H | 0.6 | 0.7 | 0.75 | 0.85 |
alpha | 0.45 | 0.4 | 0.375 | 0.325 |
gamma | 0.1 | 0.2 | 0.25 | 0.25 |
XEerrsup | 14.5509 | 5.8126 | 4.07903 | 2.36231 |
Comega | 25.3346 | 17.6205 | 16.3161 | 9.44923 |
Citation: Weiguo Liu, Yan Jiang, Zhi Li. Rate of convergence of Euler approximation of time-dependent mixed SDEs driven by Brownian motions and fractional Brownian motions[J]. AIMS Mathematics, 2020, 5(3): 2163-2195. doi: 10.3934/math.2020144
[1] | Xueqi Wen, Zhi Li . $ p $th moment exponential stability and convergence analysis of semilinear stochastic evolution equations driven by Riemann-Liouville fractional Brownian motion. AIMS Mathematics, 2022, 7(8): 14652-14671. doi: 10.3934/math.2022806 |
[2] | Kung-Chi Chen, Kuo-Shing Chen . Pricing green financial options under the mixed fractal Brownian motions with jump diffusion environment. AIMS Mathematics, 2024, 9(8): 21496-21523. doi: 10.3934/math.20241044 |
[3] | Kinda Abuasbeh, Ramsha Shafqat, Azmat Ullah Khan Niazi, Muath Awadalla . Nonlocal fuzzy fractional stochastic evolution equations with fractional Brownian motion of order (1,2). AIMS Mathematics, 2022, 7(10): 19344-19358. doi: 10.3934/math.20221062 |
[4] | Huantian Xie, Nenghui Kuang . Least squares type estimations for discretely observed nonergodic Gaussian Ornstein-Uhlenbeck processes of the second kind. AIMS Mathematics, 2022, 7(1): 1095-1114. doi: 10.3934/math.2022065 |
[5] | Weiwei Shen, Yan Zhang . Strong convergence of the Euler-Maruyama method for the stochastic volatility jump-diffusion model and financial applications. AIMS Mathematics, 2025, 10(5): 12032-12054. doi: 10.3934/math.2025545 |
[6] | Fawaz K. Alalhareth, Seham M. Al-Mekhlafi, Ahmed Boudaoui, Noura Laksaci, Mohammed H. Alharbi . Numerical treatment for a novel crossover mathematical model of the COVID-19 epidemic. AIMS Mathematics, 2024, 9(3): 5376-5393. doi: 10.3934/math.2024259 |
[7] | Xinyi Wang, Jingshen Wang, Zhidong Guo . Pricing equity warrants under the sub-mixed fractional Brownian motion regime with stochastic interest rate. AIMS Mathematics, 2022, 7(9): 16612-16631. doi: 10.3934/math.2022910 |
[8] | Yifei Wang, Haibo Gu, Ruya An . Averaging principle for space-fractional stochastic partial differential equations driven by Lévy white noise and fractional Brownian motion. AIMS Mathematics, 2025, 10(4): 9013-9033. doi: 10.3934/math.2025414 |
[9] | Mounir Zili, Eya Zougar, Mohamed Rhaima . Fractional stochastic heat equation with mixed operator and driven by fractional-type noise. AIMS Mathematics, 2024, 9(10): 28970-29000. doi: 10.3934/math.20241406 |
[10] | Fan Yang, Xiaohui Ai . Exponential stability of stochastic complex networks with multi-weights driven by second-order process based on graph theory. AIMS Mathematics, 2024, 9(4): 9847-9866. doi: 10.3934/math.2024482 |
In this paper, we consider the following time-dependent mixed stochastic differential equations(SDEs) involving independent Brownian motions and fractional Brownian motions(fBms) with Hurst parameter H∈(12,1), defined on the complete probability space (Ω,F,{Ft}t∈[0,T],P),
Xt=X0+∫t0a(s,Xs)ds+∫t0b(s,Xs)dWs+∫t0c(s,Xs)dBHs,t∈[0,T], | (1.1) |
where X0 is F0-measurable random variable, the stochastic integral with respect to Brownian motion W={Wt:t∈[0,T]} and fBm BH={BHt:t∈[0,T]} are interpreted as Itô and pathwise Riemann-Stieltjes integral respectively. a,b,c:Ω×[0,T]×R→R are measurable functions such that all integrals on the right hand side of (1.1) are well defined.
On one hand, SDEs driven only by Brownian motions has long history. We can refer to the monograph [1]. On the other hand, the increasing interest of SDEs driven only by fBms is motivated by their applications in various fields of science such as physics, chemistry, computational mathematics, financial markets e.t. (see [2,3,4]). In particular, for H>12 the increments of fBm are positively correlated and moreover the generalized spectral density behaves like λ−2H+1. The two properties have recently led to applications of them in various fields, which include the noise simulation in electronic circuits [5], the modelling of the subdiffusion of a protein molecule [6], the pricing of weather derivatives [7] and so on.
Recently, mixed stochastic models containing both Brownian motions and fBms gained a lot of attention, e.g. [8,9,10,11,12]. They allow us to model systems driven by a combination of random noises, one of which is white and another has a long memory. The motivation to consider such equations comes from some financial applications, where Brownian motion as a model is inappropriate because of the lack of memory, and fBm with H>12 is too smooth. A model driven by both processes is free of such drawbacks. For example, in financial mathematics, the underlying random noise includes a "fundamental" part, describing the economical background for a stock market, and a "trading" part, coming from the randomness inherent for the stock market. In this case, the fundamental part of the noise should have a long memory, while the second part is likely to be a white noise.
The existence and uniqueness for the solutions of mixed SDEs is discussed by an extensive literature (see [9,13,14]). However, the solution of (1.1) is rarely analytically tractable, so it is important to consider certain numerical methods to solve it. Euler approximation used in this paper usually is most popular and probably simplest among all methods of approximation of SDEs. There have been several works devoted to Euler approximation of mixed SDEs (see [13,15,16]). Guerra and Nualart [13] established the global existence and uniqueness for the solutions of multidimensional time-dependent mixed SDEs under the assumption that W and BH are independent. The proof relied on an estimate for Euler approximation of them, which was obtained by using fractional calculus and Itô integration. Mishura and Shevchenko [16] considered the following mixed SDEs involving both standard Brownian motions and fBms with Hurst parameter H>12,
Xt=X0+∫t0a(s,Xs)ds+∫t0b(s,Xs)dWs+∫t0c(Xs)dBHs,t∈[0,T]. | (1.2) |
Under the boundedness of a(t,x),b(t,x),c(x)(c(x)>0) together with their partial derivatives in x, and (2H−1)-Hölder continuity of a(t,x) and b(t,x) in t, they showed that the mean-square rate of convergence for Euler approximation of (1.2) was O(δ12∧(2H−1)), where δ is the mesh of the partition of [0,T]. We can also find that a faster convergent rate O(δ12) can be deduced if we apply the modified Euler method (see [15]).
However, there is an obstacle to discuss mixed SDEs because of the different machinery behind Itô integral with respect to W and Riemann-Stieltjes integral with respect to BH, particularly in the multidimensional and time-dependent cases. Exactly, the former integral is treated usually in a mean-square sense, while the latter is understood in a pathwise sense and all estimates are pathwise with random constants. Therefore, it is very hard to analyze with standard tools of stochastic analysis. This forces us to consider very smooth coefficients and to make delicate estimates on a suitable space. For example, the measurable space Wα,∞0([0,T]) with the following norm was introduced in [17],
‖Xt‖α,∞:=supt∈[0,T]‖Xt‖α=supt∈[0,T](|Xt|+∫t0|Xt−Xs|(t−s)α+1)<∞a.s.α∈(1−H,1/2). | (1.3) |
The method of dealing with mixed SDEs in Wα,∞0([0,T]) does solve a lot of questions (see [9,11,13,14,18,19]).
The aim of this paper is to prove that the rate of convergence of Euler approximation of (1.1) is O(δ12∧(2H−1)) in probability in the space of Wα,∞0([0,T]) (see Remark 7). Meanwhile, we also get that, for any fixed ε>0, there exist a positive constant Cε and a subset Ωε of Ω with P(Ωε)>1−ε such that
E[supt∈[0,T]‖Xt−Xδt‖2αIΩε]≤Cεδ1∧(4H−2) and E[supt∈[0,T]‖Xt−Xδt‖2αIΩ/Ωε]≤Cε1/2, |
where C is a general positive constant independent of δ and ε (see Corollary 3). Unsurprisingly, the rate of convergence appears to be equal to the worst of the rates for corresponding "pure" equations (see [18,20]). Our approach is different from [18,20] in the sense that we combine pathwise approach with Itô integration in order to handle both types of integrals by using the Garsia-Rademich-Rumsey inequality. The proof of our result combines the techniques of Malliavin calculus with classical fractional calculus. The main ideas are to estimate the pathwise Riemann-Stieltjes integral by a random constant with moments of any order (see (2.3) and (2.5) and to express it as the sum of a Skorohod integral plus a correction term which involves the trace of the Malliavin derivative (see (2.11) and (2.13).) One can read Remark 10 for details. To the best of our knowledge, up to now, there is no paper which investigates the rate of convergence of Euler approximation of (1.1). We here make a first attempt to research such problem.
The rest of this paper is organized as follows. Several important functional space and some elements of fractional calculus and Malliavin calculus on an interval are give in Section 2. Section 3 contains the results concerning the rate of convergence for Euler scheme associated to (1.1). We first give our assumptions and some priori estimates, and then prove the rate of convergence is O(δ12∧(2H−1)) in probability. In Section 4, we give a numerical example and in Section 5, we summarize the work done in this paper and look forward to the next stage of our work. Finally, in Section 6, we prove the bounded estimation (3.5) and recall a couple of technical results.
Let (Ω,F,{Ft}0≤t≤T,P) be a complete probability space equipped with a filtration {Ft}0≤t≤T satisfying standard assumptions, i.e., it is increasing and right-continuous while F0 contains all P-null sets. Denote by W={Wt:t∈[0,T]} Brownian motions and BH={BHt:t∈[0,T]} fractional Brownian motions (fBms) with Hurst parameter H∈(1/2,1). Both are defined on (Ω,F,{Ft}0≤t≤T,P), where Ft=σ{X0,Ws,BHs|s∈[0,t]}. As we know, they are mean zero centered Gaussian processes with covariance kernels R(s,t)=min{s,t} and RH(s,t)=12(s2H+t2H−|t−s|2H) for any s,t∈[0,T] respectively. fBm is different from Brownian motion, it is neither a semimartingale nor a Markov process. Moreover, it holds (E|Bt−Bs|2)1/2=|t−s|H,s,t∈[0,T], and almost all sample paths of BH are Hölder continuous of any order μ∈(0,H). Now, let us briefly recall the Malliavin calculus, fractional calculus and three important functional spaces.
1. The space of β-Hölder continuous functions: Cβ([0,T]).
Let β∈(0,1). For a function f:[0,T]→R,‖f‖0,T,β denotes the β-Hölder norm of f on [0,T], that is,
‖f‖0,T,β:=sup0≤s<t≤T|ft−fs|(t−s)β. |
If ‖f‖0,T,β<∞, then we say f∈Cβ([0,T]).
2. The functional space: Wα,1([0,T]). (see [11])
Consider the fixed interval [0,T] and α∈(1−H,1/2). We denote by Wα,1([0,T]) the space of measurable functions f on [0,T] such that
‖f‖α,0,T:=∫T0|fs|sαds+∫T0∫s0|fs−fr|(s−r)α+1drds<∞. |
It is useful for us to estimate the pathwise Riemann-Stieltjes integral with respect to fBm. (see (2.3) and (2.5)
3. The functional space: Wα,∞0([0,T]). (see [13])
Definition 2.1. Let α∈(0,1/2). For any measurable function f:[0,T]→R, define
‖f‖α,∞:=supt∈[0,T]‖ft‖α=supt∈[0,T]{|ft|+∫t0|ft−fs|(t−s)1+αds}. |
If ‖f‖α,∞<∞, then we say f∈Wα,∞0([0,T]).
Remark 2.1. Wα,∞0([0,T]) is called Besov space (see [11], [18]). Moreover, given any ε such that 0<ε<α, there exists the following inclusions (see [13]):
Cα+ε∞([0,T])⊂Wα,∞0([0,T])⊂Cα−ε∞([0,T]), |
where Cα∞([0,T]) denotes the space of sup and α-Hölder continuous functions f:[0,T]→R, equipped with the norm ‖f‖∞,0,T,α:=sup0≤t≤T|ft|+sup0≤s<t≤T|ft−fs||t−s|α. In particular, both the fractional Brownian motion BH with H>1/2, and the standard Brownian motion W, have their trajectories in Wα,∞0([0,T]).
Due to the fact that fBm is neither a semi-martingale nor a Markov process, Itô's stochastic calculus is not fit for it. In this subsection, we show the definitions of the generalized fractional integrals and derivative operators (see [21]) before introducing the pathwise Riemann-Stieltjes integral with respect to fBm.
For p≥1 and a,b∈R with a<b, we denote by Lp(a,b) the space of Lebesgue measurable functions f:[a,b]→R satisfying
‖f‖Lp(a,b):=(∫ba|fx|pdx)1/p<∞. |
If f∈L1(a,b) and 0<α<1, the left-sided and right-sided fractional Riemann-Liouville integrals with respect to f of order α are defined by, for almost all t∈(a,b),
Iαa+ft=1Γ(α)∫ta(t−s)α−1fsds |
and
Iαb−ft=eiπαΓ(α)∫bt(s−t)α−1fsds, |
where Γ(α)=∫∞0rα−1e−rdr is the Gamma function. Let Iαa+(Lp) (resp. I1−αb−(Lq)) be the image of Lp(a,b) by the operator Ia+ (resp. I1−αb−). Suppose that limε↓0f(a+ε) and limε↓0g(b−ε) exist, moreover,
fa+∈Iαa+(Lp(a,b)),gb−∈I1−αb−(Lq(a,b)), |
where p≥1,q≥1,1p+1q≤1,0<α<1,fa+(t)=ft−limε↓0f(a+ε) and gb−(t)=limε↓0g(b−ε)−gt, then the fractional (Weyl) derivatives are defined by, for almost all s∈(a,b),
(Dαa+fa+)(s)=1Γ(1−α)(fs−fa(s−a)α+α∫safs−fτ(s−τ)1+αdτ)1(a,b)(s) |
and
(D1−αb−gb−)(s)=exp(−iπα)Γ(α)(gb−gs(b−s)1−α+(1−α)∫bsgτ−gs(τ−s)2−αdτ)1(a,b)(s). |
Remark 2.2. From [13], we can see Iαa+(Dαa+fa+)=fa+ for any fa+∈Iαa+(Lp(a,b)) and I1−αb−(D1−αb−gb−)=gb− for any gb−∈I1−αb−(Lq(a,b)). Moreover, from [11], we also see Dαa+fa+∈Lp(a,b) and D1−αb−gb−∈Lq(a,b).
Now we can construct the pathwise Riemann-Stieltjes integral with respect to fBm. If f∈Cλ([a,b]) and g∈Cμ([a,b]) with λ+μ>1, then, from the classical paper [22] by Young, the Riemann-Stieltjes integral ∫bafdg exists. Furthermore, Zähle [23] provided an explicit expression for it in terms of fractional derivatives as follows.
Proposition 2.1. Suppose that f∈Cλ([a,b]) and g∈Cμ([a,b]) with λ+μ>1. Let λ>α and μ>1−α. Then the pathwise Riemann-Stieltjes integral ∫bafsdgs exists and it can be expressed as
∫bafsdgs=eiπα∫baDαa+fa+(s)D1−αb−gb−(s)ds. | (2.1) |
We know that the fBm BH is ν-Hölder continuous for ∀ν∈(0,H). Therefore, for f∈Cβ([a,b]) and 1−H<α<β<1, we can express the pathwise Riemann-Stieltjes integral with respect to fBm according to (2.1) as
∫bafsdBHs=eiπα∫baDαa+fa+(s)D1−αb−BHb−(s)ds, | (2.2) |
where BHb−(s)=BH(b)−BH(s).
Remark 2.3. (see [24]) For any α∈(1−H,1), it follows from [21] that D1−αb−BHb−(x)∈L∞(a,b). Therefore, for any f∈Iαa+(L1(a,b)), (2.2) still holds.
The stochastic integral (2.2) admits the following estimate (see [11]): for α∈(1−H,1/2) and t∈[0,T], there exists a random variable ψ(ω,α,t) with finite moments of any order such that, for any determination or random function f∈Wα,1([0,T]),
|∫t0fsdBHs|≤ψ(ω,α,t)(∫t0|fs|sαds+∫t0∫s0|fs−fr|(s−r)α+1drds)=:ψ(ω,α,t)‖f‖α,0,t. | (2.3) |
Moreover, from the classical Garsia-Rodemich-Rumsey inequality (see (1.2) of [25]), we can choose ψ(ω,α,t) as
ψ(ω,α,t)=sup0≤u<v≤t|(D1−αv−BHv−)(u)|≤Cα,H,θξ0,t(BH):=Cα,H,θ(∫t0∫t0|BHx−BHy|2/θ|x−y|2H/θdxdy)θ/2<∞a.s., | (2.4) |
where Cα,H,θ is a constant depending on the underlying arguments, θ<α+H−1. Without loss of generality we can assume that θ=(α+H−1)/2 (see, for example, [17]). It is easily obtained from (2.4) that ψ(ω,α,t) is continuous in t and ψ(ω,α,t)≤ψ(ω,α,T) for all ω,α and t∈[0,T].
Also we need the following inequality from Proposition 4.1 in [17]: for any α∈(1−H,1/2),0≤s≤t≤T and f∈Wα,1([0,T]), we have
|∫tsfsdBHs|≤ψ(ω,α,t)(∫ts|fr|(r−s)αds+∫ts∫rs|fr−fv|(r−v)α+1dvdr). | (2.5) |
Particularly, for any η∈(0,H), there exists some constant Cη depending on η such that
|BHt−BHs|≤Cηψ(ω,α,t)|t−s|H−η holds . | (2.6) |
Again applying the Garsia-Rademich-Rumsey inequality to Wt and ∫tsb(u,Xu)dWu, for any η∈(0,1/2), one can deduce
|Wt−Ws|≤ϕ(ω,η,t)|t−s|1/2−η | (2.7) |
and
|∫tsb(u,Xu)dWu|≤ϕb(ω,η,t)|t−s|1/2−η | (2.8) |
where
ϕ(ω,η,t)=Cη(∫t0∫t0|Wx−Wy|2/η|x−y|1/ηdxdy)η/2 |
and
ϕb(ω,η,t)=Kη(∫t0∫t0|∫yxb(u,Xu)dWu|2/η|x−y|1/ηdxdy)η/2 |
respectively, Cη,Kη are both constants depending on η.
Let ℑ be the set of step functions on [0,T] and consider the Hilbert space H defined as the closure of ℑ with respect to the scalar product ⟨1[0,t],1[0,s]⟩H=RH(s,t) for s,t∈[0,T]. The mapping 1[0,t]↦BHt can be extended to an isometry between H and its associated Gaussian space. This isometry will be denoted by ϕ↦BH(ϕ). Note that
⟨ϕ,ψ⟩H=∫T0(K∗Hϕ)(s)(K∗Hψ)(s)ds for ϕ,ψ∈ℑ, |
where
(K∗Hϕ)(s)=∫Tsϕ(u)∂KH∂u(u,s)du,s∈[0,T] |
and
KH(t,s)=cHs12−H∫ts(u−s)H−32uH−12du1[0,t)(s), |
here cH=(H(2H−1)B(2−2H,H−12))12 and B denotes the Beta function. Moreover, we have L1/H([0,T])⊂H and in particular
⟨φ,ψ⟩H=H(2H−1)∫T0∫T0φ(r)ψ(u)|r−u|2H−2drdu |
for φ,ψ∈L1/H(0,T)
For n≥1, let F=f(Bt1,⋯,Btn) be smooth and cylindrical random variables with ti∈[0,T] for i=1,⋯,n and f being bounded and smooth. Then, the derivative operator D in the Sobolev space D1,2 is defined by
DsF=n∑i=1∂F∂xi(Bt1,⋯,Btn)1[0,ti](s),s∈[0,T]. | (2.9) |
In particular DsBHt=1[0,t](s). As usual, D1,2 is the closure of the set of smooth random variables with respect to the norm‖F‖21,2=E|F|2+E‖D.F‖2H
If F1,F2∈D1,2 such that F1 and ‖DF1‖H are bounded, then F1F2∈D1,2 and
D(F1F2)=F2DF1+F1DF2. |
Moreover, recall also the following chain rule: For F∈D1,2 and g∈C1(R) with bounded derivative we have g(F)∈D1,2 and
Dg(F)=g′(F)DF. | (2.10) |
The divergence operator, or Skorohod integral operator δ, is the adjoint of the derivative operator and we have the duality relationship E[Fδ(u)]=E⟨DF,u⟩H for every F∈D1,2 and u∈Dom(δ). We should also note that, for 1/2<H<1,
D1,2⊆L2(Ω,H)⊆L1H(Ω,H)⊆Dom(δ), |
Here, Lp(Ω,H) denotes the space of stochastic functions with finite p−order moment.
If (ut)t∈[0,T] is a stochastic process with Hölder continuous sample paths of order β>1−H, then the Riemann-Stieltjes integral with respect to BH is well defined. If u moreover satisfies ut∈D1,2 for all t∈[0,T] and
sups∈[0,T]E|us|2+supr,s∈[0,T]E|Drus|2<∞, |
then the relation
∫T0utdBHt=∫T0utδBHt+αH∫T0∫T0Dsut|s−t|2H−2dsdt | (2.11) |
holds, here αH=H(2H−1). Set p>1H, from Remark 5 of [26], for the Skorohod integral of the process {ut:t∈[0,T]}, we have the inequality
E[supz∈[0,t]|∫z0usδBHs|p]≤C[∫t0|Eus|pds+E∫t0(∫t0|Dsur|1Hds)pHdr]. | (2.12) |
Remark 2.4. Due to Hölder inequality, it can be obtained from (2.12) that
E[supz∈[0,t]|∫z0usδBHs|p]≤C[∫t0E|us|pds+E∫t0∫t0|Dsur|pdsdr]. | (2.13) |
For any n∈N, consider the isometric partition of [0,T]: {0=t0<t1<⋯<tn=T,δ=Tn},tk=kδ,k=0,1,…,n. Define τt:=max{tk:tk<t} and nt:=max{k:tk<t}. The Euler approximation of (1.1) is expressed as
Xδt=Xtk+a(tk,Xδtk)(t−tk)+b(tk,Xδtk)(Wt−Wtk)+c(tk,Xδtk)(BHt−BHtk),t∈(tk,tk+1], |
or, in the integral form,
Xδt=X0+∫t0a(τs,Xδτs)ds+∫t0b(τs,Xδτs)dWs+∫t0c(τs,Xδτs)dBHs,t∈[0,T]. | (3.1) |
Throughout this paper, we denote by C the generic positive constants independent of δ and ω. Their values are not important for us and maybe different from line to line. The mixture of Itô integral and pathwise Riemann-Stieltjes integral makes things a lot harder, forcing us to consider very smooth coefficients. Specifically, in this paper, besides the independence of W and BH, we suppose the coefficients of (1.1) satisfy the following hypotheses almost surely.
(Hab): The coefficients a(t,x),b(t,x) together with their partial derivatives in x are bounded. Moreover, a(t,x) and b(t,x) are β-Hölder continuous in t. That is, there exist two constants C>0 and β∈(12∨(2H−1),1] such that
(1). |a(t,x)|+|b(t,x)|≤C;
(2). |ax(t,x)|+|bx(t,x)|≤C;
(3). |a(t,x)−a(s,x)|+|b(t,x)−b(s,x)|≤C|t−s|β.
(Hc): The coefficient c(t,x) is continuously differentiable in x. Moreover, there exist two constants C>0 and β∈(12∨(2H−1),1] such that
(1). |c(t,x)|+|cx(t,x)|≤C;
(2). |c(t,x)−c(s,x)|+|cx(t,x)−cx(s,x)|≤C|t−s|β;
(3). |cx(t,x)−cx(t,y)|≤C|x−y|.
Note that the above assumptions (2) of (Hab) and (1) of (Hc) imply the Lipschitz continuity, that is, there exists some constant C>0 such that, for any t∈[0,T] and x,y∈R,
|a(t,x)−a(t,y)|+|b(t,x)−b(t,y)|+|c(t,x)−c(t,y)|≤C|x−y|. |
Remark 3.1.
1. As was stated in [13], under the assumptions (Hab) and (Hc), the main SDEs (1) has a unique solution {X(t),t∈[0,T]} in the space of Wα,∞0([0,T]) with α∈(1−H,1/2).
2. Even if, instead of the boundedness of a(t,x), we assume that the coefficient a(t,x) is linear growth, all results in this paper are still true.
Now, we are going to formulate some useful properties of the Euler approximation {Xδt,t∈[0,T]}. For this, we need some additional notations. Denote ψt:=ψ(ω,α,t)∨1,ϕt:=ϕ(ω,η,t)∨ϕb(ω,η,t)∨1 and ξt:=ψt∨ϕt. Obviously, ξt is non-deceasing in t, that is, for any t∈[0,T], ξt≤ξT=ψT∨ϕT holds almost surely. Moreover, ξt has finite moments of any order. For any R>1, define a stopping time πR:=inf{t:ξt≥R}∧T. Let ΩR={ω:πR=T}. By Lemma 4.4 in [11], P(πR<T) tends to 0 as R→∞ if assumptions (Hab) and (Hc) hold.
Our first result is Hölder continuity for the processes {Xt}t∈[0,T] and {Xδt}t∈[0,T] defined in (1.1) and (3.1) respectively.
Lemma 3.1. If the coefficients of (1.1) satisfy the conditions (Hab)(1)(2) and (Hc), then, for any 1−H<α<12, it has a unique solution X such that {Xt}t∈[0,T]∈Wα,∞0([0,T],R) almost surely. Moreover, for any 0<η<12 and 0≤s≤t≤T, there exists some constant C>0 such that
|Xt−Xs|≤Cξ2TeCψ11−αT⋅|t−s|12−η. | (3.2) |
Proof. It follows from [13] that, for any 1−H<α<12, there exists a unique solution {Xt}t∈[0,T] of (1.1) belonging to Wα,∞0([0,T]) almost surely. Now, we estimate (3.2). From (2.8) and the condition (1) of (Hab) we immediately get
|Xt−Xs|≤|∫tsa(u,Xu)du|+|∫tsb(u,Xu)dWu|+|∫tsc(u,Xu)dBHu|≤C(|t−s|+ϕt|t−s|12−η+Q(s,t)). |
Using the estimation (2.5) and for s,t∈[0,T], we have
Q(s,t)≤ψt(∫tsc(u,Xu)(u−s)αdu+∫ts∫us|c(u,Xu)−c(r,Xr)|(u−r)α+1drdu)≤Cψt(|t−s|1−α+∫ts∫us(u−r)β+|Xu−Xr|(u−r)α+1drdu)≤Cψt(|t−s|1−α+∫ts∫us|Xu−Xr|(u−r)α+1drdu). |
By exchanging the order of integration and choosing 0<η<12−α, for s,u∈[0,T], we have
ζ(s,u):=∫us|Xu−Xr|(u−r)α+1dr≤C∫us1(u−r)α+1[|u−r|+ϕu|u−r|12−η+ψu(|u−r|1−α+∫ur∫qr|Xq−Xz|(q−z)α+1dzdq)]dr≤Cξu|u−s|12−η−α+Cψu∫usζ(s,q)(u−q)αdq. |
Consequently, Lemma 6.3 implies that
ζ(s,u)≤CξueCψ11−αu⋅|u−s|12−η−α. |
Therefore,
|Xt−Xs|≤Cξ2teCψ11−αt⋅|t−s|12−η |
which completes the proof.
Remark 3.2. A similar proof to Lemma 3.1, we also have
|Xδt−Xδs|≤Cξ2TeCψ11−αT⋅|t−s|12−η. |
Then, we prove the boundedness of the processes {Xt}t∈[0,T] and {Xδt}t∈[0,T].
Lemma 3.2. Let E|X0|p<∞ for p≥1. If the assumptions (Hab) and (Hc) hold, then, for any 1−H<α<12, there exists a constant C>0 such that E[supt∈[0,T]‖Xt‖pα]≤C and E[supt∈[0,T]‖Xδt‖pα]≤C, where {Xt}t∈[0,T] and {Xδt}t∈[0,T] are the solutions of (1.1) and Euler (3.1) respectively.
Proof. It can be directly derived from Theorem 4.2 of [11] that E[supt∈[0,T]‖Xt‖pα]≤C. Hence we just need to show E[supt∈[0,T]‖Xδt‖pα]≤C. Write
‖Xδt‖α=|Xδt|+∫t0|Xδt−Xδs|(t−s)α+1ds≤|X0|+3∑k=1Jk(t), |
where
J1(t)=|∫t0a(τs,Xδτs)ds|+∫t0|∫tsa(τr,Xδτr)dr|(t−s)α+1ds,J2(t)=|∫t0b(τs,Xδτs)dWs|+∫t0|∫tsb(τr,Xδτr)dWr|(t−s)α+1ds≤supt∈[0,T]|∫t0b(τs,Xδτs)dWs|+supt∈[0,T]∫t0|∫tsb(τr,Xδτr)dWr|(t−s)α+1ds=:J21+J22,J3(t)=|∫t0c(τs,Xδτs)dBHs|+∫t0|∫tsc(τr,Xδτr)dBHr|(t−s)α+1ds=:J31(t)+J32(t). |
It follows easily from (1) of (Hab) and 1−H<α<12 that J1(t)≤C.
We estimate J31(t) by using (2.3). From the definition of τt and Euler equation (3.1), we have
J31(t)≤ψt(∫t0|c(τs,Xδτs)|sαds+∫t0∫s0|c(τs,Xδτs)−c(τr,Xδτr)|(s−r)α+1drds)≤Cψt[1+∫t0∫τs0(|Xδs−Xδτs|(s−r)α+1+|Xδr−Xδτr|(s−r)α+1+|Xδs−Xδr|(s−r)α+1)drds]=:Cψt(1+J311(t)+J312(t)+J313(t)). |
In order to estimate J311(t) and J312(t), we need the following difference
|Xδt−Xδτt|≤C[(t−τt)+ϕt(t−τt)12−η+ψt(t−τt)H−η]≤Cξt(t−τt)12−η. | (3.3) |
In fact, (3.3) can be easily derived from (2.6) and (2.7).
If one takes η∈(0,12−α), it is obviously that J311(t)≤Cξt.
Note that the area of {0≤s≤t;0≤r≤τs} is equivalent to the area of {0≤r≤τt;τr+δ≤s≤t}. So, by exchanging the order of integration of J312, we have
J312(t)=∫τt0∫tτr+δ1(s−r)α+1ds|Xδr−Xδτr|dr≤1αnt−1∑k=0∫tk+1tk1(tk+1−r)α|Xδr−Xδτr|dr≤Cξt, |
where we use the difference (3.3) and the following inequality (see (4.15) of [17])
∫tk+1tk(r−tk)β(tk+1−r)−αdr≤δβ−α+1B(1−α,1+β). | (3.4) |
Evidently, J313(t)≤∫t0‖Xδs‖αds.
Now, we estimate J32(t) by using (2.5) and the following estimation
∫t0∫τr0(τr−τq)β(t−q)α(r−q)α+1dqdr≤C (see Lemma 6.1). | (3.5) |
From the definition of τt and Euler equation (3.1), we have
J32(t)≤ψt∫t0∫ts|c(τr,Xδτr)|(r−s)αdr+∫ts∫rs|c(τr,Xδτr)−c(τq,Xδτq)|(r−q)α+1dqdr(t−s)α+1ds≤Cψt(1+∫t0∫r0∫q01(t−s)α+1ds|c(τr,Xδτr)−c(τq,Xδτq)|(r−q)α+1dqdr)≤Cψt(1+∫t0∫τr0|Xδr−Xδτr|+|Xδq−Xδτq|+|Xδr−Xδq|(r−q)α+1(t−q)αdqdr)=:Cψt(1+J321(t)+J322(t)+J323(t)). |
It is obvious that
J323(t)≤∫t0‖Xδs‖α(t−s)−αds. |
From (3.3) and 0<η<12−α, we estimate J321(t) and J322(t).
J321(t)≤C∫t0∫τr0(r−q)−α−1dq|Xδr−Xδτr|(t−r)αdr≤Cξt. |
Exchanging the order of integration of J322(t), we have
J322(t)=∫τt0∫tτq+δ(r−q)−α−1dr|Xδq−Xδτq|(t−q)αdq≤∫τt0|Xδq−Xδτq|(τq+δ−q)α(t−q)αdq. |
Further, noting that 0<α<12 and then using (3.4), we have
J322(t)≤Cξt∫τt0(τq+δ−q)−α(t−q)−α(q−τq)12−ηdq≤Cξt{nt−2∑k=01(t−tk+1)α∫tk+1tk(q−tk)12−η(tk+1−q)αdq+δ12−η∫tnttnt−11(tnt−q)α(t−q)αdq}≤Cξt{nt−2∑k=0(t−tk+1)−αδ32−η−α+δ12−η(tnt−tnt−1)1−2α}≤Cξt{nt−2∑k=0(t−tk+1)−αδ32−η−α+δ32−η−2α}≤Cξt(∫t0(t−s)−αds+1)≤Cξt. |
Summing up all the above estimations, we obtain
‖Xδt‖α≤|X0|+J21+J22+Cξ2t+Cψt∫t0‖Xδs‖α(t−s)−αds. |
So, it follows from Lemma 6.3 that
‖Xδt‖α≤C(|X0|+J21+J22+ξ2t)eCψ11−αt≤C(|X0|+J21+J22+ξ2T)eCψ11−αT. |
For each p≥1,ω∈Ω, and α∈(1−H,12), we take into account that the right-hand side of the above inequality does not depend on t and arrive at
supt∈[0,T]‖Xδt‖pα≤C(|X0|p+Jp21+Jp22+ξ2pT)eCψ11−αT. |
Therefore,
E[supt∈[0,T]‖Xδt‖pα]≤C(E|X0|2p+EJ2p21+EJ2p22+Eξ4pT)12{E[eCψ11−αT]}12. |
Taking into account that 1<11−α<2, we apply Fernique theorem (see (24) in [11]) and obtain that
E[eCψ11−αT]≤C. |
Our next step is to estimate E[J2p21] with the help of the Doob martingale inequality.
E[J2p21]≤E[supt∈[0,T]|∫t0b(τs,Xδτs)dWs|2p]≤E[supt∈[0,T]∫t0|b(τs,Xδτs)|2pds]≤C. |
Applying (2.8) to J2p22 and noting that 0<η<12−α, one can easily obtain
E[J2p22]≤CE[|ϕt|2p]. |
Finally, because ϕt and ψt have bounded moments of p-order, we get
E[supt∈[0,T]‖Xδt‖pα]≤C, |
where the constant C depends on α and p, but is independent of δ and ω.
As a result of Lemma 3.2, one can easily get the following corollary.
Corollary 3.1. Let E|X0|p<∞ for p≥1. If assumptions (Hab) and (Hc) hold, then there exists a constant C>0 such that
E[supt∈[0,T]|Xδt−Xδτt|p]≤Cδp/2 |
where {Xt}t∈[0,T] and {Xδt}t∈[0,T] are the solutions of (1.1) and Euler (3.1) respectively.
Thirdly, we are ready to prove that the moments of the Malliavin derivative of Euler approximation (3.1) is bounded. We refer the reader to Nualart and Saussereau [27] for results on Malliavin regularity of the solutions of stochastic differential equations.
Lemma 3.3. Let t∈(tk,tk+1], i.e., τt=tk, Xδτt be the solution of Euler (3.1) at the point tk,k=0,1,⋯,n. If the assumptions (Hab) and (Hc) hold and X0 is independent of BH with E[|X0|p]<∞ for p≥1, then, there exist some positive constant C such that E[supt∈[0,T]|DsXδt|p]≤C for any s>τt and some constant Mp,R dependent on p and R such that E[supt∈[0,T]|DsXδτt∧πR|p]≤Mp,R for any s∈[0,τt], here, for any fixed R>1, the stopping time πR=inf{t:ξt≥R}∧T.
Proof. If s>t then it is clear that DsXδt=0.
Thanks to (2.9), (2.10) and the independence of W and BH, if τt<s≤t then we have
DsXδt=Ds[Xδτt+a(τt,Xδτt)(t−τt)+b(τt,Xδτt)(Wt−Wτt)+c(τt,Xδτt)(BHt−BHτt)]=c(τt,Xδτt). |
By the boundedness of c(t,x), we have E[supt∈[0,T]|DsXδt|p]≤C for any s>τt.
If s∈[0,τt] then we have
DsXδτt∧πR=Ds(X0+∫τt∧πR0a(τu,Xδτu)du+∫τt∧πR0b(τu,Xδτu)dWu+∫τt∧πR0c(τu,Xδτu)dBHu)=c(τs,Xδτs)+∫τt∧πRτs+δax(τu,Xδτu)DsXδτudu+∫τt∧πRτs+δbx(τu,Xδτu)DsXδτudWu+∫τt∧πRτs+δcx(τu,Xδτu)DsXδτudBHu. |
From Lemma 2.3 of [14] and the boundedness of ax(t,x), bx(t,x) and cx(t,x), for any s belonging to the interval [0,T], we obtain
E[supt∈[0,T]|DsXδτt∧πR|p]≤Mp,R for any s≤τt. |
For any s∈[0,τt], note that the equation
Ds(Xδt−Xδτt)=DsXδτt(ax(τt,Xδτt)(t−τt)+bx(τt,Xδτt)(Wt−Wτt),+cx(τt,Xδτt)(BHt−BHτt)) |
is true. Therefore, we have the following corollary by Lemma 3.3 and Cauchy-Schwarz type inequality.
Corollary 3.2. If the conditions of Lemma 3.3 are satisfied, then, for any s∈[0,τt], we have
E[supt∈[0,T]|Ds(Xδt∧πR−Xδτt∧πR)|p]≤Cδp/2. |
and
E[supt∈[0,T]|DsXδt∧πR|p]≤Mp,Rfor anys≤t. |
The aim of this subsection is to estimate the rate of convergence of Euler (3.1) to the solution of (1.1).
Theorem 3.1. Suppose that X0 is a random variable independent of W and BH with E|X0|4<∞. If assumptions (Hab) and (Hc) are satisfied then we have
limn→∞P(nγsupt∈[0,T]‖Xt−Xδt‖α>ε)=0inWα,∞0([0,T]) |
for any γ<min{12,2H−1}, ε>0 and α∈(1−H,12). Here, {Xt}t∈[0,T] and {Xδt}t∈[0,T] are the solutions of (1.1) and Euler (3.1) respectively and n=Tδ.
Remark 3.3. Theorem 3.1 shows that the rate of convergence for Euler (3.1) is equal to γ (γ<min{12,2H−1}) in probability with the norm sup‖⋅‖α, i.e. in the sense of probability, we can establish an estimate for the error of |Xt−Xδt| with the norm sup‖⋅‖α in certain Besov space Wα,∞0([0,T]) (see Definition 2.1), exactly,
for any ε>0 and any sufficiently small ρ>0, there exists δ0>0 and Ωε,δ0,ρ such that P(Ωε,δ0,ρ)>1−ε and for any ω∈Ωε,δ0,ρ,δ<δ0,
supt∈[0,T]‖Xt−Xδt‖α=supt∈[0,T]{|Xt−Xδt|+∫t0|Xt−Xδt−(Xs−Xδs)|(t−s)1+αds}<C(ω)δmin{12,2H−1}−ρ, |
here α∈(1−H,12), C(ω) does not depend on δ and ε (but depends on ρ).
Proof. of Theroem 3.1. Fix an arbitrary ε>0 and any R>1. As mentioned previously, πR=inf{t:ξt≥R}∧T. Consider
P(nγsupt∈[0,T]‖Xt−Xδt‖α>ε)≤P(πR<T)+P(nγsupt∈[0,T]‖Xt−Xδt‖α>ε,πR=T). | (3.6) |
For the second term on the right hand side of (3.6), applying Chebyshev's inequality, we have
P(nγsupt∈[0,T]‖Xt−Xδt‖α>ε,πR=T)≤P(nγsupt∈[0,T]‖Xt∧πR−Xδt∧πR‖α>ε)≤n2γE[(supt∈[0,T]‖Xt∧πR−Xδt∧πR‖α)2]ε2. | (3.7) |
Now, we estimate (3.7). According to Definition 2.1, for any t∈[0,T], we have
E[(supz∈[0,t]‖Xz∧πR−Xδz∧πR‖α)2]≤2E[supz∈[0,t]|Xz∧πR−Xδz∧πR|2]+2E[(∫t∧πR0|Xt∧πR−Xδt∧πR−Xs+Xδs|(t−s)α+1ds)2]=:2(I1(t)+I2(t)). | (3.8) |
In turn, I1(t) can be estimated as
I1(t)≤3Esupz∈[0,t]|∫z∧πR0(a(s,Xs)−a(τs,Xδτs))ds|2+3Esupz∈[0,t]|∫z∧πR0(b(s,Xs)−b(τs,Xδτs))dWs|2+3Esupz∈[0,t]|∫z∧πR0(c(s,Xs)−c(τs,Xδτs))dBHs|2=:3(I11(t)+I12(t)+I13(t)). | (3.9) |
From Corollary 3.1, we have
I11(t)≤3Esupz∈[0,t]|∫z∧πR0(a(s,Xs)−a(s,Xδs))ds|2+3Esupz∈[0,t]|∫z∧πR0(a(s,Xδs)−a(s,Xδτs))ds|2+3Esupz∈[0,t]|∫z∧πR0(a(s,Xδτs)−a(τs,Xδτs))ds|2≤C∫t0Esupz∈[0,s]|Xz∧πR−Xδz∧πR|2ds+C∫t0E|Xδs−Xδτs|2ds+Cδ2β≤C∫t0I1(s)ds+Cδ+Cδ2β. | (3.10) |
From Doob martingale inequality and Corollary 3.1, we have
I12(t)≤3Esupz∈[0,t]|∫z∧πR0(b(s,Xs)−b(s,Xδs))dWs|2+3Esupz∈[0,t]|∫z∧πR0(b(s,Xδs)−b(s,Xδτs))dWs|2+3Esupz∈[0,t]|∫z∧πR0(b(s,Xδτs)−b(τs,Xδτs))dWs|2≤C∫t0Esupz∈[0,s]|Xz∧πR−Xδz∧πR|2ds+C∫t0E|Xδs−Xδτs|2ds+Cδ2β≤C∫t0I1(s)ds+Cδ. | (3.11) |
Estimate I13(t) by dividing it into three parts.
I13(t)≤3Esupz∈[0,t]|∫z∧πR0(c(s,Xs)−c(s,Xδs))dBHs|2+3Esupz∈[0,t]|∫z∧πR0(c(s,Xδs)−c(s,Xδτs))dBHs|2+3Esupz∈[0,t]|∫z∧πR0(c(s,Xδτs)−c(τs,Xδτs))dBHs|2=:3(I131(t)+I132(t)+I133(t)). | (3.12) |
Taking into account the estimation (2.3) and the definition of stopping time πR, we have
I131(t)≤2R2Esupz∈[0,t]|∫z∧πR0|c(s,Xs)−c(s,Xδs)|sαds|2+2R2Esupz∈[0,t]|∫z∧πR0∫s0|c(s,Xs)−c(s,Xδs)−c(q,Xq)+c(q,Xδq)|(s−q)α+1dqds|2=:2R2(I1311(t)+I1312(t)). |
Noting that 1−H<α<12 and c(t,x) is Lipschitz continuous, we have
I1311(t)≤E∫t∧πR0|c(s,Xs)−c(s,Xδs)|2ds⋅∫t0s−2αds≤∫t0Esupz∈[0,s]|Xz∧πR−Xδz∧πR|2ds⋅∫t0s−2αds≤C∫t0I1(s)ds. | (3.13) |
Using Lemma 6.2, Lemma 3.1 and 1−H<α<12, we have
I1312(t)≤CEsupz∈[0,t]|∫z∧πR0∫s0|Xs−Xδs−Xq+Xδq|(s−q)α+1dqds|2+CEsupz∈[0,t]|∫z∧πR0∫s0|Xs−Xδs|(s−q)β(s−q)α+1dqds|2+CEsupz∈[0,t]|∫z∧πR0∫s0|Xs−Xδs|(|Xs−Xq|+|Xδs−Xδq|)(s−q)α+1dqds|2≤C∫t0E|∫s∧πR0|Xs∧πR−Xδs∧πR−Xq+Xδq|(s−q)α+1dq|2ds+C∫t0Esupz∈[0,s]|Xz∧πR−Xδz∧πR|2ds⋅∫t0s2(β−α)ds+C2R∫t0Esupz∈[0,s]|Xz∧πR−Xδz∧πR|2ds⋅∫t0s2γds≤C∫t0I2(s)ds+C2R∫t0I1(s)ds, | (3.14) |
where CR=CReCR11−α, γ=12−η−α>0 only if we take η∈(0,12−α).
Estimate I132(t) and I133(t) by the relation between the pathwise Riemann-Stieltjes integral and the Skorohod integral with respect to fBm. Firstly,
I132(t)≤Esupz∈[0,t]|∫z∧πR0(c(s,Xδs)−c(s,Xδτs))dBHs|2=2Esupz∈[0,t]|∫z∧πR0(c(s,Xδs)−c(s,Xδτs))δBHs|2+2α2HEsupz∈[0,t]|∫z∧πR0∫z∧πR0Du(c(s,Xδs)−c(s,Xδτs))|s−u|2H−21[0,s](u)duds|2=:2I1321(t)+2α2HI1322(t). | (3.15) |
From the estimation (2.13), Lemma 3.2 and 3.3, Corollary 3.1 and 3.2, we have
I1321(t)≤C(∫t0E|c(s,Xδs)−c(s,Xδτs)|2ds+E∫t∧πR0∫s0|Du(c(s,Xδs)−c(s,Xδτs))|2duds)≤C∫t0E|Xδs−Xδτs|2ds+C∫t∧πR0∫τs0E[|cx(s,Xδs)|2⋅|Du(Xδs−Xδτs)|2]duds+C∫t∧πR0∫sτsE[|cx(s,Xδs)|2⋅|Du(Xδs−Xδτs)|2]duds+C∫t∧πR0∫s0E[|cx(s,Xδs)−cx(s,Xδτs)|2⋅|DuXδτs|2]duds≤Cδ+C∫t0∫τs0[E|cx(s,Xδs)|4]12⋅[E|Du(Xδs∧πR−Xδτs∧πR)|4]12duds+C∫t0∫sτs[E|cx(s,Xδs)|4]12⋅[E|DuXδs∧πR|4]12duds+C∫t0∫s0[E|cx(s,Xδs)−cx(s,Xδτs)|4]1/2⋅[E|DuXδτs∧πR|4]1/2duds≤Cδ. | (3.16) |
Similarly to the estimation of I1321(t),
I1322(t)≤4E|∫t0∫τs0cx(s,Xδs)Du(Xδs∧πR−Xδτs∧πR)|s−u|2H−2duds|2+4E|∫t0∫sτscx(s,Xδs)Du(Xδs∧πR−Xδτs∧πR)|s−u|2H−2duds|2+2E|∫t0∫s0|DuXδτs∧πR|(cx(s,Xδs)−cx(s,Xδτs))|s−u|2H−2duds|2≤C∫t0∫τs0E|cx(s,Xδs)Du(Xδs∧πR−Xδτs∧πR)|2⋅|s−u|2H−2duds+Cδ2H−1∫t0∫sτsE|cx(s,Xδs)Du(Xδs∧πR−Xδτs∧πR)|2⋅|s−u|2H−2duds+C∫t0∫s0E[|Xδs−Xδτs|2⋅|DuXδτs∧πR|2]⋅|s−u|2H−2duds≤C∫t0∫τs0[E|Du(Xδs∧πR−Xδτs∧πR)|4]1/2⋅|s−u|2H−2duds+Cδ2H−1∫t0∫sτs[E|DuXδs∧πR|4]1/2⋅|s−u|2H−2duds+C∫t0∫s0[E|Xδs−Xδτs|4]1/2⋅[E|DuXδτs∧πR|4]1/2⋅|s−u|2H−2duds≤C(δ+δ4H−2). | (3.17) |
Secondly,
I133(t)≤2Esupz∈[0,t]|∫z∧πR0(c(s,Xδτs)−c(τs,Xδτs))δBHs|2+2α2HEsupz∈[0,t]|∫z∧πR0∫s∧πR0Du(c(s,Xδτs)−c(τs,Xδτs))|s−u|2H−2duds|2=:2I1331(t)+2α2HI1332(t). | (3.18) |
Further we apply (2.13) and Lemma 3.3 to I1331(t),
I1331(t)≤C∫t0E|c(s,Xδτs)−c(τs,Xδτs)|2ds+CE∫t∧πR0∫s∧πR0|Du(c(s,Xδτs)−c(τs,Xδτs))|2duds≤Cδ2β+CE∫t0∫s0|cx(s,Xδτs)−cx(τs,Xδτs)|2⋅|DuXδτs∧πR|2duds≤Cδ2β+Cδ2β∫t0∫s0E|DuXδτs∧πR|2duds≤Cδ2β. | (3.19) |
Similar to the estimation of I1331(t), we have
I1332(t)≤CE∫t∧πR0(∫s∧πR0DuXδτs(cx(s,Xδτs)−cx(τs,Xδτs))|s−u|2H−2du)2ds≤Cδ2β∫t0∫s0E|DuXδτs∧πR|2⋅|s−u|2H−2du∫s0|s−u|2H−2duds≤Cδ2β. | (3.20) |
Summing up all the estimations (3.9)-(3.20), we obtain
I1(t)≤C(δ+δ4H−2)+R2⋅C2R∫t0I1(s)ds+CR2∫t0I2(s)ds. | (3.21) |
Next step, we consider I2(t).
I2(t)≤3E(∫t∧πR0|∫t∧πRs(a(q,Xq)−a(τq,Xδτq))dq|(t−s)α+1ds)2+3E(∫t∧πR0|∫t∧πRs(b(q,Xq)−b(τq,Xδτq))dWq|(t−s)α+1ds)2+3E(∫t∧πR0|∫t∧πRs(c(q,Xq)−c(τq,Xδτq))dBHq|(t−s)α+1ds)2=:3(I21(t)+I22(t)+I23(t)). | (3.22) |
As regards I21(t), we estimate it in the following way:
I21(t)≤3E(∫t∧πR0|∫t∧πRs(a(q,Xq)−a(q,Xδq))dq|(t−s)α+1ds)2+3E(∫t∧πR0|∫t∧πRs(a(q,Xδq)−a(q,Xδτq))dq|(t−s)α+1ds)2+3E(∫t∧πR0|∫t∧πRs(a(q,Xδτq)−a(τq,Xδτq))dq|(t−s)α+1ds)2=:3(I211(t)+I212(t)+I213(t)). | (3.23) |
Choosing ϱ such that α<ϱ<12, by Hölder inequality, Lipschitz continuity of a(t,x) and exchanging the order of integration, we have
I211(t)≤CE∫t∧πR0|∫t∧πRs(a(q,Xq)−a(q,Xδq))dq|2(t−s)2α+2−2ϱds≤CE∫t∧πR0∫t∧πRs|a(q,Xq)−a(q,Xδq)|2dq(t−s)2α+1−2ϱds≤C∫t0∫tsE|Xq∧πR−Xδq∧πR|2dq(t−s)2α+1−2ϱds≤C∫t0I1(s)ds. | (3.24) |
By a similar discussion to I211(t), one can easily get
I212(t)≤Cδ, | (3.25) |
and
I213(t)≤Cδ2β. | (3.26) |
Similar to the estimation of I21(t), we estimate I22(t) by dividing it into three parts as well,
I22(t)≤3E(∫t∧πR0|∫t∧πRs(a(q,Xq)−a(q,Xδq))dWq|(t−s)α+1ds)2+3E(∫t∧πR0|∫t∧πRs(a(q,Xδq)−a(q,Xδτq))dWq|(t−s)α+1ds)2+3E(∫t∧πR0|∫t∧πRs(a(q,Xδτq)−a(τq,Xδτq))dWq|(t−s)α+1ds)2=:3(I221(t)+I222(t)+I223(t)). | (3.27) |
From Hölder inequality, Burkhölder-Davis-Gundy inequality and then exchanging the order of integration, we get
I221(t)≤CE∫t∧πR0|∫t∧πRs(a(q,Xq)−a(q,Xδq))dWq|2(t−s)α+32ds∫t0(t−s)−α−12ds≤C∫t0∫tsE|a(q∧πR,Xq∧πR)−a(q∧πR,Xδq∧πR)|2(t−s)−α−32dqds≤C∫t0E|Xq∧πR−Xδq∧πR|2∫q0(t−s)−α−32dsdq≤C∫t0I1(s)(t−s)−α−12ds. | (3.28) |
A similar discussion to I221(t), one can easily get
I222(t)≤Cδ, | (3.29) |
and
I223(t)≤Cδ2β. | (3.30) |
Now we go on with the term I23(t) including fBm.
I23(t)≤3E(∫t∧πR0|∫t∧πRs(c(q,Xq)−c(q,Xδq))dBHq|(t−s)α+1ds)2+3E(∫t∧πR0|∫t∧πRs(c(q,Xδq)−c(q,Xδτq))dBHq|(t−s)α+1ds)2+3E(∫t∧πR0|∫t∧πRs(c(q,Xδτq)−c(τq,Xδτq))dBHq|(t−s)α+1ds)2=:3(I231(t)+I232(t)+I233(t)). | (3.31) |
With the help of the estimation (2.5) and the definition of stopping time πR, we have
I231(t)≤2R2E(∫t∧πR0|∫t∧πRs(c(q,Xq)−c(q,Xδq))(q−s)−αdq|(t−s)α+1ds)2+2R2E(∫t∧πR0∫t∧πRs∫qs|c(q,Xq)−c(q,Xδq)−c(r,Xr)+c(r,Xδr)|(q−r)α+1drdq(t−s)α+1ds)2=:2R2(I2311(t)+I2312(t)). | (3.32) |
By exchanging the order of the integration we have
I2311(t)≤CE(∫t∧πR0|c(q,Xq)−c(q,Xδq)|∫q0(q−s)−α(t−s)−α−1dsdq)2≤C∫t0E|Xq∧πR−Xδq∧πR|2⋅(t−q)−2αdq≤C∫t0I1(s)(t−s)−2αds, | (3.33) |
Here we use the following estimation (see (4.15) of [17]):
∫q0(q−s)−α(t−s)−α−1dsdq≤(t−q)−2α∫qt−q0(1+s)−α−1s−αds≤B(2α,1−α)(t−q)−2α. |
According to Lemma 6.2, I2312(t) admits the following estimation:
I2312(t)≤CE∫t∧πR0|∫t∧πRs∫qs|c(q,Xq)−c(q,Xδq)−c(r,Xr)+c(r,Xδr)|(q−r)α+1drdq|2(t−s)2+2α−2ϱds≤CE∫t∧πR0|∫t∧πRs∫qs|Xq−Xδq−Xr+Xδr|(q−r)α+1drdq|2(t−s)2+2α−2ϱds+CE∫t∧πR0|∫t∧πRs∫qs|Xq−Xδq|(q−r)β−α−1drdq|2(t−s)2+2α−2ϱds+CE∫t∧πR0|∫t∧πRs∫qs|Xq−Xδq|(|Xq−Xr|+|Xδq−Xδr|)(q−r)α+1drdq|2(t−s)2+2α−2ϱds=:I23121(t)+I23122(t)+I23123(t), | (3.34) |
here α<ϱ<12.
By Hölder inequality and exchanging the order of the integration, we have
I23121(t)≤CE∫t0∫ts(∫q∧πRs|Xq∧πR−Xδq∧πR−Xr+Xδr|(q−r)α+1dr)2dq(t−s)1+2α−2ϱds≤C∫t0∫q0E(∫q∧πR0|Xq∧πR−Xδq∧πR−Xr+Xδr|(q−r)α+1dr)2(t−s)1+2α−2ϱdsdq≤C∫t0I2(s)ds | (3.35) |
and
I23122(t)=E∫t∧πR0|∫t∧πRs|Xq−Xδq|(q−s)β−αdq|2(t−s)2+2α−2ϱds≤C∫t0∫tsE|Xq∧πR−Xδq∧πR|2dq∫ts(q−s)2β−2αdq(t−s)2+2α−2ϱds≤C∫t0E|Xq∧πR−Xδq∧πR|2∫q0(t−s)2β−4α−1+2ϱdsdq≤C∫t0t2β−4α+2ϱI1(q)dq≤C∫t0I1(s)ds. | (3.36) |
According to Remark 3.2, similar to the above estimation, we have
I23123(t)≤C2RE∫t∧πR0|∫t∧πRs∫qs|Xq−Xδq|(q−r)γ−1drdq|2(t−s)2+2α−2ϱds≤C2RE∫t∧πR0|∫t∧πRs|Xq−Xδq|(q−s)γdq|2(t−s)2+2α−2ϱds≤C2RE∫t0∫ts|Xq∧πR−Xδq∧πR|2dq∫ts(q−s)2γdq(t−s)2+2α−2ϱds≤C2R∫t0∫tsE|Xq∧πR−Xδq∧πR|2(t−s)2γ−2α−1+2ϱdqds≤C2R∫t0E|Xq∧πR−Xδq∧πR|2∫q0(t−s)2γ−2α−1+2ϱdsdq≤C2R∫t0I1(s)ds, | (3.37) |
where CR=CReCR11−α (see (3.14)), α<ϱ<12, γ=12−η−α>0 only if we take η∈(0,12−α).
Then we estimate I232(t). (2.11) implies
I232(t)≤2E(∫t∧πR0|∫t∧πRs(c(q,Xδq)−c(q,Xδτq))δBHq|(t−s)α+1ds)2+2α2HE(∫t∧πR0|∫t∧πRs∫qsDr(c(q,Xδq)−c(q,Xδτq))|q−r|2H−2drdq|(t−s)α+1ds)2=:2I2321(t)+2α2HI2322(t). | (3.38) |
According to (2.13), Corollary 3.1 and 3.2 as well as Lemma 3.3, we have
I2321(t)≤C∫t∧πR0E|∫t∧πRs(c(q,Xδq)−c(q,Xδτq))δBHq|2(t−s)2+2α−2ϱds≤C∫t0∫tsE|c(q,Xδq)−c(q,Xδτq)|2(t−s)2+2α−2ϱdqds+C∫t∧πR0∫t∧πRs∫qsE|Dr(c(q,Xδq)−c(q,Xδτq))|2(t−s)2+2α−2ϱdrdqds≤C∫t0∫tsE|Xδq−Xδτq|2(t−s)2+2α−2ϱdqds+C∫t0∫ts∫qτqE|cx(q,Xδq)[Dr(Xδq∧πR−Xδτq∧πR)]|2(t−s)2+2α−2ϱdrdqds+C∫t0∫ts∫τqsE|cx(q,Xδq)[Dr(Xδq∧πR−Xδτq∧πR)]|2(t−s)2+2α−2ϱdrdqds+C∫t0∫ts∫qsE|(cx(q,Xδq)−cx(q,Xδτq))DrXδτq∧πR|2(t−s)2+2α−2ϱdrdqds≤Cδ∫t0(t−s)−1−2α+2ϱds+C∫t0∫ts∫qτqE|DrXδq∧πR|2(t−s)2+2α−2ϱdrdqds+C∫t0∫ts∫τqsE|Dr(Xδq∧πR−Xδτq∧πR)|2(t−s)2+2α−2ϱdrdqds+C∫t0∫ts∫qs(E|Xδq−Xδτq|4)12⋅(E|DrXδτq∧πR|4)12(t−s)2+2α−2ϱdrdqds≤Cδ | (3.39) |
and
I2322(t)≤CE∫t∧πR0|∫t∧πRs∫qsDr(c(q,Xδq)−c(q,Xδτq))|q−r|2H−2drdq|2(t−s)2+2α−2ϱds≤CE∫t∧πR0∫t∧πRs|∫qsDr(c(q,Xδq)−c(q,Xδτq))|q−r|2H−2dr|2dq(t−s)1+2α−2ϱds≤CE∫t0∫ts|∫qs(cx(q,Xδq)Dr(Xδq∧πR−Xδτq∧πR))|q−r|2H−2dr|2dq(t−s)1+2α−2ϱds+CE∫t0∫ts|∫qs((cx(q,Xδq)−cx(q,Xδτq))DrXδτq∧πR)|q−r|2H−2dr|2dq(t−s)1+2α−2ϱds≤C∫t0∫ts∫τqsE|cx(q,Xδq)Dr(Xδq∧πR−Xδτq∧πR)|2⋅|q−r|2H−2drdq(t−s)1+2α−2ϱds+C∫t0∫ts∫qτqE|cx(q,Xδq)DrXδq∧πR|2⋅|q−r|2H−2dr(∫qτq|q−r|2H−2dr)dq(t−s)1+2α−2ϱds+C∫t0∫ts∫qsE|(cx(q,Xδq)−cx(q,Xδτq))DrXδτq∧πR|2⋅|q−r|2H−2drdq(t−s)1+2α−2ϱds≤Cδ∫t0∫ts∫τqs|q−r|2H−2drdq(t−s)1+2α−2ϱds+Cδ2H−1∫t0∫ts∫qτq|q−r|2H−2drdq(t−s)1+2α−2ϱds+Cδ∫t0∫ts∫qs|q−r|2H−2drdq(t−s)1+2α−2ϱds≤C(δ+δ4H−2). | (3.40) |
Finally, we estimate I233(t). (2.11) implies
I233(t)≤2E(∫t∧πR0|∫t∧πRs(c(q,Xδτq)−c(τq,Xδτq))δBHq|(t−s)α+1ds)2+2α2HE(∫t∧πR0|∫t∧πRs∫qsDr(c(q,Xδτq)−c(τq,Xδτq))|q−r|2H−2drdq|(t−s)α+1ds)2=:2I2331(t)+2α2HI2332(t). | (3.41) |
From the estimation (2.13) and Lemma 3.3, we have
I2331(t)≤C∫t∧πR0E|∫t∧πRs(c(q,Xδτq)−c(τq,Xδτq))δBHq|2(t−s)2+2α−2ϱds≤C∫t0∫tsE|c(q,Xδτq)−c(τq,Xδτq)|2dq(t−s)2+2α−2ϱds+C∫t∧πR0∫t∧πRs∫qsE|Dr(c(q,Xδτq)−c(τq,Xδτq)δ)|2drdq(t−s)2+2α−2ϱds≤Cδ2β+C∫t0∫ts∫qsE|(cx(q,Xδτq)−cx(τq,Xδτq))DrXδτq∧πR|2drdq(t−s)2+2α−2ϱds≤Cδ2β | (3.42) |
and
I2332(t)≤E(∫t0∫ts∫qs|(cx(q,Xδτq)−cx(τq,Xδτq))DrXδτq∧πR|⋅|q−r|2H−2drdq(t−s)α+1ds)2≤Cδ2βE∫t0∫ts(∫qs|DrXδτq∧πR|⋅|q−r|2H−2dr)2dq(t−s)1+2α−2ϱds≤Cδ2β∫t0∫ts∫qsE|DrXδτq∧πR|2⋅(q−r)2H−2drdq(t−s)1+2α−2ϱds≤Cδ2β. | (3.43) |
Summing up all estimations (3.22)-(3.43), we have
I2(t)≤C(δ+δ4H−2)+R2⋅C2R∫t0I1(s)(t−s)12+αds+CR2∫t0I2(s)ds. | (3.44) |
Then, taking into account (3.21) and (3.44), we obtain
I1(t)+I2(t)≤C(δ+δ4H−2)+R2⋅C2R∫t0I1(s)(t−s)12+αds+CR2∫t0I2(s)ds. |
Evidently, the above estimation can be written as
I1(t)+I2(t)≤C(δ+δ4H−2)+R2⋅C2R∫t0I1(s)+I2(s)(t−s)12+αds. |
Therefore, Lemma 6.3 yields
E[supz∈[0,t]‖Xz∧πR−Xδz∧πR‖α]2≤2(I1(t)+I2(t))≤C(δ+δ4H−2)e(R2⋅C2R)21−2α. | (3.45) |
Plugging (3.45)) into (3.7), we arrive at
P(nγsupt∈[0,T]‖Xt−Xδt‖α>ε,πR=T)≤Cn2γ(δ+δ4H−2)ε2e(R2⋅C2R)21−2α. | (3.46) |
Passing to the limit as n→∞, we prove that the right hand side of (3.6) approaches 0.
Then (3.6) gives
limn→∞P(nγsupt∈[0,T]‖Xt−Xδt‖α>ε)≤P(πR<T) | (3.47) |
Letting R→∞, by Lemma 4.4 of [11], we obtain
limn→∞P(nγsupt∈[0,T]‖Xt−Xδt‖α>ε)=0. |
Corollary 3.3. If the conditions of Theorem 3.1 are satisfied, then, for any fixed ε>0, there exist a positive constant Cε and a subset Ωε of Ω with P(Ωε)>1−ε such that
E[supt∈[0,T]‖Xt−Xδt‖2αIΩε]≤Cε(δ+δ4H−2) | (3.48) |
and
E[supt∈[0,T]‖Xt−Xδt‖2αIΩ/Ωε]≤Cε1/2 | (3.49) |
where Cε=Cexp{Cε82α−1⋅exp(Cε12α−1)} and C is a general positive constant independent of δ and ε.
Proof. For any fixed ε>0, let R=2EξTε, πε=inf{t:ξt≥2EξTε}∧T and Ωε={ω:πε=T}. We have P(Ωε)>1−ε. In fact,
P(Ωε)=P(πε=T)=1−P(πε<T)≥1−P(ξT≥2EξTε)>1−ε. |
(3.48) can be derived from (3.45) immediately, and (3.49) can be from Lemma 3.2 and P(Ω/Ωε)<ε.
Remark 3.4. In [28], it is proved that, for some equation with b(t,x)=0 and c(t,x)=c(x), the error n2H−1(Xt−Xδt) almost surely converges to some stochastic process, i.e., as n→∞,
n2H−1(Xt−Xδt)→−12|∫t0c′(Xs)DsXtds|,a.s. |
In [29], it is shown that, for the Itô-SDEs with b(t,x)=b(x) and c(t,x)=0, the error nE|Xt−Xδt|2 converges to some stochastic process, i.e., as n→∞,
nE|Xt−Xδt|2→12E|Yt∫t0bb′(Xδs)Y−1sdBs|2 |
with another Brownian motion B, which is independent of the Brownian motion W, and
Ys=exp(∫s0b′(Xδu)−12bb′(Xδu)du+∫s0b′(Xδu)dWu). |
The above facts mean that the estimation of the rate of convergence in Theorem 3.1 is sharp.
Remark 3.5. In this paper we have restricted ourselves to the case of a scalar SDEs. This is only to keep our notations and computations relatively simple but the theory developed above can certainly be generalized to the multidimensional case without any difficulty. Moreover, instead of fractional Brownian motion one can take any process, which is almost surely Hölder continuous with Hölder exponent λ>12.
Remark 3.6. The proof of our main result combines the techniques of Malliavin calculus with classical fractional calculus. The main idea is to estimate the path-wise Riemann Stieltjes ∫t0c(s,Xs)dBHs flexibly by (2.3) or (2.11). Specifically,
(1) We estimate I131(t)(and I231(t)) by the properties of fractional calculus instead of Malliavin calculus, i.e., by (2.3)(and (2.5)) instead of (2.11)(and(2.13)). It is because we can hardly establish the boundedness of the Malliavin derivative DsXt for any s≤t, and the estimation for the second moment of the difference between DsXt and DsXδt for any s≤t. Indeed, for analyzing both of them, we need also the second Malliavin derivative and then the third Malliavin derivative etc., however, there is not closable formulas for them.
(2) However, we estimate I132(t),I133(t)(andI232(t),I233(t)) by (2.11)(and(2.13)) instead of (2.3)(and(2.5)) because we have little idea how to process the second term generated by (2.3) and (2.5). For example, it is very difficult to estimate the upper bound of the following expression:
∫t0E|∫s∧πR0|Xδs∧πR−Xδτs∧πR−Xδq+Xδτq|(s−q)α+1dq|2ds. |
Let us consider the following mixed SDE driven by both Brownian motion and fractional Brownian motion,
Xt=X0+∫t0μXsds+∫t0σXsdWs+∫t0XsdBHs,t∈[0,T], | (4.1) |
Here μ,σ are nonzero constants. Mixed SDE (4.1) has the explicit solution (see [30])
Xt=X0exp{(μ−σ22)t+σWt+BHt},t∈[0,T]. | (4.2) |
For any N∈N, consider the isometric partition of [0,T]: {0=t0<t1<⋯<tN=T,δ=TN}. Define τt:=max{tk:tk<t}. The Euler approximation of (4.1) is expressed as
Xδt=Xtk+μXδtk(t−tk)+σXδtk(Wt−Wtk)+Xδtk(BHt−BHtk),t∈(tk,tk+1]. | (4.3) |
or, in the integral form,
Xδt=X0+∫t0a(τs,Xδτs)ds+∫t0b(τs,Xδτs)dWs+∫t0c(τs,Xδτs)dBHs,t∈[0,T]. |
In the M-file EulerMSDE.m we set the initial state of the random number generator to be 100 with the command randn('state', 100) and consider (4.1) with μ=2,σ=1,T=1 and X0=1. We compute a discretized Brownian motion and fBm path over [0,T] with N=28 and evaluate the solutions in (4.2) as Xtrue0_T, and then apply Euler approximation using a stepsize δ. The Euler solution is stored in the 1-by-(N+1) array XE0_T. The sup‖⋅‖α-error and the constant C(ω) in Remark 7 computed as XEerrsup and Comega respectively in the M-file EulerMSDE.m. In order to compare different convergent cases, we set H=0.6,0.7,0.75,0.85,α=0.75−0.5H and ρ=γ=min{0.5,2H−1}/2. We get the following numerical results:
From the Table 1, we can see the larger H, the smaller XEerrsup and Comega, i.e. the larger H, the smaller error and dominated constant C(ω). From the Figure 1, we can see the larger H, the better the convergence, moreover, the two graphs of H=0.75 and H=0.85 are very similar. These are consistent with our conclusion because the rate of convergence is less than min{0.5,2H−1}. (see Theorem 3.1)
H | 0.6 | 0.7 | 0.75 | 0.85 |
alpha | 0.45 | 0.4 | 0.375 | 0.325 |
gamma | 0.1 | 0.2 | 0.25 | 0.25 |
XEerrsup | 14.5509 | 5.8126 | 4.07903 | 2.36231 |
Comega | 25.3346 | 17.6205 | 16.3161 | 9.44923 |
The following time-dependent mixed stochastic differential equation driven by both Brownian motion and fBm is considered in this paper.
Xt=X0+∫t0a(s,Xs)ds+∫t0b(s,Xs)dWs+∫t0c(s,Xs)dBHs,t∈[0,T]. |
We obtain that the Euler approximation has the convergent rate O(δ12∧(2H−1)) with the norm ‖⋅‖α (see Definition 2.1) in probability. We also show that it has the rate of convergence δ1∧(4H−2) in the sense of Besov type norm on some subsets of Ω with probability close to one. Meanwhile, on the complement of above subsets, the error of Euler (3.1) can be small enough correspondingly in the same norm (see Corollary 3.3). We mention that it is also true for the result of Corollary 3.3 in the sense of mean-square norm if 1∧(4H−2) is replaced by 12∧(2H−1).
On one hand, as we known, the mean-square rate of convergence for 'pure' SDE driven by single Brownian motion is O(δ12) (see [20]) and by single fractional Brownian motion is O(δ2H−1) (see [18]). For the mixed SDEs, we can only obtain the worst convergent rate of those of 'pure' SDEs because their estimates for 'pure' equations are sharp (see [20,28]).
On the other hand, Mishura and Shevchenko [16] researched the Euler approximation of the following one-dimensional mixed SDEs,
Xt=X0+∫t0a(s,Xs)ds+∫t0b(s,Xs)dWs+∫t0c(Xs)dBHs,t∈[0,T]. | (5.1) |
They derived the mean-square rate of convergence O(δ12∧(2H−1)). In [15] the authors also find that a faster rate of convergence O(δ12) to (5.1) can be obtained if one uses the modified Euler method. In a forthcoming paper we will study, whether the rate of convergence of (modified) Euler approximation to (1.1) is O(δ12∧(2H−1)) with P-a.s. (O(δ12) with mean-square norm).
In this section, we prove the bounded estimation (3.5) and recall two results from [17].
Lemma 6.1. Given α,β(1>β>α),T and any t∈[0,T],n∈N, consider the isometric partition of [0,T]: {0=t0<t1<⋯<tn=T}. Let δ=Tn and τt:=max{tk:tk<t}(see Euler equation (3.1), then we have
∫t0∫τr0(τr−τq)β(t−q)α(r−q)α+1dqdr≤C, |
here the constant C is independent of n and δ.
Proof. From the primary inequality (a+b+c)β≤C0(aβ+bβ+cβ),a≥0,b≥0,c≥0,β>0, we have
∫t0∫τr0(τr−τq)β(t−q)α(r−q)α+1dqdr≤C0∫t0∫τr0(r−τr)β+(r−q)β+(q−τq)β(t−q)α(r−q)α+1dqdr:=C0(Q1+Q2+Q3). |
For β>α, we have
Q1=∫t0∫τr0(r−τr)β(t−q)α(r−q)α+1dqdr≤∫t0(∫τr0(r−τr)β(r−q)α+1dq)1(t−r)αdr≤∫t0(r−τr)β−α(t−r)αdr≤δβ−αT1−α1−α≤C1. |
Exchanging the order of integration, we have
Q2=∫t0∫τr0(r−q)β(t−q)α(r−q)α+1dqdr=∫τt0∫tτq+δ(r−q)β−α−1dr1(t−q)αdq≤C2. |
Q3=∫t0∫τr0(q−τq)β(t−q)α(r−q)α+1dqdr=∫τt0(∫tτq+δ(r−q)−α−1dr)(q−τq)β(t−q)αdq≤1α∫τt0(q−τq)β(τq+δ−q)α(t−q)αdq≤1α(nt−2∑k=01(t−tk+1)α∫tk+1tk(q−tk)β(tk+1−q)αdq+δβ∫tnttnt−11(tnt−q)α(t−q)αdq)≤1α(1αnt−2∑k=0δβ+1−α(t−tk+1)α+δβ∫tnttnt−11(tnt−q)2αdq)≤1α(δβ−ααnt−2∑k=0∫tk+2tk+11(t−tk+1)αds+δβ+1−2α)≤1α(δβ−αα∫t01(t−s)αds+δβ+1−2α)≤1α(δβ−αT1−αα(1−α)+δβ+1−2α)≤C3. |
Let C=C0(C1+C2+C3), we complete the proof.
Lemma 6.2. (The modification of Lemma 7.1 of [17]) Let c:[0,T]×R→R be a function such that c(t,x) satisfies the assumption (Hc), then, for all x1,x2,x3,x4∈R, we have
|c(t1,x1)−c(t2,x2)−c(t1,x3)+c(t2,x4)|≤C⋅|x1−x2−x3+x4|+C⋅|x1−x3|⋅(|t2−t1|β+|x1−x2|+|x3−x4|). |
The following lemma is a generalization of Gronwall lemma.
Lemma 6.3. (Lemma 7.6 of [17]) Fix 0≤θ<1,a,b≥0. Let x:[0,∞)→[0,∞) be a continuous function such that for each t
xt≤a+btθ∫t0(t−s)−θs−θxsds. |
Then
xt≤adθecθtb1/(1−θ), |
where Γ is the Gamma function, cθ and dθ are positive constants depending only on θ (as an example, one can set cθ=2(Γ(1−θ))1/(1−θ) and dθ=4Γ(1−θ)1−θe2).
The authors are very grateful to Professor Guiwu Hu for his support and encouragement in making this work possible. We also would like to thank the referees for the careful reading of the manuscript and for their valuable suggestions. This work is supported by Project of Department of Education of Guangdong Province (No.2018KTSCX072) and Guangdong University of Finance & Economics, Big data and Educational Statistics Application Laboratory (No.2017WSYS001).
The authors declare that they have no conflict of interest.
[1] | X. Mao, Stochastic Differential Equations and Applications, Chichester, UK: Horwood, 1997. |
[2] |
P. Guasoni, No arbitrage with transaction costs, with fractional Brownian motion and beyond, Math. Financ., 16 (2006), 569-582. doi: 10.1111/j.1467-9965.2006.00283.x
![]() |
[3] | A. A. Kilbas, H. M. Srivastava, J. J. Trujillo, Theory and applications of fractional differential equations, Amsterdam: Elsevier Science B.V., 2006. |
[4] | Zh. Li, W. Zhan, L. Xu, Stochastic differential equations with time-dependent coefficients driven by fractional Brownian motion, Physica A, Volume 530, 15 September 2019, 121565. |
[5] |
G. Denk, R. Winkler, Modelling and simulation of transient noise in circuit simulation, Math. Comp. Model. Dynm., 13 (2007), 383-394. doi: 10.1080/13873950500064400
![]() |
[6] |
S. C. Kou, Stochastic modeling in nanoscale biophysics: subdiffusion within proteins, Ann. Appl. Stat., 2 (2008), 501-535. doi: 10.1214/07-AOAS149
![]() |
[7] |
F. E. Benth, On arbitrage-free pricing of weather derivatives based on fractional Brownian motion, Appl. Math. Financ., 10 (2003), 303-324. doi: 10.1080/1350486032000174628
![]() |
[8] |
P. Cheridito, Mixed fractional Brownian motion, Bernoulli, 7 (2001), 913-934. doi: 10.2307/3318626
![]() |
[9] |
J. L. da Silva, M. Erraoui, E. H. Essaky, Mixed stochastic differential equations: Existence and uniqueness result, J. Theor. Probab., 31 (2018), 1119-1141. doi: 10.1007/s10959-016-0738-9
![]() |
[10] | Y. Krvavych, Yu. Mishura, Exponential formula and Girsanov theorem for mixed semilinear stochastic differential equations, Birkhäuser Verlag Basel/Switzerland: Trends in Mathematics, 2001. |
[11] |
Yu. S. Mishura, S. V. Posashkova, Stochastic differential equations driven by a Wiener process and fractional Brownian motion: Convergence in Besov space with respect to a parameter, Comput. Math. Appl., 62 (2011), 1166-1180. doi: 10.1016/j.camwa.2011.02.032
![]() |
[12] |
G. Shevchenko, T. Shalaiko, Malliavin regularity of solutions to mixed stochastic differential equations, Stat. Probabil. Lett., 83 (2013), 2638-2646. doi: 10.1016/j.spl.2013.08.013
![]() |
[13] |
J. Guerra, D. Nualart, Stochastic differential equations driven by fractional Brownian motion and standard Brownian motion, Stoch. Anal. Appl., 26 (2008), 1053-1075. doi: 10.1080/07362990802286483
![]() |
[14] | Yu. S. Mishura, G. M. Shevchenko, Mixed stochastic differential equations with long-range dependence: Existence, uniqueness and convergence of solutions, Comput. Math. Appl., 64 (2012), 3217-3227. |
[15] |
W. G. Liu, J. W. Luo, Modified Euler approximation of stochastic differential equation driven by Brownian motion and fractional Brownian motion, Commun. Stat. Theor. M., 46 (2017), 7427-7443. doi: 10.1080/03610926.2016.1152487
![]() |
[16] | Yu. S. Mishura, G. M. Shevchenko, Rate of convergence of Euler approximation of solution to mixed stochastic differential equation involving Brownian motion and fractional Brownian motion, Rand. Opera. Stoch. Equ., 19 (2011), 387-406. |
[17] | D. Nualart, A. Răşcanu, Differential equation driven by fractional Brownian motion, Collect. Math., 53 (2002), 55-81. |
[18] |
Yu. S. Mishura, G. M. Shevchenko, The rate of convergence for Euler approximations of solutions of stochastic differential equations driven by fractional Brownian motion, Stochastics, 80 (2008), 489-511. doi: 10.1080/17442500802024892
![]() |
[19] |
G. Shevchenko, Mixed fractional stochastic differential equations with jumps, Stochastics, 86 (2014), 203-217. doi: 10.1080/17442508.2013.774404
![]() |
[20] | P. E. Kloeden, E. Platen, Numerical Solution of Stochastic Differential Equations, Berlin: Springer, 1992. |
[21] | S. G. Samko, A. A. Kilbas, O. I. Marichev, Fractional integrals and derivatives, In: Theory and Applications, Gordon and Breach Science Publishers, Yvendon, xxxvi+976 pp. ISBN: 2-88124-864-0, 1993. |
[22] |
L. C. Young, An inequality of Hölder type connected with Stieltjes integration, Acta. Math. Djursholm, 67 (1936), 251-282. doi: 10.1007/BF02401743
![]() |
[23] |
M. Zähle, Integration with respect to fractal functions and stochastic calculus I, Probab. Theory. Rel., 111 (1998), 333-374. doi: 10.1007/s004400050171
![]() |
[24] | Yu. S. Mishura, G. M. Shevchenko, Existence and uniqueness of the solution of stochastic differential equation involving wiener process and fractional Brownian motion with Hurst Index H > 1/2, Commun. Stat. Theor. M., 40 (2011), 3492-3508. |
[25] |
Y. Z. Hu, K. Le, A multiparameter Garsia-Rodemich-Rumsey inequality and some applications, Stoch. Proc. Appl., 123 (2013), 3359-3377. doi: 10.1016/j.spa.2013.04.019
![]() |
[26] |
E. Alòs, D. Nualart, Stochastic integration with respect to the fractional Brownian motion, Stoch. Stoch. Rep., 75 (2003), 129-152. doi: 10.1080/1045112031000078917
![]() |
[27] |
D. Nualart, B. Saussereau, Malliavin calculus for stochastic differential equations driven by a fractional Brownian motion, Stoch. Proc. Appl., 119 (2009), 391-409. doi: 10.1016/j.spa.2008.02.016
![]() |
[28] |
A. Neuenkirch, I. Nourdin, Exact rate of convergence of some approximation schemes associated to SDEs driven by a fractional Brownian motion, J. Theor. Probab., 20 (2007), 871-899. doi: 10.1007/s10959-007-0083-0
![]() |
[29] | S. Cambanis, Y. Hu, Exact convergence rate of the Euler-Maruyama scheme, with application to sampling design, Stoch. Stoch. Rep., 59 (1996), 211-240. |
[30] |
G. M. J. Schoenmakers, P. E. Kloeden, Robust option replication for a Black-Scholes model extended with nondeterministic trends, J. Appl. Math. Stoch. Analy., 12 (1999), 113-120. doi: 10.1155/S104895339900012X
![]() |
1. | Nasser H. Sweilam, Seham M. AL-Mekhlafi, Saleh M. Hassan, Nehaya R. Alsenaideh, Abdelaziz Elazab Radwan, New Coronavirus (2019-nCov) Mathematical Model Using Piecewise Hybrid Fractional Order Derivatives; Numerical Treatments, 2022, 10, 2227-7390, 4579, 10.3390/math10234579 | |
2. | Mohamed El-Beltagy, Ahmed Etman, Sroor Maged, Development of a fractional Wiener-Hermite expansion for analyzing the fractional stochastic models, 2022, 156, 09600779, 111847, 10.1016/j.chaos.2022.111847 | |
3. | Daniel Alpay, Palle Jorgensen, mu-Brownian Motion, Dualities, Diffusions, Transforms, and Reproducing Kernel Hilbert Spaces, 2022, 35, 0894-9840, 2757, 10.1007/s10959-021-01146-w | |
4. | N.H. Sweilam, S.M. AL-Mekhlafi, S.M. Hassan, N.R. Alsunaideh, A.E. Radwan, Crossover dynamics of climate change models: Numerical simulations, 2023, 75, 11100168, 447, 10.1016/j.aej.2023.05.093 | |
5. | Monir Chadad, Mohamed Erraoui, Reflected stochastic differential equations driven by standard and fractional Brownian motion, 2024, 24, 0219-4937, 10.1142/S0219493724500114 | |
6. | Kęstutis Kubilius, Approximation of the Fractional SDEs with Stochastic Forcing, 2024, 12, 2227-7390, 3875, 10.3390/math12243875 | |
7. | Zhaoqiang Yang, Chenglong Xu, Strong Convergence of a Modified Euler—Maruyama Method for Mixed Stochastic Fractional Integro—Differential Equations with Local Lipschitz Coefficients, 2025, 9, 2504-3110, 296, 10.3390/fractalfract9050296 |
H | 0.6 | 0.7 | 0.75 | 0.85 |
alpha | 0.45 | 0.4 | 0.375 | 0.325 |
gamma | 0.1 | 0.2 | 0.25 | 0.25 |
XEerrsup | 14.5509 | 5.8126 | 4.07903 | 2.36231 |
Comega | 25.3346 | 17.6205 | 16.3161 | 9.44923 |