Open Access

Applications of anticipated BSDEs driven by time-changing Lévy noises

Journal of Inequalities and Applications20162016:290

https://doi.org/10.1186/s13660-016-1230-x

Received: 30 April 2016

Accepted: 8 November 2016

Published: 22 November 2016

Abstract

We study a very particular anticipated BSDEs when the driver is time-changing Lévy noise. We give an estimate of the solutions in the system satisfying some non-Lipschitz conditions. Also, we state an useful comparison theorem for the solutions. At last, we establish another specific Feynman-Kac formula for a quasilinear PDE.

Keywords

anticipated BSDEan estimate of the solutionscomparison theoremnonlinear Feynman-Kac formula

1 Introduction

The theory of nonlinear BSDEs under a standard Lipschitz condition was pioneered by Pardoux and Peng in 1990 [1]. From then on it became very popular and has attracted many researchers’ attention for theoretical research. It plays an important role in many fields such as mathematical finance, the solution of PDEs, stochastic control, and pricing theory for imperfect markets. See, for example, Baños et al. [2], Cordoni et al. [36], Delong and Imkeller [7], Ekren et al. [8], Mohammed [9], Soner et al. [10] and the references therein for applications in different areas. In particular, as with applications, different noises are also gradually becoming a direction of research. Liu and Ren [11], Lu and Ren [12], Peng and Wang [13], El Karoui et al. [14] and the references therein are very useful to our study.

Especially, Peng and Yang [15] studied a special type of BSDEs, named anticipated BSDEs, such that
$$ \left \{ \textstyle\begin{array}{l} dY_{t}=-f_{t}( Y_{t}, Z_{t}, Y_{t+u(t)}, Z_{t+v(t)})\, dt+Z_{t}\, dW_{t},\quad t\in[0, T]; \\ Y_{t}=\xi_{t}, \quad t\in[T, T+K]; \\ Z_{t}=\eta_{t},\quad t\in[T, T+K], \end{array}\displaystyle \right . $$
(1)
where both \(u(\cdot)\) and \(v(\cdot)\) are positive measurable functions on \([0, T]\) and they satisfy:
  1. (a1)
    there is some positive constant K and for all \(r\in[0, T]\),
    $$r+u(r)\leq T+K, \qquad r+v(r)\leq T+K; $$
     
  2. (a2)
    there is some positive constant L, for all \(t\in[0, T]\) and every non-negative integrable \(g(\cdot)\),
    $$\int_{t}^{T}g\bigl(r+u(r)\bigr)\, dr\leq L \int_{t}^{T+K}g(r)\, dr,\qquad \int _{t}^{T}g\bigl(r+v(r)\bigr)\, dr\leq L \int_{t}^{T+K}g(r)\, dr. $$
     
In equations (1), the generator \(f(\cdot)\) includes the values of the solution of (1) of this moment and later time. The authors studied in depth the questions of existence and uniqueness, they also gave a useful comparison theorem. On this basis, they developed the connection between SDDEs and anticipated BSDEs, which is a practical idea for solving stochastic control problems (also see Cohen and Elliott [16], Li and Wu [17]). Furthermore, Lu and Ren [12] studied a class of anticipated BSDEs when the driver is Markov chains with finite state. Liu and Ren [11] addressed the classical problem of the solutions when the anticipated BSDEs’ driver is time-changing Lévy noises and provided the connection between the two kinds of equations.
On the other hand, Lévy processes are rich mathematical objects and have potential applications in many respects. The class of Lévy processes includes the Brownian case and the Poisson case. Winkel [18] dealt with the recovery problem for time-changing Lévy processes. We also refer to the literature on Lévy processes by Carr and Wu [19], David and Wim [20, 21]. Di Nunno and Sjursen [22] studied the following particular BSDEs with a time-changing Lévy noises driver, which is a framework for the independent time-change of the Brownian motion and Poisson random field:
$$ Y_{t} =\xi+ \int_{t}^{T} g_{s}(\lambda_{s},Y_{s}, \phi_{s})\,ds- \int_{t}^{T} \int_{\mathbb{R}} \phi_{s}(x)\mu(ds,dx), \quad t\in[0,T], $$
(2)
where μ is the mixture of measure between a Wiener process measure B on \([0,T]\times\{0\}\) and a doubly stochastic Poisson measure on \([0,T]\times \mathbb{R}_{0}\) with
$$\mu(\Delta):=B\bigl(\Delta\cap[0,T]\times\{0\}\bigr)+\tilde{H}\bigl(\Delta\cap [0,T]\times\mathbb{R}_{0}\bigr), \quad \Delta\subseteq X=[0,T]\times \mathbb{R}. $$
The authors proved a classical problem of the solution and addressed an application of the control problems of the BSDE. They also studied the sufficient conditions of an optimal equation.
Liu and Ren [11] discussed the following type of anticipated BSDEs when the driver is time-changing Lévy noises:
$$ \left \{ \textstyle\begin{array}{l} dY_{t} =-f(t,\lambda_{t},Y_{t},Z_{t},Y_{t+u(t)},Z_{t+v(t)})\, dt+Z_{t}(x)\mu (dt,dx), \quad t\in[0,T]; \\ Y_{t}=\xi_{t}, \quad t\in[T,T+K] ; \\ Z_{t}=\eta_{t}, \quad t\in[T,T+K]. \end{array}\displaystyle \right . $$
(3)
They addressed the classical problem of existence and uniqueness of (3). In addition, the duality between SDDEs and anticipated BSDEs was established.

The aim of this paper is to discuss some applications of anticipated BSDEs (3). As the first step, we aim to give an estimate of the solutions in the system satisfying some non-Lipschitz conditions. The comparison theorem of the solutions will be established. Also, we solve the solution of a class of quasilinear PDEs. In our following study, we will establish some other applications of mathematical finance, including utility maximization and hedging.

The rest of this paper is organized as follows. In Section 2, we present some preliminaries of time-changing Lévy noises. Section 3 is devoted to giving an estimate of the solutions. Section 4 proves a comparison theorem of the solution. In the last section, we have included an application of anticipated BSDEs to the nonlinear Feynman-Kac formula.

2 Preliminaries of time-changing Lévy noises

In this section, we review some concepts and assumptions of time-changing Lévy noises. we refer to Di Nunno and Sjursen [22], Liu and Ren [11], Peng and Yang [15], and the references therein for details.

In the following, we need the following notations. Given T, a fixed finite horizon. X can be seen as a union of \([0,T]\times\{0\}\) and \([0,T]\times\mathbb{R}_{0}\), where \(\mathbb {R}_{0}:=\mathbb{R}\backslash\{0\}\). Let \(\mathcal{F}_{t}=\sigma\{X_{s},0\leq s\leq t\}\) be the natural filtration of X, and we assume that the \(\mathcal{F}_{0}\) are the P-null sets of \(\mathcal{F}\).

Given a 2-dimensional stochastic process \(\lambda_{t}=(\lambda_{t}^{B},\lambda _{t}^{H})\), for each component \(\lambda_{t}^{i}\) (\(i=B,H\)), we set
  1. (b1)

    for each \(t\in[0,T+K]\), \(\lambda^{i}_{t}\geq0\) a.s. a.e.;

     
  2. (b2)

    for \(\forall\epsilon>0\), and \(t\in[0,T+K]\), \(\lim_{h\rightarrow0}P (|\lambda^{i}_{t+h}-\lambda^{i}_{h}|\geq\epsilon)=0\);

     
  3. (b3)

    \(\mathbb{E}[\int^{T+K}_{0}\lambda_{s}^{i}\, ds]<+\infty\).

     
For each \(\Delta\subseteq X\), we denote the mixture of a measure Λ by
$$ \Lambda(\Delta) := \int_{0}^{T}\mathbf{1}_{\{(t,0)\in\Delta\}}(t) \lambda_{t}^{B}\, dt + \int_{0}^{T} \int_{\mathbb{R}_{0}}\mathbf{1}_{\{(t,x)\in\Delta\} }(t,x)q(dx) \lambda_{t}^{H}\, dt, $$
and q is a deterministic, σ-finite measure on the Borel sets of \(\mathbb{R}_{0}\) and it satisfies \(\int_{\mathbb{R}_{0}}z^{2}q(dx)<+\infty \). Here, we denote by \(\mathcal{F}^{\Lambda}\) the σ-algebra generated by the values of Λ. \({\Lambda}^{B}\) denotes the restriction of Λ to \([0,T]\times\{0\} \) and \({\Lambda}^{H}\) denotes the restriction of Λ to \([0,T]\times\mathbb{R}_{0}\).

Now, we recall the noises driving (3); the following characterization is mainly due to Di Nunno and Sjursen [22] (see also Liu and Ren [11]).

Definition 1

B is a signed random measure on \(\mathcal{B}\{[0,T]\times\{0\}\}\) and satisfies:
  1. (c1)

    \(P (B(\Delta)\leq y|\mathcal{F}^{\Lambda} )=P (B(\Delta)\leq y|\Lambda^{B}(\Delta) )=\Phi (\frac {y}{\sqrt{\Lambda^{B}(\Delta)}} )\), \(y\in\mathbb{R}\), \(\Delta\subseteq [0,T]\times\{0\}\), where \(\Phi(y)=\int_{-\infty}^{y}\frac{1}{\sqrt{2\pi }}e^{-\frac{s^{2}}{2}}\,ds\).

     
  2. (c2)

    \(P (B(\Delta_{1})\leq x,B(\Delta_{2})\leq y|\mathcal{F}^{\Lambda} )=P (B(\Delta_{1})\leq x|\mathcal {F}^{\Lambda} )P (B(\Delta_{2})\leq y|\mathcal{F}^{\Lambda } )\), \(x, y\in\mathbb{R}\), \(\Delta_{1},\Delta_{2}\subseteq[0,T]\times\{0\}\), when \(\Delta_{1}\cap\Delta_{2}=\phi\).

     
H is a signed random measure on \(\mathcal{B}\{[0,T]\times\mathbb {R}_{0}\}\) and satisfies:
  1. (c3)

    \(P (H(\Delta)=n|\mathcal{F}^{\Lambda} )=P (H(\Delta )=n|\Lambda^{H}(\Lambda) )=\frac{[\Lambda^{H}(\Delta )]^{n}}{n!}e^{-{\Lambda^{H}(\Delta)}}\), \(n \in\mathbb{N}\), \(\Delta\subseteq [0,T]\times\mathbb{R}_{0}\).

     
  2. (c4)

    \(P (H(\Delta_{1})\leq x,H(\Delta_{2})\leq y|\mathcal{F}^{\Lambda} )=P (H(\Delta_{1})\leq x|\mathcal {F}^{\Lambda} )P (H(\Delta_{2})\leq y|\mathcal{F}^{\Lambda } )\), \(x, y\in\mathbb{R}\), \(\Delta_{1},\Delta_{2}\subseteq[0,T]\times\mathbb {R}_{0}\), when \(\Delta_{1}\cap\Delta_{2}=\phi\).

     
And let us suppose that
  1. (c5)

    \(P (B(\Delta_{1})\leq y,H(\Delta_{2})\leq y|\mathcal{F}^{\Lambda} )=P (B(\Delta_{1})\leq x|\mathcal {F}^{\Lambda} )P (H(\Delta_{2})\leq y|\mathcal{F}^{\Lambda } )\), \(y\in\mathbb{R}\), \(\Delta_{1}\subseteq[0,T]\times\{0\}\), \(\Delta _{2}\subseteq[0,T]\times\mathbb{R}_{0}\), when \(\Delta_{1}\cap\Delta_{2}=\phi\).

     
For convenience, let \(\tilde{H}:=H-\Lambda^{H}\) be the signed random measure with the following form:
$$\tilde{H}(\Delta)=H(\Delta)-\Lambda^{H}(\Delta),\quad \Delta\subset [0,T]\times\mathbb{R}_{0}. $$
From (c1)-(c5), we have
$$\begin{aligned}& \mathbb{E}\bigl[\mu(\Delta)|\mathcal{F}^{\Lambda}\bigr]=0, \qquad \mathbb{E}\bigl[B(\Delta )^{2}|\mathcal{F}^{\Lambda}\bigr]= \Lambda^{B}(\Delta), \\& \mathbb{E}\bigl[\tilde{H}(\Delta)^{2}|\mathcal{F}^{\Lambda} \bigr]=\Lambda^{H}(\Delta ),\qquad \mathbb{E}\bigl[\mu( \Delta)^{2}|\mathcal{F}^{\Lambda}\bigr]=\Lambda(\Delta), \end{aligned}$$
and
$$\mathbb{E}\bigl[\mu(\Delta_{1})\mu(\Delta_{2})| \mathcal{F}^{\Lambda}\bigr]=\mathbb {E}\bigl[\mu(\Delta_{1})| \mathcal{F}^{\Lambda}\bigr]\mathbb{E}\bigl[\mu(\Delta _{2})| \mathcal{F}^{\Lambda}\bigr]=0, $$
where \(\Delta_{1} ,\Delta_{2}\subseteq X\), \(\Delta_{1} \cap\Delta_{2}=\phi\). Hence \(\Delta_{1}\) and \(\Delta_{2}\) are conditionally orthogonal. The random measures B and H are regarded as a mixture of time-changing processes. Particularly, we also define \(\mathbb{F}=\{\mathcal {F}_{t},t\in[0,T]\}\), where \(\mathcal{F}_{t}=\bigcap_{r>t}\mathcal{F}^{\mu }_{r}\) and \(\mathcal{F}^{\mu}_{t}=\mathcal{F}^{B}_{t}\vee\mathcal{F}^{H}_{t}\vee\mathcal {F}^{\Lambda}_{t}\), \(\mathcal{F}^{B}_{t}\) is generated by \(B(\Delta)\cap [0,T]\). \(\mathcal{F}^{H}_{t}\) by \(H(\Delta)\cap[0,T]\), \(\mathcal {F}^{\Lambda}_{t}\) by \(\Lambda(\Delta)\). We set \(\mathbb{G}=\{\mathcal {G}_{t},t\in[0,T]\}\) with \(\mathcal{G}_{t}=\mathcal{F}^{\mu}_{t}\vee\mathcal {F}^{\Lambda}\), which includes time-changing noises’ information of the future values of \(\Lambda^{B}\) and \(\Lambda^{H}\).
For convenience, we list the following spaces of this paper as follows.
  • \(S^{2}_{\mathbb{G}}(0,T+K;\mathbb{R})\) = {\(Y_{t}:\Omega\times[0,T+K] | Y(t,\omega) \) is a \(\mathbb{G}\)-adapted stochastic processes satisfying that \(\mathbb{E}[\sup_{0\leq t\leq T+K}|Y_{t}|^{2}]<\infty\)}.

  • \(L^{2}_{\mathbb{G}}(0,T+K;\mathbb{R})\) = {\(Z_{t}: \Omega\times [0,T+K]\times\mathbb{R}| Z(t,x,\omega) \) is a \(\mathbb{G}\)-adapted stochastic processes satisfying \(\|Z\|=\mathbb{E}[\int_{0}^{T+K}|Z_{s}(0)|^{2}\lambda_{s}^{B}\, ds+\int_{0}^{T+K}\int _{\mathbb{R}_{0}}|Z_{s}(x)|^{2}q(dx)\lambda_{s}^{H}\, ds]<\infty\)}.

For the proof of results, we give the following assumptions.
  1. (H1)

    \(f(s,\lambda,y,z,\xi,\eta): [0,T]\times[0,+\infty)^{2}\times\Omega\times\mathbb{R}\times L^{2}_{\mathbb{G}}(s,T;\mathbb{R})\times S^{2}_{\mathbb{G}}(s,T+K;\mathbb{R})\times L^{2}_{\mathbb{G}}(s,T+K;\mathbb{R})\rightarrow L^{2}(\mathcal{G}_{s};\mathbb {R})\) is \(\mathcal{G}_{s}\)-measurable for all \(s\in[0,T]\);

     
  2. (H2)
    \(\exists C>0\), all \(s\in[0,T]\), \(y^{1},y^{2}\in\mathbb{R}\), \(z^{1},z^{2} \in L^{2}_{\mathbb{G}}(s,T;\mathbb{R})\), \(\xi^{1},\xi^{2}\in S^{2}_{\mathbb{G}}(s,T+K;\mathbb {R})\), \(\eta^{1},\eta^{2} \in L^{2}_{\mathbb{G}}(s,T+K;\mathbb{R})\), \(r^{1},r^{2}\in[s,T+K]\), we have
    $$\begin{aligned}& \bigl\vert f_{s}\bigl(\lambda,y^{1},z^{1}, \xi^{1}_{r^{1}},\eta^{1}_{r^{2} } \bigr)-f_{s}\bigl(\lambda ,y^{2},z^{2}, \xi^{2}_{r^{1}}, \eta^{2}_{r^{2} }\bigr)\bigr\vert \\& \quad \leq C \biggl[\bigl\vert y^{1}-y^{2}\bigr\vert + \bigl\vert z^{1}(0)-z^{2}(0)\bigr\vert \sqrt{ \lambda^{B}_{s}}+\sqrt{ \int _{\mathbb{R}_{0}}\bigl\vert z^{1}(x)-z^{2}(x) \bigr\vert ^{2} q(dx)}\sqrt{\lambda_{s}^{H}} \\& \qquad {}+\mathbb{E}^{\mathcal{G}_{s}}\biggl(\bigl\vert \xi^{1}_{r^{1}}- \xi^{2}_{r^{1}}\bigr\vert +\bigl\vert \eta^{1}_{r^{2} }(0)- \eta^{2}_{r^{2} }(0)\bigr\vert \sqrt{ \lambda_{r^{2} }^{B}} \\& \qquad {}+\sqrt{ \int_{\mathbb {R}_{0}}\bigl\vert \eta^{1}_{r^{2}}(x)- \eta^{2}_{r^{2} }(x)\bigr\vert ^{2} q(dx)}\sqrt {\lambda_{r^{2} }^{H}}\biggr) \biggr]; \end{aligned}$$
     
  3. (H3)

    \(f_{s}(\lambda,0,0,0,0) \in L^{2}_{\mathbb{G}}(0,T+K;\mathbb{R})\).

     

3 Estimate of the solutions to anticipated BSDEs

Liu and Ren [11] established in general existence and uniqueness of solution of (3). Now, we study these equations in depth. Our main question is to study an estimate of the solutions.

Theorem 2

Suppose (H1), (H2), and (H3) hold, \(u(t)\) and \(v(t)\) satisfying (a1) and (a2). Then, for any \(\xi_{t}\in S^{2}_{\mathbb {G}}(T,T+K;\mathbb{R})\) and \(\eta_{t}\in L^{2}_{\mathbb{G}}(T,T+K;\mathbb {R})\), the solution \((Y,Z)\in S^{2}_{\mathbb{G}}(0,T+K;\mathbb{R})\times L^{2}_{\mathbb{G}}(0,T+K;\mathbb{R})\) of the anticipated BSDEs (3) satisfies
$$\begin{aligned}& \mathbb{E}^{\mathcal{G}_{t}} \biggl[\sup_{t\leq s\leq T}\vert Y_{s}\vert ^{2} \,ds+ \int_{t}^{T} \int_{\mathbb{R}} \bigl\vert Z_{s}(x)\bigr\vert ^{2}\Lambda(dx,ds) \biggr] \\& \quad \leq C_{0}\mathbb{E}^{\mathcal{G}_{t}} \biggl[\vert \xi_{T}\vert ^{2}+ \int_{T}^{T+K}\vert \xi _{s}\vert ^{2}\,ds+ \int_{T}^{T+K} \int_{\mathbb{R}}\vert \eta_{s}\vert ^{2} \Lambda(dx,ds) \\& \qquad {}+\biggl( \int _{t}^{T}\bigl\vert f(0,0,0,0)\bigr\vert \,ds\biggr)^{2} \biggr], \end{aligned}$$
for \(t\in[0,T]\), and \(C_{0}>0\) only depends on C, L, and K.

Proof

By the Itô formula on \(e^{\beta r}|Y_{r}|^{2} \), we get
$$\begin{aligned} de^{\beta r}|Y_{r}|^{2} =& \bigl(\beta e^{\beta r}|Y_{r}|^{2}+2 e^{\beta r}Y_{r}f_{r}(Y_{r},Z_{r},Y_{r+u(r)},Z_{r+v(r)}) \bigr)\,dr \\ &{}+e^{\beta r} \int_{\mathbb {R}}|Z_{r}|^{2}(x)\Lambda(dr,dx) \\ &{}+2e^{\beta r}Y_{r} \int_{\mathbb{R}}Z_{r}(x)\mu(dr,dx). \end{aligned}$$
Iterating each side from s to T, \(s\in[0,T]\), we obtain
$$\begin{aligned}& e^{\beta T} |\xi_{T}|^{2}+2 \int_{s}^{T} e^{\beta r}Y_{r}f_{r}(Y_{r},Z_{r},Y_{r+u(r)},Z_{r+v(r)}) \,dr \\& \quad = e^{\beta s}|Y_{s}|^{2}+ \int_{s}^{T}\beta e^{\beta r}|Y_{r}|^{2} \,dr+ \int_{s}^{T} \int _{\mathbb{R}}e^{\beta r}\bigl\vert Z_{r}(x)\bigr\vert ^{2}\Lambda(dr,dx) \\& \qquad {}+2 \int_{s}^{T} \int_{\mathbb{R}}e^{\beta r}Y_{r}Z_{r}(x) \mu(dr,dx). \end{aligned}$$
(4)
Under condition (H2), we get
$$\begin{aligned}& 2 \int_{s}^{T} e^{\beta r}Y_{r}f_{r}(Y_{r},Z_{r},Y_{r+u(r)},Z_{r+v(r)}) \,dr \\& \quad = 2 \int_{s}^{T}e^{\beta r}Y_{r} \bigl(f_{r}(Y_{r},Z_{r},Y_{r+u(r)},Z_{r+v(r)})-f_{r}(Y_{r},Z_{r},Y_{r+u(r)},0) \bigr)\,dr \\& \qquad {}+2 \int_{s}^{T}e^{\beta r}Y_{r} \bigl(f_{r}(Y_{r},Z_{r},Y_{r+u(r)},0)-f_{r}(Y_{r},Z_{r},0,0) \bigr)\,dr \\& \qquad {}+2 \int_{s}^{T}e^{\beta r}Y_{r} \bigl(f_{r}(Y_{r},Z_{r},0,0)-f_{r}(Y_{r},0,0,0) \bigr)\,dr \\& \qquad {}+2 \int_{s}^{T}e^{\beta r}Y_{r} \bigl(f_{r}(Y_{r},0,0,0)-f_{r}(0,0,0,0) \bigr) \,dr+2 \int_{s}^{T}e^{\beta r}Y_{r}f_{r}(0,0,0,0) \,dr \\& \quad \leq 2C \int_{s}^{T}e^{\beta r}\mathbb{E}^{\mathcal{G}_{r}} \biggl[\bigl\vert Z_{r+v(r)}(0)\bigr\vert \sqrt{ \lambda^{B}_{r+v(r)}} +\sqrt{ \int_{\mathbb {R}_{0}}\bigl\vert Z_{r+v(r)}(x)\bigr\vert ^{2}q(dx)}\sqrt{\lambda^{H}_{r+v(r)}} \biggr] \vert Y_{r}\vert \,dr \\& \qquad {}+2C \int_{s}^{T}e^{\beta r}\mathbb{E}^{\mathcal {G}_{r}} \bigl[\vert Y_{r+u(r)}\vert \bigr]\vert Y_{r}\vert \,dr \\& \qquad {}+2C \int_{s}^{T}e^{\beta r} \biggl[\bigl\vert Z(0) \bigr\vert \sqrt{\lambda^{B}_{r}}+\sqrt{ \int _{\mathbb{R}_{0}}\bigl\vert Z_{r}(x)\bigr\vert ^{2}q(dx)}\sqrt{\lambda^{H}_{r}} \biggr] \vert Y_{r}\vert \,dr \\& \qquad {}+2C \int_{s}^{T} e^{\beta r}\vert Y_{r} \vert ^{2}\,dr+2 \int_{s}^{T} e^{\beta r}f_{r}(0,0,0,0)Y_{r} \,dr. \\& \quad \leq \frac{1}{6L} \int_{s}^{T}e^{\beta r}\mathbb{E}^{\mathcal{G}_{r}} \biggl[\bigl\vert Z_{r+v(r)}(0)\bigr\vert \sqrt{ \lambda^{B}_{r+v(r)}} +\sqrt{ \int_{\mathbb {R}_{0}}\bigl\vert Z_{r+v(r)}(x)\bigr\vert ^{2}q(dx)}\sqrt{\lambda^{H}_{r+v(r)}} \biggr]^{2}\,dr \\& \qquad {}+\frac{1}{4LT} \int_{s}^{T}e^{\beta r}\mathbb{E}^{\mathcal {G}_{r}} \bigl[\vert Y_{r+u(r)}\vert \bigr]^{2}\,dr \\& \qquad {}+\frac{1}{6} \int_{s}^{T}e^{\beta r} \biggl[\bigl\vert Z(0) \bigr\vert \sqrt{\lambda^{B}_{r}}+\sqrt { \int_{\mathbb{R}_{0}}\bigl\vert Z_{r}(x)\bigr\vert ^{2}q(dx)}\sqrt{\lambda^{H}_{r}} \biggr]^{2}\,dr \\& \qquad {}+\bigl(6LC^{2}+4LTC^{2}+6C^{2}+2C\bigr) \int_{s}^{T} e^{\beta r}\vert Y_{r} \vert ^{2}\,dr+2 \int_{s}^{T} e^{\beta r}f_{r}(0,0,0,0)Y_{r} \,dr. \\& \quad \leq \frac{1}{3} \int_{s}^{T+K} \int_{\mathbb{R}}e^{\beta r}\mathbb {E}^{\mathcal{G}_{r}}\bigl[\bigl\vert Z_{r}(x)\bigr\vert ^{2}\bigr]\Lambda(dr,dx)+ \frac{1}{4T} \int _{s}^{T+K}e^{\beta r}\mathbb{E}^{\mathcal{G}_{r}} \bigl[\vert Y_{r}\vert ^{2}\bigr]\,dr \\& \qquad {}+\frac{1}{3} \int_{s}^{T} \int_{\mathbb{R}}e^{\beta r}\bigl\vert Y_{r}(x)\bigr\vert ^{2}\Lambda (dr,dx)+2 \int_{s}^{T} e^{\beta r}f_{r}(0,0,0,0)Y_{r} \,dr \\& \qquad {}+\bigl(6LC^{2}+4LTC^{2}+6C^{2}+2C\bigr) \int_{s}^{T} e^{\beta r}\vert Y_{r} \vert ^{2}\,dr. \end{aligned}$$
From (4) we have
$$\begin{aligned}& e^{\beta s} \vert Y_{s}\vert ^{2}+ \frac{2}{3} \int_{s}^{T} \int_{\mathbb{R}}e^{\beta r}\bigl\vert Z_{r}(x)\bigr\vert ^{2}\Lambda(dr,dx) \\& \qquad {}+ \int_{s}^{T} \bigl(\beta-6LC^{2}-4LTC^{2}-C^{2}-2C \bigr)e^{\beta r}\vert Y_{r}\vert ^{2}\,dr \\& \quad \leq e^{\beta T}\vert \xi_{T}\vert ^{2}+2 \int_{s}^{T} e^{\beta r}Y_{r}f_{r}(0,0,0,0) \,dr-2 \int _{s}^{T} \int_{\mathbb{R}}e^{\beta r}Y_{r}Z_{r}(x) \mu(dr,dx) \\& \qquad {}+\frac{1}{3} \int_{s}^{T+K} \int_{\mathbb{R}}e^{\beta r}\mathbb {E}^{\mathcal{G}_{r}}\bigl[\bigl\vert Z_{r}(x)\bigr\vert ^{2}\bigr]\Lambda(dr,dx)+ \frac{1}{4T} \int _{s}^{T+K}e^{\beta r}\mathbb{E}^{\mathcal{G}_{r}} \bigl[\vert Y_{r}\vert ^{2}\bigr]\,dr. \end{aligned}$$
(5)
Taking condition expectations under \({\mathcal{G}_{s}}\), we have
$$\begin{aligned}& e^{\beta s}\vert Y_{s}\vert ^{2} + \frac{2}{3}\mathbb{E}^{\mathcal{G}_{s}}\biggl[ \int_{s}^{T} \int _{\mathbb{R}}e^{\beta r}\bigl\vert Z_{r}(x)\bigr\vert ^{2}\Lambda(dr,dx)\biggr] \\& \qquad {}+\mathbb{E}^{\mathcal{G}_{s}}\biggl[ \int_{s}^{T} \bigl(\beta -6LC^{2}-4LTC^{2}-C^{2}-2C \bigr)e^{\beta r}\vert Y_{r}\vert ^{2}\,dr\biggr] \\& \quad \leq \mathbb{E}^{\mathcal{G}_{s}}\bigl[e^{\beta T}\vert \xi_{T}\vert ^{2}\bigr]+2\mathbb {E}^{\mathcal{G}_{s}}\biggl[ \int_{s}^{T} e^{\beta r}Y_{r}f_{r}(0,0,0,0) \,dr\biggr] \\& \qquad {}+\frac{1}{3}\mathbb{E}^{\mathcal{G}_{s}}\biggl[ \int_{s}^{T+K} \int_{\mathbb {R}}e^{\beta r}\mathbb{E}^{\mathcal{G}_{r}}\bigl[\bigl\vert Z_{r}(x)\bigr\vert ^{2}\bigr]\Lambda (dr,dx) \biggr] \\& \qquad {}+\frac{1}{4T}\mathbb{E}^{\mathcal{G}_{s}}\biggl[ \int_{s}^{T+K}e^{\beta r}\mathbb{E}^{\mathcal{G}_{r}} \bigl[\vert Y_{r}\vert ^{2}\bigr]\biggr]\,dr \\& \quad \leq \mathbb{E}^{\mathcal{G}_{s}}\bigl[e^{\beta T}\vert \xi_{T}\vert ^{2}\bigr]+2\mathbb {E}^{\mathcal{G}_{s}}\biggl[ \int_{s}^{T} e^{\beta r}Y_{r}f_{r}(0,0,0,0) \,dr\biggr] \\& \qquad {}+\frac{1}{ 3}\mathbb{E}^{\mathcal{G}_{s}}\biggl[ \int_{s}^{T} \int_{\mathbb{R}}e^{\beta r}\bigl\vert Z_{r}(x)\bigr\vert ^{2}\Lambda(dr,dx)\biggr] \\& \qquad {}+\frac{1}{3}\mathbb{E}^{\mathcal{G}_{s}}\biggl[ \int_{T}^{T+K} \int_{\mathbb {R}}e^{\beta r}\bigl\vert \eta_{r}(x) \bigr\vert ^{2}\Lambda(dr,dx)\biggr]+\frac{1}{4T}\mathbb {E}^{\mathcal{G}_{s}}\biggl[ \int_{s}^{T+K}e^{\beta r}\vert Y_{r} \vert ^{2}\biggr]\,dr. \end{aligned}$$
Set \(\beta=6LC^{2}+4LTC^{2}+6C^{2}+\frac{1}{4T}+2C\), from (4) we have
$$\begin{aligned}& \mathbb{E}^{\mathcal{G}_{s}} \int_{s}^{T} \int_{\mathbb{R}}e^{\beta r}\bigl\vert Z_{r}(x)\bigr\vert ^{2}\Lambda(dr,dx) \\& \quad \leq3\mathbb{E}^{\mathcal{G}_{s}}e^{\beta T}\vert \xi_{T} \vert ^{2}+\mathbb {E}^{\mathcal{G}_{s}} \int_{T}^{T+K} \int_{\mathbb{R}}e^{\beta r}\bigl\vert \eta _{r}(x) \bigr\vert ^{2}\Lambda(dr,dx) \\& \qquad {}+\frac{3}{4T}\mathbb{E}^{\mathcal{G}_{s}} \int_{T}^{T+K}e^{\beta r}\vert \xi _{r}\vert ^{2}\,dr+6\mathbb{E}^{\mathcal{G}_{s}} \int_{s}^{T} e^{\beta r}f_{r}(0,0,0,0)Y_{r} \,dr. \end{aligned}$$
(6)
For \(\forall s\in[t,T]\),
$$\begin{aligned}& \biggl\vert \int_{s}^{T} \int_{\mathbb{R}}e^{\beta r}Y_{r}Z_{r}(x) \mu(dr,dx)\biggr\vert \\& \quad \leq\biggl\vert \int_{t}^{T} \int_{\mathbb{R}}e^{\beta r}Y_{r} Z_{r}(x) \mu(dr,dx)\biggr\vert +\biggl\vert \int _{t}^{s} \int_{\mathbb{R}}e^{\beta r}Y_{r}Z_{r}(x) \mu(dr,dx)\biggr\vert . \end{aligned}$$
By the Burkholder-Davis-Gundy inequality and the elementary inequality \(ab\leq\frac{1}{2\varepsilon}a^{2}+\frac{\varepsilon}{2}b^{2}\) for \(\varepsilon >0\), we obtain
$$\begin{aligned}& \mathbb{E}^{\mathcal{G}_{t}}\biggl[\biggl\vert \sup _{t\leq s\leq T} \int_{s}^{T} \int _{\mathbb{R}}e^{\beta r}Y_{r}Z_{r}(x) \mu(dr,dx)\biggr\vert \biggr] \\& \quad \leq2\mathbb{E}^{\mathcal{G}_{t}}\biggl[\sup_{t\leq s\leq T}\biggl\vert \int_{t}^{s} \int_{\mathbb{R}}e^{\beta r}Y_{r}Z_{r}(x) \mu(dr,dx)\biggr\vert \biggr] \\& \quad \leq2\sqrt{32}\mathbb{E}^{\mathcal{G}_{t}}\biggl[ \int_{t}^{s}e^{2\beta r}\biggl( \int _{t}^{T}Y_{r} Z_{r}(x) \mu(dr,dx)\biggr)^{2}\biggr]^{\frac{1}{2}} \\& \quad \leq\frac{1}{4}\mathbb{E}^{\mathcal{G}_{t}}\sup_{t\leq r\leq T} \bigl[e^{\beta r}\vert Y_{r}\vert ^{2}\bigr]+128 \mathbb{E}^{\mathcal{G}_{t}}\biggl[ \int_{t}^{T} \int _{\mathbb{R}} e^{\beta r}\bigl\vert Z_{r}(x) \bigr\vert ^{2}\Lambda(dr,dx)\biggr]. \end{aligned}$$
(7)
From (5) and (7), we get
$$\begin{aligned}& \mathbb{E}^{\mathcal{G}_{t}}\Bigl[\sup_{t\leq s\leq T}e^{\beta s} \vert Y_{s}\vert \Bigr] \\& \quad \leq\mathbb{E}^{\mathcal{G}_{t}}\biggl[e^{\beta T}\vert \xi_{T}\vert ^{2}+2 \int_{s}^{T} e^{\beta r}\vert Y_{r} \vert \bigl\vert f_{r}(0,0,0,0)\bigr\vert \,dr \\& \qquad {}+2\sup _{t\leq s\leq T}\biggl\vert \int_{s}^{T} \int_{\mathbb{R}}e^{\beta r}Y_{r} Z_{r}(x) \mu(dr,dx)\biggr\vert \biggr] \\& \qquad {}+\frac{1}{3}\mathbb{E}^{\mathcal{G}_{t}}\biggl\{ \int_{s}^{T+K} \int_{\mathbb {R}}e^{\beta r}\mathbb{E}^{\mathcal{G}_{r}}\bigl[\bigl\vert Z_{r}(x)\bigr\vert ^{2}\bigr]\Lambda(dr,dx)\biggr\} +\frac{1}{4T}\mathbb{E}^{\mathcal{G}_{t}}\biggl\{ \int_{s}^{T+K}e^{\beta r}\mathbb {E}^{\mathcal{G}_{r}}\bigl[\vert Y_{r}\vert ^{2}\bigr]\,dr \biggr\} \\& \quad \leq\mathbb{E}^{\mathcal{G}_{t}}\biggl[e^{\beta T}\vert \xi_{T}\vert ^{2}+2 \int_{s}^{T} e^{\beta r}\vert Y_{r} \vert \bigl\vert f_{r}(0,0,0,0)\bigr\vert \,dr\biggr]+ \frac{1}{2}\mathbb{E}^{\mathcal {G}_{t}}\sup_{t\leq r\leq T} \bigl[e^{\beta r}\vert Y_{r}\vert ^{2}\bigr] \\& \qquad {}+256\mathbb{E}^{\mathcal{G}_{t}}\biggl[ \int_{t}^{T} \int_{\mathbb{R}} e^{\beta r}\bigl\vert Z_{r}(x)\bigr\vert ^{2}\Lambda(dr,dx)\biggr]+ \frac{1}{3} \mathbb{E}^{\mathcal{G}_{t}}\biggl[ \int _{s}^{T+K} \int_{\mathbb{R}}e^{\beta r}\bigl\vert Z_{r}(x)\bigr\vert ^{2}\Lambda(dr,dx)\biggr] \\& \qquad {}+\frac{1}{4T}\mathbb{E}^{\mathcal{G}_{t}}\biggl[ \int_{s}^{T+K}e^{\beta r}\vert Y_{r} \vert ^{2}\,dr\biggr] \\& \quad \leq\mathbb{E}^{\mathcal{G}_{t}}\biggl[e^{\beta T}\vert \xi_{T}\vert ^{2}+2 \int_{s}^{T} e^{\beta r}\vert Y_{r} \vert \bigl\vert f_{r}(0,0,0,0)\bigr\vert \,dr\biggr]+ \frac{3}{4}\mathbb{E}^{\mathcal {G}_{t}}\sup_{t\leq r\leq T} \bigl[e^{\beta r}\vert Y_{r}\vert ^{2}\bigr] \\& \qquad {}+\frac{1}{3}\mathbb{E}^{\mathcal{G}_{t}}\biggl[ \int_{T}^{T+K} \int_{\mathbb {R}}e^{\beta r}\bigl\vert \eta_{r}(x) \bigr\vert ^{2}\Lambda(dr,dx)\biggr] \\& \qquad {}+\biggl(256+\frac{1}{3} \biggr)\mathbb {E}^{\mathcal{G}_{t}}\biggl[ \int_{t}^{T} \int_{\mathbb{R}} e^{\beta r}\bigl\vert Z_{r}(x)\bigr\vert ^{2}\Lambda(dr,dx)\biggr] \\& \qquad {}+\frac{1}{4T}\mathbb{E}^{\mathcal{G}_{t}}\biggl[ \int_{T}^{T+K}e^{\beta r}\vert \xi _{r}\vert ^{2}\,dr\biggr]. \end{aligned}$$
(8)
From (6) and (8), let \(C_{0} \) be a positive number that depends on T, L, and C, we get
$$\begin{aligned}& \frac{1}{4} \mathbb{E}^{\mathcal{G}_{t}}\Bigl[\sup _{t\leq s\leq T}e^{\beta s}\vert Y_{s}\vert \Bigr] \\ & \quad \leq C_{0}\mathbb{E}^{\mathcal{G}_{t}}\biggl[e^{\beta T} \vert \xi_{T}\vert ^{2}+ \int _{T}^{T+K}e^{\beta r}\vert \xi_{r}\vert ^{2}\,dr+ \int_{T}^{T+K} \int_{\mathbb{R}}e^{\beta r}\bigl\vert \eta_{r}(x) \bigr\vert ^{2}\Lambda(dr,dx)\biggr] \\ & \qquad {}+C_{0} \mathbb{E}^{\mathcal{G}_{t}}\Bigl(\sup _{t\leq r\leq T}e^{\frac{1}{2}\beta r}\vert Y_{r}\vert \Bigr) \biggl( \int_{s}^{T} e^{\frac{1}{2} \beta r} \bigl\vert f_{r}(0,0,0,0)\bigr\vert \,dr\biggr) \\ & \quad \leq C_{0}\mathbb{E}^{\mathcal{G}_{t}}\biggl[\vert \xi_{T}\vert ^{2}+ \int_{T}^{T+K}\vert \xi _{r}\vert ^{2}\,dr+ \int_{T}^{T+K} \int_{\mathbb{R}}\bigl\vert \eta_{r}(x)\bigr\vert ^{2}\Lambda (dr,dx)\biggr] \\ & \qquad {}+\frac{1}{8}\mathbb{E}^{\mathcal{G}_{t}}\Bigl[\sup _{t\leq r\leq T}e^{\frac{1}{2}\beta r}\vert Y_{r}\vert \Bigr]+2C_{0}^{2}\mathbb{E}^{\mathcal{G}_{t}}\biggl[\biggl( \int_{s}^{T} \bigl\vert f_{r}(0,0,0,0) \bigr\vert \,dr\biggr)^{2}\biggr]. \end{aligned}$$
(9)
From (8) and (9), we get
$$\begin{aligned}& \mathbb{E}^{\mathcal{G}_{t}} \biggl[\sup_{t\leq s\leq T}\vert Y_{s}\vert ^{2} \,ds+ \int_{t}^{T} \int_{\mathbb{R}} \bigl\vert Z_{s}(x)\bigr\vert ^{2}\Lambda(dx,ds) \biggr] \\ & \quad \leq C_{0}\mathbb{E}^{\mathcal{G}_{t}} \biggl[\vert \xi_{T}\vert ^{2}+ \int_{T}^{T+K}\vert \xi _{s}\vert ^{2}\,ds+ \int_{T}^{T+K} \int_{\mathbb{R}}\vert \eta_{s}\vert ^{2} \Lambda(dx,ds) \\ & \qquad {}+\biggl( \int _{t}^{T}\bigl\vert f(0,0,0,0)\bigr\vert \,ds\biggr)^{2} \biggr]. \end{aligned}$$
The proof is complete. □

4 Comparison theorem

The comparison theorem has important applications in many respects. It is one of the most important theorems in the history of BSDEs. Because of wide application it attracts more and more attention from researchers. Now we introduce the comparison theorem of anticipated BSDEs (3) that the generator \(f(\cdot) \) is linear. The linear anticipated BSDE is written in the following form:
$$ \left \{ \textstyle\begin{array}{l} -dY_{t}= [A_{t}Y_{t}+D_{t}Y_{t+u(t)}+C_{t}+E_{t}(0)Z_{t}(0)\sqrt{\lambda _{t}^{B}}+\int_{\mathbb{R}_{0}}E_{t}(x)Z_{t}(x)q(dx)\sqrt{\lambda_{t}^{H}} ]\, dt \\ \hphantom{-dY_{t}={}}{}-Z_{t}(0)\, dB_{t}-\int_{\mathbb{R}_{0}}Z_{t}(x)\tilde{H}(dt,dx); \\ Y_{t}=\xi_{T},\quad t\in[T,T+K]. \end{array}\displaystyle \right . $$
(10)

Lemma 3

Assume that we have a linear driver anticipated BSDE of (10), and we assume that:
  1. (d1)

    There exists \(C_{1}>0\) for each \(t\in[0,T]\), it hold that \(|A_{t}|< C_{1}\), and \(D_{t}< C_{1}\).

     
  2. (d2)

    \(C_{t}\in S^{2}_{\mathbb{G}}(0,T+K;\mathbb{R})\), and \(E_{t}\in L^{2}_{\mathbb{G}}(0,T+K;\mathbb{R})\).

     
  3. (d3)

    \(Y_{t}\) is increasing.

     
It is then true that \(Y_{t}\geq\mathbb{E}^{\mathcal{G}_{t}} [\xi_{T} X_{T}+\int_{t}^{T}C_{t}X_{t}\, dt ]\), \(t\in[0,T]\) a.e., a.s.
Here
$$ X_{s}= \left \{ \textstyle\begin{array}{l@{\quad}l} \exp [\int_{t}^{s}\int_{\mathbb{R}}E_{r}\mu(dr,dx)-\frac{1}{2}\int _{t}^{s}\int_{\mathbb{R}}E_{r}^{2}\Lambda(dr,dx)+\int_{t}^{s}(A_{r}+D_{r})\,dr ], & s\in[t,T]; \\ 1,& s\in[0,t]. \end{array}\displaystyle \right . $$

Proof

We can easily conclude that the anticipated BSDE of (10) satisfies conditions (H1)-(H3). Following Liu and Ren [11], we find that the anticipated BSDE of (10) has a unique solution.

Set
$$dX_{t}=X_{t} \biggl[(A_{t}+D_{t})\, dt+\frac{E_{t}(0)}{\sqrt{\lambda_{t}^{B}}}\, dB_{t}+ \int _{\mathbb{R}_{0}}\frac{E_{t}(x)}{\sqrt{\lambda_{t}^{H}}}\tilde{H}(dx,dt) \biggr]. $$
We have
$$ X_{s}= \left \{ \textstyle\begin{array}{l@{\quad}l} \exp [\int_{t}^{s}\int_{\mathbb{R}}E_{r}\mu(dr,dx)-\frac{1}{2}\int _{t}^{s}\int_{\mathbb{R}}E_{r}^{2}\Lambda(dr,dx)+\int_{t}^{s}(A_{r}+D_{r})\,dr ], & s\in[t,T]; \\ 1,& s\in[0,t]. \end{array}\displaystyle \right . $$
Applying Itô’s formula to \(X_{s}Y_{s}\) on \(s\in[t,T]\),
$$\begin{aligned} dX_{s}Y_{s} =& -X_{s} \biggl[A_{s}Y_{s}+D_{s}Y_{s+u(s)}+C_{s}+E_{s}(0)Z_{s}(0) \sqrt{\lambda _{s}^{B}}+ \int_{\mathbb{R}_{0}}E_{s}(x)Z_{s}(x)q(dx)\sqrt {\lambda_{s}^{H}} \biggr]\,ds \\ &{}+X_{s} \biggl[Z_{s}(0)\, dB_{s}+ \int_{\mathbb{R}_{0}}Z_{s}(x)\tilde{H}(ds,dx) \biggr] \\ &{}+X_{s}Y_{s} \biggl[(A_{s}+D_{s}) \,ds+\frac{E_{s}(0)}{\sqrt{\lambda_{s}^{B}}}\, dB_{s}+ \int _{\mathbb{R}_{0}}\frac{E_{s}(x)}{\sqrt{\lambda_{s}^{H}}}\tilde{H}(dx,ds) \biggr] \\ &{}+X_{s} \biggl[\frac{E_{s}(0)Z_{s}(0)}{\sqrt{\lambda_{s}^{B}}}\lambda_{s}^{B} \, ds+ \int _{\mathbb{R}_{0}}\frac{E_{s}(x)Z_{s}(x)}{\sqrt{\lambda_{s}^{H}}}q(dx)\lambda _{s}^{H} \, ds \biggr] \\ =&- [X_{s}D_{s}Y_{s+u(s)}-X_{s}D_{s}Y_{s}+X_{s}C_{s} ]\,ds+ \biggl[X_{s}Z_{s}+X_{s}Y_{s} \frac{E_{s}(x)}{\sqrt{\lambda_{s}}} \biggr]\mu(dx,ds). \end{aligned}$$
Integrating each side from t to T,
$$\begin{aligned}& X_{T}Y_{T}-X_{t}Y_{t} \\& \quad =- \int^{T}_{t} [X_{s}D_{s}Y_{s+u(s)}-X_{s}D_{s}Y_{t}+X_{s}C_{s} ]\,ds \\& \qquad {}+ \int ^{T}_{t} \int_{\mathbb{R}} \biggl[X_{s}Z_{s}+X_{s}Y_{s} \frac{E_{s}(x)}{\sqrt{\lambda _{s}}} \biggr]\mu(dx,ds). \end{aligned}$$
Taking conditional expectations, we obtain
$$\begin{aligned} Y_{t} =&\mathbb{E}^{\mathcal{G}_{t}} \biggl[X_{T} \xi_{T}+ \int_{t}^{T}C_{s}X_{s}\, ds \biggr]+\mathbb{E}^{\mathcal{G}_{t}} \int_{t}^{T} \bigl[X_{s}D_{s}(Y_{s+u(s)}-Y_{s}) \bigr]\,ds \\ \geq&\mathbb{E}^{\mathcal{G}_{t}} \biggl[X_{T}\xi_{T}+ \int_{t}^{T}C_{s}X_{s}\, ds \biggr]. \end{aligned}$$
 □

Theorem 4

Suppose that
$$g_{2}\bigl(t,y,y',z\bigr)=f \biggl(t, \lambda,y,y',z(0)E_{t}(0)\sqrt{ \lambda_{t}^{B}}, \int _{\mathbb{R}_{0}}z(x)E_{t}(x)q(dx)\sqrt{ \lambda_{t}^{H}} \biggr). $$
Let f have the form of (7), satisfying (H1)-(H3). \((Y_{t}^{(1)},Z_{t}^{(1)})\) and \((Y_{t}^{(2)},Z_{t}^{(2)})\) are, respectively, solutions to their corresponding anticipated BSDEs (10). Moreover, we suppose that:
  1. (e1)

    \(\xi_{s}^{(1)}\geq\xi_{s}^{(2)}\), \(s\in[T,T+K]\) a.s.

     
  2. (e2)

    \(g_{1}(t,\lambda,y,z,y') \geq g_{2}(t,\lambda,y,z,y')\), \(t\in[0,T]\), \(y\in\mathbb{R}\), \(y'\in S^{2}_{\mathbb{G}}(0,T+K;\mathbb{R})\), \(z\in L^{2}_{\mathbb{G}}(0,T+K;\mathbb{R})\).

     
  3. (e3)

    \(g_{2}(t,\lambda,y,\cdot ,y')\) is increasing, i.e., \(g_{2}(t,\lambda,y,z_{1},y')\geq g_{2}(t,\lambda,y,z_{2},y')\), if \(z_{1}\geq z_{2}\).

     
Then
$$Y_{t}^{(1)}\geq Y_{t}^{(2)},\quad \textit{a.e.},\textit{a.s.} $$

Proof

Set \(\tilde{Y}_{t}=Y_{t}^{(1)}-Y_{t}^{(2)}\), \(\tilde {Z}_{t}=Z_{t}^{(1)}-Z_{t}^{(2)}\), \(\tilde{\xi}_{t}=\xi_{t}^{(1)}-\xi_{t}^{(2)}\), and \(C_{t}=g_{1}(t,Y^{(1)}_{t},Z^{(1)}_{t},Y^{(1)}_{t+u(t)})-g_{2}(t,Y^{(1)}_{t},Z^{(1)}_{t},Y^{(1)}_{t+u(t)})\); by the BSDEs (10), we have
$$ \left \{ \textstyle\begin{array}{l} -d\tilde{Y}_{t}= [g_{2} (t,Y^{(1)}_{t},Z^{(1)}_{t},Y^{(1)}_{t+u(t)} )-g_{2} (t,Y^{(2)}_{t},Z^{(2)}_{t},Y^{(2)}_{t+u(t)} )+C_{t} ]\, dt \\ \hphantom{-d\tilde{Y}_{t}={}}{}-\tilde{Z}_{t}\mu(dt,dx), \quad t\in[0,T]; \\ \tilde{Y}_{T}=\tilde{\xi}_{T}. \end{array}\displaystyle \right . $$
(11)
Note that \(g_{2} (t,Y^{(1)}_{t},Z^{(1)}_{t},Y^{(1)}_{t+u(t)} )-g_{2} (t,Y^{(2)}_{t},Z^{(2)}_{t},Y^{(2)}_{t+u(t)} ) =A_{t}Y_{t}+D_{t}Y_{t+u(t)}+C_{t}+P_{t}E_{t}(0) \tilde{Z}_{t}(0)\sqrt{\lambda _{t}^{B}}+Q_{t}\int_{\mathbb{R}_{0}}E_{t}(x)\tilde{Z}_{t}(x)q(dx)\sqrt{\lambda_{t}^{H}}\).
Here
$$\begin{aligned}& A_{t}= \left \{ \textstyle\begin{array}{l@{\quad}l} \frac{g_{2} (t,Y^{(1)}_{t},Z^{(1)}_{t},Y^{(1)}_{t+u(t)} )-g_{2} (t,Y^{(2)}_{t},Z^{(1)}_{t},Y^{(1)}_{t+u(t)} )}{Y_{t}^{(1)}-Y_{t}^{(2)}},&\mbox{if } Y_{t}^{(1)}\neq Y_{t}^{(2)}; \\ 0,&\mbox{if } Y_{t}^{(1)}=Y_{t}^{(2)}, \end{array}\displaystyle \right . \\& D_{t}= \left \{ \textstyle\begin{array}{l@{\quad}l} \frac{g_{2} (t,Y^{(2)}_{t},Z^{(1)}_{t},Y^{(1)}_{t+u(t)} )-g_{2} (t,Y^{(2)}_{t},Z^{(1)}_{t},Y^{(2)}_{t+u(t)} )}{Y_{t+u(t)}^{(1)}-Y_{t+u(t)}^{(2)}},&\mbox{if } Y_{t+u(t)}^{(1)}\neq Y_{t+u(t)}^{(2)}; \\ 0,&\mbox{if } Y_{t+u(t)}^{(1)}=Y_{t+u(t)}^{(2)}, \end{array}\displaystyle \right . \\& P_{t}= \left \{ \textstyle\begin{array}{l@{\quad}l} \frac{g_{2} (t,Y^{(2)}_{t},Z^{(1,B)}_{t},Y^{(2)}_{t+u(t)} )-g_{2} (t,Y^{(2)}_{t},Z^{(2,B)}_{t},Y^{(2)}_{t+u(t)} )}{E_{t}(0)\tilde {Z}_{t}(0)\sqrt{\lambda_{t}^{B}}},&\mbox{if } E_{t}(0)\tilde{Z}_{t}(0)\sqrt{\lambda _{t}^{B}}\neq0; \\ 0,&\mbox{if } E_{t}(0)\tilde{Z}_{t}(0)\sqrt{\lambda_{t}^{B}}=0, \end{array}\displaystyle \right . \\& Q_{t}= \left \{ \textstyle\begin{array}{l@{\quad}l} \frac{g_{2} (t,Y^{(2)}_{t},Z^{(1,H)}_{t},Y^{(2)}_{t+u(t)} )-g_{2} (t,Y^{(2)}_{t},Z^{(2,H)}_{t},Y^{(2)}_{t+u(t)} )}{\int_{\mathbb {R}_{0}}E_{t}(x)\tilde{Z}_{t}(x)q(dx)\sqrt{\lambda_{t}^{H}}},&\mbox{if }\int_{\mathbb {R}_{0}}E_{t}(x)\tilde{Z}_{t}(x)q(dx)\sqrt{\lambda_{t}^{H}}\neq0; \\ 0,&\mbox{if }\int_{\mathbb{R}_{0}}E_{t}(x)\tilde{Z}_{t}(x)q(dx)\sqrt{\lambda_{t}^{H}}=0. \end{array}\displaystyle \right . \end{aligned}$$
So we get
$$ \left \{ \textstyle\begin{array}{l} -d\tilde{Y}_{t}= [A_{t}\tilde{Y}_{t}+D_{t}\tilde {Y}_{t+u(t)}+C_{t}+P_{t}E_{t}(0)\tilde{Z}_{t}(0)\sqrt{\lambda_{t}^{B}}+Q_{t}\int _{\mathbb{R}_{0}}E_{t}(x)\tilde{Z}_{t}(x)q(dx)\sqrt{\lambda_{t}^{H}} ]\, dt \\ \hphantom{-d\tilde{Y}_{t}={}}{}-\tilde{Z}_{t}(0)\, dB_{t}-\int_{\mathbb{R}_{0}}\tilde{Z}_{t}(x)\tilde {H}(dt,dx), \quad t\in[0,T]; \\ \tilde{Y}_{T}=\tilde{\xi}_{T}. \end{array}\displaystyle \right . $$
Since \(g_{2}\) satisfies (H1)-(H3), and \(A_{t}\), \(D_{t}\), \(P_{t}\), and \(Q_{t}\) satisfy the conditions of Lemma 3, the solution \(\tilde {Y}_{t}\) of the anticipated BSDE in (9) satisfies
$$\tilde{Y}_{t}\geq\mathbb{E}^{\mathcal{G}_{t}} \biggl[\tilde{ \xi}_{T} X_{T}+ \int _{t}^{T}C_{t}X_{t}\, dt \biggr]\geq0 \quad \mbox{a.e.},\mbox{a.s.} $$

Hence \(Y_{t}^{(1)}\geq Y_{t}^{(2)}\) a.e., a.s. □

5 A nonlinear Feynman-Kac formula

In order to solve the quasilinear PDE, by El Karoui et al. [14] let us denote by \(X^{x,t}_{s}\) the solution of the SDE of the following form:
$$ \left \{ \textstyle\begin{array}{l} dX_{s}^{x,t}=b(X_{s}^{x,t})\,ds+\sigma(X_{s}^{x,t},0)\, dBs+\int_{\mathbb {R}_{0}}\sigma(X_{s}^{x,t},\alpha)\tilde{H}(ds,d\alpha); \\ X_{t}^{x,t}=x. \end{array}\displaystyle \right . $$
(12)
Here \(b(x):\mathbb{R}\rightarrow\mathbb{R}\) and \(\sigma(\cdot,\cdot ):\mathbb{R}\times\mathbb{R}\rightarrow\mathbb{R}\) are given functions.
Now we set up a nonlinear Feynman-Kac formula for a special class of quasilinear PDEs in the form
$$ \left \{ \textstyle\begin{array}{l} \frac{\partial\Phi}{\partial t}+\mathcal{L}\Phi+f(t,\Phi (x,t),z(x,t),\Phi(x ,t+u(t)),z(x, t+v(t))) \\ \quad =0, \quad (t,x)\in[0,T]\times \mathbb{R}; \\ \Phi(T,x)=g(x), \quad x\in\mathbb{R}. \end{array}\displaystyle \right . $$
(13)
Here \(\mathcal{L}\) is a two order elliptic differential operator,
$$\mathcal{L}\Phi=\frac{1}{2} \sum_{i,j=1}^{n} a_{ij}(x)\frac{\partial^{2}\Phi }{\partial x_{i}\, \partial x_{j}} +\sum_{i=1}^{n} b_{i}(x)\frac{\partial\Phi}{\partial x_{i}}\quad \bigl(a_{ij}=\bigl[\sigma \sigma^{T}\bigr]_{ij}\bigr). $$
It is interesting to notice that we consider coupled BSDE with (12),
$$ \left \{ \textstyle\begin{array}{l} -dY_{s}^{x,t} =f_{s}(\lambda _{s},X^{x,t}_{s},Y^{x,t}_{s},Z^{x,t}_{s},Y^{x,t}_{s+u(s)},Z^{x,t}_{s+v(s)})\,ds \\ \hphantom{-dY_{s}^{x,t} ={}}{}-\int _{\mathbb{R}} Z^{x,t}_{s}(\alpha)\mu(ds,d\alpha), \quad s\in[t,T]; \\ Y^{x,t}_{t}=g(x), \quad t\in[T,T+K]; \\ Z^{x,t}_{t}=\eta_{t}, \quad t\in[T,T+K], \end{array}\displaystyle \right . $$
(14)
where \(u(\cdot)\) and \(v(\cdot)\) satisfy assumptions (a1) and (a2), f satisfies (H1), (H2), and (H3).

Given fixed \((t,x)\in[0, T]\times\mathbb{R}\), we note that \(\Phi (t,x)=Y^{t,x}_{t}\) is an adapted solution of the anticipated BSDE (14). Applying the Itô formula to \((\Phi (X_{s}^{x,t},s))\) on \(s\in[t,T]\), from (13) we achieve the theorem as follows.

Theorem 5

Assume that, for a given time duration \([t,T]\), the anticipated BSDEs (14) has an adapted solution \((Y_{t}^{t,x},Z_{t}^{t,x})\). Then the function \(\Phi(t, x):=Y_{t}^{t,x} \), \((t, x)\in[0, T]\times\mathbb {R}\) is a viscosity solution of the quasilinear PDE (13).

Declarations

Acknowledgements

The project of Youxin Liu is partially supported by the NSRP of Education Department of Anhui Province (No. KJ2013B347 and No. KJ2016A760).

Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.

Authors’ Affiliations

(1)
Department of Elementary Teaching, Wuhu Institute of Technology

References

  1. Pardoux, E, Peng, S: Adapted solution of a backward stochastic differential equation. Syst. Control Lett. 14, 55-61 (1990) MathSciNetView ArticleMATHGoogle Scholar
  2. Baños, DR, Cordoni, F, Nunno, G, Di Persio, L, Røse, EE: Stochastic systems with memory and jumps. arXiv:1603.00272 (2016)
  3. Cordoni, F, Di Persio, L: Backward stochastic differential equations approach to hedging, option pricing, and insurance problems. Int. J. Stoch. Anal. 2014, Article ID 152389 (2014) MathSciNetMATHGoogle Scholar
  4. Cordoni, F, Di Persio, L: A BSDE with delayed generator approach to pricing under counterparty risk and collateralization. Int. J. Stoch. Anal. 2016, Article ID 1059303 (2016) MathSciNetGoogle Scholar
  5. Cordoni, F, Di Persio, L, Maticiuc, L, Zălinescu, A: A stochastic approach to path-dependent nonlinear Kolmogorov equations via BSDEs with time-delayed generators and applications to finance. arXiv:1602.05793 (2016)
  6. Cordoni, F, Di Persio, L, Oliva, I: A nonlinear Kolmogorov equation for stochastic functional delay differential equations with jumps. arXiv:1602.03851 (2016)
  7. Delong, L, Imkeller, P: Backward stochastic differential equations with time delayed generators-results and counterexamples. Ann. Appl. Probab. 20(4), 1512-1536 (2010) MathSciNetView ArticleMATHGoogle Scholar
  8. Ekren, I, Keller, C, Touzi, N, Zhang, J: On viscosity solutions of path dependent PDEs. Ann. Probab. 42(1), 204-236 (2014) MathSciNetView ArticleMATHGoogle Scholar
  9. Mohammed, AS-E: Stochastic differential systems with memory: theory, examples and applications In: Stochastic Analysis and Related Topics VI: Proceedings of the Sixth Oslo-Silivri Workshop Geilo 1996, vol. 42, pp. 1-77 (1998) View ArticleGoogle Scholar
  10. Soner, HM, Touzi, N, Zhang, J: Wellposedness of second order backward SDEs. Probab. Theory Relat. Fields 153(1-2), 149-190 (2012) MathSciNetView ArticleMATHGoogle Scholar
  11. Liu, Y, Ren, Y: Anticipated BSDEs driven by time-changed Lévy noises. J. Korean Stat. Soc. 44(3), 403-409 (2015) MathSciNetView ArticleMATHGoogle Scholar
  12. Lu, W, Ren, Y: Anticipated backward stochastic differential equations on Markov chains. Stat. Probab. Lett. 83, 1711-1719 (2013) MathSciNetView ArticleMATHGoogle Scholar
  13. Peng, S, Wang, F: BSDE, path-dependent PDE and nonlinear Feynman-Kac formula. Sci. China Math. 59(1), 1-18 (2011) MathSciNetGoogle Scholar
  14. El Karoui, N, Peng, S, Quenez, M: Backward stochastic differential equations in finance. Math. Finance 7(1), 1-71 (1997) MathSciNetView ArticleMATHGoogle Scholar
  15. Peng, S, Yang, Z: Anticipated backward stochastic differential equations. Ann. Probab. 37, 877-902 (2009) MathSciNetView ArticleMATHGoogle Scholar
  16. Cohen, SN, Elliott, RJ: Solutions of backward stochastic differential equations on Markov chains. Commun. Stoch. Anal. 2, 251-262 (2008) MathSciNetMATHGoogle Scholar
  17. Li, N, Wu, Z: Maximum principle for anticipated recursive stochastic optimal control problem with delay and Lévy processes. Appl. Math. J. Chin. Univ. 29(1), 67-85 (2014) View ArticleMATHGoogle Scholar
  18. Winkel, M: The recovery problem for time-changed Lévy processes. Research Report Ma Phy Sto 2001-37 (2001) Google Scholar
  19. Carr, P, Wu, L: Time-changed Lévy processes and option pricing. J. Financ. Econ. 71, 113-141 (2014) View ArticleGoogle Scholar
  20. David, N, Wim, S: Chaotic and predictable representations for Lévy processes. Stoch. Process. Appl. 90, 109-122 (2000) View ArticleMATHGoogle Scholar
  21. David, N, Wim, S: BSDE’s and Feynman-Kac formula for Lévy processes with applications in finance. Bernoulli 7, 761-776 (2001) MathSciNetView ArticleMATHGoogle Scholar
  22. Di Nunno, G, Sjursen, S: BSDEs driven by time-changed Lévy noises and optimal control. Stoch. Process. Appl. 124, 1679-1709 (2014) View ArticleMATHGoogle Scholar

Copyright

© Liu 2016