Preprint
Article

The convergence and boundedness of solutions to SFDEs with the G-framework

Altmetrics

Downloads

88

Views

29

Comments

0

A peer-reviewed article of this preprint also exists.

This version is not peer-reviewed

Submitted:

22 August 2023

Posted:

22 August 2023

You are already at the latest version

Alerts
Abstract
Generally, stochastic functional differential equations (SFDEs) pose a challenge as they often lack explicit exact solutions. Consequently, it becomes necessary to seek certain favorable conditions under which numerical solutions can converge towards the exact solutions. This article aims to delve into the convergence analysis of solutions for stochastic functional differential equations by employing the framework of G-Brownian motion. To establish the goal, we find a set of useful monotone type conditions and work within the space Cr((−∞,0];Rn). The investigation conducted in this article confirms the mean square boundedness of solutions. Furthermore, this study enables us to compute both LG2 and exponential estimates.
Keywords: 
Subject: Physical Sciences  -   Other

1. Introduction

Stochastic functional differential equations (SFDEs) find applications in various fields of engineering and science, such as neural networks [15], financial assets [4,26,30], population dynamics [1,20], and gene expression [21]. A vast literature exists on moment estimates, convergence, stability, and existence of solutions for SFDEs [17,18,19,22,23]. The study of SFDEs driven by G-Brownian motion is relatively new, dating back to the invention of G-Brownian theory in 2006 [24]. In [4,25], the classical Lipschitz condition and linear growth condition were used to establish the existence-uniqueness theorem for SFDEs in the space B C ( ( , 0 ] ; R n ) . The study of SFDEs under the G-framework with non-Lipschitz conditions and mean square stability was investigated in [8]. The work in [7,9,10,11] provides insights into pth moment estimates, the Cauchy-Maruyama approximation scheme, and exponential estimates for solutions to SFDEs within the framework of G-Brownian motion. Asymptotic estimates were studied in [28], while SFDEs under the G-Lévy processes were investigated in [5]. In this article, we introduce some useful monotone type conditions to study SFDEs under the G-framework within the space C r ( ( , 0 ] ; R n ) . Our findings contribute to the growing body of research on SFDEs driven by G-Brownian motion and deepen our understanding of the role of G-framework in stochastic analysis. We study the convergence of solutions for a SFDEs using the framework of G-Brownian motion. Our analysis results in the mean square boundedness of solutions and allows us to compute both L G 2 and exponential estimates. Consider a matrix A; its transpose is denoted by A T . Let C ( ( , 0 ] ; R n ) denotes the set of continuous mappings from ( , 0 ] to R n . Define the space C r ( ( , 0 ] ; R n ) , r > 0 as
C r ( ( , 0 ] ; R n ) = { α C ( ( , 0 ] ; R n ) : lim σ e r σ α ( σ ) exists in R n } ,
Associated with norm α r = sup < σ 0 e r σ | α ( σ ) | < , the space C r ( ( , 0 ] ; R n ) is a Banach space of bounded continuous mappings. For each 0 < r 1 r 2 < , C r 1 C r 2 [14,29]. Represent the σ -algebra of C r by B ( C r ) and C r 0 = { α C r : lim σ e r σ α ( σ ) = 0 } . Let L 2 ( C r ) denote the space of all F -measurable stochastic processes ψ taking values in C r , such that E ^ | α | r 2 < . Similarly, let L 2 ( C r 0 ) denote the space of all F -measurable stochastic processes α taking values in C r 0 , such that E | α | r 2 < . Let ( Ω , F , P ) be a complete probability space, where F is a sigma-algebra of subsets. The natural filtration F t on ( Ω , F , P ) is defined as the sigma-algebra, denoted by F t = σ B ( v ) : 0 v t , where B ( v ) represents the Borel sigma-algebra of C r . We use P to represent the set of all probability measures on ( C r , B ( C r ) ) . Additionally, L b ( C r ) denotes the collection of continuous bounded functionals on C r . Finally, let Λ 0 be the collection of probability measures on ( , 0 ] satisfying 0 μ ( d σ ) = 1 for every μ Λ 0 . We define
Λ k = { μ Λ 0 : μ ( k ) = 0 e k σ μ ( d σ ) < } ,
where Λ k 0 Λ k Λ 0 for any k ( 0 , k 0 ) [29]. Let κ : C r ( ( , 0 ] ; R n ) R n , η : C r ( ( , 0 ] ; R n ) R n and γ : C r ( ( , 0 ] ; R n ) R n be Borel measurable. Consider the SFDEs driven by G-Brownian motion of the form
d z ( t ) = κ ( z t ) d t + η ( z t ) d B , B ( t ) + γ ( z t ) d B ( t ) ,
on t 0 where z t = { z ( t + θ ) : < θ 0 } . Equation (2) has the starting value z 0 = ζ C r ( ( , 0 ] ; R n ) . Let B , B ( t ) denote the quadratic variation process of the G-Brownian motion B ( t ) defined on a complete probability space ( Ω , F , P ) , where B ( t ) is a one-dimensional process under the filtration F t 0 satisfying the usual conditions. The remaining paper is arranged as follows. Section 2 presents the basic results. In section 3, some useful lemmas are given. Section 4 investigates the mean square boundedness and convergence of solutions. In section 5, we first study the L G 2 estimate and then derive the exponential estimate. Section 6 contains conclusions.

2. Basic Notions and Results

This section presents some fundamental concepts and results that we utilize in the forthcoming research [2,6,16,27]. Let H be a space of real mappings defined on a non-empty set Ω .
 Definition 1. 
x , y H , a functional E ^ : H R assuring the below given features is called a G-expectation
  • E ^ [ y ] E ^ [ x ] whenever y x .
  • E ^ [ m 1 ] = m 1 , for any m 1 R .
  • E ^ [ m 2 y ] = m 2 E ^ [ y ] , for any m 2 R + .
  • E ^ [ y + x ] E ^ [ y ] + E ^ [ x ] .
Suppose that Ω be the space of R n -valued continuous paths ( w ( t ) ) t 0 such that w ( 0 ) = 0 associated with the norm
D ( w 1 , w 2 ) = j = 1 2 j max t [ 0 , j ] | w 1 ( t ) w 2 ( t ) | 1 .
Choose Ω T = { ω T : ω Ω } . Assume the canonical process B ( t ) = B ( t , w ) where t 0 and w Ω . Let λ C b . L i p ( R n × d ) and t 1 , t 2 , . . . , t n [ 0 , T ] then
L i p 0 ( Ω T ) = λ ( B ( t 1 ) , B ( t 2 ) , . . . , B ( t n ) ) : n 1 .
Notice that L i p 0 ( Ω t ) L i p 0 ( Ω T ) , L i p 0 ( Ω ) = m = 1 L i p 0 ( Ω m ) and the completion of L i p 0 ( Ω ) associated with E ^ [ | . | p ] 1 p , p 1 is L G p ( Ω ) . Related to { B ( t ) } t 0 , we can express the filtration as F t = σ { B ( u ) , 0 u t } where F = { F t } t 0 . Let 0 t 0 t 1 . . . t N < and T = { t 0 , t 1 , . . . , t N } be a partition of [ 0 , T ] . Let p 1 then M G p , 0 ( 0 , T ) is given by
M G p , 0 ( 0 , T ) = { ρ t ( w ) = o = 0 N 1 ϑ j ( w ) I [ t o , t o + 1 ] ( t ) : ϑ o L G p ( Ω t o ) , o = 0 , 1 , . . . , N 1 } ,
The space M G p ( 0 , T ) is the completion of M G p , 0 ( 0 , T ) under the norm ρ = 0 T E [ | ρ s | p ] d s 1 / p , p 1 .
 Definition 2. 
Let ρ t M G 2 , 0 ( 0 , T ) . The G-Itô integral I ( ρ ) is defined as the stochastic integral of a function ρ with respect to G-Brownian motion given by
I ( ρ ) = 0 T ρ ( s ) d B a ( s ) = k = 0 N 1 ξ k B a ( t k + 1 ) B a ( t k ) ,
One can extend I : M G 2 , 0 ( 0 , T ) L G 2 ( F T ) to I : M G 2 ( 0 , T ) L G 2 ( F T ) , where for ρ M G 2 ( 0 , T ) we have
0 T ρ ( s ) d B a ( s ) = I ( ρ ) .
 Definition 3. 
Let B a ( 0 ) = 0 . The G-quadratic variation process { B a ( t ) } t 0 is given as follows
B a ( t ) = lim N j = 0 N 1 B a ( t j + 1 N ) B a ( t j N ) 2 = B a ( t ) 2 2 0 t B a ( s ) d B a ( s ) .
Consider a function K 0 , T : M G 0 , 1 ( 0 , T ) L G 2 ( F T ) given as
K 0 , T ( ρ ) = 0 T ρ ( s ) d B a ( s ) = i = 0 N 1 ξ i B a ( t i + 1 ) B a ( t i ) .
One can extend K 0 , T to M G 1 ( 0 , T ) . For ρ M G 1 ( 0 , T ) , it is given by
0 T ρ ( s ) d B a ( s ) = K 0 , T ( ρ ) .
 Lemma 1. 
Assume that γ M G p ( 0 , T ) and p 2 . Then
E ^ sup 0 t T | 0 t γ ( s ) d B ( s ) | p a 3 E ^ 0 t | γ ( s ) | 2 d s p 2 ,
where a 3 ( 0 , ) is a p dependent constant.
 Lemma 2. 
Assume that γ M G p ( 0 , T ) , p 1 . Then
E ^ sup 0 t T | 0 t γ ( s ) d B , B ( s ) | p a 2 E ^ 0 t | γ ( s ) | 2 d s p 2 ,
where a 2 ( 0 , ) depends on p.
 Definition 4. 
The G-Brownian motion is an n-dimensional stochastic process { B ( t ) } t 0 fulfilling the following characteristics
( i )
B ( 0 ) = 0 .
( ii )
B ( t + v ) B ( t ) is G-normally distributed and independent of B ( t 1 ) , B ( t 2 ) , . . . . . . . . B ( t n ) for any n N and 0 t 1 t 2 , . . . , t n t .
 Lemma 3. 
Let p 0 , z L G p . For each m 1 > 0 ,
ν ^ ( | z | > m 1 ) E ^ [ | z | p ] m 1 ,
where E ^ | z | p < .
The following two basic lemmas can also be utilized in forthcoming sections of this paper [17].
 Lemma 4. 
Let c , l 0 and δ ( 0 , 1 ) . Then
( c + l ) 2 c 2 δ + l 2 1 δ .
 Lemma 5. 
Let p 2 and δ ^ , c , l > 0 . Then
( i )
c p 1 l ( p 1 ) δ ^ c p p + l p p δ ^ p 1 .
( ii )
c p 2 l 2 ( p 2 ) δ ^ c p p + 2 l p p δ ^ p 2 2 .

3. Some Useful Results

In this section, we introduce and discuss some important assumptions and establish two lemmas. We consider the following hypotheses
( H )
Let a i > 0 , i = 1 , 2 , . . , 5 and α ( σ ) β ( σ ) = Ψ ( σ ) . For any α , β C r ( ( , 0 ] ; R n ) and for any probability measure μ 1 , μ 2 , μ 3 Λ 2 r the following inequalities hold
[ Ψ ( 0 ) ] T [ κ ( α ) κ ( β ) ] a 1 | Ψ ( 0 ) | 2 + a 2 0 | Ψ ( σ ) | 2 μ 1 ( d σ ) ,
[ Ψ ( 0 ) ] T [ η ( α ) η ( β ) ] a 3 | Ψ ( 0 ) | 2 + a 4 0 | Ψ ( σ ) | 2 μ 2 ( d σ ) ,
| γ ( α ) γ ( β ) | 2 a 5 0 | Ψ ( σ ) | 2 μ 3 ( d σ ) .
 Lemma 6. 
Let a < p r and p 1 . Then
z t r p e a t ζ r p + sup 0 < u t | z ( u ) | p ,
where ζ C r ( ( , 0 ] ; R n ) .
 Proof. 
Assuming that p r > a , we can obtain the following using the definition of the norm | · | r :
z t r p = sup < σ 0 e r σ | z ( t + σ ) | p sup < σ 0 e a σ | z ( t + σ ) | p sup 0 < u t e a ( t u ) | z ( u ) | p + sup < u 0 e a ( t u ) | z ( u ) | p = e a t ζ r p + e a t sup 0 < u t e a s | z ( u ) | p sup 0 < u t | z ( u ) | p + e a t ζ r p .
The proof stands completed. □
Throughout this paper we let that for any p 1 , a < p r .
 Lemma 7. 
Let a < p r , p 2 and μ i Λ k , ∀ i N . Then
0 t 0 | z ( σ + s ) | p μ i ( d σ ) d s μ i ( p r ) p r ζ r p + 0 t | z ( s ) | p d s ,
0 t 0 e a s | z ( σ + s ) | p μ i ( d σ ) d s μ i ( p r ) p r a ζ r p + μ i ( p r ) 0 t e a s | z ( s ) | p d s ,
where ζ C r ( ( , 0 ] ; R n ) .
 Proof. 
As for each i Z + , μ i Λ p r and ζ C r ( ( , 0 ] ; R n ) , by using the Fubini theorem and the definition of norm, it follows
0 t 0 | z ( σ + s ) | p μ i ( d σ ) d s = 0 t s e p r ( σ + s ) | z ( σ + s ) | p e p r ( s + σ ) μ i ( d σ ) + s 0 | z ( σ + s ) | p μ i ( d σ ) d s ζ r p 0 t e p r s d s 0 e p r σ μ i ( d σ ) + 0 μ i ( d σ ) 0 t | z ( s ) | p d s .
Observing that 0 μ i ( d σ ) = 1 and 0 e p r σ μ i ( d σ ) = μ i ( p r ) , i N , it follows
0 t 0 | z ( s + σ ) | p μ i ( d σ ) d s μ i ( p r ) p r ζ r p + 0 t | z ( s ) | p d s .
The proof of (6) is complete. Using similar arguments as used above we determine
0 t 0 e a s | z ( σ + s ) | p μ i ( d σ ) d s = 0 t e a s d s s | z ( σ + s ) | p μ i ( d σ ) + s 0 | z ( σ + s ) | p μ i ( d σ ) = 0 t e a s d s s | z ( σ + s ) | p μ i ( d σ ) + t 0 μ i ( d σ ) σ t e a s | z ( σ + s ) | p d s 0 t e a s d s s e p r ( σ + s ) | z ( s + σ ) | p e p r ( σ + s ) μ i ( d σ ) + 0 μ i ( d σ ) 0 t e a ( s σ ) | z ( s ) | p d s ζ r p 0 t e ( p r a ) s d s 0 e p r σ μ i ( d σ ) + 0 e a σ μ i ( d σ ) 0 t e a s | z ( s ) | p d s .
With reference to equation (1), and taking note that p r > a , we can conclude that
0 t 0 e a s | z ( σ + s ) | p μ i ( d σ ) d s μ i ( p r ) p r a ζ r p + μ i ( p r ) 0 t e a s | z ( s ) | p d s .
The proof of (7) is complete. □

4. Convergence and Mean Square Boundedness

Firstly, let us derive the mean square boundedness for solutions to equation (2).
 Theorem 1. 
Let the hypothesis H holds. Assume the equation (2) with initial condition ζ C r ( ( , 0 ] ; R n ) has just one solution z ( t ) . Let a i , i = 1 , 2 , . . , 5 assure 2 a 1 > 2 a 2 μ 1 ( 2 r ) + 2 b 1 a 4 μ 2 ( 2 r ) + b 1 a 5 μ 3 ( 2 r ) 2 b 1 a 3 . Then there is a ( 0 , ( 2 a 1 + 2 b 1 a 3 2 a 2 μ 1 ( 2 r ) 2 b 1 a 4 μ 2 ( 2 r ) b 1 a 5 μ 3 ( 2 r ) ) 2 r ) so that
E ^ [ | z ( t ) | 2 ] c 1 + c 2 e a t ,
where
c 1 = 1 a 1 δ | κ ( 0 ) | 2 + b 1 δ 1 | η ( 0 ) | 2 + b 1 δ 2 | γ ( 0 ) | 2
and
c 2 = E ^ | z ( 0 ) | 2 + 2 a 2 μ 1 ( 2 r ) 2 r a E ^ ζ r 2 + 2 b 1 a 4 μ 2 ( 2 r ) 2 r a E ^ ζ r 2 + b 1 a 5 μ 3 ( 2 r ) ( 2 r a ) ( 1 δ 2 ) E ^ ζ r 2 .
The values of δ, δ 1 and δ 2 are sufficiently small so that
2 a 1 δ a b 1 δ 1 + 2 b 1 a 3 2 a 2 μ 1 ( 2 r ) 2 b 1 a 4 μ 2 ( 2 r ) b 1 a 5 1 δ 2 μ 3 ( 2 r ) > 0 .
 Proof. 
By using the G-Itô formula, G-Itô integral and Lemma 2, it follows
E ^ [ e a t | z ( t ) | 2 ] E ^ | z ( 0 ) | 2 + E ^ 0 t e a s a | z ( s ) | 2 + 2 z T ( s ) κ ( z s ) d s + b 1 E ^ 0 t e a s 2 z T ( s ) η ( z s ) + | γ ( z s ) | 2 d s .
Utilizing (3), (4) and Lemma 5 we determine
z T ( t ) κ ( z t ) ( δ 2 a 1 ) | z ( t ) | 2 + 1 2 δ | κ ( 0 ) | 2 + a 2 0 | z ( t + σ ) | 2 μ 1 ( d σ ) ,
z T ( t ) η ( z t ) ( δ 1 2 a 3 ) | z ( t ) | 2 + 1 2 δ 1 | η ( 0 ) | 2 + a 4 0 | z ( t + σ ) | 2 μ 2 ( d σ ) .
It follows from Lemma 4 and the condition given in (5) that
| γ ( z t ) | 2 1 δ 2 | γ ( 0 ) | 2 + a 5 1 δ 2 0 | z ( t + σ ) | 2 μ 3 ( d σ ) .
Using the above inequalities, (9) becomes
E ^ [ e a t | z ( t ) | 2 ] E ^ | z ( 0 ) | 2 + 1 a 1 δ | κ ( 0 ) | 2 + b 1 δ 1 | η ( 0 ) | 2 + b 1 δ 2 | γ ( 0 ) | 2 ( e a t 1 ) + ( δ + a 2 a 1 + b 1 δ 1 2 b 1 a 3 ) E ^ 0 t e a s | z ( s ) | 2 d s + 2 a 2 E ^ 0 t e a s 0 | z ( s + σ ) | 2 μ 1 ( d σ ) d s + 2 b 1 a 4 E ^ 0 t e a s 0 | z ( s + σ ) | 2 μ 2 ( d σ ) d s + b 1 a 5 1 δ 2 E ^ 0 t e a s 0 | z ( s + σ ) | 2 μ 3 ( d σ ) d s
In view of Lemma 7, it follows
0 t 0 e a s | z ( s + σ ) | 2 μ i ( d σ ) d s 1 2 r a ζ r 2 μ i ( 2 r ) + μ i ( 2 r ) 0 t e a s | z ( s ) | 2 d s .
Substituting (11) in (10) we derive
E ^ [ e a t | z ( t ) | 2 ] E ^ | z ( 0 ) | 2 + 2 a 2 μ 1 ( 2 r ) 2 r a E ^ ζ r 2 + 2 b 1 a 4 μ 2 ( 2 r ) 2 r a E ^ ζ r 2 + b 1 a 5 μ 3 ( 2 r ) ( 2 r a ) ( 1 δ 2 ) E ^ ζ r 2 + 1 λ 1 δ | κ ( 0 ) | 2 + b 1 δ 1 | η ( 0 ) | 2 + b 1 δ 2 | γ ( 0 ) | 2 ( e a t 1 ) ( 2 a 1 δ a b 1 δ 1 + 2 b 1 a 3 2 a 2 μ 1 ( 2 r ) 2 b 1 a 4 μ 2 ( 2 r ) b 1 a 5 1 δ 2 μ 3 ( 2 r ) ) E ^ 0 t e a s | z ( s ) | 2 d s .
As 2 a 1 > 2 a 2 μ 1 ( 2 r ) + 2 b 1 a 4 μ 2 ( 2 r ) + b 1 a 5 μ 3 ( 2 r ) 2 b 1 a 3 and a ( 0 , ( 2 a 1 + 2 b 1 a 3 2 a 2 μ 1 ( 2 r ) 2 b 1 a 4 μ 2 ( 2 r ) b 1 a 5 μ 3 ( 2 r ) ) 2 r ) . Selecting δ , δ 1 and δ 2 sufficiently small so that
2 a 1 δ a b 1 δ 1 + 2 b 1 a 3 2 a 2 μ 1 ( 2 r ) 2 b 1 a 4 μ 2 ( 2 r ) b 1 a 5 1 δ 2 μ 3 ( 2 r ) > 0 ,
we obtain the desired result
E ^ [ | z ( t ) | 2 ] c 1 + c 2 e a t ,
where
c 1 = 1 a 1 δ | κ ( 0 ) | 2 + b 1 δ 1 | η ( 0 ) | 2 + b 1 δ 2 | γ ( 0 ) | 2
and
c 2 = E ^ | z 0 | 2 + 2 a 2 μ 1 ( 2 q ) 2 r a E ^ ζ r 2 + 2 b 1 a 4 μ 2 ( 2 r ) 2 r a E ^ ζ r 2 + b 1 a 5 μ 3 ( 2 r ) ( 2 r a ) ( 1 δ 2 ) E ^ ζ r 2 .
Theorem 1 describes that equation (2) has a mean square bounded solution. The following Theorem 2 expresses that any two distinct solutions of equation (2) are convergent.
 Theorem 2. 
Assuming that all hypotheses of Theorem 1 are satisfied, let z ( t ) and y ( t ) be two solutions of equation (2) associated with initial values ζ and ξ, respectively. Then, we have:
E ^ [ | z ( t ) y ( t ) | 2 ] c 3 E ^ | ξ ζ | r 2 e a t ,
where c 3 = 1 + 1 2 r a ( 2 a 2 μ 1 ( 2 r ) + 2 b 1 a 4 μ 2 ( 2 r ) + b 1 a 5 μ 3 ( 2 r ) ) .
 Proof. 
Define ϕ ( t ) = z ( t ) y ( t ) , γ ^ ( t ) = γ ( z t ) γ ( y t ) , κ ^ ( t ) = κ ( z t ) κ ( y t ) , and η ^ ( t ) = η ( z t ) η ( y t ) . Utilizing the G-Itô integral, Lemma 2, and G-Itô formula, it follows
E ^ [ e a t | ϕ ( t ) | 2 ] E ^ | ξ ( 0 ) ζ ( 0 ) | 2 + E ^ 0 t e a s [ a | ϕ ( s ) | 2 + 2 ϕ T ( s ) κ ^ ( s ) ] d s + b 1 E ^ 0 t e ϕ s [ 2 ϕ T ( s ) η ^ ( s ) + | γ ^ ( s ) | 2 ] d s .
From hypothesis H, we derive
ϕ T ( t ) κ ^ ( t ) a 1 | ϕ ( t ) | 2 + a 2 0 ϕ ( σ + t ) μ 1 ( d σ ) , ϕ T ( t ) η ^ ( t ) a 3 | ϕ ( t ) | 2 + a 4 0 ϕ ( σ + t ) μ 2 ( d σ )
and
| γ ^ ( t ) | 2 a 5 0 ϕ ( t + σ ) μ 3 ( d σ ) .
Utilizing the above inequalities, (13) becomes
E ^ [ e a t | ϕ ( t ) | 2 ] E ^ | ξ ( 0 ) ζ ( 0 ) | 2 + ( a 2 a 1 2 b 1 a 3 ) E ^ 0 t e a s | ϕ ( s ) | 2 d s + 2 a 2 E ^ 0 t 0 e a s ϕ ( s + σ ) μ 1 ( d σ ) d s + 2 b 1 a 4 E ^ 0 t 0 e a s ϕ ( s + σ ) μ 2 ( d σ ) d s + k 1 a 5 E ^ 0 t 0 e a s ϕ ( s + σ ) μ 3 ( d σ ) d s .
From Lemma 7 for i = 1 , 2 , 3 it follows
0 t 0 e a s | ϕ ( s + σ ) | 2 μ i ( d σ ) d s 1 2 q a ζ ξ r 2 μ i ( 2 r ) + μ i ( 2 r ) 0 t e a s | ϕ ( s ) | 2 d s
Plugging (15) in (14) we determine
e a t E ^ | ϕ ( t ) | 2 E ^ | ξ ( 0 ) ζ ( 0 ) | 2 + 1 2 r a [ 2 a 2 μ 1 ( 2 r ) + 2 b 1 a 4 μ 2 ( 2 r ) + b 1 a 5 μ 3 ( 2 r ) ] E ^ ζ ξ r 2 ( 2 a 1 + 2 b 1 a 3 a 2 a 2 μ 1 ( 2 r ) 2 k 1 a 4 μ 2 ( 2 r ) b 1 a 5 μ 3 ( 2 r ) ) E ^ 0 t e a s | ϕ ( s ) | 2 d s .
As 2 a 1 > 2 a 2 μ 1 ( 2 r ) + 2 b 1 a 4 μ 2 ( 2 r ) + b 1 a 5 μ 3 ( 2 r ) 2 b 1 a 3 and a ( 0 , ( 2 a 1 + 2 b 1 a 3 2 a 2 μ 1 ( 2 r ) 2 b 1 a 4 μ 2 ( 2 r ) b 1 a 5 μ 3 ( 2 r ) ) 2 r ) it follows
E ^ | ϕ ( t ) | 2 [ 1 + 1 2 r a ( 2 a 2 μ 1 ( 2 r ) + 2 b 1 a 4 μ 2 ( 2 r ) + b 1 a 5 μ 3 ( 2 r ) ) ] E ^ ζ ξ r 2 e a t ,
consequently, we derive the following required result
E ^ | z ( t ) y ( t ) | 2 c 3 E ^ ζ ξ r 2 e a t ,
where c 3 = 1 + 1 2 r a ( 2 a 2 μ 1 ( 2 r ) + 2 b 1 a 4 μ 2 ( 2 r ) + b 1 a 5 μ 3 ( 2 r ) ) . □
If κ ( 0 ) = η ( 0 ) = γ ( 0 ) = 0 , then from Theorem 2 we can obtain that the trivial solution of equation (2) is mean square exponentially stable.
 Example 1. 
Consider z ( t ) and y ( t ) be two solutions of the equation
d z ( t ) = z t d t + s i n ( z t ) d B ( t ) + z t d B ( t )
with initial values ζ and ξ respectively. Define ϕ ( t ) = z ( t ) y ( t ) , η ^ ( t ) = γ ^ ( t ) = z t y t and κ ^ ( t ) = s i n ( z t ) s i n ( y t ) . Under the given hypothesis one can easily derive that y ( t ) is mean square convergent to z ( t ) .

5. The Exponential Estimate

Firstly, let us determine the L G 2 estimates. Let equation (2) with initial condition ζ C r ( ( , 0 ] ; R n ) has just one solution z ( t ) on t [ 0 , ) .
 Theorem 3. 
Assume that the hypothesis H holds and E ζ r 2 < . For every t 0 ,
E ^ sup < s t | z ( t ) | 2 [ E ^ ζ r 2 + m 1 ] e m 2 t ,
where m 1 = c + 2 r [ r + a 2 μ 1 ( 2 r ) + b 1 ( a 5 μ 3 ( 2 r ) + μ 2 ( 2 r ) ) + 2 b 3 a 5 μ 3 ( 2 r ) ] E ^ ζ r 2 , c = 2 [ | κ ( 0 ) | 2 + b 1 ( | η ( 0 ) | 2 + 2 | γ ( 0 ) | 2 ) + 4 b 3 | γ ( 0 ) | 2 ] T and m 2 = 2 [ 2 a 2 2 a 1 + 1 + b 1 ( 2 a 5 2 a 3 + 3 ) + 4 b 3 a 5 ] .
 Proof. 
Using the G-It o ^ formula and properties of the G-expectation, it follows
E ^ sup 0 s t | z ( t ) | 2 E ^ | z ( 0 ) | 2 + 2 E ^ sup 0 s t 0 t z T ( s ) κ ( z s ) d s + E ^ sup 0 s t 0 t ( 2 z T ( s ) η ( z s ) + | γ ( z s ) | 2 ) d B , B ( s ) + 2 E ^ sup 0 s t 0 t z T ( s ) γ ( z s ) d B ( s )
From our assumption H, 2 a 1 a 2 i = 1 2 a i 2 and ( i = 1 2 a i ) 2 2 i = 1 2 a i 2 , it follows
z T ( t ) κ ( z t ) ( a 1 1 2 ) | z ( t ) | 2 + 1 2 | κ ( 0 ) | 2 + a 2 0 | z ( t + σ ) | 2 μ 1 ( d σ ) ,
z T ( t ) η ( z t ) ( a 3 1 2 ) | z ( t ) | 2 + 1 2 | η ( 0 ) | 2 + a 4 0 | z ( t + σ ) | 2 μ 2 ( d σ ) ,
| γ ( z t ) | 2 2 | γ ( 0 ) | 2 + 2 a 5 0 | z ( t + σ ) | 2 μ 3 ( d σ ) .
In view of (17) and (11), we determine
2 E ^ sup 0 < s t 0 t z T ( s ) κ ( z s ) d s | κ ( 0 ) | 2 T + a 2 r E ^ ζ r 2 μ 1 ( 2 r ) + ( 2 a 2 2 a 1 + 1 ) E ^ 0 t | z ( s ) | 2 d s .
Utilizing (18), (19), (11) and Lemma 2, it follows
E ^ sup 0 s t 0 t ( 2 z T ( s ) η ( z s ) + | γ ( z s ) | 2 ) d B , B ( s ) b 1 E ^ 0 t ( 2 z T ( s ) η ( z s ) + | γ ( z s ) | 2 ) d s b 1 [ | η ( 0 ) | 2 + 2 | γ ( 0 ) | 2 ] T + b 1 1 r ( a 5 μ 3 ( 2 r ) + μ 2 ( 2 r ) ) E ^ ζ r 2 + b 1 ( 2 a 5 2 a 3 + 3 ) E ^ 0 t | z ( s ) | 2 d s .
The inequality a 1 a 2 1 2 i = 1 2 a i , Lemma 1 and (19) give
2 E ^ sup 0 < s t 0 t z T ( s ) γ ( z s ) d B ( t ) 2 b 2 E ^ 0 t | z T ( s ) γ ( z s ) | 2 d s 1 2 1 2 E ^ sup 0 < s t | z ( s ) | 2 + 4 b 2 2 | γ ( 0 ) | 2 T + 4 b 2 2 a 5 E ^ 0 t 0 | z ( s + σ ) | 2 μ 3 ( d σ ) d s .
Using Lemma 7, we derive
2 E ^ sup 0 < s t 0 t z T ( s ) γ ( z s ) d B ( s ) 1 2 E ^ sup 0 < s t | z ( s ) | 2 + 4 b 3 | γ ( 0 ) | 2 T + 4 b 3 a 5 2 r μ 3 ( 2 r ) E ^ ζ r 2 + 4 b 3 a 5 E ^ 0 t | z ( s ) | 2 d s ,
where b 3 = b 2 2 . By substituting the aforementioned inequalities into equation (16), and letting m 1 = c + 2 r [ r + a 2 μ 1 ( 2 r ) + b 1 ( a 5 μ 3 ( 2 r ) + μ 2 ( 2 r ) ) + 2 b 3 a 5 μ 3 ( 2 r ) ] E ζ r 2 , m 2 = 2 [ 2 a 2 2 a 1 + 1 + b 1 ( 2 a 5 2 a 3 + 3 ) + 4 b 3 a 5 ] , we can evaluate the result
E ^ [ sup 0 s t | z ( t ) | 2 ] m 1 + m 2 0 t E ^ sup 0 s t | z ( s ) | 2 d s ,
where c = 2 | κ ( 0 ) | 2 + b 1 ( | η ( 0 ) | 2 + 2 | γ ( 0 ) | 2 ) + 4 b 3 | γ ( 0 ) | 2 T . By observing that
E ^ [ sup < s t | z ( s ) | 2 ] E ^ ζ r 2 + E ^ [ sup 0 s t | z ( s ) | 2 ] ,
it follows
E ^ sup < s t | z ( s ) | 2 E ^ ζ r 2 + m 1 + m 2 0 t E ^ sup 0 s t | z ( s ) | 2 d s E ^ ζ r 2 + m 1 + m 2 0 t E ^ sup < s t | z ( s ) | 2 d s .
Finally, the required result is obtained by using the Grownwall inequality. □
 Theorem 4. 
Under the conditions of Theorem 3, it follows
lim t sup 1 t l o g | z ( t ) | α ,
and α = 2 a 2 2 a 1 + 1 + b 1 ( 2 a 5 2 a 3 + 3 ) + 4 b 3 a 5 .
 Proof. 
Assuming that m 1 = m + 2 r [ r + a 2 μ 1 ( 2 r ) + b 1 ( a 5 μ 3 ( 2 r ) + μ 2 ( 2 r ) ) + 2 b 3 a 5 μ 3 ( 2 r ) ] E ζ r 2 and m 2 = 2 [ 2 a 2 2 a 1 + 1 + b 1 ( 2 a 5 2 a 3 + 3 ) + 4 b 3 a 5 ] then from the inequality (21), we can conclude that:
E ^ sup 0 s t | z ( s ) | 2 m 1 e m 2 t ,
where m = 2 | κ ( 0 ) | 2 + b 1 ( | η ( 0 ) | 2 + 2 | γ ( 0 ) | 2 ) + 4 b 3 | γ ( 0 ) | 2 T . For each q = 1 , 2 , 3 , . . . , from (22) it follows
E ^ sup q 1 t q | z ( t ) | 2 m 1 e m 2 q .
By utilizing Lemma 3 for every given δ > 0 , we obtain:
ν w : sup q 1 t q | z ( t ) | 2 > e ( m 2 + δ ) q E ^ sup q 1 t q | z ( t ) | 2 e ( m 2 + δ ) q m 1 e m 2 q e ( m 2 + δ ) q = m 1 e δ q .
But the Borel-Cantelli lemma gives that for almost every w Ω there is a random number q 0 = q 0 ( w ) Z in a manner that when q q 0 , then
sup q 1 t q | z ( t ) | 2 e ( m 2 + δ ) q ,
which implies
lim t sup 1 t l o g | z ( t ) | m 2 + δ 2 = 2 a 2 2 a 1 + 1 + b 1 ( 2 a 5 2 a 3 + 3 ) + 4 b 3 a 5 + δ 2 ,
as δ is arbitrary and letting α = 2 a 2 2 a 1 + 1 + b 1 ( 2 a 5 2 a 3 + 3 ) + 4 b 3 a 5 , we can conclude that
α lim t sup 1 t l o g | z ( t ) | .
The proof stands completed. □
The lemma above expresses that the second moment of the Lyapunov exponent, as defined in [13] as lim t sup 1 t log | z ( t ) | , is bounded above by M.

6. Conclusion

Several stochastic functional differential equations (SFDEs) in financial mathematics do not hold the standard Lipschitz assumption such as the Cox-Ingersoll-Ross, Heston and Ait-Sahalia models. In this article some useful monotone type conditions have been introduced. We have proved that any two solutions of SFDEs in the G-framework under distinct initial conditions are convergent. The solutions are mean square bounded. The L G 2 and exponential estimates have been calculated. We anticipate that the findings presented in this article will offer valuable insights into the analysis of equations, even when not under the constraints of standard assumptions. This contribution is poised to have a substantial positive impact on the examination of various unresolved inquiries, including the investigation into the existence, uniqueness, convergence and stability of solutions for backward and forward stochastic dynamic systems driven by G-Brownian motion with conditions of a monotone nature.

Author Contributions

”Conceptualization, F.F.; methodology, R.U.; formal analysis, Q.C.; investigation, F.F. and R.U.; All authors have read and agreed to the published version of the manuscript.”,

Funding

“This research was funded by NUST Pakistan.”

Acknowledgments

“The financial support of NUST Pakistan is acknowledged with thanks.”

Conflicts of Interest

“All authors have no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results”.

References

  1. Bahar, A.; Mao, X.; Stochastic delay Lotka-Volterra model, J. Math. Anal. Appl. 2004, 292, 364–380. [CrossRef]
  2. Bai, X.; Lin, Y.; On the existence and uniqueness of solutions to stochastic differential equations driven by G-Brownian motion with Integral-Lipschitz coefficients, Acta Mathematicae Applicatae Sinica, English Series 2014, 3(30), 589–610.
  3. Denis, L.; M. Hu, M.; Peng, S.; Function spaces and capacity related to a sublinear expectation: Application to G-Brownian motion paths, Potential Anal 2011, 34, 139–161. [CrossRef]
  4. Chang, M.H.; Youree, R.K. ; The European option with hereditary price structures: basic theory, Appl. Math. Comput. 1999, 102, 279–296. [Google Scholar] [CrossRef]
  5. Faizullah, F.; Ullah, R.; Quanxin Zhu, On the existence and mean square stability of neutral stochastic functional differential equations driven by G-Levy process, Mathematical Methods in the Applied Sciences 2022, 1-12.
  6. Faizullah, F.; Bux, M.; Rana, M.A.; Rahman, G.; Existence and stability of solutions to non-linear neutral stochastic functional differential equations in the framework of G-Brownian motion, Advances in Difference Equations 2017, 1, 350–364. [CrossRef]
  7. Faizullah, F.; Zhu, Q.; Ullah, R.; The existence-uniqueness and exponential estimate of solutions for stochastic functional differential equations driven by G-Brownian motion, Mathematical Methods in the Applied Sciences 2020, 1–12. [CrossRef]
  8. Faizullah, F.; Existence and uniqueness of solutions to SFDEs driven by G-Brownian motion with non-Lipschitz conditions, Journal of Computational Analysis and Applications 2017, 2(23), 344-354.
  9. Faizullah, F.; A note on p-th moment estimates for stochastic functional differential equations in the framework of G-Brownian motion, Iran J Sci Technol Trans Sci, 2017, 41, 1131–1138. [CrossRef]
  10. Faizullah, F. ; On the pth moment estimates of solutions to stochastic functional differential equations in the G-framework, SpringerPlus 2016, 5(872), 1–11. [CrossRef]
  11. Faizullah, F.; Existence of solutions for G-SFDEs with Cauchy-Maruyama Approximation Scheme, Abst. Appl. Anal. 2014, 1-8. [CrossRef]
  12. Gao, F.; Pathwise properties and homeomorphic flows for stochastic differential equations driven by G-Brownian motion, Stoch Proc Appl. 2009, 119, 3356–3382. [CrossRef]
  13. Kim, Y.H. ; On the pth moment extimats for the slutions of stochastic differential equations, J. Inequal. Appl. 2014, 395, 1–9. [Google Scholar]
  14. Kuang, Y.; Smith, H.L.; Global stability for infinite delay Lotka-Volterra type system, J. Differential Equations, 103 (1993) 221–246. [CrossRef]
  15. Li, X.; Fu, X.; Stability analysis of stochastic functional differential equations with infinite delay and its application to recurrent neural networks, Journal of Computational and Applied Mathematics 2010, 234, 407–417. [CrossRef]
  16. Luo, P.; Wang, F.; On the comparison theorem for multi-dimensional G-SDEs, Statistics and Probability Letters 2015, 96, 38–44. [CrossRef]
  17. Mao, X.; Stochastic Differential Equations and their Applications, Horwood Publishing Chichester, Coll House England 1997.
  18. Mao, X.; Exponential Stability of Stochastic Differential Equations, Marcel Dekker, Inc. 270 Madison Avenue, New York 1994. [CrossRef]
  19. Mao, X. ; Chenggui Yuan, Stochastic Differential Equations with Markovian Switching,Imperial College Press, London WC2H 9HE, 2006.
  20. Mao, X.; Yuan, C.; Zou, J. ; Stochastic differential delay equations of population dynamics, J. Math. Anal. Appl. 2005, 304, 296–320. [Google Scholar] [CrossRef]
  21. Miekisz, J.; Poleszczuk, J.; Bodnar, M.; Fory’s, U.; Stochastic models of gene expression with delayed degradation, Bulletin of mathematical biology 2011, 73(9), 2231–2247. [CrossRef]
  22. Mohammed, S-E.A.; Stochastic Functional Differential Equations, Research Notes in Mathematics 99, Pitman Advanced Publishing Program, Boston, MA, 1984.
  23. Mohammed, S-E.A.; Scheutzow, M,K,R.; The stable manifold theorem for non-linear stochastic systems with memory. I. Existence of the semiflow, Journal of Functional Analysis 2003, 205, 271-305. [CrossRef]
  24. Peng, S.; Multi-dimentional G-Brownian motion and related stochastic calculus under G-expectation, Stoch Proc Appl 2008, 118, 2223–2253. [CrossRef]
  25. Ren, Y.; Bi, Q.; Sakthivel, R.; Stochastic functional differential equations with infinite delay driven by G-Brownian motion, Math Method Appl Sci. 2013, 36(13), 1746-1759. [CrossRef]
  26. Takeuchi, A.; Joint distributions for stochastic functional differential equations, An International Journal of Probability and Stochastic Processes 2016 5(88), 711–736. [CrossRef]
  27. Ullah, R.; Faizullah, F.; On existence and approximate solutions for stochastic differential equations in the framework of G-Brownian motion, European Physical Journal Plus 2017, 132, 435–443. [CrossRef]
  28. Wei, W.; Zhang, M.; Luo, P.; Asympyotic estimates for the solution of stochastic differential equations driven by G-Brownian motion, Applicable Analysis 2017, 3, 1-12. [CrossRef]
  29. Wu, F.; Yin, G.; Mei, H. ; Stochastic functional differential equations with infinite delay: Existence and uniqueness of solutions, solution maps, Markov properties, and ergodicity, J. Differential Equations 2017, 262, 1226–1252. [Google Scholar] [CrossRef]
  30. Zheng, Y.; Asset pricing based on stochastic delay differential equations, Graduate Theses and Dissertations, (2015). [CrossRef]
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.
Copyright: This open access article is published under a Creative Commons CC BY 4.0 license, which permit the free download, distribution, and reuse, provided that the author and preprint are cited in any reuse.
Prerpints.org logo

Preprints.org is a free preprint server supported by MDPI in Basel, Switzerland.

Subscribe

© 2024 MDPI (Basel, Switzerland) unless otherwise stated