash514 2002
[5] C Zhu and G Yin ldquoOn competitive Lotka-Volterra model inrandom environmentsrdquo Journal of Mathematical Analysis andApplications vol 357 no 1 pp 154ndash170 2009
[6] Q Luo and X Mao ldquoStochastic population dynamics underregime switchingrdquo Journal of Mathematical Analysis and Appli-cations vol 334 no 1 pp 69ndash84 2007
[7] X Li D Jiang and X Mao ldquoPopulation dynamical behaviorof Lotka-Volterra system under regime switchingrdquo Journal ofComputational and Applied Mathematics vol 232 no 2 pp427ndash448 2009
[8] X Li A Gray D Jiang and X Mao ldquoSufficient and neces-sary conditions of stochastic permanence and extinction forstochastic logistic populations under regime switchingrdquo Journal
Journal of Applied Mathematics 11
ofMathematical Analysis and Applications vol 376 no 1 pp 11ndash28 2011
[9] M Liu and KWang ldquoAsymptotic properties and simulations ofa stochastic logistic model under regime switchingrdquoMathemat-ical and Computer Modelling vol 54 no 9-10 pp 2139ndash21542011
[10] G Hu and K Wang ldquoStability in distribution of competi-tive Lotka-Volterra system with Markovian switchingrdquo AppliedMathematical Modelling vol 35 no 7 pp 3189ndash3200 2011
[11] M Liu and K Wang ldquoAsymptotic properties and simulationsof a stochastic logistic model under regime switching IIrdquoMathematical andComputerModelling vol 55 no 3-4 pp 405ndash418 2012
[12] ZWu H Huang and LWang ldquoStochastic delay logistic modelunder regime switchingrdquo Abstract and Applied Analysis vol2012 Article ID 241702 26 pages 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
2 Journal of Applied Mathematics
For the sake of convenience we take the following two-dimensional competitive model as an example
1198891199091= 1199091[11990310
(120574 (119905)) minus 11988611
(120574 (119905)) 1199091 minus 11988612
(120574 (119905)) 1199092] 119889119905
+ 1205721(120574 (119905)) 119909
11198891198611(119905)
1198891199092= 1199092[11990320
(120574 (119905)) minus 11988621
(120574 (119905)) 1199091minus 11988622
(120574 (119905)) 1199092] 119889119905
+ 1205722(120574 (119905)) 11990921198891198612 (119905)
(6)
where 119909119894is the size of 119894th species at time 119905 119903
1198940(119895) represents
the growth rate of 119894th species in regime 119895 for 119894 = 1 2119895 isin S and 119861
1and 119861
2are independent standard Brownian
motions However there are many stochastic factors thataffect some coefficients intensely but have little impact onother coefficients in (6) For example suppose that thestochastic factor is rain falls and 119909
1is able to endure a damp
weather while 1199092is fond of a dry environment then the rain
falls will affect 1199092intensely but have little impact on 119909
1 Thus
a more appropriate model is governed by
1198891199091= 1199091[11990310
(12057410
(119905)) minus 11988611
(12057411
(119905)) 1199091minus 11988612
(12057412
(119905)) 1199092] 119889119905
+ 1205721(1205741 (119905)) 11990911198891198611 (119905)
1198891199092= 1199092[11990320
(12057420
(119905)) minus 11988621
(12057421
(119905)) 1199091minus 11988622
(12057422
(119905)) 1199092] 119889119905
+ 1205722(1205742 (119905)) 11990921198891198612 (119905)
(7)
where 120574119894119895(119905) and 120574
119896(119905) are right-continuous homogenous
Markovian chains taking values in finite state spaces S119894119895for
119894 = 1 2 119895 = 0 1 2 and S119896for 119896 = 1 2 respectively
Thus the above examples show that the study of thefollowing SDEs with multi-Markovian switchings is essentialand is of great importance fromboth theoretical and practicalpoints
119889119909 (119905) = 119891 (119909 (119905) 119905 1205741(119905)) 119889119905
+ 119892 (119909 (119905) 119905 1205742 (119905)) 119889119861 (119905) 1199050le 119905 le 119879
(8)
with initial conditions 119909(1199050) = 119909
0isin 1198712
F0
and 120574119894(1199050) where
120574119894(119905) is a right-continuous homogenous Markovian chain
on the probality space taking values in a finite state spaceS119894= 1 2 119873
119894 and isF
119905-adapted but independent of the
Brownian motion 119861119894(119905) 119894 = 1 2 and
119891 R119899times R+times S1997888rarr R
119899
119892 R119899times R+times S2997888rarr R
119899+119898
(9)
Equation (8) can be regarded as the result of the 1198731times 1198732
equations
119889119909 (119905) = 119891 (119909 (119905) 119905 119894) 119889119905
+ 119892 (119909 (119905) 119905 119895) 119889119861 (119905) 119894 isin S1 119895 isin S
2
(10)
switching among each other according to the movement ofthe Markovian chains It is important for us to discover the
properties of the system (8) and to find out whether thepresence of two Markovian switchings affects some knownresults The first step and the foundation of those studies areto establish the theorems for the existence and uniqueness ofthe solution to system (8) So in this paper we will give sometheorems for the existence and uniqueness of the solution tosystem (8) and study some properties of this solution Thetheory developed in this paper is the foundation for furtherstudy and can be applied in many different and complicatedsituations and hence the importance of the results in thispaper is clear
It should be pointed out that the theory developed in thispaper can be generalized to cope with the more general SDEswith more Markovian chains
119889119909 (119905) = 119891 (119909 (119905) 119905 1205741 (119905) 120574119899 (119905)) 119889119905
+ 119892 (119909 (119905) 119905 120574119899+1
(119905) 120574119899+119898
(119905)) 119889119861 (119905)
(11)
The reason we concentrate on (8) rather than (11) is to avoidthe notations becoming too complicated Once the theorydeveloped in this paper is established the reader should beable to cope with the more general (11) without any difficulty
The remaining part of this paper is as follows In Section 2the sufficient criteria for existence anduniqueness of solutionlocal solution andmaximal local solution will be establishedrespectively In Section 3 the 119871
119875-estimates of the solutionwill be given In Section 4 we will introduce an example toillustrate our main result Finally we will close the paper withconclusions in Section 5
2 SDEs with Markovian Chains
Throughout this paper let (ΩF F119905119905isin119877+
P) be a completeprobability space Let 119861(119905) = (119861
1(119905) 119861
119898(119905))119879 be an 119898-
dimensional Brownian motion defined on the probabilityspace
In this section we will consider (8) Let 120574(119905) = (1205741(119905)
1205742(119905)) We impose a hypothesis(H1) 120574
1(119905) is independent of 120574
2(119905)
Then 120574(119905) is a homogenous vector Markovian chain withtransition probabilities
119875 120574 (119904 + 119905) = (1198951 1198952) | 120574 (119904) = (119894
1 1198942)
= 119875 120574 (119905) = (1198951 1198952) | 120574 (0) = (119894
1 1198942)
= 1198751198941119894211989511198952
(119905)
(12)
where (1198941 1198942) (1198951 1198952) isin S = (1 1) (1 2) (1119873
2) (2 1)
(1198731 1198732)
Now we will prepare some lemmas which are importantfor further study
Lemma 1 1198751198941119894211989511198952
(119905) has the following properties
(i) 1198751198941119894211989511198952
(119905) ge 0 for (1198941 1198942) (1198951 1198952) isin S
(ii) sum(11989511198952)isinS 1198751198941119894211989511198952
(119905) = 1 for (1198941 1198942) isin S
(iii) 1198751198941119894211989511198952
(0) = 12057511989411198951
sdot 12057511989421198952
where 120575119894119896119895119896
= 1 if 119894119896
= 119895119896
otherwise 120575119894119896119895119896
= 0 119896 = 1 2
Journal of Applied Mathematics 3
(iv) (Chapman-Kolmogorov equation) For 119904 119905 ge 0 and(1198941 1198942) (1198951 1198952) isin S
1198751198941119894211989511198952
(119904 + 119905)
= sum
(1198961 1198962)isinS
1198751198941119894211989611198962
(119905) 1198751198961119896211989511198952
(119904) (13)
Proof The proofs of (i) (ii) and (iii) are obvious Now let usprove (iv)
1198751198941119894211989511198952
(119904 + 119905)
= 119875 120574 (119904 + 119905) = (1198951 1198952) | 120574 (0) = (119894
1 1198942)
= sum
(11989611198962)isinS
119875 120574 (119904 + 119905) = (1198951 1198952)
120574 (119905) = (1198961 1198962) | 120574 (0) = (119894
1 1198942)
= sum
(11989611198962)isinS
(119875 120574 (119904 + 119905) = (1198951 1198952)
120574 (119905) = (1198961 1198962) 120574 (0) = (119894
1 1198942)
times(119875 120574 (119905) = (1198961 1198962) 120574 (0) = (119894
1 1198942))minus1
)
times119875 120574 (119905) = (119896
1 1198962) 120574 (0) = (119894
1 1198942)
119875 120574 (0) = (1198941 1198942)
= sum
(11989611198962)isinS
119875 120574 (119904 + 119905) = (1198951 1198952) | 120574 (119905) = (119896
1 1198962)
120574 (0) = (1198941 1198942)
times 119875 120574 (119905) = (1198961 1198962) | 120574 (0) = (119894
1 1198942)
= sum
(1198961 1198962)isinS
119875 120574 (119904 + 119905) = (1198951 1198952) | 120574 (119905) = (119896
1 1198962)
times 119875 120574 (119905) = (1198961 1198962) | 120574 (0) = (119894
1 1198942)
= sum
(1198961 1198962)isinS
1198751198961119896211989511198952
(119904) 1198751198941119894211989611198962
(119905)
(14)
This completes the proof
Now we impose another hypothesis which is calledstandard condition
(H2) lim119905rarr0
1198751198941119894211989511198952
(119905) = 1205751198941119894211989511198952
Lemma 2 Under Assumption (H2) for all 119905 ge 0 (1198941 1198942) isin S
one has 1198751198941119894211989411198942
(119905) gt 0
Proof From 1198751198941119894211989411198942
(0) gt 0 and (H2) we know that forarbitrary fixed 119905 gt 0 we have
1198751198941119894211989411198942
(119905
119899) gt 0 (15)
for sufficient large 119899 Then making use of Chapman-Kolm-ogorov equation
1198751198941119894211989411198942
(119904 + 119905) = sum
(1198961 1198962)isinS
11987511989411198942119896111198962
(119905) 1198751198961119896211989411198942
(119904)
ge 1198751198941119894211989411198942
(119905) 1198751198941119894211989411198942
(119904)
(16)
gives
1198751198941119894211989411198942
(119905) ge (1198751198941119894211989411198942
(119905
119899))
119899
gt 0 (17)
which is the desired assertion
Lemma 3 Under Assumption (H2) for all (1198941 1198942) isin S
minus1199021198941119894211989411198942
= lim119905rarr0
1 minus 1198751198941119894211989411198942
(119905)
119905(18)
exists (but may be infin)
Proof Define 120601(119905) = minus ln1198751198941119894211989411198942
(119905) ge 0 Then making use of(16) gives
120601 (119904 + 119905) le 120601 (119904) + 120601 (119905) (19)
Set minus1199021198941119894211989411198942
= sup119905gt0
(120601(119905)119905) It is easy to see that
0 le minus1199021198941119894211989411198942
le infin lim sup119905rarr0
120601 (119905)
119905le minus1199021198941119894211989411198942
(20)
Now we will assert
lim inf119905rarr0
120601 (119905)
119905ge minus1199021198941119894211989411198942
(21)
In fact for 0 lt ℎ lt 119905 exist119899 0 le 120576 le ℎ such that 119905 = 119899ℎ + 120576Applying (19) yields
120601 (119905)
119905ge
119899ℎ
120576
120601 (ℎ)
ℎ+
120601 (120576)
119905 (22)
Note that 120576 rarr 0 119899ℎ120576 rarr 1 120601(120576) rarr 0 whenever ℎ rarr 0+
then for all 119905 gt 0 we have
120601 (119905)
119905le lim infℎrarr0
120601 (ℎ)
ℎ (23)
This implies minus1199021198941119894211989411198942
le lim inf119905rarr0
(120601(119905)119905) Thus
minus1199021198941119894211989411198942
= lim119905rarr0
120601 (119905)
119905 (24)
Using the definition of 120601(119905) gives
lim119905rarr0
1 minus 1198751198941119894211989411198942
(119905)
119905= lim119905rarr0
1 minus exp 120601 (119905)
120601 (119905)
120601 (119905)
119905= minus1199021198941119894211989411198942
(25)
which is the required assertion
4 Journal of Applied Mathematics
Lemma 4 Under Assumption (H2) for (1198941 1198942) (1198951 1198952) isin
S (1198941 1198942) = (1198951 1198952)
1199021198941119894211989511198952
= 1198751015840
1198941119894211989511198952
(0) = lim119905rarr0
1198751198941119894211989511198952
(119905)
119905(26)
exists and is finite
Proof By (H2) we note that for all 0 lt 120576 lt 13 exist0 lt 120575 lt 1such that
1198751198941119894211989411198942
(119905) gt 1 minus 120576 1198751198951119895211989511198952
(119905) gt 1 minus 120576
1198751198951119895211989411198942
(119905) lt 120576
(27)
provided 0 lt 119905 le 120575For forall0 le ℎ lt 119905 set 119899 = ⟨119905ℎ⟩ where ⟨119886⟩ = max
119899le119886119899 isin
Z Let
119875(11989511198952)
1198941119894211989611198962
(ℎ) = 1198751198941119894211989611198962
(ℎ)
119875(11989511198952)
1198941119894211989611198962
(119898ℎ)
= sum
(11990311199032) = (1198951 1198952)
119875(11989511198952)
1198941119894211990311199032
((119898 minus 1) ℎ) 1198751199031119903211989611198962
(ℎ)
(28)
where 119875(11989511198952)
1198941119894211989611198962
(119898ℎ)means that the probability of the 120574(119905)willnot reach to (119895
1 1198952) at times ℎ 2ℎ (119898 minus 1)ℎ but will reach
to (1198961 1198962) at time 119898ℎ Note that if ℎ le 119905 le 120575 then
120576 gt 1 minus 1198751198941119894211989411198942
(119905)
= sum
(1198961 1198962) = (1198941 1198942)
1198751198941119894211989611198962
(119905) ge 1198751198941119894211989511198952
(119905)
ge
119899
sum
119898=1
119875(11989511198952)
1198941119894211989511198952
(119898ℎ) 1198751198951119895211989511198952
(119905 minus 119898ℎ)
ge (1 minus 120576)
119899
sum
119898=1
119875(11989511198952)
1198941119894211989511198952
(119898ℎ)
(29)
which indicates119899
sum
119898=1
119875(11989511198952)
1198941119894211989511198952
(119898ℎ) le120576
1 minus 120576 (30)
Then making use of
1198751198941119894211989411198942
(119898ℎ) = 119875(11989511198952)
1198941119894211989411198942
(119898ℎ)
+
119898minus1
sum
119897=1
119875(11989511198952)
1198941119894211989511198952
(119897ℎ) 1198751198951119895211989411198942
((119898 minus 119897) ℎ)
(31)
we obtain
119875(11989511198952)
1198941119894211989411198942
(119898ℎ) ge 1198751198941119894211989411198942
(119898ℎ) minus
119898minus1
sum
119897=1
119875(11989511198952)
1198941119894211989511198952
(119897ℎ)
ge 1 minus 120576 minus120576
1 minus 120576
(32)
Consequently
1198751198941119894211989511198952
(119905)
gt
119899
sum
119898=1
119875(11989511198952)
1198941119894211989411198942
((119898 minus 1) ℎ) 1198751198941119894211989511198952
(ℎ) 1198751198951119895211989511198952
(119905 minus 119898ℎ)
ge 119899 (1 minus 120576 minus120576
1 minus 120576)1198751198941119894211989511198952
(ℎ) (1 minus 120576)
ge 119899 (1 minus 3120576) 1198751198941119894211989511198952
(ℎ)
(33)
Dividing both sides of the above inequality by ℎ and noting119899ℎ rarr 119905 whenever ℎ rarr 0 yield
lim supℎrarr0
1198751198941119894211989511198952
(ℎ)
ℎle
1
1 minus 3120576
1198751198941119894211989511198952
(119905)
119905lt infin (34)
Then letting 119905 rarr 0 gives
lim supℎrarr0
1198751198941119894211989511198952
(ℎ)
ℎle
1
1 minus 3120576lim inf119905rarr0
1198751198941119894211989511198952
(119905)
119905 (35)
and the required assertion follows immediately by letting120576 rarr 0 This completes the proof
Set 120574(119905) = (1205741(119905) 1205742(119905)) then it is easy to see that
almost every sample path of 120574(119905) is a right continuous stepfunction Now letting P(119905) = (119875
1198941119894211989511198952
(119905))11987311198732times11987311198732
Q =
(1199021198941119894211989511198952
)11987311198732times11987311198732
= P1015840(0) Then by Chapman-Kolmogorovequation
P (119905 + ℎ) = P (119905)P (ℎ) = P (ℎ)P (119905) (36)
we have
P (119905 + ℎ) minus P (119905)
ℎ= P (119905) [
P (ℎ) minus Iℎ
] = [P (ℎ) minus I
ℎ]P (119905)
(37)
Letting ℎ rarr 0 and taking limits give
P1015840(119905) = P (119905)Q
P1015840(119905) = QP (119905)
(38)
Note that
P (0) = I (39)
Then by solving the ordinary differential equations (38) and(39) we obtain the following lemma
Lemma 5 For P(119905) and Q one has
P (119905) = exp Q119905 (40)
We are now in the position to give the sufficient condi-tions for the existence and uniqueness of the solution of (8)For this end let us first give the definition of the solution
Journal of Applied Mathematics 5
Definition 6 An R119899-valued stochastic process 119909(119905)1199050le119905le119879
iscalled a solution of (8) if it has the following properties
(i) 119909(119905)1199050le119905le119879
is continuous andF119905-adapted
(ii) 119891(119909(119905) 119905 1205741(119905))1199050le119905le119879
isinL1([1199050 119879]R119899)while 119892(119909(119905)
119905 1205742(119905))1199050le119905le119879
isin L2([1199050 119879]R119899times119898)
(iii) for any 119905 isin [1199050 119879] equation
119909 (119905) = 119909 (1199050) + int
119905
1199050
119891 (119909 (119904) 119904 1205741(119904)) 119889119904
+ int
119905
1199050
119892 (119909 (119904) 119904 1205742(119904)) 119889119861 (119904)
(41)
holds with probability 1
A solution 119909(119905)1199050le119905le119879
is said to be unique if any othersolution (119905)
1199050le119905le119879
is indistinguishable from 119909(119905)1199050le119905le119879
Now we can give our main results in this section
Theorem 7 Assume that there exist two positive constants 119870
and 119870 such that(Lipschitz condition) for all 119909 119910 isin R119899 119905 isin [119905
0 119879] and
(1198941 1198942) isin S
1003816100381610038161003816119891 (119909 119905 1198941) minus 119891 (119910 119905 119894
1)10038161003816100381610038162
or1003816100381610038161003816119892 (119909 119905 119894
2) minus 119892 (119910 119905 119894
2)10038161003816100381610038162le 119870
1003816100381610038161003816119909 minus 11991010038161003816100381610038162
(42)
(Linear growth condition) for all (119909 119905 (1198941 1198942)) isin R119899 times
[1199050 119879] times S
1003816100381610038161003816119891 (119909 119905 1198941)10038161003816100381610038162or
1003816100381610038161003816119892 (119909 119905 1198942)10038161003816100381610038162le 119870(1 + |119909|)
2 (43)
Then there exists a unique solution 119909(119905) to (8) and moreoverthe solution obeys
119864( sup1199050le119905le119879
|119909 (119905)|2)
le (1 + 311986410038161003816100381610038161199090
10038161003816100381610038162) exp 3119870 (119879 minus 119905
0) (119879 minus 119905
0+ 4)
(44)
Proof Recall that almost every sample path of 120574(119905) =
(1205741(119905) 1205742(119905)) is a right continuous step function with a finite
number of jumps on [1199050 119879] Thus there exists a sequence of
stopping times 120591119896119896ge0
such that
(i) for almost every 120596 isin Ω there is a finite 119896120596for 1199050
=
1205910lt 1205911lt sdot sdot sdot lt 120591
119896= 119879 and 120591
119896= 119879 if 119896 gt 119896
120596
(ii) both 1205741(sdot) and 120574
2(sdot) in 120574(sdot) are constants on interval
[[120591119896 120591119896+1
[[ namely
1205741 (119905) = 120574
1(120591119896) on 120591
119896le 119905 le 120591
119896+1forall119896 ge 0
1205742(119905) = 120574
2(120591119896) on 120591
119896le 119905 le 120591
119896+1forall119896 ge 0
(45)
First of all let us consider (8) on 119905 isin [[1205910 1205911[[ then (8)
becomes119889119909 (119905) = 119891 (119909 (119905) 119905 120574
1(1199050)) 119889119905
+ 119892 (119909 (119905) 119905 1205742(1199050)) 119889119861 (119905)
(46)
with initial conditions 119909(1199050) = 119909
0 120574(1199050) = (120574
1(1199050) 1205741(1199050))
Then by the theory of SDEs we obtain that (46) has a uniquesolution which obeys 119909(120591
1) isin 1198712
F1205911
(ΩR119899) We next consider(8) on 119905 isin [[120591
1 1205912[[ which becomes
119889119909 (119905) = 119891 (119909 (119905) 119905 1205741 (1205911)) 119889119905
+ 119892 (119909 (119905) 119905 1205742(1205911)) 119889119861 (119905)
(47)
Again by the theory of SDEs (47) has a unique solutionwhichobeys 119909(120591
1) isin 119871
2
F1205912
(ΩR119899) Repeating this procedure weconclude that (8) has a unique solution 119909(119905) on [119905
0 119879]
Now let us prove (44) For every 119896 ge 1 define thestopping time
120591119896= 119879 and inf 119905 isin [119905
0 119879] |119909 (119905)| ge 119896 (48)
It is obvious that 120591119896uarr 119879 as Set 119909
119896(119905) = 119909(119905and120591
119896) for 119905 isin [119905
0 119879]
Then 119909119896(119905) obeys the equation
119909119896(119905) = 119909
0+ int
119905
1199050
119891 (119909119896(119904) 119904 120574
1(119904)) 119868[[1199050120591119896]](119904) 119889119904
+ int
119905
1199050
119892 (119909119896(119904) 119904 120574
2(119904)) 119868[[1199050120591119896]](119904) 119889119861 (119904)
(49)
Making use of the elementary inequality |119886+119887+119888|2le 3(|119886|
2+
|119887|2+ |119888|2) the Holder inequality and (43) we can see that
1003816100381610038161003816119909119896 (119905)10038161003816100381610038162= 3
1003816100381610038161003816119909010038161003816100381610038162+ 3119870 (119879 minus 119905
0) int
119905
1199050
(1 +1003816100381610038161003816119909119896 (119904)
10038161003816100381610038162) 119889119904
+ 3
100381610038161003816100381610038161003816100381610038161003816
int
119905
1199050
119892 (119909119896(119904) 119904 120574
2(119904)) 119868[[1199050120591119896]](119904) 119889119861 (119904)
100381610038161003816100381610038161003816100381610038161003816
2
(50)
Thus applying the Doob martingale inequality and (43) wecan further show that
119864( sup1199050le119904le119905
1003816100381610038161003816119909119896 (119904)10038161003816100381610038162)
= 311986410038161003816100381610038161199090
10038161003816100381610038162+ 3119870 (119879 minus 119905
0) int
119905
1199050
(1 + 1198641003816100381610038161003816119909119896 (119904)
10038161003816100381610038162) 119889119904
+ 12119864int
119905
1199050
1003816100381610038161003816119892 (119909119896(119904) 119904 120574
2(119904))
10038161003816100381610038162119868[[1199050120591119896]](119904) 119889119904
le 311986410038161003816100381610038161199090
10038161003816100381610038162+ 3119870 (119879 minus 119905
0+ 4)int
119905
1199050
(1 + 1198641003816100381610038161003816119909119896 (119904)
10038161003816100381610038162) 119889119904
(51)
That is to say
1 + 119864( sup1199050le119904le119905
1003816100381610038161003816119909119896 (119904)10038161003816100381610038162) le 1 + 3119864
1003816100381610038161003816119909010038161003816100381610038162+ 3119870 (119879 minus 119905
0+ 4)
times int
119905
1199050
[1 + 119864 sup1199050le119903le119904
1003816100381610038161003816119909119896 (119903)10038161003816100381610038162]119889119904
(52)
6 Journal of Applied Mathematics
Using the Gronwall inequality leads to
1 + 119864( sup1199050le119904le119905
1003816100381610038161003816119909119896 (119904)10038161003816100381610038162)
le (1 + 311986410038161003816100381610038161199090
10038161003816100381610038162) exp 3119870 (119879 minus 119905
0) (119879 minus 119905
0+ 4)
(53)
Consequently
1 + 119864( sup1199050le119905le120591119896
|119909 (119905)|2)
le (1 + 311986410038161003816100381610038161199090
10038161003816100381610038162) exp 3119870 (119879 minus 119905
0) (119879 minus 119905
0+ 4)
(54)
Then the required inequality (44) follows immediately byletting 119896 rarr infin
Condition (42) indicates that the coefficients 119891(119909 119905 1198941)
and 119892(119909 119905 1198942) do not change faster than a linear function of
119909 as change in 119909 This means in particular the continuity of119891(119909 119905 119894
1) and 119892(119909 119905 119894
2) in 119909 for all 119905 isin [119905
0 119879] Then functions
that are discontinuous with respect to 119909 are excluded as thecoefficients Besides there are many functions that do notsatisfy the Lipschitz conditionThese imply that the Lipschitzcondition is too restrictive To improve this Lipschitz condi-tion let us introduce the concept of local solution
Definition 8 Let120590infinbe a stopping time such that 119905
0le 120590infin
le 119879
as An R119899-valuedF119905-adapted continuous stochastic process
119909(119905)1199050le119905lt120590infin
is called a local solution of (8) if 119909(1199050) = 1199090and
moreover there is a nondecreasing sequence 120590119896119896ge1
such that1199050le 120590119896 uarr 120590infin
as and
119909 (119905) = 119909 (1199050) + int
119905and120590119896
1199050
119891 (119909 (119904) 119904 1205741(119904)) 119889119904
+ int
119905and120590119896
1199050
119892 (119909 (119904) 119904 1205742(119904)) 119889119861 (119904)
(55)
holds for any 119905 isin [1199050 119879) and 119896 ge 1 with probability one If
furthermore
lim sup119905rarr+120590
infin
|119909 (119905)| = infin whenever 120590infin
lt 119879 (56)
then it is called a maximal local solution and 120590infinis called the
explosion time Amaximal local solution 119909(119905) 1199050le 119905 lt 120590
infin
is said to be unique if any other maximal local solution 119909(119905)
1199050le 119905 lt 120590
infin is indistinguishable from it namely 120590
infin= 120590infin
and 119909(119905) = 119909(119905) for 1199050le 119905 lt 120590
infinwith probability one
Definition 9 (local Lipschitz condition) For every integer119896 ge 1 there exists a positive constant ℎ
119896such that for all
119905 isin [1199050 119879] 119894 = (119894
1 1198942) isin S and those 119909 119910 isin R119899 with
|119909| or |119910| le 119896
1003816100381610038161003816119891 (119909 119905 1198941) minus 119891 (119910 119905 119894
1)10038161003816100381610038162
or1003816100381610038161003816119892 (119909 119905 119894
2) minus 119892 (119910 119905 119894
2)10038161003816100381610038162le ℎ119896
1003816100381610038161003816119909 minus 11991010038161003816100381610038162
(57)
The following theorem shows the existence of uniquemaximal local solution under the local Lipschitz conditionwithout the linear growth condition
Theorem 10 Under condition (57) there exists a uniquemaximal local solution of (8)
Proof Define functions
119891119896(119909 119905 119894
1) =
119891 (119909 119905 1198941) if |119909| le 119896
119891(119896119909
|119909| 119905 1198941) if |119909| gt 119896
119892119896(119909 119905 119894
2) =
119892 (119909 119905 1198942) if |119909| le 119896
119892 (119896119909
|119909| 119905 1198942) if |119909| gt 119896
(58)
Then 119891119896and 119892
119896satisfy the Lipschitz condition and the linear
growth condition Thus by Theorem 7 there is a uniquesolution 119909
119896(119905) of the equation
119889119909119896(119905) = 119891
119896(119909119896(119905) 119905 120574
1(119905)) 119889119905
+ 119892119896(119909119896(119905) 119905 120574
2(119905)) 119889119861 (119905)
(59)
with the initial conditions 119909119896(1199050) = 119909
0and 120574(119905
0) = (120574
1(1199050)
1205742(1199050)) Define the stopping times
120590119896= 119879 and inf 119905 isin [119905
0 119879]
1003816100381610038161003816119909119896 (119905)1003816100381610038161003816 ge 119896 (60)
Clearly if 1199050le 119905 le 120590
119896
119909119896(119905) = 119909
119896+1(119905) (61)
which indicates that 120590119896is increasing so 120590
119896has its limit 120590
infin=
lim119896rarr+infin
120590119896 Define 119909(119905) 119905
0le 119905 lt 120590
infin by
119909 (119905) = 119909119896(119905) 119905 isin [[120590
119896minus1 120590119896[[ 119896 ge 1 (62)
where 1205900= 1199050 Applying (61) one can show that 119909(119905 and 120590
119896) =
119909119896(119905 and 120590119896) It then follows from (59) that
119909 (119905 and 120590119896) = 1199090+ int
119905and120590119896
1199050
119891119896(119909 (119904) 119904 120574
1(119904)) 119889119904
+ int
119905and120590119896
1199050
119892119896(119909 (119904) 119904 1205742 (119904)) 119889119861 (119904)
= 1199090+ int
119905and120590119896
1199050
119891 (119909 (119904) 119904 1205741(119904)) 119889119904
+ int
119905and120590119896
1199050
119892 (119909 (119904) 119904 1205742 (119904)) 119889119861 (119904)
(63)
for any 119905 isin [1199050 119879) and 119896 ge 1 It is easy to see that if 120590
infinle 119879
then
lim sup119905rarr120590infin
|119909 (119905)|
ge lim sup119896rarr+infin
1003816100381610038161003816119909 (120590119896)1003816100381610038161003816 = lim sup119896rarr+infin
1003816100381610038161003816119909119896 (120590119896)1003816100381610038161003816 = infin
(64)
Therefore 119909(119905) 1199050le 119905 lt 120590
infin is a maximal local solution
Now we will prove the uniqueness Let 119909(119905) 1199050
le 119905 lt
120590infin
be another maximal local solution Define
120590119896= 120590infin
and inf 119905 isin [[1199050 120590infin
[[ 119909 (119905) ge 119896 (65)
Journal of Applied Mathematics 7
Then 120590119896
rarr 120590infin
as and
119875 119909 (119905) = 119909 (119905) 119905 isin [[1199050 120590119896and 120590119896]] = 1 forall119896 ge 1 (66)
Letting 119896 rarr infin gives
119875 119909 (119905) = 119909 (119905) 119905 isin [[1199050 120590infin
and 120590infin
[[ = 1 (67)
In order to complete the proof we need only to show that120590infin
= 120590infin
as In fact for almost every 120596 isin 120590infin
lt 120590infin
wehave
1003816100381610038161003816119909 (120590infin
120596)1003816100381610038161003816
= lim119896rarr+infin
1003816100381610038161003816119909 (120590119896 120596)
1003816100381610038161003816 = lim119896rarr+infin
1003816100381610038161003816119909 (120590119896 120596)
1003816100381610038161003816 = infin(68)
which contradicts the fact that 119909(119905 120596) is continuous on 119905 isin
[1199050 120590infin
(120596)) This implies 120590infin
ge 120590infin
as In the same way onecan show 120590
infinle 120590infinasThus wemust have 120590
infin= 120590infinasThis
completes the proof
In many situations we often consider an SDE on [1199050infin)
119889119909 (119905) = 119891 (119909 (119905) 119905 1205741(119905)) 119889119905
+ 119892 (119909 (119905) 119905 1205742(119905)) 119889119861 (119905) 119905
0le 119905 lt infin
(69)
with initial data 119909(1199050) = 119909
0and 120574(119905
0) If the assumption of
the existence-and-uniqueness theorem holds on every finitesubinterval [119905
0 119879] of [119905
0infin) then (69) has a unique solution
119909(119905) on the entire interval [1199050infin) Such a solution is called
a global solution To establish a more general result aboutglobal solution we need more notations To this end weintroduce an operator 119871119881 from R119899 times R
+times S to R which is
given by
119871119881 (119909 119905 (1198941 1198942))
= 119881119905(119909 119905 (119894
1 1198942)) + 119881
119909(119909 119905 (119894
1 1198942)) 119891 (119909 119905 119894
1)
+ 05 trace [119892119879 (119909 119905 1198942) 119881119909119909
(119909 119905 (1198941 1198942)) 119892 (119909 119905 119894
2)]
+ sum
(11989511198952)isinS
1199021198941119894211989511198952
119881 (119909 119905 (1198951 1198952))
(70)
where 119881(119909 119905 (1198941 1198942)) isin 119862
21(R119899 times R
+times SR
+) and
119881119905(119909 119905 (119894
1 1198942)) =
120597119881 (119909 119905 (1198941 1198942))
120597119905
119881119909(119909 119905 (119894
1 1198942))
= (120597119881 (119909 119905 (119894
1 1198942))
1205971199091
120597119881 (119909 119905 (119894
1 1198942))
120597119909119899
)
119881119909119909
(119909 119905 (1198941 1198942)) = (
1205972119881 (119909 119905 (119894
1 1198942))
120597119909119894120597119909119895
)
119899times119899
(71)
Theorem 11 Assume that the local Lipschitz condition (57)holds Assume also that there is a function 119881(119909 119905 (119894
1 1198942)) isin
11986221
(R119899 times R+times SR
+) and a constant 120579 gt 0 such that
lim|119909|rarrinfin
( inf(119905(1198941 1198942))isinR+timesS
119881 (119909 119905 (1198941 1198942))) = infin (72)
119871119881 (119909 119905 (1198941 1198942)) le 120579 (1 + 119881 (119909 119905 (119894
1 1198942)))
forall (119909 119905 (1198941 1198942)) isin R
119899times R+times S
(73)
Then there exists a unique global solution 119909(119905) to (69)
Proof We need only to prove the theorem for any initialcondition 119909
0isin R119899 and (120574
1(1199050) 1205742(1199050)) isin S FromTheorem 10
we know that the local Lipschitz condition guarantees theexistence of the unique maximal solution 119909(119905) on [119905
0 120590infin
)where 120590
infinis the explosion time We need only to show 120590
infin=
infin as If this is not true then we can find a pair of positiveconstants 120576 and 119879 such that
119875 120590infin
le 119879 gt 2120576 (74)
For each integer 119896 ge 1 define the stopping time
120590119896= inf 119905 ge 119905
0 |119909 (119905)| ge 119896 (75)
Since 120590119896
rarr 120590infinalmost surely we can find a sufficiently large
integer 1198960for
119875 120590119896le 119879 gt 120576 forall119896 ge 119896
0 (76)
Fix any 119896 ge 1198960 then for any 119905
0le 119905 le 119879 by virtue of the
generalized Ito formula (see eg [1])
119864119881 (119909 (119905 and 120590119896) 119905 and 120590
119896 120574 (119905 and 120590
119896))
= 119881 (1199090 1199050 1205740) + 119864int
119905and120590119896
1199050
119871119881 (119909 (119904) 119904 120574 (119904)) 119889119904
le 119881 (1199090 1199050 1205740) + 120579 (119879 minus 119905
0)
+ 120579int
119905
1199050
119864119881 (119909 (119904 and 120590119896) 119904 and 120590
119896 120574 (119904 and 120590
119896)) 119889119904
(77)
Making use of the Gronwall inequality gives
119864119881 (119909 (119905 and 120590119896) 119905 and 120590
119896 120574 (119905 and 120590
119896))
le [119881 (1199090 1199050 1205740) + 120579 (119879 minus 119905
0)] exp 120579 (119879 minus 119905
0)
(78)
Therefore
119864 (119868120590119896le119879
119881 (119909 (120590119896) 120590119896 120574 (120590119896)))
le [119881 (1199090 1199050 1205740) + 120579 (119879 minus 119905
0)] exp 120579 (119879 minus 119905
0)
(79)
At the same time set
119892119896= inf 119881 (119909 119905 (119894
1 1198942)) |119909| ge 119896 119905 isin [119905
0 119879] (119894
1 1198942) isin S
(80)
8 Journal of Applied Mathematics
Then (72) means 119892119896
rarr infin It follows from (76) and (79) that
[119881 (1199090 1199050 1205740) + 120579 (119879 minus 119905
0)] exp 120579 (119879 minus 119905
0)
ge 119892119896119875 120590119896le 119879 ge 120576119892
119896
(81)
Letting 119896 rarr infin yields a contradiction that is to say120590infin
= infinThe proof is complete
3 119871119875-Estimates
In the previous section we have investigated the existenceand uniqueness of the solution to (8) In this section as abovelet 119909(119905) 119905
0le 119905 le 119879 be the unique solution of (8) with initial
conditions 119909(1199050) = 1199090and 120574(119905
0) and we will estimate the 119901th
moment of the solution
Theorem 12 Assume that there is a function 119881(119909 119905 (1198941 1198942)) isin
11986221
(R119899 times R+times SR
+) and positive constants 119901 120578 120579 such that
for all (119909 119905 (1198941 1198942)) isin R119899 times R
+times S
120578|119909|119901le 119881 (119909 119905 (119894
1 1198942)) (82)
119871119881 (119909 119905 (1198941 1198942)) le 120579119881 (119909 119905 (119894
1 1198942)) (83)
Assume also the initial condition 119909(1199050) = 119909
0and 120574(119905
0) obeys
119864119881(1199090 1199050 120574(1199050)) lt infin then one has
119864|119909 (119905)|119901
le119864119881 (119909
0 1199050 120574 (1199050)) exp 120579 (119905 minus 119905
0)
120578 forall119905 isin [119905
0 119879]
(84)
Proof For each integer 119896 ge 1 define the stopping time
120590119896= 119879 and inf 119905 ge 119905
0 |119909 (119905)| ge 119896 (85)
Thus 120590119896
rarr 119879 as Using the generalized Itorsquos formula and(83) we obtain that for 119905 isin [119905
0 119879]
119864119881 (119909 (119905 and 120590119896) 119905 and 120590
119896 120574 (119905 and 120590
119896))
le 119864119881 (1199090 1199050 120574 (1199050))
+ 120579int
119905
1199050
119864119881 (119909 (119904 and 120590119896) 119904 and 120590
119896 120574 (119904 and 120590
119896)) 119889119904
(86)
Then the Gronwall inequality indicates
119864119881 (119909 (119905 and 120590119896) 119905 and 120590
119896 120574 (119905 and 120590
119896))
le 119864119881 (1199090 1199050 120574 (1199050)) exp 120579 (119905 minus 119905
0)
(87)
for all 119905 isin [1199050 119879] By virtue of condition (83) we obtain the
required assertion (84)
Corollary 13 Assume 119901 ge 2 and 1199090isin 119871119901
F1199050
(ΩR119899) Assumealso that there exists a constant 120579 gt 0 such that for all(119909 119905 (119894
1 1198942)) isin R119899 times R
+times S
119909119879119891 (119909 119905 119894
1) +
119901 minus 1
2
1003816100381610038161003816119892 (119909 119905 1198942)10038161003816100381610038162le 120579 (1 + |119909|
2) (88)
Then one has
119864|119909 (119905)|119901
le 2(119901minus2)2
(1 + 11986410038161003816100381610038161199090
1003816100381610038161003816119901) exp 119901120579 (119905 minus 119905
0) forall119905 isin [119905
0 119879]
(89)
Proof Define 119881(119909 119905 (1198941 1198942)) = (1 + |119909|
2)1199012 Making use of
(88) yields
119871119881 (119909 119905 (1198941 1198942))
= 119901(1 + |119909|2)(119901minus2)2
119909119879119891 (119909 119905 119894
1)
+119901
2(1 + |119909|
2)(119901minus2)21003816100381610038161003816119892 (119909 119905 119894
2)10038161003816100381610038162
+119901 (119901 minus 2)
2(1 + |119909|
2)(119901minus4)210038161003816100381610038161003816
119909119879119892 (119909 119905 119894
2)10038161003816100381610038161003816
2
le 119901(1 + |119909|2)(119901minus2)2
[119909119879119891 (119909 119905 119894
1) +
119901 minus 1
2
1003816100381610038161003816119892 (119909 119905 1198942)10038161003816100381610038162]
le 119901120579(1 + |119909|2)1199012
(90)
Then byTheorem 12 we get
119864(1 + |119909 (119905)|2)1199012
le 119864(1 +10038161003816100381610038161199090
10038161003816100381610038162)1199012
exp 119901120579 (119905 minus 1199050)
(91)
and the required assertion (89) followsIt is useful to point out that if the linear growth condition
(43) is satisfied then (88) is fulfilled with 120579 = radic119870 + 119870(119901 minus
1)2 Now we will show the other important properties of thesolution
Theorem 14 Let 119901 ge 2 and 1199090
isin 119871119901
F1199050
(ΩR119899) Assume alsothat the linear growth condition (43) holds Then one has
119864|119909 (119905) minus 119909 (119904)|119901le 119862(119905 minus 119904)
1199012 (92)
where
119862 = 2119901minus2
(1 + 11986410038161003816100381610038161199090
1003816100381610038161003816119901) exp 119901120579 (119879 minus 119905
0)
times ([2 (119879 minus 1199050)]1199012
+ [119901 (119901 minus 1)]1199012
)
(93)
and 120579 = radic119870+119870(119901minus1)2 Particularly the 119901th moment of thesolution is continuous on [119905
0 119879]
Journal of Applied Mathematics 9
Proof Applying the elementary inequality |119886 + 119887|119901
le
2119901minus1
(|119886|119901+ |119887|119901) the Holder inequality and the linear growth
condition we can derive that
119864|119909 (119905) minus 119909 (119904)|119901
le 2119901minus1
119864
10038161003816100381610038161003816100381610038161003816int
119905
119904
119891 (119909 (119906) 119906 1205741(119906)) 119889119906
10038161003816100381610038161003816100381610038161003816
119901
+ 2119901minus1
119864
10038161003816100381610038161003816100381610038161003816int
119905
119904
119892 (119909 (119906) 119906 1205742(119906)) 119889119861 (119906)
10038161003816100381610038161003816100381610038161003816
119901
le [2 (119905 minus 119904)]119901minus1
119864int
119905
119904
1003816100381610038161003816119891 (119909 (119906) 119906 1205741(119906))
1003816100381610038161003816119901119889119906
+ 05[2119901 (119901 minus 1)]1199012
(119905 minus 119904)(119901minus2)2
119864
times int
119905
119904
1003816100381610038161003816119892 (119909 (119906) 119906 1205742 (119906))1003816100381610038161003816119901119889119906
le 1198621(119905 minus 119904)
(119901minus2)2int
119905
119904
119864(1 + |119909 (119906)|2)1199012
119889119906
(94)
where1198621= 2(119901minus2)2
1198701199012
([2(119879minus 1199050)]1199012
+ [119901(119901minus1)]1199012
) Using(91) yields
119864|119909 (119905) minus 119909 (119904)|119901
le 1198621(119905 minus 119904)
(119901minus2)2
times int
119905
119904
2(119901minus2)2
(1 + 11986410038161003816100381610038161199090
1003816100381610038161003816119901) exp 119901120579 (119906 minus 119905
0) 119889119906
le 11986212(119901minus2)2
(1 + 11986410038161003816100381610038161199090
1003816100381610038161003816119901)
times exp 119901120579 (119879 minus 1199050) (119905 minus 119904)
(1199012)
(95)
which is the desired inequality
Theorem 15 Let 119901 ge 2 and 1199090
isin 119871119901
F1199050
(ΩR119899) Assume thatthere is a119870 gt 0 such that for all (119909 119905 (119894
1 1198942)) isin R119899times[119905
0 119879]timesS
119909119879 1003816100381610038161003816119891 (119909 119905 119894
1)1003816100381610038161003816 or
1003816100381610038161003816119892 (119909 119905 1198942)10038161003816100381610038162le 119870(1 + |119909|)
2 (96)
Then
119864( sup1199050le119905le119879
|119909 (119905)|119901)
le (1 + 211986410038161003816100381610038161199090
1003816100381610038161003816119901) exp 2119901 (10119901 + 1)119870 (119879 minus 119905
0)
(97)
Proof Making use of the generalized Itorsquos formula andcondition (96) we have
|119909 (119905)|119901
le10038161003816100381610038161199090
1003816100381610038161003816119901+ 119872(119905)
+ int
119905
1199050
119901|119909 (119904)|119901minus2
[119909119879(119904) 119891 (119909 (119904) 119904 120574
1(119904))
+119901 minus 1
2
1003816100381610038161003816119892 (119909 (119904) 119904 1205742 (119904))10038161003816100381610038162] 119889119904
le10038161003816100381610038161199090
1003816100381610038161003816119901+ 119872(119905) + 05119901 (119901 + 1)119870
times int
119905
1199050
|119909 (119904)|119901minus2
(1 + |119909 (119904)|2) 119889119904
le10038161003816100381610038161199090
1003816100381610038161003816119901+ 119872(119905) + 119901 (119901 + 1)119870int
119905
1199050
(1 + |119909 (119904)|119901) 119889119904
(98)
where
119872(119905) = int
119905
1199050
119901|119909 (119904)|119901minus2
119909119879(119904) 119892 (119909 (119904) 119904 1205742 (119904)) 119889119861 (119904) (99)
Therefore for any 1199051isin [1199050 119879]
119864( sup1199050le119905le1199051
|119909 (119905)|119901) le 119864
100381610038161003816100381611990901003816100381610038161003816119901+ 119864( sup
1199050le119905le1199051
|119872 (119905)|)
+ 119901 (119901 + 1)119870119864int
1199051
1199050
(1 + |119909 (119904)|119901) 119889119904
(100)
At the same time applying the well-known Burkholder-Davis-Gundy inequality (see eg [2]) gives
119864( sup1199050le119905le1199051
|119872 (119905)|)
le 3119864(int
1199051
1199050
1199012|119909 (119904)|
2119901minus21003816100381610038161003816119892 (119909 (119904) 119904 1205742 (119904))10038161003816100381610038162119889119904)
05
le 3119864( sup1199050le119905le1199051
|119909 (119905)|119901
timesint
1199051
1199050
1199012|119909 (119904)|
119901minus21003816100381610038161003816119892 (119909 (119904) 119904 1205742(119904))
10038161003816100381610038162119889119904)
05
le 05119864( sup1199050le119905le1199051
|119909 (119905)|119901) + 45119864
times int
1199051
1199050
1199012119870|119909 (119904)|
119901minus2(1 + |119909 (119904)|
2) 119889119904
le 05119864( sup1199050le119905le1199051
|119909 (119905)|119901) + 9119901
2119870119864int
1199051
1199050
(1 + |119909 (119904)|119901) 119889119904
(101)
10 Journal of Applied Mathematics
Substituting the above inequality into (100) gives
119864( sup1199050le119905le1199051
|119909 (119905)|119901) le 2119864
100381610038161003816100381611990901003816100381610038161003816119901
+ 2119901 (10119901 + 1)119870int
1199051
1199050
(1 + |119909 (119904)|119901) 119889119904
(102)
Thus
1 + 119864( sup1199050le119905le1199051
|119909 (119905)|119901)
le 1 + 211986410038161003816100381610038161199090
1003816100381610038161003816119901+ 2119901 (10119901 + 1)119870
times int
1199051
1199050
(1 + sup1199050le119905le119904
|119909 (119905)|119901)119889119904
(103)
Then the required assertion follows from the Gronwallinequality
Up to now we have discussed the 119871119901-estimates for the
solution in the case when 119901 ge 2 As for 0 lt 119901 lt 2 the similarresults can be given without any difficulty as long as we notethat the Holder inequality implies
119864|119909 (119905)|119901le [119864|119909 (119905)|
2]05119901
(104)
4 Example
Consider the following Black-Scholes model
119889119883 (119905) = 120583 (1205741(119905))119883 (119905) 119889119905 + ] (120574
2(119905))119883 (119905) 119889119861 (119905) (105)
where 1205741(119905) is a right-continuous homogenous Markovian
chain taking values in finite state spaces S1= 1 2 and 120574
2(119905)
is a right-continuous homogenous Markovian chain takingvalues in finite state spaces S
2= 1 2 3 120583(119894) = 119894 119894 = 1 2
](119895) = 119895 + 1 119895 = 1 2 3 Taking 119870 = 16 119870 = 9 then (42)and (43) hold Therefore by Theorem 7 (105) has a uniquesolution
5 Conclusions and Further Research
This paper is devoted to studying the existence and unique-ness of solution of SDEs with multi-Markovian switchingsand estimating the119901thmoment of the solutionWe have usedtwo continuous-time Markovian chains to model the SDEsThis area is becoming increasingly useful in engineeringeconomics communication theory active networking andso forth The sufficient criteria for existence and unique-ness of solution local solution and maximal local solutionwere established Those results indicate that (8) keeps manyproperties that (89) owns At the same time although thehypothesis (H1) is used in this paper wewant to point out thatthis hypothesis is not essential In fact (H1) can be replacedby the following generalized hypothesis
(H1)1015840 both 1205741(119905) and 120574
2(119905) are right-continuous homoge-
nous Markovian chains such that 120574(119905) = (1205741(119905) 1205742(119905)) is a
homogenous vector chainUnder hypothesis (H1)1015840 the results given in this paper
can be established similarly It is easy to see that if 1205741(119905) equiv
1205742(119905) and 120574
1(119905) is a right-continuous homogenous Markovian
chain then (H1)1015840 is fulfilled immediately At the same timeif 1205741(119905) equiv 120574
2(119905) (8) will reduce to the classical SDEs with
single Markovian chain that is to say the classical theoryabout SDEs with single Markovian chain is a special caseof our theory On the other hand many theorems in thispaper will play important roles in further study For exampleTheorem 15 will be useful when one studies the approximatesolutions
Some important and interesting questions can be furtherinvestigated using the results in this paper For exampleapproximate solutions boundedness and stability stochas-tic functional differential equations with vector Markovianswitching and their applications In particular the stability of(8) is one of the most important and interesting topics andthose investigations are in progress
Acknowledgments
The authors thank the editor and referees for their veryimportant and helpful comments and suggestions Theauthors also thank Dr C Zhang for helping them to improvethe English exposition This research is supported by theNSFC of China (nos 11171081 and 11171056)
References
[1] A V SkorokhodAsymptoticMethods in theTheory of StochasticDifferential Equations vol 78 of Translations of MathematicalMonographs American Mathematical Society Providence RIUSA 1989
[2] X Mao and C Yuan Stochastic Differential Euations withMarkovian Switching Imperial College Press London UK2006
[3] G Yin and X Y Zhou ldquoMarkowitzrsquos mean-variance portfolioselection with regime switching from discrete-time models totheir continuous-time limitsrdquo IEEE Transactions on AutomaticControl vol 49 no 3 pp 349ndash360 2004
[4] J Buffington and R J Elliott ldquoAmerican options with regimeswitchingrdquo International Journal of Theoretical and AppliedFinance vol 5 no 5 pp 497ndash514 2002
[5] C Zhu and G Yin ldquoOn competitive Lotka-Volterra model inrandom environmentsrdquo Journal of Mathematical Analysis andApplications vol 357 no 1 pp 154ndash170 2009
[6] Q Luo and X Mao ldquoStochastic population dynamics underregime switchingrdquo Journal of Mathematical Analysis and Appli-cations vol 334 no 1 pp 69ndash84 2007
[7] X Li D Jiang and X Mao ldquoPopulation dynamical behaviorof Lotka-Volterra system under regime switchingrdquo Journal ofComputational and Applied Mathematics vol 232 no 2 pp427ndash448 2009
[8] X Li A Gray D Jiang and X Mao ldquoSufficient and neces-sary conditions of stochastic permanence and extinction forstochastic logistic populations under regime switchingrdquo Journal
Journal of Applied Mathematics 11
ofMathematical Analysis and Applications vol 376 no 1 pp 11ndash28 2011
[9] M Liu and KWang ldquoAsymptotic properties and simulations ofa stochastic logistic model under regime switchingrdquoMathemat-ical and Computer Modelling vol 54 no 9-10 pp 2139ndash21542011
[10] G Hu and K Wang ldquoStability in distribution of competi-tive Lotka-Volterra system with Markovian switchingrdquo AppliedMathematical Modelling vol 35 no 7 pp 3189ndash3200 2011
[11] M Liu and K Wang ldquoAsymptotic properties and simulationsof a stochastic logistic model under regime switching IIrdquoMathematical andComputerModelling vol 55 no 3-4 pp 405ndash418 2012
[12] ZWu H Huang and LWang ldquoStochastic delay logistic modelunder regime switchingrdquo Abstract and Applied Analysis vol2012 Article ID 241702 26 pages 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Journal of Applied Mathematics 3
(iv) (Chapman-Kolmogorov equation) For 119904 119905 ge 0 and(1198941 1198942) (1198951 1198952) isin S
1198751198941119894211989511198952
(119904 + 119905)
= sum
(1198961 1198962)isinS
1198751198941119894211989611198962
(119905) 1198751198961119896211989511198952
(119904) (13)
Proof The proofs of (i) (ii) and (iii) are obvious Now let usprove (iv)
1198751198941119894211989511198952
(119904 + 119905)
= 119875 120574 (119904 + 119905) = (1198951 1198952) | 120574 (0) = (119894
1 1198942)
= sum
(11989611198962)isinS
119875 120574 (119904 + 119905) = (1198951 1198952)
120574 (119905) = (1198961 1198962) | 120574 (0) = (119894
1 1198942)
= sum
(11989611198962)isinS
(119875 120574 (119904 + 119905) = (1198951 1198952)
120574 (119905) = (1198961 1198962) 120574 (0) = (119894
1 1198942)
times(119875 120574 (119905) = (1198961 1198962) 120574 (0) = (119894
1 1198942))minus1
)
times119875 120574 (119905) = (119896
1 1198962) 120574 (0) = (119894
1 1198942)
119875 120574 (0) = (1198941 1198942)
= sum
(11989611198962)isinS
119875 120574 (119904 + 119905) = (1198951 1198952) | 120574 (119905) = (119896
1 1198962)
120574 (0) = (1198941 1198942)
times 119875 120574 (119905) = (1198961 1198962) | 120574 (0) = (119894
1 1198942)
= sum
(1198961 1198962)isinS
119875 120574 (119904 + 119905) = (1198951 1198952) | 120574 (119905) = (119896
1 1198962)
times 119875 120574 (119905) = (1198961 1198962) | 120574 (0) = (119894
1 1198942)
= sum
(1198961 1198962)isinS
1198751198961119896211989511198952
(119904) 1198751198941119894211989611198962
(119905)
(14)
This completes the proof
Now we impose another hypothesis which is calledstandard condition
(H2) lim119905rarr0
1198751198941119894211989511198952
(119905) = 1205751198941119894211989511198952
Lemma 2 Under Assumption (H2) for all 119905 ge 0 (1198941 1198942) isin S
one has 1198751198941119894211989411198942
(119905) gt 0
Proof From 1198751198941119894211989411198942
(0) gt 0 and (H2) we know that forarbitrary fixed 119905 gt 0 we have
1198751198941119894211989411198942
(119905
119899) gt 0 (15)
for sufficient large 119899 Then making use of Chapman-Kolm-ogorov equation
1198751198941119894211989411198942
(119904 + 119905) = sum
(1198961 1198962)isinS
11987511989411198942119896111198962
(119905) 1198751198961119896211989411198942
(119904)
ge 1198751198941119894211989411198942
(119905) 1198751198941119894211989411198942
(119904)
(16)
gives
1198751198941119894211989411198942
(119905) ge (1198751198941119894211989411198942
(119905
119899))
119899
gt 0 (17)
which is the desired assertion
Lemma 3 Under Assumption (H2) for all (1198941 1198942) isin S
minus1199021198941119894211989411198942
= lim119905rarr0
1 minus 1198751198941119894211989411198942
(119905)
119905(18)
exists (but may be infin)
Proof Define 120601(119905) = minus ln1198751198941119894211989411198942
(119905) ge 0 Then making use of(16) gives
120601 (119904 + 119905) le 120601 (119904) + 120601 (119905) (19)
Set minus1199021198941119894211989411198942
= sup119905gt0
(120601(119905)119905) It is easy to see that
0 le minus1199021198941119894211989411198942
le infin lim sup119905rarr0
120601 (119905)
119905le minus1199021198941119894211989411198942
(20)
Now we will assert
lim inf119905rarr0
120601 (119905)
119905ge minus1199021198941119894211989411198942
(21)
In fact for 0 lt ℎ lt 119905 exist119899 0 le 120576 le ℎ such that 119905 = 119899ℎ + 120576Applying (19) yields
120601 (119905)
119905ge
119899ℎ
120576
120601 (ℎ)
ℎ+
120601 (120576)
119905 (22)
Note that 120576 rarr 0 119899ℎ120576 rarr 1 120601(120576) rarr 0 whenever ℎ rarr 0+
then for all 119905 gt 0 we have
120601 (119905)
119905le lim infℎrarr0
120601 (ℎ)
ℎ (23)
This implies minus1199021198941119894211989411198942
le lim inf119905rarr0
(120601(119905)119905) Thus
minus1199021198941119894211989411198942
= lim119905rarr0
120601 (119905)
119905 (24)
Using the definition of 120601(119905) gives
lim119905rarr0
1 minus 1198751198941119894211989411198942
(119905)
119905= lim119905rarr0
1 minus exp 120601 (119905)
120601 (119905)
120601 (119905)
119905= minus1199021198941119894211989411198942
(25)
which is the required assertion
4 Journal of Applied Mathematics
Lemma 4 Under Assumption (H2) for (1198941 1198942) (1198951 1198952) isin
S (1198941 1198942) = (1198951 1198952)
1199021198941119894211989511198952
= 1198751015840
1198941119894211989511198952
(0) = lim119905rarr0
1198751198941119894211989511198952
(119905)
119905(26)
exists and is finite
Proof By (H2) we note that for all 0 lt 120576 lt 13 exist0 lt 120575 lt 1such that
1198751198941119894211989411198942
(119905) gt 1 minus 120576 1198751198951119895211989511198952
(119905) gt 1 minus 120576
1198751198951119895211989411198942
(119905) lt 120576
(27)
provided 0 lt 119905 le 120575For forall0 le ℎ lt 119905 set 119899 = ⟨119905ℎ⟩ where ⟨119886⟩ = max
119899le119886119899 isin
Z Let
119875(11989511198952)
1198941119894211989611198962
(ℎ) = 1198751198941119894211989611198962
(ℎ)
119875(11989511198952)
1198941119894211989611198962
(119898ℎ)
= sum
(11990311199032) = (1198951 1198952)
119875(11989511198952)
1198941119894211990311199032
((119898 minus 1) ℎ) 1198751199031119903211989611198962
(ℎ)
(28)
where 119875(11989511198952)
1198941119894211989611198962
(119898ℎ)means that the probability of the 120574(119905)willnot reach to (119895
1 1198952) at times ℎ 2ℎ (119898 minus 1)ℎ but will reach
to (1198961 1198962) at time 119898ℎ Note that if ℎ le 119905 le 120575 then
120576 gt 1 minus 1198751198941119894211989411198942
(119905)
= sum
(1198961 1198962) = (1198941 1198942)
1198751198941119894211989611198962
(119905) ge 1198751198941119894211989511198952
(119905)
ge
119899
sum
119898=1
119875(11989511198952)
1198941119894211989511198952
(119898ℎ) 1198751198951119895211989511198952
(119905 minus 119898ℎ)
ge (1 minus 120576)
119899
sum
119898=1
119875(11989511198952)
1198941119894211989511198952
(119898ℎ)
(29)
which indicates119899
sum
119898=1
119875(11989511198952)
1198941119894211989511198952
(119898ℎ) le120576
1 minus 120576 (30)
Then making use of
1198751198941119894211989411198942
(119898ℎ) = 119875(11989511198952)
1198941119894211989411198942
(119898ℎ)
+
119898minus1
sum
119897=1
119875(11989511198952)
1198941119894211989511198952
(119897ℎ) 1198751198951119895211989411198942
((119898 minus 119897) ℎ)
(31)
we obtain
119875(11989511198952)
1198941119894211989411198942
(119898ℎ) ge 1198751198941119894211989411198942
(119898ℎ) minus
119898minus1
sum
119897=1
119875(11989511198952)
1198941119894211989511198952
(119897ℎ)
ge 1 minus 120576 minus120576
1 minus 120576
(32)
Consequently
1198751198941119894211989511198952
(119905)
gt
119899
sum
119898=1
119875(11989511198952)
1198941119894211989411198942
((119898 minus 1) ℎ) 1198751198941119894211989511198952
(ℎ) 1198751198951119895211989511198952
(119905 minus 119898ℎ)
ge 119899 (1 minus 120576 minus120576
1 minus 120576)1198751198941119894211989511198952
(ℎ) (1 minus 120576)
ge 119899 (1 minus 3120576) 1198751198941119894211989511198952
(ℎ)
(33)
Dividing both sides of the above inequality by ℎ and noting119899ℎ rarr 119905 whenever ℎ rarr 0 yield
lim supℎrarr0
1198751198941119894211989511198952
(ℎ)
ℎle
1
1 minus 3120576
1198751198941119894211989511198952
(119905)
119905lt infin (34)
Then letting 119905 rarr 0 gives
lim supℎrarr0
1198751198941119894211989511198952
(ℎ)
ℎle
1
1 minus 3120576lim inf119905rarr0
1198751198941119894211989511198952
(119905)
119905 (35)
and the required assertion follows immediately by letting120576 rarr 0 This completes the proof
Set 120574(119905) = (1205741(119905) 1205742(119905)) then it is easy to see that
almost every sample path of 120574(119905) is a right continuous stepfunction Now letting P(119905) = (119875
1198941119894211989511198952
(119905))11987311198732times11987311198732
Q =
(1199021198941119894211989511198952
)11987311198732times11987311198732
= P1015840(0) Then by Chapman-Kolmogorovequation
P (119905 + ℎ) = P (119905)P (ℎ) = P (ℎ)P (119905) (36)
we have
P (119905 + ℎ) minus P (119905)
ℎ= P (119905) [
P (ℎ) minus Iℎ
] = [P (ℎ) minus I
ℎ]P (119905)
(37)
Letting ℎ rarr 0 and taking limits give
P1015840(119905) = P (119905)Q
P1015840(119905) = QP (119905)
(38)
Note that
P (0) = I (39)
Then by solving the ordinary differential equations (38) and(39) we obtain the following lemma
Lemma 5 For P(119905) and Q one has
P (119905) = exp Q119905 (40)
We are now in the position to give the sufficient condi-tions for the existence and uniqueness of the solution of (8)For this end let us first give the definition of the solution
Journal of Applied Mathematics 5
Definition 6 An R119899-valued stochastic process 119909(119905)1199050le119905le119879
iscalled a solution of (8) if it has the following properties
(i) 119909(119905)1199050le119905le119879
is continuous andF119905-adapted
(ii) 119891(119909(119905) 119905 1205741(119905))1199050le119905le119879
isinL1([1199050 119879]R119899)while 119892(119909(119905)
119905 1205742(119905))1199050le119905le119879
isin L2([1199050 119879]R119899times119898)
(iii) for any 119905 isin [1199050 119879] equation
119909 (119905) = 119909 (1199050) + int
119905
1199050
119891 (119909 (119904) 119904 1205741(119904)) 119889119904
+ int
119905
1199050
119892 (119909 (119904) 119904 1205742(119904)) 119889119861 (119904)
(41)
holds with probability 1
A solution 119909(119905)1199050le119905le119879
is said to be unique if any othersolution (119905)
1199050le119905le119879
is indistinguishable from 119909(119905)1199050le119905le119879
Now we can give our main results in this section
Theorem 7 Assume that there exist two positive constants 119870
and 119870 such that(Lipschitz condition) for all 119909 119910 isin R119899 119905 isin [119905
0 119879] and
(1198941 1198942) isin S
1003816100381610038161003816119891 (119909 119905 1198941) minus 119891 (119910 119905 119894
1)10038161003816100381610038162
or1003816100381610038161003816119892 (119909 119905 119894
2) minus 119892 (119910 119905 119894
2)10038161003816100381610038162le 119870
1003816100381610038161003816119909 minus 11991010038161003816100381610038162
(42)
(Linear growth condition) for all (119909 119905 (1198941 1198942)) isin R119899 times
[1199050 119879] times S
1003816100381610038161003816119891 (119909 119905 1198941)10038161003816100381610038162or
1003816100381610038161003816119892 (119909 119905 1198942)10038161003816100381610038162le 119870(1 + |119909|)
2 (43)
Then there exists a unique solution 119909(119905) to (8) and moreoverthe solution obeys
119864( sup1199050le119905le119879
|119909 (119905)|2)
le (1 + 311986410038161003816100381610038161199090
10038161003816100381610038162) exp 3119870 (119879 minus 119905
0) (119879 minus 119905
0+ 4)
(44)
Proof Recall that almost every sample path of 120574(119905) =
(1205741(119905) 1205742(119905)) is a right continuous step function with a finite
number of jumps on [1199050 119879] Thus there exists a sequence of
stopping times 120591119896119896ge0
such that
(i) for almost every 120596 isin Ω there is a finite 119896120596for 1199050
=
1205910lt 1205911lt sdot sdot sdot lt 120591
119896= 119879 and 120591
119896= 119879 if 119896 gt 119896
120596
(ii) both 1205741(sdot) and 120574
2(sdot) in 120574(sdot) are constants on interval
[[120591119896 120591119896+1
[[ namely
1205741 (119905) = 120574
1(120591119896) on 120591
119896le 119905 le 120591
119896+1forall119896 ge 0
1205742(119905) = 120574
2(120591119896) on 120591
119896le 119905 le 120591
119896+1forall119896 ge 0
(45)
First of all let us consider (8) on 119905 isin [[1205910 1205911[[ then (8)
becomes119889119909 (119905) = 119891 (119909 (119905) 119905 120574
1(1199050)) 119889119905
+ 119892 (119909 (119905) 119905 1205742(1199050)) 119889119861 (119905)
(46)
with initial conditions 119909(1199050) = 119909
0 120574(1199050) = (120574
1(1199050) 1205741(1199050))
Then by the theory of SDEs we obtain that (46) has a uniquesolution which obeys 119909(120591
1) isin 1198712
F1205911
(ΩR119899) We next consider(8) on 119905 isin [[120591
1 1205912[[ which becomes
119889119909 (119905) = 119891 (119909 (119905) 119905 1205741 (1205911)) 119889119905
+ 119892 (119909 (119905) 119905 1205742(1205911)) 119889119861 (119905)
(47)
Again by the theory of SDEs (47) has a unique solutionwhichobeys 119909(120591
1) isin 119871
2
F1205912
(ΩR119899) Repeating this procedure weconclude that (8) has a unique solution 119909(119905) on [119905
0 119879]
Now let us prove (44) For every 119896 ge 1 define thestopping time
120591119896= 119879 and inf 119905 isin [119905
0 119879] |119909 (119905)| ge 119896 (48)
It is obvious that 120591119896uarr 119879 as Set 119909
119896(119905) = 119909(119905and120591
119896) for 119905 isin [119905
0 119879]
Then 119909119896(119905) obeys the equation
119909119896(119905) = 119909
0+ int
119905
1199050
119891 (119909119896(119904) 119904 120574
1(119904)) 119868[[1199050120591119896]](119904) 119889119904
+ int
119905
1199050
119892 (119909119896(119904) 119904 120574
2(119904)) 119868[[1199050120591119896]](119904) 119889119861 (119904)
(49)
Making use of the elementary inequality |119886+119887+119888|2le 3(|119886|
2+
|119887|2+ |119888|2) the Holder inequality and (43) we can see that
1003816100381610038161003816119909119896 (119905)10038161003816100381610038162= 3
1003816100381610038161003816119909010038161003816100381610038162+ 3119870 (119879 minus 119905
0) int
119905
1199050
(1 +1003816100381610038161003816119909119896 (119904)
10038161003816100381610038162) 119889119904
+ 3
100381610038161003816100381610038161003816100381610038161003816
int
119905
1199050
119892 (119909119896(119904) 119904 120574
2(119904)) 119868[[1199050120591119896]](119904) 119889119861 (119904)
100381610038161003816100381610038161003816100381610038161003816
2
(50)
Thus applying the Doob martingale inequality and (43) wecan further show that
119864( sup1199050le119904le119905
1003816100381610038161003816119909119896 (119904)10038161003816100381610038162)
= 311986410038161003816100381610038161199090
10038161003816100381610038162+ 3119870 (119879 minus 119905
0) int
119905
1199050
(1 + 1198641003816100381610038161003816119909119896 (119904)
10038161003816100381610038162) 119889119904
+ 12119864int
119905
1199050
1003816100381610038161003816119892 (119909119896(119904) 119904 120574
2(119904))
10038161003816100381610038162119868[[1199050120591119896]](119904) 119889119904
le 311986410038161003816100381610038161199090
10038161003816100381610038162+ 3119870 (119879 minus 119905
0+ 4)int
119905
1199050
(1 + 1198641003816100381610038161003816119909119896 (119904)
10038161003816100381610038162) 119889119904
(51)
That is to say
1 + 119864( sup1199050le119904le119905
1003816100381610038161003816119909119896 (119904)10038161003816100381610038162) le 1 + 3119864
1003816100381610038161003816119909010038161003816100381610038162+ 3119870 (119879 minus 119905
0+ 4)
times int
119905
1199050
[1 + 119864 sup1199050le119903le119904
1003816100381610038161003816119909119896 (119903)10038161003816100381610038162]119889119904
(52)
6 Journal of Applied Mathematics
Using the Gronwall inequality leads to
1 + 119864( sup1199050le119904le119905
1003816100381610038161003816119909119896 (119904)10038161003816100381610038162)
le (1 + 311986410038161003816100381610038161199090
10038161003816100381610038162) exp 3119870 (119879 minus 119905
0) (119879 minus 119905
0+ 4)
(53)
Consequently
1 + 119864( sup1199050le119905le120591119896
|119909 (119905)|2)
le (1 + 311986410038161003816100381610038161199090
10038161003816100381610038162) exp 3119870 (119879 minus 119905
0) (119879 minus 119905
0+ 4)
(54)
Then the required inequality (44) follows immediately byletting 119896 rarr infin
Condition (42) indicates that the coefficients 119891(119909 119905 1198941)
and 119892(119909 119905 1198942) do not change faster than a linear function of
119909 as change in 119909 This means in particular the continuity of119891(119909 119905 119894
1) and 119892(119909 119905 119894
2) in 119909 for all 119905 isin [119905
0 119879] Then functions
that are discontinuous with respect to 119909 are excluded as thecoefficients Besides there are many functions that do notsatisfy the Lipschitz conditionThese imply that the Lipschitzcondition is too restrictive To improve this Lipschitz condi-tion let us introduce the concept of local solution
Definition 8 Let120590infinbe a stopping time such that 119905
0le 120590infin
le 119879
as An R119899-valuedF119905-adapted continuous stochastic process
119909(119905)1199050le119905lt120590infin
is called a local solution of (8) if 119909(1199050) = 1199090and
moreover there is a nondecreasing sequence 120590119896119896ge1
such that1199050le 120590119896 uarr 120590infin
as and
119909 (119905) = 119909 (1199050) + int
119905and120590119896
1199050
119891 (119909 (119904) 119904 1205741(119904)) 119889119904
+ int
119905and120590119896
1199050
119892 (119909 (119904) 119904 1205742(119904)) 119889119861 (119904)
(55)
holds for any 119905 isin [1199050 119879) and 119896 ge 1 with probability one If
furthermore
lim sup119905rarr+120590
infin
|119909 (119905)| = infin whenever 120590infin
lt 119879 (56)
then it is called a maximal local solution and 120590infinis called the
explosion time Amaximal local solution 119909(119905) 1199050le 119905 lt 120590
infin
is said to be unique if any other maximal local solution 119909(119905)
1199050le 119905 lt 120590
infin is indistinguishable from it namely 120590
infin= 120590infin
and 119909(119905) = 119909(119905) for 1199050le 119905 lt 120590
infinwith probability one
Definition 9 (local Lipschitz condition) For every integer119896 ge 1 there exists a positive constant ℎ
119896such that for all
119905 isin [1199050 119879] 119894 = (119894
1 1198942) isin S and those 119909 119910 isin R119899 with
|119909| or |119910| le 119896
1003816100381610038161003816119891 (119909 119905 1198941) minus 119891 (119910 119905 119894
1)10038161003816100381610038162
or1003816100381610038161003816119892 (119909 119905 119894
2) minus 119892 (119910 119905 119894
2)10038161003816100381610038162le ℎ119896
1003816100381610038161003816119909 minus 11991010038161003816100381610038162
(57)
The following theorem shows the existence of uniquemaximal local solution under the local Lipschitz conditionwithout the linear growth condition
Theorem 10 Under condition (57) there exists a uniquemaximal local solution of (8)
Proof Define functions
119891119896(119909 119905 119894
1) =
119891 (119909 119905 1198941) if |119909| le 119896
119891(119896119909
|119909| 119905 1198941) if |119909| gt 119896
119892119896(119909 119905 119894
2) =
119892 (119909 119905 1198942) if |119909| le 119896
119892 (119896119909
|119909| 119905 1198942) if |119909| gt 119896
(58)
Then 119891119896and 119892
119896satisfy the Lipschitz condition and the linear
growth condition Thus by Theorem 7 there is a uniquesolution 119909
119896(119905) of the equation
119889119909119896(119905) = 119891
119896(119909119896(119905) 119905 120574
1(119905)) 119889119905
+ 119892119896(119909119896(119905) 119905 120574
2(119905)) 119889119861 (119905)
(59)
with the initial conditions 119909119896(1199050) = 119909
0and 120574(119905
0) = (120574
1(1199050)
1205742(1199050)) Define the stopping times
120590119896= 119879 and inf 119905 isin [119905
0 119879]
1003816100381610038161003816119909119896 (119905)1003816100381610038161003816 ge 119896 (60)
Clearly if 1199050le 119905 le 120590
119896
119909119896(119905) = 119909
119896+1(119905) (61)
which indicates that 120590119896is increasing so 120590
119896has its limit 120590
infin=
lim119896rarr+infin
120590119896 Define 119909(119905) 119905
0le 119905 lt 120590
infin by
119909 (119905) = 119909119896(119905) 119905 isin [[120590
119896minus1 120590119896[[ 119896 ge 1 (62)
where 1205900= 1199050 Applying (61) one can show that 119909(119905 and 120590
119896) =
119909119896(119905 and 120590119896) It then follows from (59) that
119909 (119905 and 120590119896) = 1199090+ int
119905and120590119896
1199050
119891119896(119909 (119904) 119904 120574
1(119904)) 119889119904
+ int
119905and120590119896
1199050
119892119896(119909 (119904) 119904 1205742 (119904)) 119889119861 (119904)
= 1199090+ int
119905and120590119896
1199050
119891 (119909 (119904) 119904 1205741(119904)) 119889119904
+ int
119905and120590119896
1199050
119892 (119909 (119904) 119904 1205742 (119904)) 119889119861 (119904)
(63)
for any 119905 isin [1199050 119879) and 119896 ge 1 It is easy to see that if 120590
infinle 119879
then
lim sup119905rarr120590infin
|119909 (119905)|
ge lim sup119896rarr+infin
1003816100381610038161003816119909 (120590119896)1003816100381610038161003816 = lim sup119896rarr+infin
1003816100381610038161003816119909119896 (120590119896)1003816100381610038161003816 = infin
(64)
Therefore 119909(119905) 1199050le 119905 lt 120590
infin is a maximal local solution
Now we will prove the uniqueness Let 119909(119905) 1199050
le 119905 lt
120590infin
be another maximal local solution Define
120590119896= 120590infin
and inf 119905 isin [[1199050 120590infin
[[ 119909 (119905) ge 119896 (65)
Journal of Applied Mathematics 7
Then 120590119896
rarr 120590infin
as and
119875 119909 (119905) = 119909 (119905) 119905 isin [[1199050 120590119896and 120590119896]] = 1 forall119896 ge 1 (66)
Letting 119896 rarr infin gives
119875 119909 (119905) = 119909 (119905) 119905 isin [[1199050 120590infin
and 120590infin
[[ = 1 (67)
In order to complete the proof we need only to show that120590infin
= 120590infin
as In fact for almost every 120596 isin 120590infin
lt 120590infin
wehave
1003816100381610038161003816119909 (120590infin
120596)1003816100381610038161003816
= lim119896rarr+infin
1003816100381610038161003816119909 (120590119896 120596)
1003816100381610038161003816 = lim119896rarr+infin
1003816100381610038161003816119909 (120590119896 120596)
1003816100381610038161003816 = infin(68)
which contradicts the fact that 119909(119905 120596) is continuous on 119905 isin
[1199050 120590infin
(120596)) This implies 120590infin
ge 120590infin
as In the same way onecan show 120590
infinle 120590infinasThus wemust have 120590
infin= 120590infinasThis
completes the proof
In many situations we often consider an SDE on [1199050infin)
119889119909 (119905) = 119891 (119909 (119905) 119905 1205741(119905)) 119889119905
+ 119892 (119909 (119905) 119905 1205742(119905)) 119889119861 (119905) 119905
0le 119905 lt infin
(69)
with initial data 119909(1199050) = 119909
0and 120574(119905
0) If the assumption of
the existence-and-uniqueness theorem holds on every finitesubinterval [119905
0 119879] of [119905
0infin) then (69) has a unique solution
119909(119905) on the entire interval [1199050infin) Such a solution is called
a global solution To establish a more general result aboutglobal solution we need more notations To this end weintroduce an operator 119871119881 from R119899 times R
+times S to R which is
given by
119871119881 (119909 119905 (1198941 1198942))
= 119881119905(119909 119905 (119894
1 1198942)) + 119881
119909(119909 119905 (119894
1 1198942)) 119891 (119909 119905 119894
1)
+ 05 trace [119892119879 (119909 119905 1198942) 119881119909119909
(119909 119905 (1198941 1198942)) 119892 (119909 119905 119894
2)]
+ sum
(11989511198952)isinS
1199021198941119894211989511198952
119881 (119909 119905 (1198951 1198952))
(70)
where 119881(119909 119905 (1198941 1198942)) isin 119862
21(R119899 times R
+times SR
+) and
119881119905(119909 119905 (119894
1 1198942)) =
120597119881 (119909 119905 (1198941 1198942))
120597119905
119881119909(119909 119905 (119894
1 1198942))
= (120597119881 (119909 119905 (119894
1 1198942))
1205971199091
120597119881 (119909 119905 (119894
1 1198942))
120597119909119899
)
119881119909119909
(119909 119905 (1198941 1198942)) = (
1205972119881 (119909 119905 (119894
1 1198942))
120597119909119894120597119909119895
)
119899times119899
(71)
Theorem 11 Assume that the local Lipschitz condition (57)holds Assume also that there is a function 119881(119909 119905 (119894
1 1198942)) isin
11986221
(R119899 times R+times SR
+) and a constant 120579 gt 0 such that
lim|119909|rarrinfin
( inf(119905(1198941 1198942))isinR+timesS
119881 (119909 119905 (1198941 1198942))) = infin (72)
119871119881 (119909 119905 (1198941 1198942)) le 120579 (1 + 119881 (119909 119905 (119894
1 1198942)))
forall (119909 119905 (1198941 1198942)) isin R
119899times R+times S
(73)
Then there exists a unique global solution 119909(119905) to (69)
Proof We need only to prove the theorem for any initialcondition 119909
0isin R119899 and (120574
1(1199050) 1205742(1199050)) isin S FromTheorem 10
we know that the local Lipschitz condition guarantees theexistence of the unique maximal solution 119909(119905) on [119905
0 120590infin
)where 120590
infinis the explosion time We need only to show 120590
infin=
infin as If this is not true then we can find a pair of positiveconstants 120576 and 119879 such that
119875 120590infin
le 119879 gt 2120576 (74)
For each integer 119896 ge 1 define the stopping time
120590119896= inf 119905 ge 119905
0 |119909 (119905)| ge 119896 (75)
Since 120590119896
rarr 120590infinalmost surely we can find a sufficiently large
integer 1198960for
119875 120590119896le 119879 gt 120576 forall119896 ge 119896
0 (76)
Fix any 119896 ge 1198960 then for any 119905
0le 119905 le 119879 by virtue of the
generalized Ito formula (see eg [1])
119864119881 (119909 (119905 and 120590119896) 119905 and 120590
119896 120574 (119905 and 120590
119896))
= 119881 (1199090 1199050 1205740) + 119864int
119905and120590119896
1199050
119871119881 (119909 (119904) 119904 120574 (119904)) 119889119904
le 119881 (1199090 1199050 1205740) + 120579 (119879 minus 119905
0)
+ 120579int
119905
1199050
119864119881 (119909 (119904 and 120590119896) 119904 and 120590
119896 120574 (119904 and 120590
119896)) 119889119904
(77)
Making use of the Gronwall inequality gives
119864119881 (119909 (119905 and 120590119896) 119905 and 120590
119896 120574 (119905 and 120590
119896))
le [119881 (1199090 1199050 1205740) + 120579 (119879 minus 119905
0)] exp 120579 (119879 minus 119905
0)
(78)
Therefore
119864 (119868120590119896le119879
119881 (119909 (120590119896) 120590119896 120574 (120590119896)))
le [119881 (1199090 1199050 1205740) + 120579 (119879 minus 119905
0)] exp 120579 (119879 minus 119905
0)
(79)
At the same time set
119892119896= inf 119881 (119909 119905 (119894
1 1198942)) |119909| ge 119896 119905 isin [119905
0 119879] (119894
1 1198942) isin S
(80)
8 Journal of Applied Mathematics
Then (72) means 119892119896
rarr infin It follows from (76) and (79) that
[119881 (1199090 1199050 1205740) + 120579 (119879 minus 119905
0)] exp 120579 (119879 minus 119905
0)
ge 119892119896119875 120590119896le 119879 ge 120576119892
119896
(81)
Letting 119896 rarr infin yields a contradiction that is to say120590infin
= infinThe proof is complete
3 119871119875-Estimates
In the previous section we have investigated the existenceand uniqueness of the solution to (8) In this section as abovelet 119909(119905) 119905
0le 119905 le 119879 be the unique solution of (8) with initial
conditions 119909(1199050) = 1199090and 120574(119905
0) and we will estimate the 119901th
moment of the solution
Theorem 12 Assume that there is a function 119881(119909 119905 (1198941 1198942)) isin
11986221
(R119899 times R+times SR
+) and positive constants 119901 120578 120579 such that
for all (119909 119905 (1198941 1198942)) isin R119899 times R
+times S
120578|119909|119901le 119881 (119909 119905 (119894
1 1198942)) (82)
119871119881 (119909 119905 (1198941 1198942)) le 120579119881 (119909 119905 (119894
1 1198942)) (83)
Assume also the initial condition 119909(1199050) = 119909
0and 120574(119905
0) obeys
119864119881(1199090 1199050 120574(1199050)) lt infin then one has
119864|119909 (119905)|119901
le119864119881 (119909
0 1199050 120574 (1199050)) exp 120579 (119905 minus 119905
0)
120578 forall119905 isin [119905
0 119879]
(84)
Proof For each integer 119896 ge 1 define the stopping time
120590119896= 119879 and inf 119905 ge 119905
0 |119909 (119905)| ge 119896 (85)
Thus 120590119896
rarr 119879 as Using the generalized Itorsquos formula and(83) we obtain that for 119905 isin [119905
0 119879]
119864119881 (119909 (119905 and 120590119896) 119905 and 120590
119896 120574 (119905 and 120590
119896))
le 119864119881 (1199090 1199050 120574 (1199050))
+ 120579int
119905
1199050
119864119881 (119909 (119904 and 120590119896) 119904 and 120590
119896 120574 (119904 and 120590
119896)) 119889119904
(86)
Then the Gronwall inequality indicates
119864119881 (119909 (119905 and 120590119896) 119905 and 120590
119896 120574 (119905 and 120590
119896))
le 119864119881 (1199090 1199050 120574 (1199050)) exp 120579 (119905 minus 119905
0)
(87)
for all 119905 isin [1199050 119879] By virtue of condition (83) we obtain the
required assertion (84)
Corollary 13 Assume 119901 ge 2 and 1199090isin 119871119901
F1199050
(ΩR119899) Assumealso that there exists a constant 120579 gt 0 such that for all(119909 119905 (119894
1 1198942)) isin R119899 times R
+times S
119909119879119891 (119909 119905 119894
1) +
119901 minus 1
2
1003816100381610038161003816119892 (119909 119905 1198942)10038161003816100381610038162le 120579 (1 + |119909|
2) (88)
Then one has
119864|119909 (119905)|119901
le 2(119901minus2)2
(1 + 11986410038161003816100381610038161199090
1003816100381610038161003816119901) exp 119901120579 (119905 minus 119905
0) forall119905 isin [119905
0 119879]
(89)
Proof Define 119881(119909 119905 (1198941 1198942)) = (1 + |119909|
2)1199012 Making use of
(88) yields
119871119881 (119909 119905 (1198941 1198942))
= 119901(1 + |119909|2)(119901minus2)2
119909119879119891 (119909 119905 119894
1)
+119901
2(1 + |119909|
2)(119901minus2)21003816100381610038161003816119892 (119909 119905 119894
2)10038161003816100381610038162
+119901 (119901 minus 2)
2(1 + |119909|
2)(119901minus4)210038161003816100381610038161003816
119909119879119892 (119909 119905 119894
2)10038161003816100381610038161003816
2
le 119901(1 + |119909|2)(119901minus2)2
[119909119879119891 (119909 119905 119894
1) +
119901 minus 1
2
1003816100381610038161003816119892 (119909 119905 1198942)10038161003816100381610038162]
le 119901120579(1 + |119909|2)1199012
(90)
Then byTheorem 12 we get
119864(1 + |119909 (119905)|2)1199012
le 119864(1 +10038161003816100381610038161199090
10038161003816100381610038162)1199012
exp 119901120579 (119905 minus 1199050)
(91)
and the required assertion (89) followsIt is useful to point out that if the linear growth condition
(43) is satisfied then (88) is fulfilled with 120579 = radic119870 + 119870(119901 minus
1)2 Now we will show the other important properties of thesolution
Theorem 14 Let 119901 ge 2 and 1199090
isin 119871119901
F1199050
(ΩR119899) Assume alsothat the linear growth condition (43) holds Then one has
119864|119909 (119905) minus 119909 (119904)|119901le 119862(119905 minus 119904)
1199012 (92)
where
119862 = 2119901minus2
(1 + 11986410038161003816100381610038161199090
1003816100381610038161003816119901) exp 119901120579 (119879 minus 119905
0)
times ([2 (119879 minus 1199050)]1199012
+ [119901 (119901 minus 1)]1199012
)
(93)
and 120579 = radic119870+119870(119901minus1)2 Particularly the 119901th moment of thesolution is continuous on [119905
0 119879]
Journal of Applied Mathematics 9
Proof Applying the elementary inequality |119886 + 119887|119901
le
2119901minus1
(|119886|119901+ |119887|119901) the Holder inequality and the linear growth
condition we can derive that
119864|119909 (119905) minus 119909 (119904)|119901
le 2119901minus1
119864
10038161003816100381610038161003816100381610038161003816int
119905
119904
119891 (119909 (119906) 119906 1205741(119906)) 119889119906
10038161003816100381610038161003816100381610038161003816
119901
+ 2119901minus1
119864
10038161003816100381610038161003816100381610038161003816int
119905
119904
119892 (119909 (119906) 119906 1205742(119906)) 119889119861 (119906)
10038161003816100381610038161003816100381610038161003816
119901
le [2 (119905 minus 119904)]119901minus1
119864int
119905
119904
1003816100381610038161003816119891 (119909 (119906) 119906 1205741(119906))
1003816100381610038161003816119901119889119906
+ 05[2119901 (119901 minus 1)]1199012
(119905 minus 119904)(119901minus2)2
119864
times int
119905
119904
1003816100381610038161003816119892 (119909 (119906) 119906 1205742 (119906))1003816100381610038161003816119901119889119906
le 1198621(119905 minus 119904)
(119901minus2)2int
119905
119904
119864(1 + |119909 (119906)|2)1199012
119889119906
(94)
where1198621= 2(119901minus2)2
1198701199012
([2(119879minus 1199050)]1199012
+ [119901(119901minus1)]1199012
) Using(91) yields
119864|119909 (119905) minus 119909 (119904)|119901
le 1198621(119905 minus 119904)
(119901minus2)2
times int
119905
119904
2(119901minus2)2
(1 + 11986410038161003816100381610038161199090
1003816100381610038161003816119901) exp 119901120579 (119906 minus 119905
0) 119889119906
le 11986212(119901minus2)2
(1 + 11986410038161003816100381610038161199090
1003816100381610038161003816119901)
times exp 119901120579 (119879 minus 1199050) (119905 minus 119904)
(1199012)
(95)
which is the desired inequality
Theorem 15 Let 119901 ge 2 and 1199090
isin 119871119901
F1199050
(ΩR119899) Assume thatthere is a119870 gt 0 such that for all (119909 119905 (119894
1 1198942)) isin R119899times[119905
0 119879]timesS
119909119879 1003816100381610038161003816119891 (119909 119905 119894
1)1003816100381610038161003816 or
1003816100381610038161003816119892 (119909 119905 1198942)10038161003816100381610038162le 119870(1 + |119909|)
2 (96)
Then
119864( sup1199050le119905le119879
|119909 (119905)|119901)
le (1 + 211986410038161003816100381610038161199090
1003816100381610038161003816119901) exp 2119901 (10119901 + 1)119870 (119879 minus 119905
0)
(97)
Proof Making use of the generalized Itorsquos formula andcondition (96) we have
|119909 (119905)|119901
le10038161003816100381610038161199090
1003816100381610038161003816119901+ 119872(119905)
+ int
119905
1199050
119901|119909 (119904)|119901minus2
[119909119879(119904) 119891 (119909 (119904) 119904 120574
1(119904))
+119901 minus 1
2
1003816100381610038161003816119892 (119909 (119904) 119904 1205742 (119904))10038161003816100381610038162] 119889119904
le10038161003816100381610038161199090
1003816100381610038161003816119901+ 119872(119905) + 05119901 (119901 + 1)119870
times int
119905
1199050
|119909 (119904)|119901minus2
(1 + |119909 (119904)|2) 119889119904
le10038161003816100381610038161199090
1003816100381610038161003816119901+ 119872(119905) + 119901 (119901 + 1)119870int
119905
1199050
(1 + |119909 (119904)|119901) 119889119904
(98)
where
119872(119905) = int
119905
1199050
119901|119909 (119904)|119901minus2
119909119879(119904) 119892 (119909 (119904) 119904 1205742 (119904)) 119889119861 (119904) (99)
Therefore for any 1199051isin [1199050 119879]
119864( sup1199050le119905le1199051
|119909 (119905)|119901) le 119864
100381610038161003816100381611990901003816100381610038161003816119901+ 119864( sup
1199050le119905le1199051
|119872 (119905)|)
+ 119901 (119901 + 1)119870119864int
1199051
1199050
(1 + |119909 (119904)|119901) 119889119904
(100)
At the same time applying the well-known Burkholder-Davis-Gundy inequality (see eg [2]) gives
119864( sup1199050le119905le1199051
|119872 (119905)|)
le 3119864(int
1199051
1199050
1199012|119909 (119904)|
2119901minus21003816100381610038161003816119892 (119909 (119904) 119904 1205742 (119904))10038161003816100381610038162119889119904)
05
le 3119864( sup1199050le119905le1199051
|119909 (119905)|119901
timesint
1199051
1199050
1199012|119909 (119904)|
119901minus21003816100381610038161003816119892 (119909 (119904) 119904 1205742(119904))
10038161003816100381610038162119889119904)
05
le 05119864( sup1199050le119905le1199051
|119909 (119905)|119901) + 45119864
times int
1199051
1199050
1199012119870|119909 (119904)|
119901minus2(1 + |119909 (119904)|
2) 119889119904
le 05119864( sup1199050le119905le1199051
|119909 (119905)|119901) + 9119901
2119870119864int
1199051
1199050
(1 + |119909 (119904)|119901) 119889119904
(101)
10 Journal of Applied Mathematics
Substituting the above inequality into (100) gives
119864( sup1199050le119905le1199051
|119909 (119905)|119901) le 2119864
100381610038161003816100381611990901003816100381610038161003816119901
+ 2119901 (10119901 + 1)119870int
1199051
1199050
(1 + |119909 (119904)|119901) 119889119904
(102)
Thus
1 + 119864( sup1199050le119905le1199051
|119909 (119905)|119901)
le 1 + 211986410038161003816100381610038161199090
1003816100381610038161003816119901+ 2119901 (10119901 + 1)119870
times int
1199051
1199050
(1 + sup1199050le119905le119904
|119909 (119905)|119901)119889119904
(103)
Then the required assertion follows from the Gronwallinequality
Up to now we have discussed the 119871119901-estimates for the
solution in the case when 119901 ge 2 As for 0 lt 119901 lt 2 the similarresults can be given without any difficulty as long as we notethat the Holder inequality implies
119864|119909 (119905)|119901le [119864|119909 (119905)|
2]05119901
(104)
4 Example
Consider the following Black-Scholes model
119889119883 (119905) = 120583 (1205741(119905))119883 (119905) 119889119905 + ] (120574
2(119905))119883 (119905) 119889119861 (119905) (105)
where 1205741(119905) is a right-continuous homogenous Markovian
chain taking values in finite state spaces S1= 1 2 and 120574
2(119905)
is a right-continuous homogenous Markovian chain takingvalues in finite state spaces S
2= 1 2 3 120583(119894) = 119894 119894 = 1 2
](119895) = 119895 + 1 119895 = 1 2 3 Taking 119870 = 16 119870 = 9 then (42)and (43) hold Therefore by Theorem 7 (105) has a uniquesolution
5 Conclusions and Further Research
This paper is devoted to studying the existence and unique-ness of solution of SDEs with multi-Markovian switchingsand estimating the119901thmoment of the solutionWe have usedtwo continuous-time Markovian chains to model the SDEsThis area is becoming increasingly useful in engineeringeconomics communication theory active networking andso forth The sufficient criteria for existence and unique-ness of solution local solution and maximal local solutionwere established Those results indicate that (8) keeps manyproperties that (89) owns At the same time although thehypothesis (H1) is used in this paper wewant to point out thatthis hypothesis is not essential In fact (H1) can be replacedby the following generalized hypothesis
(H1)1015840 both 1205741(119905) and 120574
2(119905) are right-continuous homoge-
nous Markovian chains such that 120574(119905) = (1205741(119905) 1205742(119905)) is a
homogenous vector chainUnder hypothesis (H1)1015840 the results given in this paper
can be established similarly It is easy to see that if 1205741(119905) equiv
1205742(119905) and 120574
1(119905) is a right-continuous homogenous Markovian
chain then (H1)1015840 is fulfilled immediately At the same timeif 1205741(119905) equiv 120574
2(119905) (8) will reduce to the classical SDEs with
single Markovian chain that is to say the classical theoryabout SDEs with single Markovian chain is a special caseof our theory On the other hand many theorems in thispaper will play important roles in further study For exampleTheorem 15 will be useful when one studies the approximatesolutions
Some important and interesting questions can be furtherinvestigated using the results in this paper For exampleapproximate solutions boundedness and stability stochas-tic functional differential equations with vector Markovianswitching and their applications In particular the stability of(8) is one of the most important and interesting topics andthose investigations are in progress
Acknowledgments
The authors thank the editor and referees for their veryimportant and helpful comments and suggestions Theauthors also thank Dr C Zhang for helping them to improvethe English exposition This research is supported by theNSFC of China (nos 11171081 and 11171056)
References
[1] A V SkorokhodAsymptoticMethods in theTheory of StochasticDifferential Equations vol 78 of Translations of MathematicalMonographs American Mathematical Society Providence RIUSA 1989
[2] X Mao and C Yuan Stochastic Differential Euations withMarkovian Switching Imperial College Press London UK2006
[3] G Yin and X Y Zhou ldquoMarkowitzrsquos mean-variance portfolioselection with regime switching from discrete-time models totheir continuous-time limitsrdquo IEEE Transactions on AutomaticControl vol 49 no 3 pp 349ndash360 2004
[4] J Buffington and R J Elliott ldquoAmerican options with regimeswitchingrdquo International Journal of Theoretical and AppliedFinance vol 5 no 5 pp 497ndash514 2002
[5] C Zhu and G Yin ldquoOn competitive Lotka-Volterra model inrandom environmentsrdquo Journal of Mathematical Analysis andApplications vol 357 no 1 pp 154ndash170 2009
[6] Q Luo and X Mao ldquoStochastic population dynamics underregime switchingrdquo Journal of Mathematical Analysis and Appli-cations vol 334 no 1 pp 69ndash84 2007
[7] X Li D Jiang and X Mao ldquoPopulation dynamical behaviorof Lotka-Volterra system under regime switchingrdquo Journal ofComputational and Applied Mathematics vol 232 no 2 pp427ndash448 2009
[8] X Li A Gray D Jiang and X Mao ldquoSufficient and neces-sary conditions of stochastic permanence and extinction forstochastic logistic populations under regime switchingrdquo Journal
Journal of Applied Mathematics 11
ofMathematical Analysis and Applications vol 376 no 1 pp 11ndash28 2011
[9] M Liu and KWang ldquoAsymptotic properties and simulations ofa stochastic logistic model under regime switchingrdquoMathemat-ical and Computer Modelling vol 54 no 9-10 pp 2139ndash21542011
[10] G Hu and K Wang ldquoStability in distribution of competi-tive Lotka-Volterra system with Markovian switchingrdquo AppliedMathematical Modelling vol 35 no 7 pp 3189ndash3200 2011
[11] M Liu and K Wang ldquoAsymptotic properties and simulationsof a stochastic logistic model under regime switching IIrdquoMathematical andComputerModelling vol 55 no 3-4 pp 405ndash418 2012
[12] ZWu H Huang and LWang ldquoStochastic delay logistic modelunder regime switchingrdquo Abstract and Applied Analysis vol2012 Article ID 241702 26 pages 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
4 Journal of Applied Mathematics
Lemma 4 Under Assumption (H2) for (1198941 1198942) (1198951 1198952) isin
S (1198941 1198942) = (1198951 1198952)
1199021198941119894211989511198952
= 1198751015840
1198941119894211989511198952
(0) = lim119905rarr0
1198751198941119894211989511198952
(119905)
119905(26)
exists and is finite
Proof By (H2) we note that for all 0 lt 120576 lt 13 exist0 lt 120575 lt 1such that
1198751198941119894211989411198942
(119905) gt 1 minus 120576 1198751198951119895211989511198952
(119905) gt 1 minus 120576
1198751198951119895211989411198942
(119905) lt 120576
(27)
provided 0 lt 119905 le 120575For forall0 le ℎ lt 119905 set 119899 = ⟨119905ℎ⟩ where ⟨119886⟩ = max
119899le119886119899 isin
Z Let
119875(11989511198952)
1198941119894211989611198962
(ℎ) = 1198751198941119894211989611198962
(ℎ)
119875(11989511198952)
1198941119894211989611198962
(119898ℎ)
= sum
(11990311199032) = (1198951 1198952)
119875(11989511198952)
1198941119894211990311199032
((119898 minus 1) ℎ) 1198751199031119903211989611198962
(ℎ)
(28)
where 119875(11989511198952)
1198941119894211989611198962
(119898ℎ)means that the probability of the 120574(119905)willnot reach to (119895
1 1198952) at times ℎ 2ℎ (119898 minus 1)ℎ but will reach
to (1198961 1198962) at time 119898ℎ Note that if ℎ le 119905 le 120575 then
120576 gt 1 minus 1198751198941119894211989411198942
(119905)
= sum
(1198961 1198962) = (1198941 1198942)
1198751198941119894211989611198962
(119905) ge 1198751198941119894211989511198952
(119905)
ge
119899
sum
119898=1
119875(11989511198952)
1198941119894211989511198952
(119898ℎ) 1198751198951119895211989511198952
(119905 minus 119898ℎ)
ge (1 minus 120576)
119899
sum
119898=1
119875(11989511198952)
1198941119894211989511198952
(119898ℎ)
(29)
which indicates119899
sum
119898=1
119875(11989511198952)
1198941119894211989511198952
(119898ℎ) le120576
1 minus 120576 (30)
Then making use of
1198751198941119894211989411198942
(119898ℎ) = 119875(11989511198952)
1198941119894211989411198942
(119898ℎ)
+
119898minus1
sum
119897=1
119875(11989511198952)
1198941119894211989511198952
(119897ℎ) 1198751198951119895211989411198942
((119898 minus 119897) ℎ)
(31)
we obtain
119875(11989511198952)
1198941119894211989411198942
(119898ℎ) ge 1198751198941119894211989411198942
(119898ℎ) minus
119898minus1
sum
119897=1
119875(11989511198952)
1198941119894211989511198952
(119897ℎ)
ge 1 minus 120576 minus120576
1 minus 120576
(32)
Consequently
1198751198941119894211989511198952
(119905)
gt
119899
sum
119898=1
119875(11989511198952)
1198941119894211989411198942
((119898 minus 1) ℎ) 1198751198941119894211989511198952
(ℎ) 1198751198951119895211989511198952
(119905 minus 119898ℎ)
ge 119899 (1 minus 120576 minus120576
1 minus 120576)1198751198941119894211989511198952
(ℎ) (1 minus 120576)
ge 119899 (1 minus 3120576) 1198751198941119894211989511198952
(ℎ)
(33)
Dividing both sides of the above inequality by ℎ and noting119899ℎ rarr 119905 whenever ℎ rarr 0 yield
lim supℎrarr0
1198751198941119894211989511198952
(ℎ)
ℎle
1
1 minus 3120576
1198751198941119894211989511198952
(119905)
119905lt infin (34)
Then letting 119905 rarr 0 gives
lim supℎrarr0
1198751198941119894211989511198952
(ℎ)
ℎle
1
1 minus 3120576lim inf119905rarr0
1198751198941119894211989511198952
(119905)
119905 (35)
and the required assertion follows immediately by letting120576 rarr 0 This completes the proof
Set 120574(119905) = (1205741(119905) 1205742(119905)) then it is easy to see that
almost every sample path of 120574(119905) is a right continuous stepfunction Now letting P(119905) = (119875
1198941119894211989511198952
(119905))11987311198732times11987311198732
Q =
(1199021198941119894211989511198952
)11987311198732times11987311198732
= P1015840(0) Then by Chapman-Kolmogorovequation
P (119905 + ℎ) = P (119905)P (ℎ) = P (ℎ)P (119905) (36)
we have
P (119905 + ℎ) minus P (119905)
ℎ= P (119905) [
P (ℎ) minus Iℎ
] = [P (ℎ) minus I
ℎ]P (119905)
(37)
Letting ℎ rarr 0 and taking limits give
P1015840(119905) = P (119905)Q
P1015840(119905) = QP (119905)
(38)
Note that
P (0) = I (39)
Then by solving the ordinary differential equations (38) and(39) we obtain the following lemma
Lemma 5 For P(119905) and Q one has
P (119905) = exp Q119905 (40)
We are now in the position to give the sufficient condi-tions for the existence and uniqueness of the solution of (8)For this end let us first give the definition of the solution
Journal of Applied Mathematics 5
Definition 6 An R119899-valued stochastic process 119909(119905)1199050le119905le119879
iscalled a solution of (8) if it has the following properties
(i) 119909(119905)1199050le119905le119879
is continuous andF119905-adapted
(ii) 119891(119909(119905) 119905 1205741(119905))1199050le119905le119879
isinL1([1199050 119879]R119899)while 119892(119909(119905)
119905 1205742(119905))1199050le119905le119879
isin L2([1199050 119879]R119899times119898)
(iii) for any 119905 isin [1199050 119879] equation
119909 (119905) = 119909 (1199050) + int
119905
1199050
119891 (119909 (119904) 119904 1205741(119904)) 119889119904
+ int
119905
1199050
119892 (119909 (119904) 119904 1205742(119904)) 119889119861 (119904)
(41)
holds with probability 1
A solution 119909(119905)1199050le119905le119879
is said to be unique if any othersolution (119905)
1199050le119905le119879
is indistinguishable from 119909(119905)1199050le119905le119879
Now we can give our main results in this section
Theorem 7 Assume that there exist two positive constants 119870
and 119870 such that(Lipschitz condition) for all 119909 119910 isin R119899 119905 isin [119905
0 119879] and
(1198941 1198942) isin S
1003816100381610038161003816119891 (119909 119905 1198941) minus 119891 (119910 119905 119894
1)10038161003816100381610038162
or1003816100381610038161003816119892 (119909 119905 119894
2) minus 119892 (119910 119905 119894
2)10038161003816100381610038162le 119870
1003816100381610038161003816119909 minus 11991010038161003816100381610038162
(42)
(Linear growth condition) for all (119909 119905 (1198941 1198942)) isin R119899 times
[1199050 119879] times S
1003816100381610038161003816119891 (119909 119905 1198941)10038161003816100381610038162or
1003816100381610038161003816119892 (119909 119905 1198942)10038161003816100381610038162le 119870(1 + |119909|)
2 (43)
Then there exists a unique solution 119909(119905) to (8) and moreoverthe solution obeys
119864( sup1199050le119905le119879
|119909 (119905)|2)
le (1 + 311986410038161003816100381610038161199090
10038161003816100381610038162) exp 3119870 (119879 minus 119905
0) (119879 minus 119905
0+ 4)
(44)
Proof Recall that almost every sample path of 120574(119905) =
(1205741(119905) 1205742(119905)) is a right continuous step function with a finite
number of jumps on [1199050 119879] Thus there exists a sequence of
stopping times 120591119896119896ge0
such that
(i) for almost every 120596 isin Ω there is a finite 119896120596for 1199050
=
1205910lt 1205911lt sdot sdot sdot lt 120591
119896= 119879 and 120591
119896= 119879 if 119896 gt 119896
120596
(ii) both 1205741(sdot) and 120574
2(sdot) in 120574(sdot) are constants on interval
[[120591119896 120591119896+1
[[ namely
1205741 (119905) = 120574
1(120591119896) on 120591
119896le 119905 le 120591
119896+1forall119896 ge 0
1205742(119905) = 120574
2(120591119896) on 120591
119896le 119905 le 120591
119896+1forall119896 ge 0
(45)
First of all let us consider (8) on 119905 isin [[1205910 1205911[[ then (8)
becomes119889119909 (119905) = 119891 (119909 (119905) 119905 120574
1(1199050)) 119889119905
+ 119892 (119909 (119905) 119905 1205742(1199050)) 119889119861 (119905)
(46)
with initial conditions 119909(1199050) = 119909
0 120574(1199050) = (120574
1(1199050) 1205741(1199050))
Then by the theory of SDEs we obtain that (46) has a uniquesolution which obeys 119909(120591
1) isin 1198712
F1205911
(ΩR119899) We next consider(8) on 119905 isin [[120591
1 1205912[[ which becomes
119889119909 (119905) = 119891 (119909 (119905) 119905 1205741 (1205911)) 119889119905
+ 119892 (119909 (119905) 119905 1205742(1205911)) 119889119861 (119905)
(47)
Again by the theory of SDEs (47) has a unique solutionwhichobeys 119909(120591
1) isin 119871
2
F1205912
(ΩR119899) Repeating this procedure weconclude that (8) has a unique solution 119909(119905) on [119905
0 119879]
Now let us prove (44) For every 119896 ge 1 define thestopping time
120591119896= 119879 and inf 119905 isin [119905
0 119879] |119909 (119905)| ge 119896 (48)
It is obvious that 120591119896uarr 119879 as Set 119909
119896(119905) = 119909(119905and120591
119896) for 119905 isin [119905
0 119879]
Then 119909119896(119905) obeys the equation
119909119896(119905) = 119909
0+ int
119905
1199050
119891 (119909119896(119904) 119904 120574
1(119904)) 119868[[1199050120591119896]](119904) 119889119904
+ int
119905
1199050
119892 (119909119896(119904) 119904 120574
2(119904)) 119868[[1199050120591119896]](119904) 119889119861 (119904)
(49)
Making use of the elementary inequality |119886+119887+119888|2le 3(|119886|
2+
|119887|2+ |119888|2) the Holder inequality and (43) we can see that
1003816100381610038161003816119909119896 (119905)10038161003816100381610038162= 3
1003816100381610038161003816119909010038161003816100381610038162+ 3119870 (119879 minus 119905
0) int
119905
1199050
(1 +1003816100381610038161003816119909119896 (119904)
10038161003816100381610038162) 119889119904
+ 3
100381610038161003816100381610038161003816100381610038161003816
int
119905
1199050
119892 (119909119896(119904) 119904 120574
2(119904)) 119868[[1199050120591119896]](119904) 119889119861 (119904)
100381610038161003816100381610038161003816100381610038161003816
2
(50)
Thus applying the Doob martingale inequality and (43) wecan further show that
119864( sup1199050le119904le119905
1003816100381610038161003816119909119896 (119904)10038161003816100381610038162)
= 311986410038161003816100381610038161199090
10038161003816100381610038162+ 3119870 (119879 minus 119905
0) int
119905
1199050
(1 + 1198641003816100381610038161003816119909119896 (119904)
10038161003816100381610038162) 119889119904
+ 12119864int
119905
1199050
1003816100381610038161003816119892 (119909119896(119904) 119904 120574
2(119904))
10038161003816100381610038162119868[[1199050120591119896]](119904) 119889119904
le 311986410038161003816100381610038161199090
10038161003816100381610038162+ 3119870 (119879 minus 119905
0+ 4)int
119905
1199050
(1 + 1198641003816100381610038161003816119909119896 (119904)
10038161003816100381610038162) 119889119904
(51)
That is to say
1 + 119864( sup1199050le119904le119905
1003816100381610038161003816119909119896 (119904)10038161003816100381610038162) le 1 + 3119864
1003816100381610038161003816119909010038161003816100381610038162+ 3119870 (119879 minus 119905
0+ 4)
times int
119905
1199050
[1 + 119864 sup1199050le119903le119904
1003816100381610038161003816119909119896 (119903)10038161003816100381610038162]119889119904
(52)
6 Journal of Applied Mathematics
Using the Gronwall inequality leads to
1 + 119864( sup1199050le119904le119905
1003816100381610038161003816119909119896 (119904)10038161003816100381610038162)
le (1 + 311986410038161003816100381610038161199090
10038161003816100381610038162) exp 3119870 (119879 minus 119905
0) (119879 minus 119905
0+ 4)
(53)
Consequently
1 + 119864( sup1199050le119905le120591119896
|119909 (119905)|2)
le (1 + 311986410038161003816100381610038161199090
10038161003816100381610038162) exp 3119870 (119879 minus 119905
0) (119879 minus 119905
0+ 4)
(54)
Then the required inequality (44) follows immediately byletting 119896 rarr infin
Condition (42) indicates that the coefficients 119891(119909 119905 1198941)
and 119892(119909 119905 1198942) do not change faster than a linear function of
119909 as change in 119909 This means in particular the continuity of119891(119909 119905 119894
1) and 119892(119909 119905 119894
2) in 119909 for all 119905 isin [119905
0 119879] Then functions
that are discontinuous with respect to 119909 are excluded as thecoefficients Besides there are many functions that do notsatisfy the Lipschitz conditionThese imply that the Lipschitzcondition is too restrictive To improve this Lipschitz condi-tion let us introduce the concept of local solution
Definition 8 Let120590infinbe a stopping time such that 119905
0le 120590infin
le 119879
as An R119899-valuedF119905-adapted continuous stochastic process
119909(119905)1199050le119905lt120590infin
is called a local solution of (8) if 119909(1199050) = 1199090and
moreover there is a nondecreasing sequence 120590119896119896ge1
such that1199050le 120590119896 uarr 120590infin
as and
119909 (119905) = 119909 (1199050) + int
119905and120590119896
1199050
119891 (119909 (119904) 119904 1205741(119904)) 119889119904
+ int
119905and120590119896
1199050
119892 (119909 (119904) 119904 1205742(119904)) 119889119861 (119904)
(55)
holds for any 119905 isin [1199050 119879) and 119896 ge 1 with probability one If
furthermore
lim sup119905rarr+120590
infin
|119909 (119905)| = infin whenever 120590infin
lt 119879 (56)
then it is called a maximal local solution and 120590infinis called the
explosion time Amaximal local solution 119909(119905) 1199050le 119905 lt 120590
infin
is said to be unique if any other maximal local solution 119909(119905)
1199050le 119905 lt 120590
infin is indistinguishable from it namely 120590
infin= 120590infin
and 119909(119905) = 119909(119905) for 1199050le 119905 lt 120590
infinwith probability one
Definition 9 (local Lipschitz condition) For every integer119896 ge 1 there exists a positive constant ℎ
119896such that for all
119905 isin [1199050 119879] 119894 = (119894
1 1198942) isin S and those 119909 119910 isin R119899 with
|119909| or |119910| le 119896
1003816100381610038161003816119891 (119909 119905 1198941) minus 119891 (119910 119905 119894
1)10038161003816100381610038162
or1003816100381610038161003816119892 (119909 119905 119894
2) minus 119892 (119910 119905 119894
2)10038161003816100381610038162le ℎ119896
1003816100381610038161003816119909 minus 11991010038161003816100381610038162
(57)
The following theorem shows the existence of uniquemaximal local solution under the local Lipschitz conditionwithout the linear growth condition
Theorem 10 Under condition (57) there exists a uniquemaximal local solution of (8)
Proof Define functions
119891119896(119909 119905 119894
1) =
119891 (119909 119905 1198941) if |119909| le 119896
119891(119896119909
|119909| 119905 1198941) if |119909| gt 119896
119892119896(119909 119905 119894
2) =
119892 (119909 119905 1198942) if |119909| le 119896
119892 (119896119909
|119909| 119905 1198942) if |119909| gt 119896
(58)
Then 119891119896and 119892
119896satisfy the Lipschitz condition and the linear
growth condition Thus by Theorem 7 there is a uniquesolution 119909
119896(119905) of the equation
119889119909119896(119905) = 119891
119896(119909119896(119905) 119905 120574
1(119905)) 119889119905
+ 119892119896(119909119896(119905) 119905 120574
2(119905)) 119889119861 (119905)
(59)
with the initial conditions 119909119896(1199050) = 119909
0and 120574(119905
0) = (120574
1(1199050)
1205742(1199050)) Define the stopping times
120590119896= 119879 and inf 119905 isin [119905
0 119879]
1003816100381610038161003816119909119896 (119905)1003816100381610038161003816 ge 119896 (60)
Clearly if 1199050le 119905 le 120590
119896
119909119896(119905) = 119909
119896+1(119905) (61)
which indicates that 120590119896is increasing so 120590
119896has its limit 120590
infin=
lim119896rarr+infin
120590119896 Define 119909(119905) 119905
0le 119905 lt 120590
infin by
119909 (119905) = 119909119896(119905) 119905 isin [[120590
119896minus1 120590119896[[ 119896 ge 1 (62)
where 1205900= 1199050 Applying (61) one can show that 119909(119905 and 120590
119896) =
119909119896(119905 and 120590119896) It then follows from (59) that
119909 (119905 and 120590119896) = 1199090+ int
119905and120590119896
1199050
119891119896(119909 (119904) 119904 120574
1(119904)) 119889119904
+ int
119905and120590119896
1199050
119892119896(119909 (119904) 119904 1205742 (119904)) 119889119861 (119904)
= 1199090+ int
119905and120590119896
1199050
119891 (119909 (119904) 119904 1205741(119904)) 119889119904
+ int
119905and120590119896
1199050
119892 (119909 (119904) 119904 1205742 (119904)) 119889119861 (119904)
(63)
for any 119905 isin [1199050 119879) and 119896 ge 1 It is easy to see that if 120590
infinle 119879
then
lim sup119905rarr120590infin
|119909 (119905)|
ge lim sup119896rarr+infin
1003816100381610038161003816119909 (120590119896)1003816100381610038161003816 = lim sup119896rarr+infin
1003816100381610038161003816119909119896 (120590119896)1003816100381610038161003816 = infin
(64)
Therefore 119909(119905) 1199050le 119905 lt 120590
infin is a maximal local solution
Now we will prove the uniqueness Let 119909(119905) 1199050
le 119905 lt
120590infin
be another maximal local solution Define
120590119896= 120590infin
and inf 119905 isin [[1199050 120590infin
[[ 119909 (119905) ge 119896 (65)
Journal of Applied Mathematics 7
Then 120590119896
rarr 120590infin
as and
119875 119909 (119905) = 119909 (119905) 119905 isin [[1199050 120590119896and 120590119896]] = 1 forall119896 ge 1 (66)
Letting 119896 rarr infin gives
119875 119909 (119905) = 119909 (119905) 119905 isin [[1199050 120590infin
and 120590infin
[[ = 1 (67)
In order to complete the proof we need only to show that120590infin
= 120590infin
as In fact for almost every 120596 isin 120590infin
lt 120590infin
wehave
1003816100381610038161003816119909 (120590infin
120596)1003816100381610038161003816
= lim119896rarr+infin
1003816100381610038161003816119909 (120590119896 120596)
1003816100381610038161003816 = lim119896rarr+infin
1003816100381610038161003816119909 (120590119896 120596)
1003816100381610038161003816 = infin(68)
which contradicts the fact that 119909(119905 120596) is continuous on 119905 isin
[1199050 120590infin
(120596)) This implies 120590infin
ge 120590infin
as In the same way onecan show 120590
infinle 120590infinasThus wemust have 120590
infin= 120590infinasThis
completes the proof
In many situations we often consider an SDE on [1199050infin)
119889119909 (119905) = 119891 (119909 (119905) 119905 1205741(119905)) 119889119905
+ 119892 (119909 (119905) 119905 1205742(119905)) 119889119861 (119905) 119905
0le 119905 lt infin
(69)
with initial data 119909(1199050) = 119909
0and 120574(119905
0) If the assumption of
the existence-and-uniqueness theorem holds on every finitesubinterval [119905
0 119879] of [119905
0infin) then (69) has a unique solution
119909(119905) on the entire interval [1199050infin) Such a solution is called
a global solution To establish a more general result aboutglobal solution we need more notations To this end weintroduce an operator 119871119881 from R119899 times R
+times S to R which is
given by
119871119881 (119909 119905 (1198941 1198942))
= 119881119905(119909 119905 (119894
1 1198942)) + 119881
119909(119909 119905 (119894
1 1198942)) 119891 (119909 119905 119894
1)
+ 05 trace [119892119879 (119909 119905 1198942) 119881119909119909
(119909 119905 (1198941 1198942)) 119892 (119909 119905 119894
2)]
+ sum
(11989511198952)isinS
1199021198941119894211989511198952
119881 (119909 119905 (1198951 1198952))
(70)
where 119881(119909 119905 (1198941 1198942)) isin 119862
21(R119899 times R
+times SR
+) and
119881119905(119909 119905 (119894
1 1198942)) =
120597119881 (119909 119905 (1198941 1198942))
120597119905
119881119909(119909 119905 (119894
1 1198942))
= (120597119881 (119909 119905 (119894
1 1198942))
1205971199091
120597119881 (119909 119905 (119894
1 1198942))
120597119909119899
)
119881119909119909
(119909 119905 (1198941 1198942)) = (
1205972119881 (119909 119905 (119894
1 1198942))
120597119909119894120597119909119895
)
119899times119899
(71)
Theorem 11 Assume that the local Lipschitz condition (57)holds Assume also that there is a function 119881(119909 119905 (119894
1 1198942)) isin
11986221
(R119899 times R+times SR
+) and a constant 120579 gt 0 such that
lim|119909|rarrinfin
( inf(119905(1198941 1198942))isinR+timesS
119881 (119909 119905 (1198941 1198942))) = infin (72)
119871119881 (119909 119905 (1198941 1198942)) le 120579 (1 + 119881 (119909 119905 (119894
1 1198942)))
forall (119909 119905 (1198941 1198942)) isin R
119899times R+times S
(73)
Then there exists a unique global solution 119909(119905) to (69)
Proof We need only to prove the theorem for any initialcondition 119909
0isin R119899 and (120574
1(1199050) 1205742(1199050)) isin S FromTheorem 10
we know that the local Lipschitz condition guarantees theexistence of the unique maximal solution 119909(119905) on [119905
0 120590infin
)where 120590
infinis the explosion time We need only to show 120590
infin=
infin as If this is not true then we can find a pair of positiveconstants 120576 and 119879 such that
119875 120590infin
le 119879 gt 2120576 (74)
For each integer 119896 ge 1 define the stopping time
120590119896= inf 119905 ge 119905
0 |119909 (119905)| ge 119896 (75)
Since 120590119896
rarr 120590infinalmost surely we can find a sufficiently large
integer 1198960for
119875 120590119896le 119879 gt 120576 forall119896 ge 119896
0 (76)
Fix any 119896 ge 1198960 then for any 119905
0le 119905 le 119879 by virtue of the
generalized Ito formula (see eg [1])
119864119881 (119909 (119905 and 120590119896) 119905 and 120590
119896 120574 (119905 and 120590
119896))
= 119881 (1199090 1199050 1205740) + 119864int
119905and120590119896
1199050
119871119881 (119909 (119904) 119904 120574 (119904)) 119889119904
le 119881 (1199090 1199050 1205740) + 120579 (119879 minus 119905
0)
+ 120579int
119905
1199050
119864119881 (119909 (119904 and 120590119896) 119904 and 120590
119896 120574 (119904 and 120590
119896)) 119889119904
(77)
Making use of the Gronwall inequality gives
119864119881 (119909 (119905 and 120590119896) 119905 and 120590
119896 120574 (119905 and 120590
119896))
le [119881 (1199090 1199050 1205740) + 120579 (119879 minus 119905
0)] exp 120579 (119879 minus 119905
0)
(78)
Therefore
119864 (119868120590119896le119879
119881 (119909 (120590119896) 120590119896 120574 (120590119896)))
le [119881 (1199090 1199050 1205740) + 120579 (119879 minus 119905
0)] exp 120579 (119879 minus 119905
0)
(79)
At the same time set
119892119896= inf 119881 (119909 119905 (119894
1 1198942)) |119909| ge 119896 119905 isin [119905
0 119879] (119894
1 1198942) isin S
(80)
8 Journal of Applied Mathematics
Then (72) means 119892119896
rarr infin It follows from (76) and (79) that
[119881 (1199090 1199050 1205740) + 120579 (119879 minus 119905
0)] exp 120579 (119879 minus 119905
0)
ge 119892119896119875 120590119896le 119879 ge 120576119892
119896
(81)
Letting 119896 rarr infin yields a contradiction that is to say120590infin
= infinThe proof is complete
3 119871119875-Estimates
In the previous section we have investigated the existenceand uniqueness of the solution to (8) In this section as abovelet 119909(119905) 119905
0le 119905 le 119879 be the unique solution of (8) with initial
conditions 119909(1199050) = 1199090and 120574(119905
0) and we will estimate the 119901th
moment of the solution
Theorem 12 Assume that there is a function 119881(119909 119905 (1198941 1198942)) isin
11986221
(R119899 times R+times SR
+) and positive constants 119901 120578 120579 such that
for all (119909 119905 (1198941 1198942)) isin R119899 times R
+times S
120578|119909|119901le 119881 (119909 119905 (119894
1 1198942)) (82)
119871119881 (119909 119905 (1198941 1198942)) le 120579119881 (119909 119905 (119894
1 1198942)) (83)
Assume also the initial condition 119909(1199050) = 119909
0and 120574(119905
0) obeys
119864119881(1199090 1199050 120574(1199050)) lt infin then one has
119864|119909 (119905)|119901
le119864119881 (119909
0 1199050 120574 (1199050)) exp 120579 (119905 minus 119905
0)
120578 forall119905 isin [119905
0 119879]
(84)
Proof For each integer 119896 ge 1 define the stopping time
120590119896= 119879 and inf 119905 ge 119905
0 |119909 (119905)| ge 119896 (85)
Thus 120590119896
rarr 119879 as Using the generalized Itorsquos formula and(83) we obtain that for 119905 isin [119905
0 119879]
119864119881 (119909 (119905 and 120590119896) 119905 and 120590
119896 120574 (119905 and 120590
119896))
le 119864119881 (1199090 1199050 120574 (1199050))
+ 120579int
119905
1199050
119864119881 (119909 (119904 and 120590119896) 119904 and 120590
119896 120574 (119904 and 120590
119896)) 119889119904
(86)
Then the Gronwall inequality indicates
119864119881 (119909 (119905 and 120590119896) 119905 and 120590
119896 120574 (119905 and 120590
119896))
le 119864119881 (1199090 1199050 120574 (1199050)) exp 120579 (119905 minus 119905
0)
(87)
for all 119905 isin [1199050 119879] By virtue of condition (83) we obtain the
required assertion (84)
Corollary 13 Assume 119901 ge 2 and 1199090isin 119871119901
F1199050
(ΩR119899) Assumealso that there exists a constant 120579 gt 0 such that for all(119909 119905 (119894
1 1198942)) isin R119899 times R
+times S
119909119879119891 (119909 119905 119894
1) +
119901 minus 1
2
1003816100381610038161003816119892 (119909 119905 1198942)10038161003816100381610038162le 120579 (1 + |119909|
2) (88)
Then one has
119864|119909 (119905)|119901
le 2(119901minus2)2
(1 + 11986410038161003816100381610038161199090
1003816100381610038161003816119901) exp 119901120579 (119905 minus 119905
0) forall119905 isin [119905
0 119879]
(89)
Proof Define 119881(119909 119905 (1198941 1198942)) = (1 + |119909|
2)1199012 Making use of
(88) yields
119871119881 (119909 119905 (1198941 1198942))
= 119901(1 + |119909|2)(119901minus2)2
119909119879119891 (119909 119905 119894
1)
+119901
2(1 + |119909|
2)(119901minus2)21003816100381610038161003816119892 (119909 119905 119894
2)10038161003816100381610038162
+119901 (119901 minus 2)
2(1 + |119909|
2)(119901minus4)210038161003816100381610038161003816
119909119879119892 (119909 119905 119894
2)10038161003816100381610038161003816
2
le 119901(1 + |119909|2)(119901minus2)2
[119909119879119891 (119909 119905 119894
1) +
119901 minus 1
2
1003816100381610038161003816119892 (119909 119905 1198942)10038161003816100381610038162]
le 119901120579(1 + |119909|2)1199012
(90)
Then byTheorem 12 we get
119864(1 + |119909 (119905)|2)1199012
le 119864(1 +10038161003816100381610038161199090
10038161003816100381610038162)1199012
exp 119901120579 (119905 minus 1199050)
(91)
and the required assertion (89) followsIt is useful to point out that if the linear growth condition
(43) is satisfied then (88) is fulfilled with 120579 = radic119870 + 119870(119901 minus
1)2 Now we will show the other important properties of thesolution
Theorem 14 Let 119901 ge 2 and 1199090
isin 119871119901
F1199050
(ΩR119899) Assume alsothat the linear growth condition (43) holds Then one has
119864|119909 (119905) minus 119909 (119904)|119901le 119862(119905 minus 119904)
1199012 (92)
where
119862 = 2119901minus2
(1 + 11986410038161003816100381610038161199090
1003816100381610038161003816119901) exp 119901120579 (119879 minus 119905
0)
times ([2 (119879 minus 1199050)]1199012
+ [119901 (119901 minus 1)]1199012
)
(93)
and 120579 = radic119870+119870(119901minus1)2 Particularly the 119901th moment of thesolution is continuous on [119905
0 119879]
Journal of Applied Mathematics 9
Proof Applying the elementary inequality |119886 + 119887|119901
le
2119901minus1
(|119886|119901+ |119887|119901) the Holder inequality and the linear growth
condition we can derive that
119864|119909 (119905) minus 119909 (119904)|119901
le 2119901minus1
119864
10038161003816100381610038161003816100381610038161003816int
119905
119904
119891 (119909 (119906) 119906 1205741(119906)) 119889119906
10038161003816100381610038161003816100381610038161003816
119901
+ 2119901minus1
119864
10038161003816100381610038161003816100381610038161003816int
119905
119904
119892 (119909 (119906) 119906 1205742(119906)) 119889119861 (119906)
10038161003816100381610038161003816100381610038161003816
119901
le [2 (119905 minus 119904)]119901minus1
119864int
119905
119904
1003816100381610038161003816119891 (119909 (119906) 119906 1205741(119906))
1003816100381610038161003816119901119889119906
+ 05[2119901 (119901 minus 1)]1199012
(119905 minus 119904)(119901minus2)2
119864
times int
119905
119904
1003816100381610038161003816119892 (119909 (119906) 119906 1205742 (119906))1003816100381610038161003816119901119889119906
le 1198621(119905 minus 119904)
(119901minus2)2int
119905
119904
119864(1 + |119909 (119906)|2)1199012
119889119906
(94)
where1198621= 2(119901minus2)2
1198701199012
([2(119879minus 1199050)]1199012
+ [119901(119901minus1)]1199012
) Using(91) yields
119864|119909 (119905) minus 119909 (119904)|119901
le 1198621(119905 minus 119904)
(119901minus2)2
times int
119905
119904
2(119901minus2)2
(1 + 11986410038161003816100381610038161199090
1003816100381610038161003816119901) exp 119901120579 (119906 minus 119905
0) 119889119906
le 11986212(119901minus2)2
(1 + 11986410038161003816100381610038161199090
1003816100381610038161003816119901)
times exp 119901120579 (119879 minus 1199050) (119905 minus 119904)
(1199012)
(95)
which is the desired inequality
Theorem 15 Let 119901 ge 2 and 1199090
isin 119871119901
F1199050
(ΩR119899) Assume thatthere is a119870 gt 0 such that for all (119909 119905 (119894
1 1198942)) isin R119899times[119905
0 119879]timesS
119909119879 1003816100381610038161003816119891 (119909 119905 119894
1)1003816100381610038161003816 or
1003816100381610038161003816119892 (119909 119905 1198942)10038161003816100381610038162le 119870(1 + |119909|)
2 (96)
Then
119864( sup1199050le119905le119879
|119909 (119905)|119901)
le (1 + 211986410038161003816100381610038161199090
1003816100381610038161003816119901) exp 2119901 (10119901 + 1)119870 (119879 minus 119905
0)
(97)
Proof Making use of the generalized Itorsquos formula andcondition (96) we have
|119909 (119905)|119901
le10038161003816100381610038161199090
1003816100381610038161003816119901+ 119872(119905)
+ int
119905
1199050
119901|119909 (119904)|119901minus2
[119909119879(119904) 119891 (119909 (119904) 119904 120574
1(119904))
+119901 minus 1
2
1003816100381610038161003816119892 (119909 (119904) 119904 1205742 (119904))10038161003816100381610038162] 119889119904
le10038161003816100381610038161199090
1003816100381610038161003816119901+ 119872(119905) + 05119901 (119901 + 1)119870
times int
119905
1199050
|119909 (119904)|119901minus2
(1 + |119909 (119904)|2) 119889119904
le10038161003816100381610038161199090
1003816100381610038161003816119901+ 119872(119905) + 119901 (119901 + 1)119870int
119905
1199050
(1 + |119909 (119904)|119901) 119889119904
(98)
where
119872(119905) = int
119905
1199050
119901|119909 (119904)|119901minus2
119909119879(119904) 119892 (119909 (119904) 119904 1205742 (119904)) 119889119861 (119904) (99)
Therefore for any 1199051isin [1199050 119879]
119864( sup1199050le119905le1199051
|119909 (119905)|119901) le 119864
100381610038161003816100381611990901003816100381610038161003816119901+ 119864( sup
1199050le119905le1199051
|119872 (119905)|)
+ 119901 (119901 + 1)119870119864int
1199051
1199050
(1 + |119909 (119904)|119901) 119889119904
(100)
At the same time applying the well-known Burkholder-Davis-Gundy inequality (see eg [2]) gives
119864( sup1199050le119905le1199051
|119872 (119905)|)
le 3119864(int
1199051
1199050
1199012|119909 (119904)|
2119901minus21003816100381610038161003816119892 (119909 (119904) 119904 1205742 (119904))10038161003816100381610038162119889119904)
05
le 3119864( sup1199050le119905le1199051
|119909 (119905)|119901
timesint
1199051
1199050
1199012|119909 (119904)|
119901minus21003816100381610038161003816119892 (119909 (119904) 119904 1205742(119904))
10038161003816100381610038162119889119904)
05
le 05119864( sup1199050le119905le1199051
|119909 (119905)|119901) + 45119864
times int
1199051
1199050
1199012119870|119909 (119904)|
119901minus2(1 + |119909 (119904)|
2) 119889119904
le 05119864( sup1199050le119905le1199051
|119909 (119905)|119901) + 9119901
2119870119864int
1199051
1199050
(1 + |119909 (119904)|119901) 119889119904
(101)
10 Journal of Applied Mathematics
Substituting the above inequality into (100) gives
119864( sup1199050le119905le1199051
|119909 (119905)|119901) le 2119864
100381610038161003816100381611990901003816100381610038161003816119901
+ 2119901 (10119901 + 1)119870int
1199051
1199050
(1 + |119909 (119904)|119901) 119889119904
(102)
Thus
1 + 119864( sup1199050le119905le1199051
|119909 (119905)|119901)
le 1 + 211986410038161003816100381610038161199090
1003816100381610038161003816119901+ 2119901 (10119901 + 1)119870
times int
1199051
1199050
(1 + sup1199050le119905le119904
|119909 (119905)|119901)119889119904
(103)
Then the required assertion follows from the Gronwallinequality
Up to now we have discussed the 119871119901-estimates for the
solution in the case when 119901 ge 2 As for 0 lt 119901 lt 2 the similarresults can be given without any difficulty as long as we notethat the Holder inequality implies
119864|119909 (119905)|119901le [119864|119909 (119905)|
2]05119901
(104)
4 Example
Consider the following Black-Scholes model
119889119883 (119905) = 120583 (1205741(119905))119883 (119905) 119889119905 + ] (120574
2(119905))119883 (119905) 119889119861 (119905) (105)
where 1205741(119905) is a right-continuous homogenous Markovian
chain taking values in finite state spaces S1= 1 2 and 120574
2(119905)
is a right-continuous homogenous Markovian chain takingvalues in finite state spaces S
2= 1 2 3 120583(119894) = 119894 119894 = 1 2
](119895) = 119895 + 1 119895 = 1 2 3 Taking 119870 = 16 119870 = 9 then (42)and (43) hold Therefore by Theorem 7 (105) has a uniquesolution
5 Conclusions and Further Research
This paper is devoted to studying the existence and unique-ness of solution of SDEs with multi-Markovian switchingsand estimating the119901thmoment of the solutionWe have usedtwo continuous-time Markovian chains to model the SDEsThis area is becoming increasingly useful in engineeringeconomics communication theory active networking andso forth The sufficient criteria for existence and unique-ness of solution local solution and maximal local solutionwere established Those results indicate that (8) keeps manyproperties that (89) owns At the same time although thehypothesis (H1) is used in this paper wewant to point out thatthis hypothesis is not essential In fact (H1) can be replacedby the following generalized hypothesis
(H1)1015840 both 1205741(119905) and 120574
2(119905) are right-continuous homoge-
nous Markovian chains such that 120574(119905) = (1205741(119905) 1205742(119905)) is a
homogenous vector chainUnder hypothesis (H1)1015840 the results given in this paper
can be established similarly It is easy to see that if 1205741(119905) equiv
1205742(119905) and 120574
1(119905) is a right-continuous homogenous Markovian
chain then (H1)1015840 is fulfilled immediately At the same timeif 1205741(119905) equiv 120574
2(119905) (8) will reduce to the classical SDEs with
single Markovian chain that is to say the classical theoryabout SDEs with single Markovian chain is a special caseof our theory On the other hand many theorems in thispaper will play important roles in further study For exampleTheorem 15 will be useful when one studies the approximatesolutions
Some important and interesting questions can be furtherinvestigated using the results in this paper For exampleapproximate solutions boundedness and stability stochas-tic functional differential equations with vector Markovianswitching and their applications In particular the stability of(8) is one of the most important and interesting topics andthose investigations are in progress
Acknowledgments
The authors thank the editor and referees for their veryimportant and helpful comments and suggestions Theauthors also thank Dr C Zhang for helping them to improvethe English exposition This research is supported by theNSFC of China (nos 11171081 and 11171056)
References
[1] A V SkorokhodAsymptoticMethods in theTheory of StochasticDifferential Equations vol 78 of Translations of MathematicalMonographs American Mathematical Society Providence RIUSA 1989
[2] X Mao and C Yuan Stochastic Differential Euations withMarkovian Switching Imperial College Press London UK2006
[3] G Yin and X Y Zhou ldquoMarkowitzrsquos mean-variance portfolioselection with regime switching from discrete-time models totheir continuous-time limitsrdquo IEEE Transactions on AutomaticControl vol 49 no 3 pp 349ndash360 2004
[4] J Buffington and R J Elliott ldquoAmerican options with regimeswitchingrdquo International Journal of Theoretical and AppliedFinance vol 5 no 5 pp 497ndash514 2002
[5] C Zhu and G Yin ldquoOn competitive Lotka-Volterra model inrandom environmentsrdquo Journal of Mathematical Analysis andApplications vol 357 no 1 pp 154ndash170 2009
[6] Q Luo and X Mao ldquoStochastic population dynamics underregime switchingrdquo Journal of Mathematical Analysis and Appli-cations vol 334 no 1 pp 69ndash84 2007
[7] X Li D Jiang and X Mao ldquoPopulation dynamical behaviorof Lotka-Volterra system under regime switchingrdquo Journal ofComputational and Applied Mathematics vol 232 no 2 pp427ndash448 2009
[8] X Li A Gray D Jiang and X Mao ldquoSufficient and neces-sary conditions of stochastic permanence and extinction forstochastic logistic populations under regime switchingrdquo Journal
Journal of Applied Mathematics 11
ofMathematical Analysis and Applications vol 376 no 1 pp 11ndash28 2011
[9] M Liu and KWang ldquoAsymptotic properties and simulations ofa stochastic logistic model under regime switchingrdquoMathemat-ical and Computer Modelling vol 54 no 9-10 pp 2139ndash21542011
[10] G Hu and K Wang ldquoStability in distribution of competi-tive Lotka-Volterra system with Markovian switchingrdquo AppliedMathematical Modelling vol 35 no 7 pp 3189ndash3200 2011
[11] M Liu and K Wang ldquoAsymptotic properties and simulationsof a stochastic logistic model under regime switching IIrdquoMathematical andComputerModelling vol 55 no 3-4 pp 405ndash418 2012
[12] ZWu H Huang and LWang ldquoStochastic delay logistic modelunder regime switchingrdquo Abstract and Applied Analysis vol2012 Article ID 241702 26 pages 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Journal of Applied Mathematics 5
Definition 6 An R119899-valued stochastic process 119909(119905)1199050le119905le119879
iscalled a solution of (8) if it has the following properties
(i) 119909(119905)1199050le119905le119879
is continuous andF119905-adapted
(ii) 119891(119909(119905) 119905 1205741(119905))1199050le119905le119879
isinL1([1199050 119879]R119899)while 119892(119909(119905)
119905 1205742(119905))1199050le119905le119879
isin L2([1199050 119879]R119899times119898)
(iii) for any 119905 isin [1199050 119879] equation
119909 (119905) = 119909 (1199050) + int
119905
1199050
119891 (119909 (119904) 119904 1205741(119904)) 119889119904
+ int
119905
1199050
119892 (119909 (119904) 119904 1205742(119904)) 119889119861 (119904)
(41)
holds with probability 1
A solution 119909(119905)1199050le119905le119879
is said to be unique if any othersolution (119905)
1199050le119905le119879
is indistinguishable from 119909(119905)1199050le119905le119879
Now we can give our main results in this section
Theorem 7 Assume that there exist two positive constants 119870
and 119870 such that(Lipschitz condition) for all 119909 119910 isin R119899 119905 isin [119905
0 119879] and
(1198941 1198942) isin S
1003816100381610038161003816119891 (119909 119905 1198941) minus 119891 (119910 119905 119894
1)10038161003816100381610038162
or1003816100381610038161003816119892 (119909 119905 119894
2) minus 119892 (119910 119905 119894
2)10038161003816100381610038162le 119870
1003816100381610038161003816119909 minus 11991010038161003816100381610038162
(42)
(Linear growth condition) for all (119909 119905 (1198941 1198942)) isin R119899 times
[1199050 119879] times S
1003816100381610038161003816119891 (119909 119905 1198941)10038161003816100381610038162or
1003816100381610038161003816119892 (119909 119905 1198942)10038161003816100381610038162le 119870(1 + |119909|)
2 (43)
Then there exists a unique solution 119909(119905) to (8) and moreoverthe solution obeys
119864( sup1199050le119905le119879
|119909 (119905)|2)
le (1 + 311986410038161003816100381610038161199090
10038161003816100381610038162) exp 3119870 (119879 minus 119905
0) (119879 minus 119905
0+ 4)
(44)
Proof Recall that almost every sample path of 120574(119905) =
(1205741(119905) 1205742(119905)) is a right continuous step function with a finite
number of jumps on [1199050 119879] Thus there exists a sequence of
stopping times 120591119896119896ge0
such that
(i) for almost every 120596 isin Ω there is a finite 119896120596for 1199050
=
1205910lt 1205911lt sdot sdot sdot lt 120591
119896= 119879 and 120591
119896= 119879 if 119896 gt 119896
120596
(ii) both 1205741(sdot) and 120574
2(sdot) in 120574(sdot) are constants on interval
[[120591119896 120591119896+1
[[ namely
1205741 (119905) = 120574
1(120591119896) on 120591
119896le 119905 le 120591
119896+1forall119896 ge 0
1205742(119905) = 120574
2(120591119896) on 120591
119896le 119905 le 120591
119896+1forall119896 ge 0
(45)
First of all let us consider (8) on 119905 isin [[1205910 1205911[[ then (8)
becomes119889119909 (119905) = 119891 (119909 (119905) 119905 120574
1(1199050)) 119889119905
+ 119892 (119909 (119905) 119905 1205742(1199050)) 119889119861 (119905)
(46)
with initial conditions 119909(1199050) = 119909
0 120574(1199050) = (120574
1(1199050) 1205741(1199050))
Then by the theory of SDEs we obtain that (46) has a uniquesolution which obeys 119909(120591
1) isin 1198712
F1205911
(ΩR119899) We next consider(8) on 119905 isin [[120591
1 1205912[[ which becomes
119889119909 (119905) = 119891 (119909 (119905) 119905 1205741 (1205911)) 119889119905
+ 119892 (119909 (119905) 119905 1205742(1205911)) 119889119861 (119905)
(47)
Again by the theory of SDEs (47) has a unique solutionwhichobeys 119909(120591
1) isin 119871
2
F1205912
(ΩR119899) Repeating this procedure weconclude that (8) has a unique solution 119909(119905) on [119905
0 119879]
Now let us prove (44) For every 119896 ge 1 define thestopping time
120591119896= 119879 and inf 119905 isin [119905
0 119879] |119909 (119905)| ge 119896 (48)
It is obvious that 120591119896uarr 119879 as Set 119909
119896(119905) = 119909(119905and120591
119896) for 119905 isin [119905
0 119879]
Then 119909119896(119905) obeys the equation
119909119896(119905) = 119909
0+ int
119905
1199050
119891 (119909119896(119904) 119904 120574
1(119904)) 119868[[1199050120591119896]](119904) 119889119904
+ int
119905
1199050
119892 (119909119896(119904) 119904 120574
2(119904)) 119868[[1199050120591119896]](119904) 119889119861 (119904)
(49)
Making use of the elementary inequality |119886+119887+119888|2le 3(|119886|
2+
|119887|2+ |119888|2) the Holder inequality and (43) we can see that
1003816100381610038161003816119909119896 (119905)10038161003816100381610038162= 3
1003816100381610038161003816119909010038161003816100381610038162+ 3119870 (119879 minus 119905
0) int
119905
1199050
(1 +1003816100381610038161003816119909119896 (119904)
10038161003816100381610038162) 119889119904
+ 3
100381610038161003816100381610038161003816100381610038161003816
int
119905
1199050
119892 (119909119896(119904) 119904 120574
2(119904)) 119868[[1199050120591119896]](119904) 119889119861 (119904)
100381610038161003816100381610038161003816100381610038161003816
2
(50)
Thus applying the Doob martingale inequality and (43) wecan further show that
119864( sup1199050le119904le119905
1003816100381610038161003816119909119896 (119904)10038161003816100381610038162)
= 311986410038161003816100381610038161199090
10038161003816100381610038162+ 3119870 (119879 minus 119905
0) int
119905
1199050
(1 + 1198641003816100381610038161003816119909119896 (119904)
10038161003816100381610038162) 119889119904
+ 12119864int
119905
1199050
1003816100381610038161003816119892 (119909119896(119904) 119904 120574
2(119904))
10038161003816100381610038162119868[[1199050120591119896]](119904) 119889119904
le 311986410038161003816100381610038161199090
10038161003816100381610038162+ 3119870 (119879 minus 119905
0+ 4)int
119905
1199050
(1 + 1198641003816100381610038161003816119909119896 (119904)
10038161003816100381610038162) 119889119904
(51)
That is to say
1 + 119864( sup1199050le119904le119905
1003816100381610038161003816119909119896 (119904)10038161003816100381610038162) le 1 + 3119864
1003816100381610038161003816119909010038161003816100381610038162+ 3119870 (119879 minus 119905
0+ 4)
times int
119905
1199050
[1 + 119864 sup1199050le119903le119904
1003816100381610038161003816119909119896 (119903)10038161003816100381610038162]119889119904
(52)
6 Journal of Applied Mathematics
Using the Gronwall inequality leads to
1 + 119864( sup1199050le119904le119905
1003816100381610038161003816119909119896 (119904)10038161003816100381610038162)
le (1 + 311986410038161003816100381610038161199090
10038161003816100381610038162) exp 3119870 (119879 minus 119905
0) (119879 minus 119905
0+ 4)
(53)
Consequently
1 + 119864( sup1199050le119905le120591119896
|119909 (119905)|2)
le (1 + 311986410038161003816100381610038161199090
10038161003816100381610038162) exp 3119870 (119879 minus 119905
0) (119879 minus 119905
0+ 4)
(54)
Then the required inequality (44) follows immediately byletting 119896 rarr infin
Condition (42) indicates that the coefficients 119891(119909 119905 1198941)
and 119892(119909 119905 1198942) do not change faster than a linear function of
119909 as change in 119909 This means in particular the continuity of119891(119909 119905 119894
1) and 119892(119909 119905 119894
2) in 119909 for all 119905 isin [119905
0 119879] Then functions
that are discontinuous with respect to 119909 are excluded as thecoefficients Besides there are many functions that do notsatisfy the Lipschitz conditionThese imply that the Lipschitzcondition is too restrictive To improve this Lipschitz condi-tion let us introduce the concept of local solution
Definition 8 Let120590infinbe a stopping time such that 119905
0le 120590infin
le 119879
as An R119899-valuedF119905-adapted continuous stochastic process
119909(119905)1199050le119905lt120590infin
is called a local solution of (8) if 119909(1199050) = 1199090and
moreover there is a nondecreasing sequence 120590119896119896ge1
such that1199050le 120590119896 uarr 120590infin
as and
119909 (119905) = 119909 (1199050) + int
119905and120590119896
1199050
119891 (119909 (119904) 119904 1205741(119904)) 119889119904
+ int
119905and120590119896
1199050
119892 (119909 (119904) 119904 1205742(119904)) 119889119861 (119904)
(55)
holds for any 119905 isin [1199050 119879) and 119896 ge 1 with probability one If
furthermore
lim sup119905rarr+120590
infin
|119909 (119905)| = infin whenever 120590infin
lt 119879 (56)
then it is called a maximal local solution and 120590infinis called the
explosion time Amaximal local solution 119909(119905) 1199050le 119905 lt 120590
infin
is said to be unique if any other maximal local solution 119909(119905)
1199050le 119905 lt 120590
infin is indistinguishable from it namely 120590
infin= 120590infin
and 119909(119905) = 119909(119905) for 1199050le 119905 lt 120590
infinwith probability one
Definition 9 (local Lipschitz condition) For every integer119896 ge 1 there exists a positive constant ℎ
119896such that for all
119905 isin [1199050 119879] 119894 = (119894
1 1198942) isin S and those 119909 119910 isin R119899 with
|119909| or |119910| le 119896
1003816100381610038161003816119891 (119909 119905 1198941) minus 119891 (119910 119905 119894
1)10038161003816100381610038162
or1003816100381610038161003816119892 (119909 119905 119894
2) minus 119892 (119910 119905 119894
2)10038161003816100381610038162le ℎ119896
1003816100381610038161003816119909 minus 11991010038161003816100381610038162
(57)
The following theorem shows the existence of uniquemaximal local solution under the local Lipschitz conditionwithout the linear growth condition
Theorem 10 Under condition (57) there exists a uniquemaximal local solution of (8)
Proof Define functions
119891119896(119909 119905 119894
1) =
119891 (119909 119905 1198941) if |119909| le 119896
119891(119896119909
|119909| 119905 1198941) if |119909| gt 119896
119892119896(119909 119905 119894
2) =
119892 (119909 119905 1198942) if |119909| le 119896
119892 (119896119909
|119909| 119905 1198942) if |119909| gt 119896
(58)
Then 119891119896and 119892
119896satisfy the Lipschitz condition and the linear
growth condition Thus by Theorem 7 there is a uniquesolution 119909
119896(119905) of the equation
119889119909119896(119905) = 119891
119896(119909119896(119905) 119905 120574
1(119905)) 119889119905
+ 119892119896(119909119896(119905) 119905 120574
2(119905)) 119889119861 (119905)
(59)
with the initial conditions 119909119896(1199050) = 119909
0and 120574(119905
0) = (120574
1(1199050)
1205742(1199050)) Define the stopping times
120590119896= 119879 and inf 119905 isin [119905
0 119879]
1003816100381610038161003816119909119896 (119905)1003816100381610038161003816 ge 119896 (60)
Clearly if 1199050le 119905 le 120590
119896
119909119896(119905) = 119909
119896+1(119905) (61)
which indicates that 120590119896is increasing so 120590
119896has its limit 120590
infin=
lim119896rarr+infin
120590119896 Define 119909(119905) 119905
0le 119905 lt 120590
infin by
119909 (119905) = 119909119896(119905) 119905 isin [[120590
119896minus1 120590119896[[ 119896 ge 1 (62)
where 1205900= 1199050 Applying (61) one can show that 119909(119905 and 120590
119896) =
119909119896(119905 and 120590119896) It then follows from (59) that
119909 (119905 and 120590119896) = 1199090+ int
119905and120590119896
1199050
119891119896(119909 (119904) 119904 120574
1(119904)) 119889119904
+ int
119905and120590119896
1199050
119892119896(119909 (119904) 119904 1205742 (119904)) 119889119861 (119904)
= 1199090+ int
119905and120590119896
1199050
119891 (119909 (119904) 119904 1205741(119904)) 119889119904
+ int
119905and120590119896
1199050
119892 (119909 (119904) 119904 1205742 (119904)) 119889119861 (119904)
(63)
for any 119905 isin [1199050 119879) and 119896 ge 1 It is easy to see that if 120590
infinle 119879
then
lim sup119905rarr120590infin
|119909 (119905)|
ge lim sup119896rarr+infin
1003816100381610038161003816119909 (120590119896)1003816100381610038161003816 = lim sup119896rarr+infin
1003816100381610038161003816119909119896 (120590119896)1003816100381610038161003816 = infin
(64)
Therefore 119909(119905) 1199050le 119905 lt 120590
infin is a maximal local solution
Now we will prove the uniqueness Let 119909(119905) 1199050
le 119905 lt
120590infin
be another maximal local solution Define
120590119896= 120590infin
and inf 119905 isin [[1199050 120590infin
[[ 119909 (119905) ge 119896 (65)
Journal of Applied Mathematics 7
Then 120590119896
rarr 120590infin
as and
119875 119909 (119905) = 119909 (119905) 119905 isin [[1199050 120590119896and 120590119896]] = 1 forall119896 ge 1 (66)
Letting 119896 rarr infin gives
119875 119909 (119905) = 119909 (119905) 119905 isin [[1199050 120590infin
and 120590infin
[[ = 1 (67)
In order to complete the proof we need only to show that120590infin
= 120590infin
as In fact for almost every 120596 isin 120590infin
lt 120590infin
wehave
1003816100381610038161003816119909 (120590infin
120596)1003816100381610038161003816
= lim119896rarr+infin
1003816100381610038161003816119909 (120590119896 120596)
1003816100381610038161003816 = lim119896rarr+infin
1003816100381610038161003816119909 (120590119896 120596)
1003816100381610038161003816 = infin(68)
which contradicts the fact that 119909(119905 120596) is continuous on 119905 isin
[1199050 120590infin
(120596)) This implies 120590infin
ge 120590infin
as In the same way onecan show 120590
infinle 120590infinasThus wemust have 120590
infin= 120590infinasThis
completes the proof
In many situations we often consider an SDE on [1199050infin)
119889119909 (119905) = 119891 (119909 (119905) 119905 1205741(119905)) 119889119905
+ 119892 (119909 (119905) 119905 1205742(119905)) 119889119861 (119905) 119905
0le 119905 lt infin
(69)
with initial data 119909(1199050) = 119909
0and 120574(119905
0) If the assumption of
the existence-and-uniqueness theorem holds on every finitesubinterval [119905
0 119879] of [119905
0infin) then (69) has a unique solution
119909(119905) on the entire interval [1199050infin) Such a solution is called
a global solution To establish a more general result aboutglobal solution we need more notations To this end weintroduce an operator 119871119881 from R119899 times R
+times S to R which is
given by
119871119881 (119909 119905 (1198941 1198942))
= 119881119905(119909 119905 (119894
1 1198942)) + 119881
119909(119909 119905 (119894
1 1198942)) 119891 (119909 119905 119894
1)
+ 05 trace [119892119879 (119909 119905 1198942) 119881119909119909
(119909 119905 (1198941 1198942)) 119892 (119909 119905 119894
2)]
+ sum
(11989511198952)isinS
1199021198941119894211989511198952
119881 (119909 119905 (1198951 1198952))
(70)
where 119881(119909 119905 (1198941 1198942)) isin 119862
21(R119899 times R
+times SR
+) and
119881119905(119909 119905 (119894
1 1198942)) =
120597119881 (119909 119905 (1198941 1198942))
120597119905
119881119909(119909 119905 (119894
1 1198942))
= (120597119881 (119909 119905 (119894
1 1198942))
1205971199091
120597119881 (119909 119905 (119894
1 1198942))
120597119909119899
)
119881119909119909
(119909 119905 (1198941 1198942)) = (
1205972119881 (119909 119905 (119894
1 1198942))
120597119909119894120597119909119895
)
119899times119899
(71)
Theorem 11 Assume that the local Lipschitz condition (57)holds Assume also that there is a function 119881(119909 119905 (119894
1 1198942)) isin
11986221
(R119899 times R+times SR
+) and a constant 120579 gt 0 such that
lim|119909|rarrinfin
( inf(119905(1198941 1198942))isinR+timesS
119881 (119909 119905 (1198941 1198942))) = infin (72)
119871119881 (119909 119905 (1198941 1198942)) le 120579 (1 + 119881 (119909 119905 (119894
1 1198942)))
forall (119909 119905 (1198941 1198942)) isin R
119899times R+times S
(73)
Then there exists a unique global solution 119909(119905) to (69)
Proof We need only to prove the theorem for any initialcondition 119909
0isin R119899 and (120574
1(1199050) 1205742(1199050)) isin S FromTheorem 10
we know that the local Lipschitz condition guarantees theexistence of the unique maximal solution 119909(119905) on [119905
0 120590infin
)where 120590
infinis the explosion time We need only to show 120590
infin=
infin as If this is not true then we can find a pair of positiveconstants 120576 and 119879 such that
119875 120590infin
le 119879 gt 2120576 (74)
For each integer 119896 ge 1 define the stopping time
120590119896= inf 119905 ge 119905
0 |119909 (119905)| ge 119896 (75)
Since 120590119896
rarr 120590infinalmost surely we can find a sufficiently large
integer 1198960for
119875 120590119896le 119879 gt 120576 forall119896 ge 119896
0 (76)
Fix any 119896 ge 1198960 then for any 119905
0le 119905 le 119879 by virtue of the
generalized Ito formula (see eg [1])
119864119881 (119909 (119905 and 120590119896) 119905 and 120590
119896 120574 (119905 and 120590
119896))
= 119881 (1199090 1199050 1205740) + 119864int
119905and120590119896
1199050
119871119881 (119909 (119904) 119904 120574 (119904)) 119889119904
le 119881 (1199090 1199050 1205740) + 120579 (119879 minus 119905
0)
+ 120579int
119905
1199050
119864119881 (119909 (119904 and 120590119896) 119904 and 120590
119896 120574 (119904 and 120590
119896)) 119889119904
(77)
Making use of the Gronwall inequality gives
119864119881 (119909 (119905 and 120590119896) 119905 and 120590
119896 120574 (119905 and 120590
119896))
le [119881 (1199090 1199050 1205740) + 120579 (119879 minus 119905
0)] exp 120579 (119879 minus 119905
0)
(78)
Therefore
119864 (119868120590119896le119879
119881 (119909 (120590119896) 120590119896 120574 (120590119896)))
le [119881 (1199090 1199050 1205740) + 120579 (119879 minus 119905
0)] exp 120579 (119879 minus 119905
0)
(79)
At the same time set
119892119896= inf 119881 (119909 119905 (119894
1 1198942)) |119909| ge 119896 119905 isin [119905
0 119879] (119894
1 1198942) isin S
(80)
8 Journal of Applied Mathematics
Then (72) means 119892119896
rarr infin It follows from (76) and (79) that
[119881 (1199090 1199050 1205740) + 120579 (119879 minus 119905
0)] exp 120579 (119879 minus 119905
0)
ge 119892119896119875 120590119896le 119879 ge 120576119892
119896
(81)
Letting 119896 rarr infin yields a contradiction that is to say120590infin
= infinThe proof is complete
3 119871119875-Estimates
In the previous section we have investigated the existenceand uniqueness of the solution to (8) In this section as abovelet 119909(119905) 119905
0le 119905 le 119879 be the unique solution of (8) with initial
conditions 119909(1199050) = 1199090and 120574(119905
0) and we will estimate the 119901th
moment of the solution
Theorem 12 Assume that there is a function 119881(119909 119905 (1198941 1198942)) isin
11986221
(R119899 times R+times SR
+) and positive constants 119901 120578 120579 such that
for all (119909 119905 (1198941 1198942)) isin R119899 times R
+times S
120578|119909|119901le 119881 (119909 119905 (119894
1 1198942)) (82)
119871119881 (119909 119905 (1198941 1198942)) le 120579119881 (119909 119905 (119894
1 1198942)) (83)
Assume also the initial condition 119909(1199050) = 119909
0and 120574(119905
0) obeys
119864119881(1199090 1199050 120574(1199050)) lt infin then one has
119864|119909 (119905)|119901
le119864119881 (119909
0 1199050 120574 (1199050)) exp 120579 (119905 minus 119905
0)
120578 forall119905 isin [119905
0 119879]
(84)
Proof For each integer 119896 ge 1 define the stopping time
120590119896= 119879 and inf 119905 ge 119905
0 |119909 (119905)| ge 119896 (85)
Thus 120590119896
rarr 119879 as Using the generalized Itorsquos formula and(83) we obtain that for 119905 isin [119905
0 119879]
119864119881 (119909 (119905 and 120590119896) 119905 and 120590
119896 120574 (119905 and 120590
119896))
le 119864119881 (1199090 1199050 120574 (1199050))
+ 120579int
119905
1199050
119864119881 (119909 (119904 and 120590119896) 119904 and 120590
119896 120574 (119904 and 120590
119896)) 119889119904
(86)
Then the Gronwall inequality indicates
119864119881 (119909 (119905 and 120590119896) 119905 and 120590
119896 120574 (119905 and 120590
119896))
le 119864119881 (1199090 1199050 120574 (1199050)) exp 120579 (119905 minus 119905
0)
(87)
for all 119905 isin [1199050 119879] By virtue of condition (83) we obtain the
required assertion (84)
Corollary 13 Assume 119901 ge 2 and 1199090isin 119871119901
F1199050
(ΩR119899) Assumealso that there exists a constant 120579 gt 0 such that for all(119909 119905 (119894
1 1198942)) isin R119899 times R
+times S
119909119879119891 (119909 119905 119894
1) +
119901 minus 1
2
1003816100381610038161003816119892 (119909 119905 1198942)10038161003816100381610038162le 120579 (1 + |119909|
2) (88)
Then one has
119864|119909 (119905)|119901
le 2(119901minus2)2
(1 + 11986410038161003816100381610038161199090
1003816100381610038161003816119901) exp 119901120579 (119905 minus 119905
0) forall119905 isin [119905
0 119879]
(89)
Proof Define 119881(119909 119905 (1198941 1198942)) = (1 + |119909|
2)1199012 Making use of
(88) yields
119871119881 (119909 119905 (1198941 1198942))
= 119901(1 + |119909|2)(119901minus2)2
119909119879119891 (119909 119905 119894
1)
+119901
2(1 + |119909|
2)(119901minus2)21003816100381610038161003816119892 (119909 119905 119894
2)10038161003816100381610038162
+119901 (119901 minus 2)
2(1 + |119909|
2)(119901minus4)210038161003816100381610038161003816
119909119879119892 (119909 119905 119894
2)10038161003816100381610038161003816
2
le 119901(1 + |119909|2)(119901minus2)2
[119909119879119891 (119909 119905 119894
1) +
119901 minus 1
2
1003816100381610038161003816119892 (119909 119905 1198942)10038161003816100381610038162]
le 119901120579(1 + |119909|2)1199012
(90)
Then byTheorem 12 we get
119864(1 + |119909 (119905)|2)1199012
le 119864(1 +10038161003816100381610038161199090
10038161003816100381610038162)1199012
exp 119901120579 (119905 minus 1199050)
(91)
and the required assertion (89) followsIt is useful to point out that if the linear growth condition
(43) is satisfied then (88) is fulfilled with 120579 = radic119870 + 119870(119901 minus
1)2 Now we will show the other important properties of thesolution
Theorem 14 Let 119901 ge 2 and 1199090
isin 119871119901
F1199050
(ΩR119899) Assume alsothat the linear growth condition (43) holds Then one has
119864|119909 (119905) minus 119909 (119904)|119901le 119862(119905 minus 119904)
1199012 (92)
where
119862 = 2119901minus2
(1 + 11986410038161003816100381610038161199090
1003816100381610038161003816119901) exp 119901120579 (119879 minus 119905
0)
times ([2 (119879 minus 1199050)]1199012
+ [119901 (119901 minus 1)]1199012
)
(93)
and 120579 = radic119870+119870(119901minus1)2 Particularly the 119901th moment of thesolution is continuous on [119905
0 119879]
Journal of Applied Mathematics 9
Proof Applying the elementary inequality |119886 + 119887|119901
le
2119901minus1
(|119886|119901+ |119887|119901) the Holder inequality and the linear growth
condition we can derive that
119864|119909 (119905) minus 119909 (119904)|119901
le 2119901minus1
119864
10038161003816100381610038161003816100381610038161003816int
119905
119904
119891 (119909 (119906) 119906 1205741(119906)) 119889119906
10038161003816100381610038161003816100381610038161003816
119901
+ 2119901minus1
119864
10038161003816100381610038161003816100381610038161003816int
119905
119904
119892 (119909 (119906) 119906 1205742(119906)) 119889119861 (119906)
10038161003816100381610038161003816100381610038161003816
119901
le [2 (119905 minus 119904)]119901minus1
119864int
119905
119904
1003816100381610038161003816119891 (119909 (119906) 119906 1205741(119906))
1003816100381610038161003816119901119889119906
+ 05[2119901 (119901 minus 1)]1199012
(119905 minus 119904)(119901minus2)2
119864
times int
119905
119904
1003816100381610038161003816119892 (119909 (119906) 119906 1205742 (119906))1003816100381610038161003816119901119889119906
le 1198621(119905 minus 119904)
(119901minus2)2int
119905
119904
119864(1 + |119909 (119906)|2)1199012
119889119906
(94)
where1198621= 2(119901minus2)2
1198701199012
([2(119879minus 1199050)]1199012
+ [119901(119901minus1)]1199012
) Using(91) yields
119864|119909 (119905) minus 119909 (119904)|119901
le 1198621(119905 minus 119904)
(119901minus2)2
times int
119905
119904
2(119901minus2)2
(1 + 11986410038161003816100381610038161199090
1003816100381610038161003816119901) exp 119901120579 (119906 minus 119905
0) 119889119906
le 11986212(119901minus2)2
(1 + 11986410038161003816100381610038161199090
1003816100381610038161003816119901)
times exp 119901120579 (119879 minus 1199050) (119905 minus 119904)
(1199012)
(95)
which is the desired inequality
Theorem 15 Let 119901 ge 2 and 1199090
isin 119871119901
F1199050
(ΩR119899) Assume thatthere is a119870 gt 0 such that for all (119909 119905 (119894
1 1198942)) isin R119899times[119905
0 119879]timesS
119909119879 1003816100381610038161003816119891 (119909 119905 119894
1)1003816100381610038161003816 or
1003816100381610038161003816119892 (119909 119905 1198942)10038161003816100381610038162le 119870(1 + |119909|)
2 (96)
Then
119864( sup1199050le119905le119879
|119909 (119905)|119901)
le (1 + 211986410038161003816100381610038161199090
1003816100381610038161003816119901) exp 2119901 (10119901 + 1)119870 (119879 minus 119905
0)
(97)
Proof Making use of the generalized Itorsquos formula andcondition (96) we have
|119909 (119905)|119901
le10038161003816100381610038161199090
1003816100381610038161003816119901+ 119872(119905)
+ int
119905
1199050
119901|119909 (119904)|119901minus2
[119909119879(119904) 119891 (119909 (119904) 119904 120574
1(119904))
+119901 minus 1
2
1003816100381610038161003816119892 (119909 (119904) 119904 1205742 (119904))10038161003816100381610038162] 119889119904
le10038161003816100381610038161199090
1003816100381610038161003816119901+ 119872(119905) + 05119901 (119901 + 1)119870
times int
119905
1199050
|119909 (119904)|119901minus2
(1 + |119909 (119904)|2) 119889119904
le10038161003816100381610038161199090
1003816100381610038161003816119901+ 119872(119905) + 119901 (119901 + 1)119870int
119905
1199050
(1 + |119909 (119904)|119901) 119889119904
(98)
where
119872(119905) = int
119905
1199050
119901|119909 (119904)|119901minus2
119909119879(119904) 119892 (119909 (119904) 119904 1205742 (119904)) 119889119861 (119904) (99)
Therefore for any 1199051isin [1199050 119879]
119864( sup1199050le119905le1199051
|119909 (119905)|119901) le 119864
100381610038161003816100381611990901003816100381610038161003816119901+ 119864( sup
1199050le119905le1199051
|119872 (119905)|)
+ 119901 (119901 + 1)119870119864int
1199051
1199050
(1 + |119909 (119904)|119901) 119889119904
(100)
At the same time applying the well-known Burkholder-Davis-Gundy inequality (see eg [2]) gives
119864( sup1199050le119905le1199051
|119872 (119905)|)
le 3119864(int
1199051
1199050
1199012|119909 (119904)|
2119901minus21003816100381610038161003816119892 (119909 (119904) 119904 1205742 (119904))10038161003816100381610038162119889119904)
05
le 3119864( sup1199050le119905le1199051
|119909 (119905)|119901
timesint
1199051
1199050
1199012|119909 (119904)|
119901minus21003816100381610038161003816119892 (119909 (119904) 119904 1205742(119904))
10038161003816100381610038162119889119904)
05
le 05119864( sup1199050le119905le1199051
|119909 (119905)|119901) + 45119864
times int
1199051
1199050
1199012119870|119909 (119904)|
119901minus2(1 + |119909 (119904)|
2) 119889119904
le 05119864( sup1199050le119905le1199051
|119909 (119905)|119901) + 9119901
2119870119864int
1199051
1199050
(1 + |119909 (119904)|119901) 119889119904
(101)
10 Journal of Applied Mathematics
Substituting the above inequality into (100) gives
119864( sup1199050le119905le1199051
|119909 (119905)|119901) le 2119864
100381610038161003816100381611990901003816100381610038161003816119901
+ 2119901 (10119901 + 1)119870int
1199051
1199050
(1 + |119909 (119904)|119901) 119889119904
(102)
Thus
1 + 119864( sup1199050le119905le1199051
|119909 (119905)|119901)
le 1 + 211986410038161003816100381610038161199090
1003816100381610038161003816119901+ 2119901 (10119901 + 1)119870
times int
1199051
1199050
(1 + sup1199050le119905le119904
|119909 (119905)|119901)119889119904
(103)
Then the required assertion follows from the Gronwallinequality
Up to now we have discussed the 119871119901-estimates for the
solution in the case when 119901 ge 2 As for 0 lt 119901 lt 2 the similarresults can be given without any difficulty as long as we notethat the Holder inequality implies
119864|119909 (119905)|119901le [119864|119909 (119905)|
2]05119901
(104)
4 Example
Consider the following Black-Scholes model
119889119883 (119905) = 120583 (1205741(119905))119883 (119905) 119889119905 + ] (120574
2(119905))119883 (119905) 119889119861 (119905) (105)
where 1205741(119905) is a right-continuous homogenous Markovian
chain taking values in finite state spaces S1= 1 2 and 120574
2(119905)
is a right-continuous homogenous Markovian chain takingvalues in finite state spaces S
2= 1 2 3 120583(119894) = 119894 119894 = 1 2
](119895) = 119895 + 1 119895 = 1 2 3 Taking 119870 = 16 119870 = 9 then (42)and (43) hold Therefore by Theorem 7 (105) has a uniquesolution
5 Conclusions and Further Research
This paper is devoted to studying the existence and unique-ness of solution of SDEs with multi-Markovian switchingsand estimating the119901thmoment of the solutionWe have usedtwo continuous-time Markovian chains to model the SDEsThis area is becoming increasingly useful in engineeringeconomics communication theory active networking andso forth The sufficient criteria for existence and unique-ness of solution local solution and maximal local solutionwere established Those results indicate that (8) keeps manyproperties that (89) owns At the same time although thehypothesis (H1) is used in this paper wewant to point out thatthis hypothesis is not essential In fact (H1) can be replacedby the following generalized hypothesis
(H1)1015840 both 1205741(119905) and 120574
2(119905) are right-continuous homoge-
nous Markovian chains such that 120574(119905) = (1205741(119905) 1205742(119905)) is a
homogenous vector chainUnder hypothesis (H1)1015840 the results given in this paper
can be established similarly It is easy to see that if 1205741(119905) equiv
1205742(119905) and 120574
1(119905) is a right-continuous homogenous Markovian
chain then (H1)1015840 is fulfilled immediately At the same timeif 1205741(119905) equiv 120574
2(119905) (8) will reduce to the classical SDEs with
single Markovian chain that is to say the classical theoryabout SDEs with single Markovian chain is a special caseof our theory On the other hand many theorems in thispaper will play important roles in further study For exampleTheorem 15 will be useful when one studies the approximatesolutions
Some important and interesting questions can be furtherinvestigated using the results in this paper For exampleapproximate solutions boundedness and stability stochas-tic functional differential equations with vector Markovianswitching and their applications In particular the stability of(8) is one of the most important and interesting topics andthose investigations are in progress
Acknowledgments
The authors thank the editor and referees for their veryimportant and helpful comments and suggestions Theauthors also thank Dr C Zhang for helping them to improvethe English exposition This research is supported by theNSFC of China (nos 11171081 and 11171056)
References
[1] A V SkorokhodAsymptoticMethods in theTheory of StochasticDifferential Equations vol 78 of Translations of MathematicalMonographs American Mathematical Society Providence RIUSA 1989
[2] X Mao and C Yuan Stochastic Differential Euations withMarkovian Switching Imperial College Press London UK2006
[3] G Yin and X Y Zhou ldquoMarkowitzrsquos mean-variance portfolioselection with regime switching from discrete-time models totheir continuous-time limitsrdquo IEEE Transactions on AutomaticControl vol 49 no 3 pp 349ndash360 2004
[4] J Buffington and R J Elliott ldquoAmerican options with regimeswitchingrdquo International Journal of Theoretical and AppliedFinance vol 5 no 5 pp 497ndash514 2002
[5] C Zhu and G Yin ldquoOn competitive Lotka-Volterra model inrandom environmentsrdquo Journal of Mathematical Analysis andApplications vol 357 no 1 pp 154ndash170 2009
[6] Q Luo and X Mao ldquoStochastic population dynamics underregime switchingrdquo Journal of Mathematical Analysis and Appli-cations vol 334 no 1 pp 69ndash84 2007
[7] X Li D Jiang and X Mao ldquoPopulation dynamical behaviorof Lotka-Volterra system under regime switchingrdquo Journal ofComputational and Applied Mathematics vol 232 no 2 pp427ndash448 2009
[8] X Li A Gray D Jiang and X Mao ldquoSufficient and neces-sary conditions of stochastic permanence and extinction forstochastic logistic populations under regime switchingrdquo Journal
Journal of Applied Mathematics 11
ofMathematical Analysis and Applications vol 376 no 1 pp 11ndash28 2011
[9] M Liu and KWang ldquoAsymptotic properties and simulations ofa stochastic logistic model under regime switchingrdquoMathemat-ical and Computer Modelling vol 54 no 9-10 pp 2139ndash21542011
[10] G Hu and K Wang ldquoStability in distribution of competi-tive Lotka-Volterra system with Markovian switchingrdquo AppliedMathematical Modelling vol 35 no 7 pp 3189ndash3200 2011
[11] M Liu and K Wang ldquoAsymptotic properties and simulationsof a stochastic logistic model under regime switching IIrdquoMathematical andComputerModelling vol 55 no 3-4 pp 405ndash418 2012
[12] ZWu H Huang and LWang ldquoStochastic delay logistic modelunder regime switchingrdquo Abstract and Applied Analysis vol2012 Article ID 241702 26 pages 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
6 Journal of Applied Mathematics
Using the Gronwall inequality leads to
1 + 119864( sup1199050le119904le119905
1003816100381610038161003816119909119896 (119904)10038161003816100381610038162)
le (1 + 311986410038161003816100381610038161199090
10038161003816100381610038162) exp 3119870 (119879 minus 119905
0) (119879 minus 119905
0+ 4)
(53)
Consequently
1 + 119864( sup1199050le119905le120591119896
|119909 (119905)|2)
le (1 + 311986410038161003816100381610038161199090
10038161003816100381610038162) exp 3119870 (119879 minus 119905
0) (119879 minus 119905
0+ 4)
(54)
Then the required inequality (44) follows immediately byletting 119896 rarr infin
Condition (42) indicates that the coefficients 119891(119909 119905 1198941)
and 119892(119909 119905 1198942) do not change faster than a linear function of
119909 as change in 119909 This means in particular the continuity of119891(119909 119905 119894
1) and 119892(119909 119905 119894
2) in 119909 for all 119905 isin [119905
0 119879] Then functions
that are discontinuous with respect to 119909 are excluded as thecoefficients Besides there are many functions that do notsatisfy the Lipschitz conditionThese imply that the Lipschitzcondition is too restrictive To improve this Lipschitz condi-tion let us introduce the concept of local solution
Definition 8 Let120590infinbe a stopping time such that 119905
0le 120590infin
le 119879
as An R119899-valuedF119905-adapted continuous stochastic process
119909(119905)1199050le119905lt120590infin
is called a local solution of (8) if 119909(1199050) = 1199090and
moreover there is a nondecreasing sequence 120590119896119896ge1
such that1199050le 120590119896 uarr 120590infin
as and
119909 (119905) = 119909 (1199050) + int
119905and120590119896
1199050
119891 (119909 (119904) 119904 1205741(119904)) 119889119904
+ int
119905and120590119896
1199050
119892 (119909 (119904) 119904 1205742(119904)) 119889119861 (119904)
(55)
holds for any 119905 isin [1199050 119879) and 119896 ge 1 with probability one If
furthermore
lim sup119905rarr+120590
infin
|119909 (119905)| = infin whenever 120590infin
lt 119879 (56)
then it is called a maximal local solution and 120590infinis called the
explosion time Amaximal local solution 119909(119905) 1199050le 119905 lt 120590
infin
is said to be unique if any other maximal local solution 119909(119905)
1199050le 119905 lt 120590
infin is indistinguishable from it namely 120590
infin= 120590infin
and 119909(119905) = 119909(119905) for 1199050le 119905 lt 120590
infinwith probability one
Definition 9 (local Lipschitz condition) For every integer119896 ge 1 there exists a positive constant ℎ
119896such that for all
119905 isin [1199050 119879] 119894 = (119894
1 1198942) isin S and those 119909 119910 isin R119899 with
|119909| or |119910| le 119896
1003816100381610038161003816119891 (119909 119905 1198941) minus 119891 (119910 119905 119894
1)10038161003816100381610038162
or1003816100381610038161003816119892 (119909 119905 119894
2) minus 119892 (119910 119905 119894
2)10038161003816100381610038162le ℎ119896
1003816100381610038161003816119909 minus 11991010038161003816100381610038162
(57)
The following theorem shows the existence of uniquemaximal local solution under the local Lipschitz conditionwithout the linear growth condition
Theorem 10 Under condition (57) there exists a uniquemaximal local solution of (8)
Proof Define functions
119891119896(119909 119905 119894
1) =
119891 (119909 119905 1198941) if |119909| le 119896
119891(119896119909
|119909| 119905 1198941) if |119909| gt 119896
119892119896(119909 119905 119894
2) =
119892 (119909 119905 1198942) if |119909| le 119896
119892 (119896119909
|119909| 119905 1198942) if |119909| gt 119896
(58)
Then 119891119896and 119892
119896satisfy the Lipschitz condition and the linear
growth condition Thus by Theorem 7 there is a uniquesolution 119909
119896(119905) of the equation
119889119909119896(119905) = 119891
119896(119909119896(119905) 119905 120574
1(119905)) 119889119905
+ 119892119896(119909119896(119905) 119905 120574
2(119905)) 119889119861 (119905)
(59)
with the initial conditions 119909119896(1199050) = 119909
0and 120574(119905
0) = (120574
1(1199050)
1205742(1199050)) Define the stopping times
120590119896= 119879 and inf 119905 isin [119905
0 119879]
1003816100381610038161003816119909119896 (119905)1003816100381610038161003816 ge 119896 (60)
Clearly if 1199050le 119905 le 120590
119896
119909119896(119905) = 119909
119896+1(119905) (61)
which indicates that 120590119896is increasing so 120590
119896has its limit 120590
infin=
lim119896rarr+infin
120590119896 Define 119909(119905) 119905
0le 119905 lt 120590
infin by
119909 (119905) = 119909119896(119905) 119905 isin [[120590
119896minus1 120590119896[[ 119896 ge 1 (62)
where 1205900= 1199050 Applying (61) one can show that 119909(119905 and 120590
119896) =
119909119896(119905 and 120590119896) It then follows from (59) that
119909 (119905 and 120590119896) = 1199090+ int
119905and120590119896
1199050
119891119896(119909 (119904) 119904 120574
1(119904)) 119889119904
+ int
119905and120590119896
1199050
119892119896(119909 (119904) 119904 1205742 (119904)) 119889119861 (119904)
= 1199090+ int
119905and120590119896
1199050
119891 (119909 (119904) 119904 1205741(119904)) 119889119904
+ int
119905and120590119896
1199050
119892 (119909 (119904) 119904 1205742 (119904)) 119889119861 (119904)
(63)
for any 119905 isin [1199050 119879) and 119896 ge 1 It is easy to see that if 120590
infinle 119879
then
lim sup119905rarr120590infin
|119909 (119905)|
ge lim sup119896rarr+infin
1003816100381610038161003816119909 (120590119896)1003816100381610038161003816 = lim sup119896rarr+infin
1003816100381610038161003816119909119896 (120590119896)1003816100381610038161003816 = infin
(64)
Therefore 119909(119905) 1199050le 119905 lt 120590
infin is a maximal local solution
Now we will prove the uniqueness Let 119909(119905) 1199050
le 119905 lt
120590infin
be another maximal local solution Define
120590119896= 120590infin
and inf 119905 isin [[1199050 120590infin
[[ 119909 (119905) ge 119896 (65)
Journal of Applied Mathematics 7
Then 120590119896
rarr 120590infin
as and
119875 119909 (119905) = 119909 (119905) 119905 isin [[1199050 120590119896and 120590119896]] = 1 forall119896 ge 1 (66)
Letting 119896 rarr infin gives
119875 119909 (119905) = 119909 (119905) 119905 isin [[1199050 120590infin
and 120590infin
[[ = 1 (67)
In order to complete the proof we need only to show that120590infin
= 120590infin
as In fact for almost every 120596 isin 120590infin
lt 120590infin
wehave
1003816100381610038161003816119909 (120590infin
120596)1003816100381610038161003816
= lim119896rarr+infin
1003816100381610038161003816119909 (120590119896 120596)
1003816100381610038161003816 = lim119896rarr+infin
1003816100381610038161003816119909 (120590119896 120596)
1003816100381610038161003816 = infin(68)
which contradicts the fact that 119909(119905 120596) is continuous on 119905 isin
[1199050 120590infin
(120596)) This implies 120590infin
ge 120590infin
as In the same way onecan show 120590
infinle 120590infinasThus wemust have 120590
infin= 120590infinasThis
completes the proof
In many situations we often consider an SDE on [1199050infin)
119889119909 (119905) = 119891 (119909 (119905) 119905 1205741(119905)) 119889119905
+ 119892 (119909 (119905) 119905 1205742(119905)) 119889119861 (119905) 119905
0le 119905 lt infin
(69)
with initial data 119909(1199050) = 119909
0and 120574(119905
0) If the assumption of
the existence-and-uniqueness theorem holds on every finitesubinterval [119905
0 119879] of [119905
0infin) then (69) has a unique solution
119909(119905) on the entire interval [1199050infin) Such a solution is called
a global solution To establish a more general result aboutglobal solution we need more notations To this end weintroduce an operator 119871119881 from R119899 times R
+times S to R which is
given by
119871119881 (119909 119905 (1198941 1198942))
= 119881119905(119909 119905 (119894
1 1198942)) + 119881
119909(119909 119905 (119894
1 1198942)) 119891 (119909 119905 119894
1)
+ 05 trace [119892119879 (119909 119905 1198942) 119881119909119909
(119909 119905 (1198941 1198942)) 119892 (119909 119905 119894
2)]
+ sum
(11989511198952)isinS
1199021198941119894211989511198952
119881 (119909 119905 (1198951 1198952))
(70)
where 119881(119909 119905 (1198941 1198942)) isin 119862
21(R119899 times R
+times SR
+) and
119881119905(119909 119905 (119894
1 1198942)) =
120597119881 (119909 119905 (1198941 1198942))
120597119905
119881119909(119909 119905 (119894
1 1198942))
= (120597119881 (119909 119905 (119894
1 1198942))
1205971199091
120597119881 (119909 119905 (119894
1 1198942))
120597119909119899
)
119881119909119909
(119909 119905 (1198941 1198942)) = (
1205972119881 (119909 119905 (119894
1 1198942))
120597119909119894120597119909119895
)
119899times119899
(71)
Theorem 11 Assume that the local Lipschitz condition (57)holds Assume also that there is a function 119881(119909 119905 (119894
1 1198942)) isin
11986221
(R119899 times R+times SR
+) and a constant 120579 gt 0 such that
lim|119909|rarrinfin
( inf(119905(1198941 1198942))isinR+timesS
119881 (119909 119905 (1198941 1198942))) = infin (72)
119871119881 (119909 119905 (1198941 1198942)) le 120579 (1 + 119881 (119909 119905 (119894
1 1198942)))
forall (119909 119905 (1198941 1198942)) isin R
119899times R+times S
(73)
Then there exists a unique global solution 119909(119905) to (69)
Proof We need only to prove the theorem for any initialcondition 119909
0isin R119899 and (120574
1(1199050) 1205742(1199050)) isin S FromTheorem 10
we know that the local Lipschitz condition guarantees theexistence of the unique maximal solution 119909(119905) on [119905
0 120590infin
)where 120590
infinis the explosion time We need only to show 120590
infin=
infin as If this is not true then we can find a pair of positiveconstants 120576 and 119879 such that
119875 120590infin
le 119879 gt 2120576 (74)
For each integer 119896 ge 1 define the stopping time
120590119896= inf 119905 ge 119905
0 |119909 (119905)| ge 119896 (75)
Since 120590119896
rarr 120590infinalmost surely we can find a sufficiently large
integer 1198960for
119875 120590119896le 119879 gt 120576 forall119896 ge 119896
0 (76)
Fix any 119896 ge 1198960 then for any 119905
0le 119905 le 119879 by virtue of the
generalized Ito formula (see eg [1])
119864119881 (119909 (119905 and 120590119896) 119905 and 120590
119896 120574 (119905 and 120590
119896))
= 119881 (1199090 1199050 1205740) + 119864int
119905and120590119896
1199050
119871119881 (119909 (119904) 119904 120574 (119904)) 119889119904
le 119881 (1199090 1199050 1205740) + 120579 (119879 minus 119905
0)
+ 120579int
119905
1199050
119864119881 (119909 (119904 and 120590119896) 119904 and 120590
119896 120574 (119904 and 120590
119896)) 119889119904
(77)
Making use of the Gronwall inequality gives
119864119881 (119909 (119905 and 120590119896) 119905 and 120590
119896 120574 (119905 and 120590
119896))
le [119881 (1199090 1199050 1205740) + 120579 (119879 minus 119905
0)] exp 120579 (119879 minus 119905
0)
(78)
Therefore
119864 (119868120590119896le119879
119881 (119909 (120590119896) 120590119896 120574 (120590119896)))
le [119881 (1199090 1199050 1205740) + 120579 (119879 minus 119905
0)] exp 120579 (119879 minus 119905
0)
(79)
At the same time set
119892119896= inf 119881 (119909 119905 (119894
1 1198942)) |119909| ge 119896 119905 isin [119905
0 119879] (119894
1 1198942) isin S
(80)
8 Journal of Applied Mathematics
Then (72) means 119892119896
rarr infin It follows from (76) and (79) that
[119881 (1199090 1199050 1205740) + 120579 (119879 minus 119905
0)] exp 120579 (119879 minus 119905
0)
ge 119892119896119875 120590119896le 119879 ge 120576119892
119896
(81)
Letting 119896 rarr infin yields a contradiction that is to say120590infin
= infinThe proof is complete
3 119871119875-Estimates
In the previous section we have investigated the existenceand uniqueness of the solution to (8) In this section as abovelet 119909(119905) 119905
0le 119905 le 119879 be the unique solution of (8) with initial
conditions 119909(1199050) = 1199090and 120574(119905
0) and we will estimate the 119901th
moment of the solution
Theorem 12 Assume that there is a function 119881(119909 119905 (1198941 1198942)) isin
11986221
(R119899 times R+times SR
+) and positive constants 119901 120578 120579 such that
for all (119909 119905 (1198941 1198942)) isin R119899 times R
+times S
120578|119909|119901le 119881 (119909 119905 (119894
1 1198942)) (82)
119871119881 (119909 119905 (1198941 1198942)) le 120579119881 (119909 119905 (119894
1 1198942)) (83)
Assume also the initial condition 119909(1199050) = 119909
0and 120574(119905
0) obeys
119864119881(1199090 1199050 120574(1199050)) lt infin then one has
119864|119909 (119905)|119901
le119864119881 (119909
0 1199050 120574 (1199050)) exp 120579 (119905 minus 119905
0)
120578 forall119905 isin [119905
0 119879]
(84)
Proof For each integer 119896 ge 1 define the stopping time
120590119896= 119879 and inf 119905 ge 119905
0 |119909 (119905)| ge 119896 (85)
Thus 120590119896
rarr 119879 as Using the generalized Itorsquos formula and(83) we obtain that for 119905 isin [119905
0 119879]
119864119881 (119909 (119905 and 120590119896) 119905 and 120590
119896 120574 (119905 and 120590
119896))
le 119864119881 (1199090 1199050 120574 (1199050))
+ 120579int
119905
1199050
119864119881 (119909 (119904 and 120590119896) 119904 and 120590
119896 120574 (119904 and 120590
119896)) 119889119904
(86)
Then the Gronwall inequality indicates
119864119881 (119909 (119905 and 120590119896) 119905 and 120590
119896 120574 (119905 and 120590
119896))
le 119864119881 (1199090 1199050 120574 (1199050)) exp 120579 (119905 minus 119905
0)
(87)
for all 119905 isin [1199050 119879] By virtue of condition (83) we obtain the
required assertion (84)
Corollary 13 Assume 119901 ge 2 and 1199090isin 119871119901
F1199050
(ΩR119899) Assumealso that there exists a constant 120579 gt 0 such that for all(119909 119905 (119894
1 1198942)) isin R119899 times R
+times S
119909119879119891 (119909 119905 119894
1) +
119901 minus 1
2
1003816100381610038161003816119892 (119909 119905 1198942)10038161003816100381610038162le 120579 (1 + |119909|
2) (88)
Then one has
119864|119909 (119905)|119901
le 2(119901minus2)2
(1 + 11986410038161003816100381610038161199090
1003816100381610038161003816119901) exp 119901120579 (119905 minus 119905
0) forall119905 isin [119905
0 119879]
(89)
Proof Define 119881(119909 119905 (1198941 1198942)) = (1 + |119909|
2)1199012 Making use of
(88) yields
119871119881 (119909 119905 (1198941 1198942))
= 119901(1 + |119909|2)(119901minus2)2
119909119879119891 (119909 119905 119894
1)
+119901
2(1 + |119909|
2)(119901minus2)21003816100381610038161003816119892 (119909 119905 119894
2)10038161003816100381610038162
+119901 (119901 minus 2)
2(1 + |119909|
2)(119901minus4)210038161003816100381610038161003816
119909119879119892 (119909 119905 119894
2)10038161003816100381610038161003816
2
le 119901(1 + |119909|2)(119901minus2)2
[119909119879119891 (119909 119905 119894
1) +
119901 minus 1
2
1003816100381610038161003816119892 (119909 119905 1198942)10038161003816100381610038162]
le 119901120579(1 + |119909|2)1199012
(90)
Then byTheorem 12 we get
119864(1 + |119909 (119905)|2)1199012
le 119864(1 +10038161003816100381610038161199090
10038161003816100381610038162)1199012
exp 119901120579 (119905 minus 1199050)
(91)
and the required assertion (89) followsIt is useful to point out that if the linear growth condition
(43) is satisfied then (88) is fulfilled with 120579 = radic119870 + 119870(119901 minus
1)2 Now we will show the other important properties of thesolution
Theorem 14 Let 119901 ge 2 and 1199090
isin 119871119901
F1199050
(ΩR119899) Assume alsothat the linear growth condition (43) holds Then one has
119864|119909 (119905) minus 119909 (119904)|119901le 119862(119905 minus 119904)
1199012 (92)
where
119862 = 2119901minus2
(1 + 11986410038161003816100381610038161199090
1003816100381610038161003816119901) exp 119901120579 (119879 minus 119905
0)
times ([2 (119879 minus 1199050)]1199012
+ [119901 (119901 minus 1)]1199012
)
(93)
and 120579 = radic119870+119870(119901minus1)2 Particularly the 119901th moment of thesolution is continuous on [119905
0 119879]
Journal of Applied Mathematics 9
Proof Applying the elementary inequality |119886 + 119887|119901
le
2119901minus1
(|119886|119901+ |119887|119901) the Holder inequality and the linear growth
condition we can derive that
119864|119909 (119905) minus 119909 (119904)|119901
le 2119901minus1
119864
10038161003816100381610038161003816100381610038161003816int
119905
119904
119891 (119909 (119906) 119906 1205741(119906)) 119889119906
10038161003816100381610038161003816100381610038161003816
119901
+ 2119901minus1
119864
10038161003816100381610038161003816100381610038161003816int
119905
119904
119892 (119909 (119906) 119906 1205742(119906)) 119889119861 (119906)
10038161003816100381610038161003816100381610038161003816
119901
le [2 (119905 minus 119904)]119901minus1
119864int
119905
119904
1003816100381610038161003816119891 (119909 (119906) 119906 1205741(119906))
1003816100381610038161003816119901119889119906
+ 05[2119901 (119901 minus 1)]1199012
(119905 minus 119904)(119901minus2)2
119864
times int
119905
119904
1003816100381610038161003816119892 (119909 (119906) 119906 1205742 (119906))1003816100381610038161003816119901119889119906
le 1198621(119905 minus 119904)
(119901minus2)2int
119905
119904
119864(1 + |119909 (119906)|2)1199012
119889119906
(94)
where1198621= 2(119901minus2)2
1198701199012
([2(119879minus 1199050)]1199012
+ [119901(119901minus1)]1199012
) Using(91) yields
119864|119909 (119905) minus 119909 (119904)|119901
le 1198621(119905 minus 119904)
(119901minus2)2
times int
119905
119904
2(119901minus2)2
(1 + 11986410038161003816100381610038161199090
1003816100381610038161003816119901) exp 119901120579 (119906 minus 119905
0) 119889119906
le 11986212(119901minus2)2
(1 + 11986410038161003816100381610038161199090
1003816100381610038161003816119901)
times exp 119901120579 (119879 minus 1199050) (119905 minus 119904)
(1199012)
(95)
which is the desired inequality
Theorem 15 Let 119901 ge 2 and 1199090
isin 119871119901
F1199050
(ΩR119899) Assume thatthere is a119870 gt 0 such that for all (119909 119905 (119894
1 1198942)) isin R119899times[119905
0 119879]timesS
119909119879 1003816100381610038161003816119891 (119909 119905 119894
1)1003816100381610038161003816 or
1003816100381610038161003816119892 (119909 119905 1198942)10038161003816100381610038162le 119870(1 + |119909|)
2 (96)
Then
119864( sup1199050le119905le119879
|119909 (119905)|119901)
le (1 + 211986410038161003816100381610038161199090
1003816100381610038161003816119901) exp 2119901 (10119901 + 1)119870 (119879 minus 119905
0)
(97)
Proof Making use of the generalized Itorsquos formula andcondition (96) we have
|119909 (119905)|119901
le10038161003816100381610038161199090
1003816100381610038161003816119901+ 119872(119905)
+ int
119905
1199050
119901|119909 (119904)|119901minus2
[119909119879(119904) 119891 (119909 (119904) 119904 120574
1(119904))
+119901 minus 1
2
1003816100381610038161003816119892 (119909 (119904) 119904 1205742 (119904))10038161003816100381610038162] 119889119904
le10038161003816100381610038161199090
1003816100381610038161003816119901+ 119872(119905) + 05119901 (119901 + 1)119870
times int
119905
1199050
|119909 (119904)|119901minus2
(1 + |119909 (119904)|2) 119889119904
le10038161003816100381610038161199090
1003816100381610038161003816119901+ 119872(119905) + 119901 (119901 + 1)119870int
119905
1199050
(1 + |119909 (119904)|119901) 119889119904
(98)
where
119872(119905) = int
119905
1199050
119901|119909 (119904)|119901minus2
119909119879(119904) 119892 (119909 (119904) 119904 1205742 (119904)) 119889119861 (119904) (99)
Therefore for any 1199051isin [1199050 119879]
119864( sup1199050le119905le1199051
|119909 (119905)|119901) le 119864
100381610038161003816100381611990901003816100381610038161003816119901+ 119864( sup
1199050le119905le1199051
|119872 (119905)|)
+ 119901 (119901 + 1)119870119864int
1199051
1199050
(1 + |119909 (119904)|119901) 119889119904
(100)
At the same time applying the well-known Burkholder-Davis-Gundy inequality (see eg [2]) gives
119864( sup1199050le119905le1199051
|119872 (119905)|)
le 3119864(int
1199051
1199050
1199012|119909 (119904)|
2119901minus21003816100381610038161003816119892 (119909 (119904) 119904 1205742 (119904))10038161003816100381610038162119889119904)
05
le 3119864( sup1199050le119905le1199051
|119909 (119905)|119901
timesint
1199051
1199050
1199012|119909 (119904)|
119901minus21003816100381610038161003816119892 (119909 (119904) 119904 1205742(119904))
10038161003816100381610038162119889119904)
05
le 05119864( sup1199050le119905le1199051
|119909 (119905)|119901) + 45119864
times int
1199051
1199050
1199012119870|119909 (119904)|
119901minus2(1 + |119909 (119904)|
2) 119889119904
le 05119864( sup1199050le119905le1199051
|119909 (119905)|119901) + 9119901
2119870119864int
1199051
1199050
(1 + |119909 (119904)|119901) 119889119904
(101)
10 Journal of Applied Mathematics
Substituting the above inequality into (100) gives
119864( sup1199050le119905le1199051
|119909 (119905)|119901) le 2119864
100381610038161003816100381611990901003816100381610038161003816119901
+ 2119901 (10119901 + 1)119870int
1199051
1199050
(1 + |119909 (119904)|119901) 119889119904
(102)
Thus
1 + 119864( sup1199050le119905le1199051
|119909 (119905)|119901)
le 1 + 211986410038161003816100381610038161199090
1003816100381610038161003816119901+ 2119901 (10119901 + 1)119870
times int
1199051
1199050
(1 + sup1199050le119905le119904
|119909 (119905)|119901)119889119904
(103)
Then the required assertion follows from the Gronwallinequality
Up to now we have discussed the 119871119901-estimates for the
solution in the case when 119901 ge 2 As for 0 lt 119901 lt 2 the similarresults can be given without any difficulty as long as we notethat the Holder inequality implies
119864|119909 (119905)|119901le [119864|119909 (119905)|
2]05119901
(104)
4 Example
Consider the following Black-Scholes model
119889119883 (119905) = 120583 (1205741(119905))119883 (119905) 119889119905 + ] (120574
2(119905))119883 (119905) 119889119861 (119905) (105)
where 1205741(119905) is a right-continuous homogenous Markovian
chain taking values in finite state spaces S1= 1 2 and 120574
2(119905)
is a right-continuous homogenous Markovian chain takingvalues in finite state spaces S
2= 1 2 3 120583(119894) = 119894 119894 = 1 2
](119895) = 119895 + 1 119895 = 1 2 3 Taking 119870 = 16 119870 = 9 then (42)and (43) hold Therefore by Theorem 7 (105) has a uniquesolution
5 Conclusions and Further Research
This paper is devoted to studying the existence and unique-ness of solution of SDEs with multi-Markovian switchingsand estimating the119901thmoment of the solutionWe have usedtwo continuous-time Markovian chains to model the SDEsThis area is becoming increasingly useful in engineeringeconomics communication theory active networking andso forth The sufficient criteria for existence and unique-ness of solution local solution and maximal local solutionwere established Those results indicate that (8) keeps manyproperties that (89) owns At the same time although thehypothesis (H1) is used in this paper wewant to point out thatthis hypothesis is not essential In fact (H1) can be replacedby the following generalized hypothesis
(H1)1015840 both 1205741(119905) and 120574
2(119905) are right-continuous homoge-
nous Markovian chains such that 120574(119905) = (1205741(119905) 1205742(119905)) is a
homogenous vector chainUnder hypothesis (H1)1015840 the results given in this paper
can be established similarly It is easy to see that if 1205741(119905) equiv
1205742(119905) and 120574
1(119905) is a right-continuous homogenous Markovian
chain then (H1)1015840 is fulfilled immediately At the same timeif 1205741(119905) equiv 120574
2(119905) (8) will reduce to the classical SDEs with
single Markovian chain that is to say the classical theoryabout SDEs with single Markovian chain is a special caseof our theory On the other hand many theorems in thispaper will play important roles in further study For exampleTheorem 15 will be useful when one studies the approximatesolutions
Some important and interesting questions can be furtherinvestigated using the results in this paper For exampleapproximate solutions boundedness and stability stochas-tic functional differential equations with vector Markovianswitching and their applications In particular the stability of(8) is one of the most important and interesting topics andthose investigations are in progress
Acknowledgments
The authors thank the editor and referees for their veryimportant and helpful comments and suggestions Theauthors also thank Dr C Zhang for helping them to improvethe English exposition This research is supported by theNSFC of China (nos 11171081 and 11171056)
References
[1] A V SkorokhodAsymptoticMethods in theTheory of StochasticDifferential Equations vol 78 of Translations of MathematicalMonographs American Mathematical Society Providence RIUSA 1989
[2] X Mao and C Yuan Stochastic Differential Euations withMarkovian Switching Imperial College Press London UK2006
[3] G Yin and X Y Zhou ldquoMarkowitzrsquos mean-variance portfolioselection with regime switching from discrete-time models totheir continuous-time limitsrdquo IEEE Transactions on AutomaticControl vol 49 no 3 pp 349ndash360 2004
[4] J Buffington and R J Elliott ldquoAmerican options with regimeswitchingrdquo International Journal of Theoretical and AppliedFinance vol 5 no 5 pp 497ndash514 2002
[5] C Zhu and G Yin ldquoOn competitive Lotka-Volterra model inrandom environmentsrdquo Journal of Mathematical Analysis andApplications vol 357 no 1 pp 154ndash170 2009
[6] Q Luo and X Mao ldquoStochastic population dynamics underregime switchingrdquo Journal of Mathematical Analysis and Appli-cations vol 334 no 1 pp 69ndash84 2007
[7] X Li D Jiang and X Mao ldquoPopulation dynamical behaviorof Lotka-Volterra system under regime switchingrdquo Journal ofComputational and Applied Mathematics vol 232 no 2 pp427ndash448 2009
[8] X Li A Gray D Jiang and X Mao ldquoSufficient and neces-sary conditions of stochastic permanence and extinction forstochastic logistic populations under regime switchingrdquo Journal
Journal of Applied Mathematics 11
ofMathematical Analysis and Applications vol 376 no 1 pp 11ndash28 2011
[9] M Liu and KWang ldquoAsymptotic properties and simulations ofa stochastic logistic model under regime switchingrdquoMathemat-ical and Computer Modelling vol 54 no 9-10 pp 2139ndash21542011
[10] G Hu and K Wang ldquoStability in distribution of competi-tive Lotka-Volterra system with Markovian switchingrdquo AppliedMathematical Modelling vol 35 no 7 pp 3189ndash3200 2011
[11] M Liu and K Wang ldquoAsymptotic properties and simulationsof a stochastic logistic model under regime switching IIrdquoMathematical andComputerModelling vol 55 no 3-4 pp 405ndash418 2012
[12] ZWu H Huang and LWang ldquoStochastic delay logistic modelunder regime switchingrdquo Abstract and Applied Analysis vol2012 Article ID 241702 26 pages 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Journal of Applied Mathematics 7
Then 120590119896
rarr 120590infin
as and
119875 119909 (119905) = 119909 (119905) 119905 isin [[1199050 120590119896and 120590119896]] = 1 forall119896 ge 1 (66)
Letting 119896 rarr infin gives
119875 119909 (119905) = 119909 (119905) 119905 isin [[1199050 120590infin
and 120590infin
[[ = 1 (67)
In order to complete the proof we need only to show that120590infin
= 120590infin
as In fact for almost every 120596 isin 120590infin
lt 120590infin
wehave
1003816100381610038161003816119909 (120590infin
120596)1003816100381610038161003816
= lim119896rarr+infin
1003816100381610038161003816119909 (120590119896 120596)
1003816100381610038161003816 = lim119896rarr+infin
1003816100381610038161003816119909 (120590119896 120596)
1003816100381610038161003816 = infin(68)
which contradicts the fact that 119909(119905 120596) is continuous on 119905 isin
[1199050 120590infin
(120596)) This implies 120590infin
ge 120590infin
as In the same way onecan show 120590
infinle 120590infinasThus wemust have 120590
infin= 120590infinasThis
completes the proof
In many situations we often consider an SDE on [1199050infin)
119889119909 (119905) = 119891 (119909 (119905) 119905 1205741(119905)) 119889119905
+ 119892 (119909 (119905) 119905 1205742(119905)) 119889119861 (119905) 119905
0le 119905 lt infin
(69)
with initial data 119909(1199050) = 119909
0and 120574(119905
0) If the assumption of
the existence-and-uniqueness theorem holds on every finitesubinterval [119905
0 119879] of [119905
0infin) then (69) has a unique solution
119909(119905) on the entire interval [1199050infin) Such a solution is called
a global solution To establish a more general result aboutglobal solution we need more notations To this end weintroduce an operator 119871119881 from R119899 times R
+times S to R which is
given by
119871119881 (119909 119905 (1198941 1198942))
= 119881119905(119909 119905 (119894
1 1198942)) + 119881
119909(119909 119905 (119894
1 1198942)) 119891 (119909 119905 119894
1)
+ 05 trace [119892119879 (119909 119905 1198942) 119881119909119909
(119909 119905 (1198941 1198942)) 119892 (119909 119905 119894
2)]
+ sum
(11989511198952)isinS
1199021198941119894211989511198952
119881 (119909 119905 (1198951 1198952))
(70)
where 119881(119909 119905 (1198941 1198942)) isin 119862
21(R119899 times R
+times SR
+) and
119881119905(119909 119905 (119894
1 1198942)) =
120597119881 (119909 119905 (1198941 1198942))
120597119905
119881119909(119909 119905 (119894
1 1198942))
= (120597119881 (119909 119905 (119894
1 1198942))
1205971199091
120597119881 (119909 119905 (119894
1 1198942))
120597119909119899
)
119881119909119909
(119909 119905 (1198941 1198942)) = (
1205972119881 (119909 119905 (119894
1 1198942))
120597119909119894120597119909119895
)
119899times119899
(71)
Theorem 11 Assume that the local Lipschitz condition (57)holds Assume also that there is a function 119881(119909 119905 (119894
1 1198942)) isin
11986221
(R119899 times R+times SR
+) and a constant 120579 gt 0 such that
lim|119909|rarrinfin
( inf(119905(1198941 1198942))isinR+timesS
119881 (119909 119905 (1198941 1198942))) = infin (72)
119871119881 (119909 119905 (1198941 1198942)) le 120579 (1 + 119881 (119909 119905 (119894
1 1198942)))
forall (119909 119905 (1198941 1198942)) isin R
119899times R+times S
(73)
Then there exists a unique global solution 119909(119905) to (69)
Proof We need only to prove the theorem for any initialcondition 119909
0isin R119899 and (120574
1(1199050) 1205742(1199050)) isin S FromTheorem 10
we know that the local Lipschitz condition guarantees theexistence of the unique maximal solution 119909(119905) on [119905
0 120590infin
)where 120590
infinis the explosion time We need only to show 120590
infin=
infin as If this is not true then we can find a pair of positiveconstants 120576 and 119879 such that
119875 120590infin
le 119879 gt 2120576 (74)
For each integer 119896 ge 1 define the stopping time
120590119896= inf 119905 ge 119905
0 |119909 (119905)| ge 119896 (75)
Since 120590119896
rarr 120590infinalmost surely we can find a sufficiently large
integer 1198960for
119875 120590119896le 119879 gt 120576 forall119896 ge 119896
0 (76)
Fix any 119896 ge 1198960 then for any 119905
0le 119905 le 119879 by virtue of the
generalized Ito formula (see eg [1])
119864119881 (119909 (119905 and 120590119896) 119905 and 120590
119896 120574 (119905 and 120590
119896))
= 119881 (1199090 1199050 1205740) + 119864int
119905and120590119896
1199050
119871119881 (119909 (119904) 119904 120574 (119904)) 119889119904
le 119881 (1199090 1199050 1205740) + 120579 (119879 minus 119905
0)
+ 120579int
119905
1199050
119864119881 (119909 (119904 and 120590119896) 119904 and 120590
119896 120574 (119904 and 120590
119896)) 119889119904
(77)
Making use of the Gronwall inequality gives
119864119881 (119909 (119905 and 120590119896) 119905 and 120590
119896 120574 (119905 and 120590
119896))
le [119881 (1199090 1199050 1205740) + 120579 (119879 minus 119905
0)] exp 120579 (119879 minus 119905
0)
(78)
Therefore
119864 (119868120590119896le119879
119881 (119909 (120590119896) 120590119896 120574 (120590119896)))
le [119881 (1199090 1199050 1205740) + 120579 (119879 minus 119905
0)] exp 120579 (119879 minus 119905
0)
(79)
At the same time set
119892119896= inf 119881 (119909 119905 (119894
1 1198942)) |119909| ge 119896 119905 isin [119905
0 119879] (119894
1 1198942) isin S
(80)
8 Journal of Applied Mathematics
Then (72) means 119892119896
rarr infin It follows from (76) and (79) that
[119881 (1199090 1199050 1205740) + 120579 (119879 minus 119905
0)] exp 120579 (119879 minus 119905
0)
ge 119892119896119875 120590119896le 119879 ge 120576119892
119896
(81)
Letting 119896 rarr infin yields a contradiction that is to say120590infin
= infinThe proof is complete
3 119871119875-Estimates
In the previous section we have investigated the existenceand uniqueness of the solution to (8) In this section as abovelet 119909(119905) 119905
0le 119905 le 119879 be the unique solution of (8) with initial
conditions 119909(1199050) = 1199090and 120574(119905
0) and we will estimate the 119901th
moment of the solution
Theorem 12 Assume that there is a function 119881(119909 119905 (1198941 1198942)) isin
11986221
(R119899 times R+times SR
+) and positive constants 119901 120578 120579 such that
for all (119909 119905 (1198941 1198942)) isin R119899 times R
+times S
120578|119909|119901le 119881 (119909 119905 (119894
1 1198942)) (82)
119871119881 (119909 119905 (1198941 1198942)) le 120579119881 (119909 119905 (119894
1 1198942)) (83)
Assume also the initial condition 119909(1199050) = 119909
0and 120574(119905
0) obeys
119864119881(1199090 1199050 120574(1199050)) lt infin then one has
119864|119909 (119905)|119901
le119864119881 (119909
0 1199050 120574 (1199050)) exp 120579 (119905 minus 119905
0)
120578 forall119905 isin [119905
0 119879]
(84)
Proof For each integer 119896 ge 1 define the stopping time
120590119896= 119879 and inf 119905 ge 119905
0 |119909 (119905)| ge 119896 (85)
Thus 120590119896
rarr 119879 as Using the generalized Itorsquos formula and(83) we obtain that for 119905 isin [119905
0 119879]
119864119881 (119909 (119905 and 120590119896) 119905 and 120590
119896 120574 (119905 and 120590
119896))
le 119864119881 (1199090 1199050 120574 (1199050))
+ 120579int
119905
1199050
119864119881 (119909 (119904 and 120590119896) 119904 and 120590
119896 120574 (119904 and 120590
119896)) 119889119904
(86)
Then the Gronwall inequality indicates
119864119881 (119909 (119905 and 120590119896) 119905 and 120590
119896 120574 (119905 and 120590
119896))
le 119864119881 (1199090 1199050 120574 (1199050)) exp 120579 (119905 minus 119905
0)
(87)
for all 119905 isin [1199050 119879] By virtue of condition (83) we obtain the
required assertion (84)
Corollary 13 Assume 119901 ge 2 and 1199090isin 119871119901
F1199050
(ΩR119899) Assumealso that there exists a constant 120579 gt 0 such that for all(119909 119905 (119894
1 1198942)) isin R119899 times R
+times S
119909119879119891 (119909 119905 119894
1) +
119901 minus 1
2
1003816100381610038161003816119892 (119909 119905 1198942)10038161003816100381610038162le 120579 (1 + |119909|
2) (88)
Then one has
119864|119909 (119905)|119901
le 2(119901minus2)2
(1 + 11986410038161003816100381610038161199090
1003816100381610038161003816119901) exp 119901120579 (119905 minus 119905
0) forall119905 isin [119905
0 119879]
(89)
Proof Define 119881(119909 119905 (1198941 1198942)) = (1 + |119909|
2)1199012 Making use of
(88) yields
119871119881 (119909 119905 (1198941 1198942))
= 119901(1 + |119909|2)(119901minus2)2
119909119879119891 (119909 119905 119894
1)
+119901
2(1 + |119909|
2)(119901minus2)21003816100381610038161003816119892 (119909 119905 119894
2)10038161003816100381610038162
+119901 (119901 minus 2)
2(1 + |119909|
2)(119901minus4)210038161003816100381610038161003816
119909119879119892 (119909 119905 119894
2)10038161003816100381610038161003816
2
le 119901(1 + |119909|2)(119901minus2)2
[119909119879119891 (119909 119905 119894
1) +
119901 minus 1
2
1003816100381610038161003816119892 (119909 119905 1198942)10038161003816100381610038162]
le 119901120579(1 + |119909|2)1199012
(90)
Then byTheorem 12 we get
119864(1 + |119909 (119905)|2)1199012
le 119864(1 +10038161003816100381610038161199090
10038161003816100381610038162)1199012
exp 119901120579 (119905 minus 1199050)
(91)
and the required assertion (89) followsIt is useful to point out that if the linear growth condition
(43) is satisfied then (88) is fulfilled with 120579 = radic119870 + 119870(119901 minus
1)2 Now we will show the other important properties of thesolution
Theorem 14 Let 119901 ge 2 and 1199090
isin 119871119901
F1199050
(ΩR119899) Assume alsothat the linear growth condition (43) holds Then one has
119864|119909 (119905) minus 119909 (119904)|119901le 119862(119905 minus 119904)
1199012 (92)
where
119862 = 2119901minus2
(1 + 11986410038161003816100381610038161199090
1003816100381610038161003816119901) exp 119901120579 (119879 minus 119905
0)
times ([2 (119879 minus 1199050)]1199012
+ [119901 (119901 minus 1)]1199012
)
(93)
and 120579 = radic119870+119870(119901minus1)2 Particularly the 119901th moment of thesolution is continuous on [119905
0 119879]
Journal of Applied Mathematics 9
Proof Applying the elementary inequality |119886 + 119887|119901
le
2119901minus1
(|119886|119901+ |119887|119901) the Holder inequality and the linear growth
condition we can derive that
119864|119909 (119905) minus 119909 (119904)|119901
le 2119901minus1
119864
10038161003816100381610038161003816100381610038161003816int
119905
119904
119891 (119909 (119906) 119906 1205741(119906)) 119889119906
10038161003816100381610038161003816100381610038161003816
119901
+ 2119901minus1
119864
10038161003816100381610038161003816100381610038161003816int
119905
119904
119892 (119909 (119906) 119906 1205742(119906)) 119889119861 (119906)
10038161003816100381610038161003816100381610038161003816
119901
le [2 (119905 minus 119904)]119901minus1
119864int
119905
119904
1003816100381610038161003816119891 (119909 (119906) 119906 1205741(119906))
1003816100381610038161003816119901119889119906
+ 05[2119901 (119901 minus 1)]1199012
(119905 minus 119904)(119901minus2)2
119864
times int
119905
119904
1003816100381610038161003816119892 (119909 (119906) 119906 1205742 (119906))1003816100381610038161003816119901119889119906
le 1198621(119905 minus 119904)
(119901minus2)2int
119905
119904
119864(1 + |119909 (119906)|2)1199012
119889119906
(94)
where1198621= 2(119901minus2)2
1198701199012
([2(119879minus 1199050)]1199012
+ [119901(119901minus1)]1199012
) Using(91) yields
119864|119909 (119905) minus 119909 (119904)|119901
le 1198621(119905 minus 119904)
(119901minus2)2
times int
119905
119904
2(119901minus2)2
(1 + 11986410038161003816100381610038161199090
1003816100381610038161003816119901) exp 119901120579 (119906 minus 119905
0) 119889119906
le 11986212(119901minus2)2
(1 + 11986410038161003816100381610038161199090
1003816100381610038161003816119901)
times exp 119901120579 (119879 minus 1199050) (119905 minus 119904)
(1199012)
(95)
which is the desired inequality
Theorem 15 Let 119901 ge 2 and 1199090
isin 119871119901
F1199050
(ΩR119899) Assume thatthere is a119870 gt 0 such that for all (119909 119905 (119894
1 1198942)) isin R119899times[119905
0 119879]timesS
119909119879 1003816100381610038161003816119891 (119909 119905 119894
1)1003816100381610038161003816 or
1003816100381610038161003816119892 (119909 119905 1198942)10038161003816100381610038162le 119870(1 + |119909|)
2 (96)
Then
119864( sup1199050le119905le119879
|119909 (119905)|119901)
le (1 + 211986410038161003816100381610038161199090
1003816100381610038161003816119901) exp 2119901 (10119901 + 1)119870 (119879 minus 119905
0)
(97)
Proof Making use of the generalized Itorsquos formula andcondition (96) we have
|119909 (119905)|119901
le10038161003816100381610038161199090
1003816100381610038161003816119901+ 119872(119905)
+ int
119905
1199050
119901|119909 (119904)|119901minus2
[119909119879(119904) 119891 (119909 (119904) 119904 120574
1(119904))
+119901 minus 1
2
1003816100381610038161003816119892 (119909 (119904) 119904 1205742 (119904))10038161003816100381610038162] 119889119904
le10038161003816100381610038161199090
1003816100381610038161003816119901+ 119872(119905) + 05119901 (119901 + 1)119870
times int
119905
1199050
|119909 (119904)|119901minus2
(1 + |119909 (119904)|2) 119889119904
le10038161003816100381610038161199090
1003816100381610038161003816119901+ 119872(119905) + 119901 (119901 + 1)119870int
119905
1199050
(1 + |119909 (119904)|119901) 119889119904
(98)
where
119872(119905) = int
119905
1199050
119901|119909 (119904)|119901minus2
119909119879(119904) 119892 (119909 (119904) 119904 1205742 (119904)) 119889119861 (119904) (99)
Therefore for any 1199051isin [1199050 119879]
119864( sup1199050le119905le1199051
|119909 (119905)|119901) le 119864
100381610038161003816100381611990901003816100381610038161003816119901+ 119864( sup
1199050le119905le1199051
|119872 (119905)|)
+ 119901 (119901 + 1)119870119864int
1199051
1199050
(1 + |119909 (119904)|119901) 119889119904
(100)
At the same time applying the well-known Burkholder-Davis-Gundy inequality (see eg [2]) gives
119864( sup1199050le119905le1199051
|119872 (119905)|)
le 3119864(int
1199051
1199050
1199012|119909 (119904)|
2119901minus21003816100381610038161003816119892 (119909 (119904) 119904 1205742 (119904))10038161003816100381610038162119889119904)
05
le 3119864( sup1199050le119905le1199051
|119909 (119905)|119901
timesint
1199051
1199050
1199012|119909 (119904)|
119901minus21003816100381610038161003816119892 (119909 (119904) 119904 1205742(119904))
10038161003816100381610038162119889119904)
05
le 05119864( sup1199050le119905le1199051
|119909 (119905)|119901) + 45119864
times int
1199051
1199050
1199012119870|119909 (119904)|
119901minus2(1 + |119909 (119904)|
2) 119889119904
le 05119864( sup1199050le119905le1199051
|119909 (119905)|119901) + 9119901
2119870119864int
1199051
1199050
(1 + |119909 (119904)|119901) 119889119904
(101)
10 Journal of Applied Mathematics
Substituting the above inequality into (100) gives
119864( sup1199050le119905le1199051
|119909 (119905)|119901) le 2119864
100381610038161003816100381611990901003816100381610038161003816119901
+ 2119901 (10119901 + 1)119870int
1199051
1199050
(1 + |119909 (119904)|119901) 119889119904
(102)
Thus
1 + 119864( sup1199050le119905le1199051
|119909 (119905)|119901)
le 1 + 211986410038161003816100381610038161199090
1003816100381610038161003816119901+ 2119901 (10119901 + 1)119870
times int
1199051
1199050
(1 + sup1199050le119905le119904
|119909 (119905)|119901)119889119904
(103)
Then the required assertion follows from the Gronwallinequality
Up to now we have discussed the 119871119901-estimates for the
solution in the case when 119901 ge 2 As for 0 lt 119901 lt 2 the similarresults can be given without any difficulty as long as we notethat the Holder inequality implies
119864|119909 (119905)|119901le [119864|119909 (119905)|
2]05119901
(104)
4 Example
Consider the following Black-Scholes model
119889119883 (119905) = 120583 (1205741(119905))119883 (119905) 119889119905 + ] (120574
2(119905))119883 (119905) 119889119861 (119905) (105)
where 1205741(119905) is a right-continuous homogenous Markovian
chain taking values in finite state spaces S1= 1 2 and 120574
2(119905)
is a right-continuous homogenous Markovian chain takingvalues in finite state spaces S
2= 1 2 3 120583(119894) = 119894 119894 = 1 2
](119895) = 119895 + 1 119895 = 1 2 3 Taking 119870 = 16 119870 = 9 then (42)and (43) hold Therefore by Theorem 7 (105) has a uniquesolution
5 Conclusions and Further Research
This paper is devoted to studying the existence and unique-ness of solution of SDEs with multi-Markovian switchingsand estimating the119901thmoment of the solutionWe have usedtwo continuous-time Markovian chains to model the SDEsThis area is becoming increasingly useful in engineeringeconomics communication theory active networking andso forth The sufficient criteria for existence and unique-ness of solution local solution and maximal local solutionwere established Those results indicate that (8) keeps manyproperties that (89) owns At the same time although thehypothesis (H1) is used in this paper wewant to point out thatthis hypothesis is not essential In fact (H1) can be replacedby the following generalized hypothesis
(H1)1015840 both 1205741(119905) and 120574
2(119905) are right-continuous homoge-
nous Markovian chains such that 120574(119905) = (1205741(119905) 1205742(119905)) is a
homogenous vector chainUnder hypothesis (H1)1015840 the results given in this paper
can be established similarly It is easy to see that if 1205741(119905) equiv
1205742(119905) and 120574
1(119905) is a right-continuous homogenous Markovian
chain then (H1)1015840 is fulfilled immediately At the same timeif 1205741(119905) equiv 120574
2(119905) (8) will reduce to the classical SDEs with
single Markovian chain that is to say the classical theoryabout SDEs with single Markovian chain is a special caseof our theory On the other hand many theorems in thispaper will play important roles in further study For exampleTheorem 15 will be useful when one studies the approximatesolutions
Some important and interesting questions can be furtherinvestigated using the results in this paper For exampleapproximate solutions boundedness and stability stochas-tic functional differential equations with vector Markovianswitching and their applications In particular the stability of(8) is one of the most important and interesting topics andthose investigations are in progress
Acknowledgments
The authors thank the editor and referees for their veryimportant and helpful comments and suggestions Theauthors also thank Dr C Zhang for helping them to improvethe English exposition This research is supported by theNSFC of China (nos 11171081 and 11171056)
References
[1] A V SkorokhodAsymptoticMethods in theTheory of StochasticDifferential Equations vol 78 of Translations of MathematicalMonographs American Mathematical Society Providence RIUSA 1989
[2] X Mao and C Yuan Stochastic Differential Euations withMarkovian Switching Imperial College Press London UK2006
[3] G Yin and X Y Zhou ldquoMarkowitzrsquos mean-variance portfolioselection with regime switching from discrete-time models totheir continuous-time limitsrdquo IEEE Transactions on AutomaticControl vol 49 no 3 pp 349ndash360 2004
[4] J Buffington and R J Elliott ldquoAmerican options with regimeswitchingrdquo International Journal of Theoretical and AppliedFinance vol 5 no 5 pp 497ndash514 2002
[5] C Zhu and G Yin ldquoOn competitive Lotka-Volterra model inrandom environmentsrdquo Journal of Mathematical Analysis andApplications vol 357 no 1 pp 154ndash170 2009
[6] Q Luo and X Mao ldquoStochastic population dynamics underregime switchingrdquo Journal of Mathematical Analysis and Appli-cations vol 334 no 1 pp 69ndash84 2007
[7] X Li D Jiang and X Mao ldquoPopulation dynamical behaviorof Lotka-Volterra system under regime switchingrdquo Journal ofComputational and Applied Mathematics vol 232 no 2 pp427ndash448 2009
[8] X Li A Gray D Jiang and X Mao ldquoSufficient and neces-sary conditions of stochastic permanence and extinction forstochastic logistic populations under regime switchingrdquo Journal
Journal of Applied Mathematics 11
ofMathematical Analysis and Applications vol 376 no 1 pp 11ndash28 2011
[9] M Liu and KWang ldquoAsymptotic properties and simulations ofa stochastic logistic model under regime switchingrdquoMathemat-ical and Computer Modelling vol 54 no 9-10 pp 2139ndash21542011
[10] G Hu and K Wang ldquoStability in distribution of competi-tive Lotka-Volterra system with Markovian switchingrdquo AppliedMathematical Modelling vol 35 no 7 pp 3189ndash3200 2011
[11] M Liu and K Wang ldquoAsymptotic properties and simulationsof a stochastic logistic model under regime switching IIrdquoMathematical andComputerModelling vol 55 no 3-4 pp 405ndash418 2012
[12] ZWu H Huang and LWang ldquoStochastic delay logistic modelunder regime switchingrdquo Abstract and Applied Analysis vol2012 Article ID 241702 26 pages 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
8 Journal of Applied Mathematics
Then (72) means 119892119896
rarr infin It follows from (76) and (79) that
[119881 (1199090 1199050 1205740) + 120579 (119879 minus 119905
0)] exp 120579 (119879 minus 119905
0)
ge 119892119896119875 120590119896le 119879 ge 120576119892
119896
(81)
Letting 119896 rarr infin yields a contradiction that is to say120590infin
= infinThe proof is complete
3 119871119875-Estimates
In the previous section we have investigated the existenceand uniqueness of the solution to (8) In this section as abovelet 119909(119905) 119905
0le 119905 le 119879 be the unique solution of (8) with initial
conditions 119909(1199050) = 1199090and 120574(119905
0) and we will estimate the 119901th
moment of the solution
Theorem 12 Assume that there is a function 119881(119909 119905 (1198941 1198942)) isin
11986221
(R119899 times R+times SR
+) and positive constants 119901 120578 120579 such that
for all (119909 119905 (1198941 1198942)) isin R119899 times R
+times S
120578|119909|119901le 119881 (119909 119905 (119894
1 1198942)) (82)
119871119881 (119909 119905 (1198941 1198942)) le 120579119881 (119909 119905 (119894
1 1198942)) (83)
Assume also the initial condition 119909(1199050) = 119909
0and 120574(119905
0) obeys
119864119881(1199090 1199050 120574(1199050)) lt infin then one has
119864|119909 (119905)|119901
le119864119881 (119909
0 1199050 120574 (1199050)) exp 120579 (119905 minus 119905
0)
120578 forall119905 isin [119905
0 119879]
(84)
Proof For each integer 119896 ge 1 define the stopping time
120590119896= 119879 and inf 119905 ge 119905
0 |119909 (119905)| ge 119896 (85)
Thus 120590119896
rarr 119879 as Using the generalized Itorsquos formula and(83) we obtain that for 119905 isin [119905
0 119879]
119864119881 (119909 (119905 and 120590119896) 119905 and 120590
119896 120574 (119905 and 120590
119896))
le 119864119881 (1199090 1199050 120574 (1199050))
+ 120579int
119905
1199050
119864119881 (119909 (119904 and 120590119896) 119904 and 120590
119896 120574 (119904 and 120590
119896)) 119889119904
(86)
Then the Gronwall inequality indicates
119864119881 (119909 (119905 and 120590119896) 119905 and 120590
119896 120574 (119905 and 120590
119896))
le 119864119881 (1199090 1199050 120574 (1199050)) exp 120579 (119905 minus 119905
0)
(87)
for all 119905 isin [1199050 119879] By virtue of condition (83) we obtain the
required assertion (84)
Corollary 13 Assume 119901 ge 2 and 1199090isin 119871119901
F1199050
(ΩR119899) Assumealso that there exists a constant 120579 gt 0 such that for all(119909 119905 (119894
1 1198942)) isin R119899 times R
+times S
119909119879119891 (119909 119905 119894
1) +
119901 minus 1
2
1003816100381610038161003816119892 (119909 119905 1198942)10038161003816100381610038162le 120579 (1 + |119909|
2) (88)
Then one has
119864|119909 (119905)|119901
le 2(119901minus2)2
(1 + 11986410038161003816100381610038161199090
1003816100381610038161003816119901) exp 119901120579 (119905 minus 119905
0) forall119905 isin [119905
0 119879]
(89)
Proof Define 119881(119909 119905 (1198941 1198942)) = (1 + |119909|
2)1199012 Making use of
(88) yields
119871119881 (119909 119905 (1198941 1198942))
= 119901(1 + |119909|2)(119901minus2)2
119909119879119891 (119909 119905 119894
1)
+119901
2(1 + |119909|
2)(119901minus2)21003816100381610038161003816119892 (119909 119905 119894
2)10038161003816100381610038162
+119901 (119901 minus 2)
2(1 + |119909|
2)(119901minus4)210038161003816100381610038161003816
119909119879119892 (119909 119905 119894
2)10038161003816100381610038161003816
2
le 119901(1 + |119909|2)(119901minus2)2
[119909119879119891 (119909 119905 119894
1) +
119901 minus 1
2
1003816100381610038161003816119892 (119909 119905 1198942)10038161003816100381610038162]
le 119901120579(1 + |119909|2)1199012
(90)
Then byTheorem 12 we get
119864(1 + |119909 (119905)|2)1199012
le 119864(1 +10038161003816100381610038161199090
10038161003816100381610038162)1199012
exp 119901120579 (119905 minus 1199050)
(91)
and the required assertion (89) followsIt is useful to point out that if the linear growth condition
(43) is satisfied then (88) is fulfilled with 120579 = radic119870 + 119870(119901 minus
1)2 Now we will show the other important properties of thesolution
Theorem 14 Let 119901 ge 2 and 1199090
isin 119871119901
F1199050
(ΩR119899) Assume alsothat the linear growth condition (43) holds Then one has
119864|119909 (119905) minus 119909 (119904)|119901le 119862(119905 minus 119904)
1199012 (92)
where
119862 = 2119901minus2
(1 + 11986410038161003816100381610038161199090
1003816100381610038161003816119901) exp 119901120579 (119879 minus 119905
0)
times ([2 (119879 minus 1199050)]1199012
+ [119901 (119901 minus 1)]1199012
)
(93)
and 120579 = radic119870+119870(119901minus1)2 Particularly the 119901th moment of thesolution is continuous on [119905
0 119879]
Journal of Applied Mathematics 9
Proof Applying the elementary inequality |119886 + 119887|119901
le
2119901minus1
(|119886|119901+ |119887|119901) the Holder inequality and the linear growth
condition we can derive that
119864|119909 (119905) minus 119909 (119904)|119901
le 2119901minus1
119864
10038161003816100381610038161003816100381610038161003816int
119905
119904
119891 (119909 (119906) 119906 1205741(119906)) 119889119906
10038161003816100381610038161003816100381610038161003816
119901
+ 2119901minus1
119864
10038161003816100381610038161003816100381610038161003816int
119905
119904
119892 (119909 (119906) 119906 1205742(119906)) 119889119861 (119906)
10038161003816100381610038161003816100381610038161003816
119901
le [2 (119905 minus 119904)]119901minus1
119864int
119905
119904
1003816100381610038161003816119891 (119909 (119906) 119906 1205741(119906))
1003816100381610038161003816119901119889119906
+ 05[2119901 (119901 minus 1)]1199012
(119905 minus 119904)(119901minus2)2
119864
times int
119905
119904
1003816100381610038161003816119892 (119909 (119906) 119906 1205742 (119906))1003816100381610038161003816119901119889119906
le 1198621(119905 minus 119904)
(119901minus2)2int
119905
119904
119864(1 + |119909 (119906)|2)1199012
119889119906
(94)
where1198621= 2(119901minus2)2
1198701199012
([2(119879minus 1199050)]1199012
+ [119901(119901minus1)]1199012
) Using(91) yields
119864|119909 (119905) minus 119909 (119904)|119901
le 1198621(119905 minus 119904)
(119901minus2)2
times int
119905
119904
2(119901minus2)2
(1 + 11986410038161003816100381610038161199090
1003816100381610038161003816119901) exp 119901120579 (119906 minus 119905
0) 119889119906
le 11986212(119901minus2)2
(1 + 11986410038161003816100381610038161199090
1003816100381610038161003816119901)
times exp 119901120579 (119879 minus 1199050) (119905 minus 119904)
(1199012)
(95)
which is the desired inequality
Theorem 15 Let 119901 ge 2 and 1199090
isin 119871119901
F1199050
(ΩR119899) Assume thatthere is a119870 gt 0 such that for all (119909 119905 (119894
1 1198942)) isin R119899times[119905
0 119879]timesS
119909119879 1003816100381610038161003816119891 (119909 119905 119894
1)1003816100381610038161003816 or
1003816100381610038161003816119892 (119909 119905 1198942)10038161003816100381610038162le 119870(1 + |119909|)
2 (96)
Then
119864( sup1199050le119905le119879
|119909 (119905)|119901)
le (1 + 211986410038161003816100381610038161199090
1003816100381610038161003816119901) exp 2119901 (10119901 + 1)119870 (119879 minus 119905
0)
(97)
Proof Making use of the generalized Itorsquos formula andcondition (96) we have
|119909 (119905)|119901
le10038161003816100381610038161199090
1003816100381610038161003816119901+ 119872(119905)
+ int
119905
1199050
119901|119909 (119904)|119901minus2
[119909119879(119904) 119891 (119909 (119904) 119904 120574
1(119904))
+119901 minus 1
2
1003816100381610038161003816119892 (119909 (119904) 119904 1205742 (119904))10038161003816100381610038162] 119889119904
le10038161003816100381610038161199090
1003816100381610038161003816119901+ 119872(119905) + 05119901 (119901 + 1)119870
times int
119905
1199050
|119909 (119904)|119901minus2
(1 + |119909 (119904)|2) 119889119904
le10038161003816100381610038161199090
1003816100381610038161003816119901+ 119872(119905) + 119901 (119901 + 1)119870int
119905
1199050
(1 + |119909 (119904)|119901) 119889119904
(98)
where
119872(119905) = int
119905
1199050
119901|119909 (119904)|119901minus2
119909119879(119904) 119892 (119909 (119904) 119904 1205742 (119904)) 119889119861 (119904) (99)
Therefore for any 1199051isin [1199050 119879]
119864( sup1199050le119905le1199051
|119909 (119905)|119901) le 119864
100381610038161003816100381611990901003816100381610038161003816119901+ 119864( sup
1199050le119905le1199051
|119872 (119905)|)
+ 119901 (119901 + 1)119870119864int
1199051
1199050
(1 + |119909 (119904)|119901) 119889119904
(100)
At the same time applying the well-known Burkholder-Davis-Gundy inequality (see eg [2]) gives
119864( sup1199050le119905le1199051
|119872 (119905)|)
le 3119864(int
1199051
1199050
1199012|119909 (119904)|
2119901minus21003816100381610038161003816119892 (119909 (119904) 119904 1205742 (119904))10038161003816100381610038162119889119904)
05
le 3119864( sup1199050le119905le1199051
|119909 (119905)|119901
timesint
1199051
1199050
1199012|119909 (119904)|
119901minus21003816100381610038161003816119892 (119909 (119904) 119904 1205742(119904))
10038161003816100381610038162119889119904)
05
le 05119864( sup1199050le119905le1199051
|119909 (119905)|119901) + 45119864
times int
1199051
1199050
1199012119870|119909 (119904)|
119901minus2(1 + |119909 (119904)|
2) 119889119904
le 05119864( sup1199050le119905le1199051
|119909 (119905)|119901) + 9119901
2119870119864int
1199051
1199050
(1 + |119909 (119904)|119901) 119889119904
(101)
10 Journal of Applied Mathematics
Substituting the above inequality into (100) gives
119864( sup1199050le119905le1199051
|119909 (119905)|119901) le 2119864
100381610038161003816100381611990901003816100381610038161003816119901
+ 2119901 (10119901 + 1)119870int
1199051
1199050
(1 + |119909 (119904)|119901) 119889119904
(102)
Thus
1 + 119864( sup1199050le119905le1199051
|119909 (119905)|119901)
le 1 + 211986410038161003816100381610038161199090
1003816100381610038161003816119901+ 2119901 (10119901 + 1)119870
times int
1199051
1199050
(1 + sup1199050le119905le119904
|119909 (119905)|119901)119889119904
(103)
Then the required assertion follows from the Gronwallinequality
Up to now we have discussed the 119871119901-estimates for the
solution in the case when 119901 ge 2 As for 0 lt 119901 lt 2 the similarresults can be given without any difficulty as long as we notethat the Holder inequality implies
119864|119909 (119905)|119901le [119864|119909 (119905)|
2]05119901
(104)
4 Example
Consider the following Black-Scholes model
119889119883 (119905) = 120583 (1205741(119905))119883 (119905) 119889119905 + ] (120574
2(119905))119883 (119905) 119889119861 (119905) (105)
where 1205741(119905) is a right-continuous homogenous Markovian
chain taking values in finite state spaces S1= 1 2 and 120574
2(119905)
is a right-continuous homogenous Markovian chain takingvalues in finite state spaces S
2= 1 2 3 120583(119894) = 119894 119894 = 1 2
](119895) = 119895 + 1 119895 = 1 2 3 Taking 119870 = 16 119870 = 9 then (42)and (43) hold Therefore by Theorem 7 (105) has a uniquesolution
5 Conclusions and Further Research
This paper is devoted to studying the existence and unique-ness of solution of SDEs with multi-Markovian switchingsand estimating the119901thmoment of the solutionWe have usedtwo continuous-time Markovian chains to model the SDEsThis area is becoming increasingly useful in engineeringeconomics communication theory active networking andso forth The sufficient criteria for existence and unique-ness of solution local solution and maximal local solutionwere established Those results indicate that (8) keeps manyproperties that (89) owns At the same time although thehypothesis (H1) is used in this paper wewant to point out thatthis hypothesis is not essential In fact (H1) can be replacedby the following generalized hypothesis
(H1)1015840 both 1205741(119905) and 120574
2(119905) are right-continuous homoge-
nous Markovian chains such that 120574(119905) = (1205741(119905) 1205742(119905)) is a
homogenous vector chainUnder hypothesis (H1)1015840 the results given in this paper
can be established similarly It is easy to see that if 1205741(119905) equiv
1205742(119905) and 120574
1(119905) is a right-continuous homogenous Markovian
chain then (H1)1015840 is fulfilled immediately At the same timeif 1205741(119905) equiv 120574
2(119905) (8) will reduce to the classical SDEs with
single Markovian chain that is to say the classical theoryabout SDEs with single Markovian chain is a special caseof our theory On the other hand many theorems in thispaper will play important roles in further study For exampleTheorem 15 will be useful when one studies the approximatesolutions
Some important and interesting questions can be furtherinvestigated using the results in this paper For exampleapproximate solutions boundedness and stability stochas-tic functional differential equations with vector Markovianswitching and their applications In particular the stability of(8) is one of the most important and interesting topics andthose investigations are in progress
Acknowledgments
The authors thank the editor and referees for their veryimportant and helpful comments and suggestions Theauthors also thank Dr C Zhang for helping them to improvethe English exposition This research is supported by theNSFC of China (nos 11171081 and 11171056)
References
[1] A V SkorokhodAsymptoticMethods in theTheory of StochasticDifferential Equations vol 78 of Translations of MathematicalMonographs American Mathematical Society Providence RIUSA 1989
[2] X Mao and C Yuan Stochastic Differential Euations withMarkovian Switching Imperial College Press London UK2006
[3] G Yin and X Y Zhou ldquoMarkowitzrsquos mean-variance portfolioselection with regime switching from discrete-time models totheir continuous-time limitsrdquo IEEE Transactions on AutomaticControl vol 49 no 3 pp 349ndash360 2004
[4] J Buffington and R J Elliott ldquoAmerican options with regimeswitchingrdquo International Journal of Theoretical and AppliedFinance vol 5 no 5 pp 497ndash514 2002
[5] C Zhu and G Yin ldquoOn competitive Lotka-Volterra model inrandom environmentsrdquo Journal of Mathematical Analysis andApplications vol 357 no 1 pp 154ndash170 2009
[6] Q Luo and X Mao ldquoStochastic population dynamics underregime switchingrdquo Journal of Mathematical Analysis and Appli-cations vol 334 no 1 pp 69ndash84 2007
[7] X Li D Jiang and X Mao ldquoPopulation dynamical behaviorof Lotka-Volterra system under regime switchingrdquo Journal ofComputational and Applied Mathematics vol 232 no 2 pp427ndash448 2009
[8] X Li A Gray D Jiang and X Mao ldquoSufficient and neces-sary conditions of stochastic permanence and extinction forstochastic logistic populations under regime switchingrdquo Journal
Journal of Applied Mathematics 11
ofMathematical Analysis and Applications vol 376 no 1 pp 11ndash28 2011
[9] M Liu and KWang ldquoAsymptotic properties and simulations ofa stochastic logistic model under regime switchingrdquoMathemat-ical and Computer Modelling vol 54 no 9-10 pp 2139ndash21542011
[10] G Hu and K Wang ldquoStability in distribution of competi-tive Lotka-Volterra system with Markovian switchingrdquo AppliedMathematical Modelling vol 35 no 7 pp 3189ndash3200 2011
[11] M Liu and K Wang ldquoAsymptotic properties and simulationsof a stochastic logistic model under regime switching IIrdquoMathematical andComputerModelling vol 55 no 3-4 pp 405ndash418 2012
[12] ZWu H Huang and LWang ldquoStochastic delay logistic modelunder regime switchingrdquo Abstract and Applied Analysis vol2012 Article ID 241702 26 pages 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Journal of Applied Mathematics 9
Proof Applying the elementary inequality |119886 + 119887|119901
le
2119901minus1
(|119886|119901+ |119887|119901) the Holder inequality and the linear growth
condition we can derive that
119864|119909 (119905) minus 119909 (119904)|119901
le 2119901minus1
119864
10038161003816100381610038161003816100381610038161003816int
119905
119904
119891 (119909 (119906) 119906 1205741(119906)) 119889119906
10038161003816100381610038161003816100381610038161003816
119901
+ 2119901minus1
119864
10038161003816100381610038161003816100381610038161003816int
119905
119904
119892 (119909 (119906) 119906 1205742(119906)) 119889119861 (119906)
10038161003816100381610038161003816100381610038161003816
119901
le [2 (119905 minus 119904)]119901minus1
119864int
119905
119904
1003816100381610038161003816119891 (119909 (119906) 119906 1205741(119906))
1003816100381610038161003816119901119889119906
+ 05[2119901 (119901 minus 1)]1199012
(119905 minus 119904)(119901minus2)2
119864
times int
119905
119904
1003816100381610038161003816119892 (119909 (119906) 119906 1205742 (119906))1003816100381610038161003816119901119889119906
le 1198621(119905 minus 119904)
(119901minus2)2int
119905
119904
119864(1 + |119909 (119906)|2)1199012
119889119906
(94)
where1198621= 2(119901minus2)2
1198701199012
([2(119879minus 1199050)]1199012
+ [119901(119901minus1)]1199012
) Using(91) yields
119864|119909 (119905) minus 119909 (119904)|119901
le 1198621(119905 minus 119904)
(119901minus2)2
times int
119905
119904
2(119901minus2)2
(1 + 11986410038161003816100381610038161199090
1003816100381610038161003816119901) exp 119901120579 (119906 minus 119905
0) 119889119906
le 11986212(119901minus2)2
(1 + 11986410038161003816100381610038161199090
1003816100381610038161003816119901)
times exp 119901120579 (119879 minus 1199050) (119905 minus 119904)
(1199012)
(95)
which is the desired inequality
Theorem 15 Let 119901 ge 2 and 1199090
isin 119871119901
F1199050
(ΩR119899) Assume thatthere is a119870 gt 0 such that for all (119909 119905 (119894
1 1198942)) isin R119899times[119905
0 119879]timesS
119909119879 1003816100381610038161003816119891 (119909 119905 119894
1)1003816100381610038161003816 or
1003816100381610038161003816119892 (119909 119905 1198942)10038161003816100381610038162le 119870(1 + |119909|)
2 (96)
Then
119864( sup1199050le119905le119879
|119909 (119905)|119901)
le (1 + 211986410038161003816100381610038161199090
1003816100381610038161003816119901) exp 2119901 (10119901 + 1)119870 (119879 minus 119905
0)
(97)
Proof Making use of the generalized Itorsquos formula andcondition (96) we have
|119909 (119905)|119901
le10038161003816100381610038161199090
1003816100381610038161003816119901+ 119872(119905)
+ int
119905
1199050
119901|119909 (119904)|119901minus2
[119909119879(119904) 119891 (119909 (119904) 119904 120574
1(119904))
+119901 minus 1
2
1003816100381610038161003816119892 (119909 (119904) 119904 1205742 (119904))10038161003816100381610038162] 119889119904
le10038161003816100381610038161199090
1003816100381610038161003816119901+ 119872(119905) + 05119901 (119901 + 1)119870
times int
119905
1199050
|119909 (119904)|119901minus2
(1 + |119909 (119904)|2) 119889119904
le10038161003816100381610038161199090
1003816100381610038161003816119901+ 119872(119905) + 119901 (119901 + 1)119870int
119905
1199050
(1 + |119909 (119904)|119901) 119889119904
(98)
where
119872(119905) = int
119905
1199050
119901|119909 (119904)|119901minus2
119909119879(119904) 119892 (119909 (119904) 119904 1205742 (119904)) 119889119861 (119904) (99)
Therefore for any 1199051isin [1199050 119879]
119864( sup1199050le119905le1199051
|119909 (119905)|119901) le 119864
100381610038161003816100381611990901003816100381610038161003816119901+ 119864( sup
1199050le119905le1199051
|119872 (119905)|)
+ 119901 (119901 + 1)119870119864int
1199051
1199050
(1 + |119909 (119904)|119901) 119889119904
(100)
At the same time applying the well-known Burkholder-Davis-Gundy inequality (see eg [2]) gives
119864( sup1199050le119905le1199051
|119872 (119905)|)
le 3119864(int
1199051
1199050
1199012|119909 (119904)|
2119901minus21003816100381610038161003816119892 (119909 (119904) 119904 1205742 (119904))10038161003816100381610038162119889119904)
05
le 3119864( sup1199050le119905le1199051
|119909 (119905)|119901
timesint
1199051
1199050
1199012|119909 (119904)|
119901minus21003816100381610038161003816119892 (119909 (119904) 119904 1205742(119904))
10038161003816100381610038162119889119904)
05
le 05119864( sup1199050le119905le1199051
|119909 (119905)|119901) + 45119864
times int
1199051
1199050
1199012119870|119909 (119904)|
119901minus2(1 + |119909 (119904)|
2) 119889119904
le 05119864( sup1199050le119905le1199051
|119909 (119905)|119901) + 9119901
2119870119864int
1199051
1199050
(1 + |119909 (119904)|119901) 119889119904
(101)
10 Journal of Applied Mathematics
Substituting the above inequality into (100) gives
119864( sup1199050le119905le1199051
|119909 (119905)|119901) le 2119864
100381610038161003816100381611990901003816100381610038161003816119901
+ 2119901 (10119901 + 1)119870int
1199051
1199050
(1 + |119909 (119904)|119901) 119889119904
(102)
Thus
1 + 119864( sup1199050le119905le1199051
|119909 (119905)|119901)
le 1 + 211986410038161003816100381610038161199090
1003816100381610038161003816119901+ 2119901 (10119901 + 1)119870
times int
1199051
1199050
(1 + sup1199050le119905le119904
|119909 (119905)|119901)119889119904
(103)
Then the required assertion follows from the Gronwallinequality
Up to now we have discussed the 119871119901-estimates for the
solution in the case when 119901 ge 2 As for 0 lt 119901 lt 2 the similarresults can be given without any difficulty as long as we notethat the Holder inequality implies
119864|119909 (119905)|119901le [119864|119909 (119905)|
2]05119901
(104)
4 Example
Consider the following Black-Scholes model
119889119883 (119905) = 120583 (1205741(119905))119883 (119905) 119889119905 + ] (120574
2(119905))119883 (119905) 119889119861 (119905) (105)
where 1205741(119905) is a right-continuous homogenous Markovian
chain taking values in finite state spaces S1= 1 2 and 120574
2(119905)
is a right-continuous homogenous Markovian chain takingvalues in finite state spaces S
2= 1 2 3 120583(119894) = 119894 119894 = 1 2
](119895) = 119895 + 1 119895 = 1 2 3 Taking 119870 = 16 119870 = 9 then (42)and (43) hold Therefore by Theorem 7 (105) has a uniquesolution
5 Conclusions and Further Research
This paper is devoted to studying the existence and unique-ness of solution of SDEs with multi-Markovian switchingsand estimating the119901thmoment of the solutionWe have usedtwo continuous-time Markovian chains to model the SDEsThis area is becoming increasingly useful in engineeringeconomics communication theory active networking andso forth The sufficient criteria for existence and unique-ness of solution local solution and maximal local solutionwere established Those results indicate that (8) keeps manyproperties that (89) owns At the same time although thehypothesis (H1) is used in this paper wewant to point out thatthis hypothesis is not essential In fact (H1) can be replacedby the following generalized hypothesis
(H1)1015840 both 1205741(119905) and 120574
2(119905) are right-continuous homoge-
nous Markovian chains such that 120574(119905) = (1205741(119905) 1205742(119905)) is a
homogenous vector chainUnder hypothesis (H1)1015840 the results given in this paper
can be established similarly It is easy to see that if 1205741(119905) equiv
1205742(119905) and 120574
1(119905) is a right-continuous homogenous Markovian
chain then (H1)1015840 is fulfilled immediately At the same timeif 1205741(119905) equiv 120574
2(119905) (8) will reduce to the classical SDEs with
single Markovian chain that is to say the classical theoryabout SDEs with single Markovian chain is a special caseof our theory On the other hand many theorems in thispaper will play important roles in further study For exampleTheorem 15 will be useful when one studies the approximatesolutions
Some important and interesting questions can be furtherinvestigated using the results in this paper For exampleapproximate solutions boundedness and stability stochas-tic functional differential equations with vector Markovianswitching and their applications In particular the stability of(8) is one of the most important and interesting topics andthose investigations are in progress
Acknowledgments
The authors thank the editor and referees for their veryimportant and helpful comments and suggestions Theauthors also thank Dr C Zhang for helping them to improvethe English exposition This research is supported by theNSFC of China (nos 11171081 and 11171056)
References
[1] A V SkorokhodAsymptoticMethods in theTheory of StochasticDifferential Equations vol 78 of Translations of MathematicalMonographs American Mathematical Society Providence RIUSA 1989
[2] X Mao and C Yuan Stochastic Differential Euations withMarkovian Switching Imperial College Press London UK2006
[3] G Yin and X Y Zhou ldquoMarkowitzrsquos mean-variance portfolioselection with regime switching from discrete-time models totheir continuous-time limitsrdquo IEEE Transactions on AutomaticControl vol 49 no 3 pp 349ndash360 2004
[4] J Buffington and R J Elliott ldquoAmerican options with regimeswitchingrdquo International Journal of Theoretical and AppliedFinance vol 5 no 5 pp 497ndash514 2002
[5] C Zhu and G Yin ldquoOn competitive Lotka-Volterra model inrandom environmentsrdquo Journal of Mathematical Analysis andApplications vol 357 no 1 pp 154ndash170 2009
[6] Q Luo and X Mao ldquoStochastic population dynamics underregime switchingrdquo Journal of Mathematical Analysis and Appli-cations vol 334 no 1 pp 69ndash84 2007
[7] X Li D Jiang and X Mao ldquoPopulation dynamical behaviorof Lotka-Volterra system under regime switchingrdquo Journal ofComputational and Applied Mathematics vol 232 no 2 pp427ndash448 2009
[8] X Li A Gray D Jiang and X Mao ldquoSufficient and neces-sary conditions of stochastic permanence and extinction forstochastic logistic populations under regime switchingrdquo Journal
Journal of Applied Mathematics 11
ofMathematical Analysis and Applications vol 376 no 1 pp 11ndash28 2011
[9] M Liu and KWang ldquoAsymptotic properties and simulations ofa stochastic logistic model under regime switchingrdquoMathemat-ical and Computer Modelling vol 54 no 9-10 pp 2139ndash21542011
[10] G Hu and K Wang ldquoStability in distribution of competi-tive Lotka-Volterra system with Markovian switchingrdquo AppliedMathematical Modelling vol 35 no 7 pp 3189ndash3200 2011
[11] M Liu and K Wang ldquoAsymptotic properties and simulationsof a stochastic logistic model under regime switching IIrdquoMathematical andComputerModelling vol 55 no 3-4 pp 405ndash418 2012
[12] ZWu H Huang and LWang ldquoStochastic delay logistic modelunder regime switchingrdquo Abstract and Applied Analysis vol2012 Article ID 241702 26 pages 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
10 Journal of Applied Mathematics
Substituting the above inequality into (100) gives
119864( sup1199050le119905le1199051
|119909 (119905)|119901) le 2119864
100381610038161003816100381611990901003816100381610038161003816119901
+ 2119901 (10119901 + 1)119870int
1199051
1199050
(1 + |119909 (119904)|119901) 119889119904
(102)
Thus
1 + 119864( sup1199050le119905le1199051
|119909 (119905)|119901)
le 1 + 211986410038161003816100381610038161199090
1003816100381610038161003816119901+ 2119901 (10119901 + 1)119870
times int
1199051
1199050
(1 + sup1199050le119905le119904
|119909 (119905)|119901)119889119904
(103)
Then the required assertion follows from the Gronwallinequality
Up to now we have discussed the 119871119901-estimates for the
solution in the case when 119901 ge 2 As for 0 lt 119901 lt 2 the similarresults can be given without any difficulty as long as we notethat the Holder inequality implies
119864|119909 (119905)|119901le [119864|119909 (119905)|
2]05119901
(104)
4 Example
Consider the following Black-Scholes model
119889119883 (119905) = 120583 (1205741(119905))119883 (119905) 119889119905 + ] (120574
2(119905))119883 (119905) 119889119861 (119905) (105)
where 1205741(119905) is a right-continuous homogenous Markovian
chain taking values in finite state spaces S1= 1 2 and 120574
2(119905)
is a right-continuous homogenous Markovian chain takingvalues in finite state spaces S
2= 1 2 3 120583(119894) = 119894 119894 = 1 2
](119895) = 119895 + 1 119895 = 1 2 3 Taking 119870 = 16 119870 = 9 then (42)and (43) hold Therefore by Theorem 7 (105) has a uniquesolution
5 Conclusions and Further Research
This paper is devoted to studying the existence and unique-ness of solution of SDEs with multi-Markovian switchingsand estimating the119901thmoment of the solutionWe have usedtwo continuous-time Markovian chains to model the SDEsThis area is becoming increasingly useful in engineeringeconomics communication theory active networking andso forth The sufficient criteria for existence and unique-ness of solution local solution and maximal local solutionwere established Those results indicate that (8) keeps manyproperties that (89) owns At the same time although thehypothesis (H1) is used in this paper wewant to point out thatthis hypothesis is not essential In fact (H1) can be replacedby the following generalized hypothesis
(H1)1015840 both 1205741(119905) and 120574
2(119905) are right-continuous homoge-
nous Markovian chains such that 120574(119905) = (1205741(119905) 1205742(119905)) is a
homogenous vector chainUnder hypothesis (H1)1015840 the results given in this paper
can be established similarly It is easy to see that if 1205741(119905) equiv
1205742(119905) and 120574
1(119905) is a right-continuous homogenous Markovian
chain then (H1)1015840 is fulfilled immediately At the same timeif 1205741(119905) equiv 120574
2(119905) (8) will reduce to the classical SDEs with
single Markovian chain that is to say the classical theoryabout SDEs with single Markovian chain is a special caseof our theory On the other hand many theorems in thispaper will play important roles in further study For exampleTheorem 15 will be useful when one studies the approximatesolutions
Some important and interesting questions can be furtherinvestigated using the results in this paper For exampleapproximate solutions boundedness and stability stochas-tic functional differential equations with vector Markovianswitching and their applications In particular the stability of(8) is one of the most important and interesting topics andthose investigations are in progress
Acknowledgments
The authors thank the editor and referees for their veryimportant and helpful comments and suggestions Theauthors also thank Dr C Zhang for helping them to improvethe English exposition This research is supported by theNSFC of China (nos 11171081 and 11171056)
References
[1] A V SkorokhodAsymptoticMethods in theTheory of StochasticDifferential Equations vol 78 of Translations of MathematicalMonographs American Mathematical Society Providence RIUSA 1989
[2] X Mao and C Yuan Stochastic Differential Euations withMarkovian Switching Imperial College Press London UK2006
[3] G Yin and X Y Zhou ldquoMarkowitzrsquos mean-variance portfolioselection with regime switching from discrete-time models totheir continuous-time limitsrdquo IEEE Transactions on AutomaticControl vol 49 no 3 pp 349ndash360 2004
[4] J Buffington and R J Elliott ldquoAmerican options with regimeswitchingrdquo International Journal of Theoretical and AppliedFinance vol 5 no 5 pp 497ndash514 2002
[5] C Zhu and G Yin ldquoOn competitive Lotka-Volterra model inrandom environmentsrdquo Journal of Mathematical Analysis andApplications vol 357 no 1 pp 154ndash170 2009
[6] Q Luo and X Mao ldquoStochastic population dynamics underregime switchingrdquo Journal of Mathematical Analysis and Appli-cations vol 334 no 1 pp 69ndash84 2007
[7] X Li D Jiang and X Mao ldquoPopulation dynamical behaviorof Lotka-Volterra system under regime switchingrdquo Journal ofComputational and Applied Mathematics vol 232 no 2 pp427ndash448 2009
[8] X Li A Gray D Jiang and X Mao ldquoSufficient and neces-sary conditions of stochastic permanence and extinction forstochastic logistic populations under regime switchingrdquo Journal
Journal of Applied Mathematics 11
ofMathematical Analysis and Applications vol 376 no 1 pp 11ndash28 2011
[9] M Liu and KWang ldquoAsymptotic properties and simulations ofa stochastic logistic model under regime switchingrdquoMathemat-ical and Computer Modelling vol 54 no 9-10 pp 2139ndash21542011
[10] G Hu and K Wang ldquoStability in distribution of competi-tive Lotka-Volterra system with Markovian switchingrdquo AppliedMathematical Modelling vol 35 no 7 pp 3189ndash3200 2011
[11] M Liu and K Wang ldquoAsymptotic properties and simulationsof a stochastic logistic model under regime switching IIrdquoMathematical andComputerModelling vol 55 no 3-4 pp 405ndash418 2012
[12] ZWu H Huang and LWang ldquoStochastic delay logistic modelunder regime switchingrdquo Abstract and Applied Analysis vol2012 Article ID 241702 26 pages 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Journal of Applied Mathematics 11
ofMathematical Analysis and Applications vol 376 no 1 pp 11ndash28 2011
[9] M Liu and KWang ldquoAsymptotic properties and simulations ofa stochastic logistic model under regime switchingrdquoMathemat-ical and Computer Modelling vol 54 no 9-10 pp 2139ndash21542011
[10] G Hu and K Wang ldquoStability in distribution of competi-tive Lotka-Volterra system with Markovian switchingrdquo AppliedMathematical Modelling vol 35 no 7 pp 3189ndash3200 2011
[11] M Liu and K Wang ldquoAsymptotic properties and simulationsof a stochastic logistic model under regime switching IIrdquoMathematical andComputerModelling vol 55 no 3-4 pp 405ndash418 2012
[12] ZWu H Huang and LWang ldquoStochastic delay logistic modelunder regime switchingrdquo Abstract and Applied Analysis vol2012 Article ID 241702 26 pages 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of