BSDEs with terminal conditions that have bounded Malliavin derivative

Patrick Cheridito, Kihun Nam

Abstract

This paper establishes the existence and uniqueness of solutions to Backward Stochastic Differential Equations (BSDEs) of the form $$ Y_t = \xi + \int_t^T f(s, Y_s, Z_s) ds - \int_t^T Z_s dW_s $$ in cases where the terminal condition $\xi$ possesses a bounded Malliavin derivative. While traditional BSDE theory requires the driver function $f(s, y, z)$ to be Lipschitz continuous in $z$, this work demonstrates that if the terminal condition is sufficiently "smooth" in the Malliavin sense, the driver can exhibit arbitrary growth in $z$, requiring only local Lipschitz continuity. Furthermore, if the terminal condition is both bounded and has a bounded Malliavin derivative, the driver's requirement for Lipschitz continuity in $y$ can also be relaxed to local Lipschitz continuity. These results extend to the Markovian setting, providing existence and uniqueness for semilinear parabolic Partial Differential Equations (PDEs) with non-Lipschitz nonlinearities and various boundary conditions.

Introduction: The Lipschitz Barrier

Since the seminal work of Pardoux and Peng in the early 1990s, Backward Stochastic Differential Equations (BSDEs) have become a cornerstone of stochastic analysis, with applications ranging from financial derivative pricing and risk management to stochastic optimal control. The standard existence and uniqueness theory, however, relies heavily on the assumption that the driver function $f$ is Lipschitz continuous in its variables $y$ and $z$.

In many practical applications, particularly those involving superquadratic growth or aggressive nonlinearities, this Lipschitz requirement is too restrictive. When the driver grows faster than the square of the control variable $z$, standard techniques often fail to guarantee unique solutions, or solutions may explode in finite time. Previous attempts to circumvent this "Lipschitz barrier" typically required the terminal condition $\xi$ to be bounded or the driver to possess a specific convex structure. This research proposes a different path: leveraging the regularity of the terminal condition via Malliavin calculus to control the behavior of the solution.

Methodology: Malliavin Calculus and A Priori Estimates

The core technical innovation of this paper is the use of Malliavin calculus—a stochastic calculus of variations—to provide a priori bounds on the $Z$ component of the BSDE solution. The $Z$ process represents the "control" or "hedging" part of the solution, and its stability is crucial for existence.

The Role of the Malliavin Derivative

The authors assume that the terminal condition $\xi$ has a bounded Malliavin derivative ($D\xi$). The "aha moment" of the study is the realization that the complexity or growth rate of the driver $f$ can be offset by the regularity of $\xi$. If $D\xi$ is bounded, the authors prove that the $Z$ process remains bounded ($Z \in H^\infty$), even if the driver $f$ grows arbitrarily fast in $z$.

Proof Strategy

To establish existence and uniqueness, the authors employ a multi-step approach:

  1. A Priori Estimates: They derive quantitative bounds for $Z$. Specifically, they show: $$|Z_t^i| \le \left( A_i + \int_t^T q_i(s)e^{-B(T-s)}ds \right) e^{B(T-t)}$$ where $A_i$ is the bound on the Malliavin derivative and $B$ relates to the Lipschitz constant of the driver in $y$.

  2. Mollification: The non-Lipschitz driver is approximated by a sequence of well-behaved Lipschitz drivers.

  3. Convergence: Using the uniform bounds on $Z$, the authors demonstrate that the solutions to the approximated equations converge to a unique limit $(Y, Z)$ in the appropriate functional spaces ($S^4 \times H^\infty$).
  4. Comparison Theorems: These are used to ensure the stability of the solution and to "sandwich" it between known bounds.

Key Results and Theoretical Contributions

Existence and Uniqueness (Theorem 2.2)

The primary result is the proof of a unique solution $(Y, Z)$ when the terminal condition has a bounded Malliavin derivative and the driver is Lipschitz in $y$ but only locally Lipschitz in $z$. This allows the driver to grow at any rate (e.g., exponential or high-order polynomial) relative to $z$.

Relaxation of $y$-Lipschitz Requirements (Corollary 2.8)

If the terminal condition $\xi$ is bounded in addition to having a bounded Malliavin derivative, the driver $f$ only needs to be locally Lipschitz in both $y$ and $z$. This significantly broadens the class of applicable drivers in stochastic modeling.

Applications to Semilinear Parabolic PDEs

In the Markovian setting, where the terminal condition is a function of a forward SDE, the BSDE solution $Y_t$ corresponds to the solution of a semilinear parabolic PDE. This research provides a framework for solving PDEs with non-Lipschitz nonlinearities. The authors successfully extend these results to include lateral boundary conditions of both Dirichlet and Neumann types, which is often difficult because boundaries can introduce discontinuities that interfere with Malliavin differentiability.

Discussion and Limitations

While this work provides a powerful alternative to the standard Lipschitz framework, it introduces its own constraints. The requirement for a bounded Malliavin derivative is quite strict; it essentially implies that the terminal condition must be a Lipschitz function of the underlying Brownian motion. This excludes common "kinked" financial payoffs, such as standard European call options, unless they are first smoothed via mollification.

Furthermore, while the driver can grow arbitrarily fast, it must still be locally Lipschitz. The paper does not address drivers that are fundamentally non-Lipschitz, such as those that are only Hölder continuous in $z$.

Conclusion

By shifting the focus from the driver's growth to the terminal condition's regularity, this paper breaks the traditional Lipschitz barrier in BSDE theory. The result is a robust framework for handling highly nonlinear stochastic processes and PDEs, providing new mathematical tools for complex modeling in finance and physics. Future research may look toward relaxing the boundedness of the Malliavin derivative to specific integrability conditions or exploring BSDEs driven by more general Lévy processes.

Summary updated

Abstract

We show existence and uniqueness of solutions to BSDEs of the form $$ Y_t = ξ+ \int_t^T f(s,Y_s,Z_s)ds - \int_t^T Z_s dW_s$$ in the case where the terminal condition $ξ$ has bounded Malliavin derivative. The driver $f(s,y,z)$ is assumed to be Lipschitz continuous in $y$ but only locally Lipschitz continuous in $z$. In particular, it can grow arbitrarily fast in $z$. If in addition to having bounded Malliavin derivative, $ξ$ is bounded, the driver needs only be locally Lipschitz continuous in $y$. In the special case where the BSDE is Markovian, we obtain existence and uniqueness results for semilinear parabolic PDEs with non-Lipschitz nonlinearities. We discuss the case where there is no lateral boundary as well as lateral boundary conditions of Dirichlet and Neumann type.
BibTeX
@article{cheridito2013bsdes,
  author = {Patrick Cheridito and Kihun Nam},
  title = {{BSDEs with terminal conditions that have bounded Malliavin derivative}},
  journal = {Journal of Functional Analysis},
  year = {2013},
  doi = {10.1016/j.jfa.2013.12.004},
}
Figure