Abstract
We focus on a class of path-dependent problems which include path-dependent PDEs and Integro PDEs (in short IPDEs), and their representation via BSDEs driven by a cadlag martingale. For those equations we introduce the notion of decoupled mild solution for which, under general assumptions, we study existence and uniqueness and its representation via the aforementioned BSDEs. This concept generalizes a similar notion introduced by the authors in recent papers in the framework of classical PDEs and IPDEs. For every initial condition (s, η), where s is an initial time and η an initial path, the solution of such BSDE produces a couple of processes (Ys, η, Zs, η). In the classical (Markovian or not) literature the function \(u(s,\eta ):= Y^{s,\eta }_{s}\) constitutes a viscosity type solution of an associated PDE (resp. IPDE); our approach allows not only to identify u as the unique decoupled mild solution, but also to solve quite generally the so called identification problem, i.e. to also characterize the (Zs, η)s, η processes in term of a deterministic function v associated to the (above decoupled mild) solution u.
Similar content being viewed by others
References
Aliprantis, C.D., Border, K.C.: Infinite-Dimensional Analysis, 2nd edn. Springer, Berlin (1999). A hitchhiker’s guide
Barrasso, A., Russo, F.: Backward stochastic differential equations with no driving martingale, Markov processes and associated pseudo partial differential equations. Preprint, hal-01431559, v2 (2017)
Barrasso, A., Russo, F.: Backward stochastic differential equations with no driving martingale, Markov processes and associated pseudo partial differential equations. part II: Decoupled mild solutions and examples. Preprint, hal-01505974 (2017)
Barrasso, A., Russo, F.: Martingale driven BSDEs, PDEs and other related deterministic problems. Preprint, hal-01566883 (2017)
Barrasso, A., Russo, F.: Gâteaux type path-dependent PDEs and BSDEs with Gaussian forward processes. In preparation (2019)
Barrasso, A., Russo, F.: Path-dependent martingale problems and additive functionals. Stochastics and Dynamics, 19 no 1, Preprint, hal-01775200 (2019)
Bion-Nadal, J.: Dynamic risk reasures and path-dependent second order PDEs. Stoch. Environ. Financ. Econ. 138, 147–178 (2016)
Bismut, J.M.: Conjugate convex functions in optimal stochastic control. J. Math. Anal. Appl. 44, 384–404 (1973)
Carbone, R., Ferrario, B., Santacroce, M.: Backward stochastic differential equations driven by càdlàg martingales. Veroyatn. Primen. 52(2), 375–385 (2007)
Cont, R., Fournié, D.-A.: Change of variable formulas for non-anticipative functionals on path space. J. Funct Anal. 259(4), 1043–1072 (2010)
Cont, R., Fournié, D.-A.: Functional Itô calculus and stochastic integral representation of martingales. Ann. Probab. 41(1), 109–133 (2013)
Cosso, A., Russo, F.: Strong-viscosity solutions: semilinear parabolic PDEs and path-dependent PDEs. To appear: Osaka Journal of Mathematics Preprint HAL-01145301 (2015)
Cosso, A., Russo, F.: Functional Itô versus Banach space stochastic calculus and strict solutions of semilinear path-dependent equations. Infin. Dimens. Anal. Quantum Probab. Relat. Top. 19(4), 1650024, 44 (2016)
Cruzeiro, A.B., Qian, Z.M.: Backward stochastic differential equations associated with the vorticity equations. J. Funct. Anal. 267(3), 660–677 (2014)
Di Girolami, C., Russo, F.: Infinite dimensional stochastic calculus via regularization and applications. Preprint HAL-INRIA, inria-00473947 version 1 (Unpublished) (2010)
Dupire, B.: Functional Itô calculus. Portfolio Research Paper, Bloomberg (2009)
Ekren, I., Keller, C., Touzi, N., Zhang, J.: On viscosity solutions of path dependent PDEs. Ann. Probab. 42(1), 204–236 (2014)
Ekren, I., Touzi, N., Zhang, J.: Viscosity solutions of fully nonlinear parabolic path dependent PDEs. Part I. To appear in Annals of Probability (2013)
El Karoui, N., Peng, S., Quenez, M.C.: Backward stochastic differential equations in finance. Math. Financ. 7(1), 1–71 (1997)
Ethier, S.N., Kurtz, T.G.: Markov Processes. Wiley Series in Probability and Mathematical Statistics: Probability and Mathematical Statistics. Wiley, New York (1986). Characterization and convergence
Flandoli, F., Zanco, G.: An infinite-dimensional approach to path-dependent Kolmogorov equations. Ann. Probab. 44(4), 2643–2693 (2016)
Fuhrman, M., Masiero, F., Tessitore, G.: Stochastic equations with delay: optimal control via BSDEs and regular solutions of Hamilton-Jacobi-Bellman equations. SIAM J. Control Optim. 48(7), 4624–4651 (2010)
Fuhrman, M., Tessitore, G.: Nonlinear Kolmogorov equations in infinite dimensional spaces: the backward stochastic differential equations approach and applications to optimal control. Ann. Probab. 30(3), 1397–1465 (2002)
Jacod, J.: Calcul Stochastique et problèmes de Martingales, Volume 714 of Lecture Notes in Mathematics. Springer, Berlin (1979)
Jacod, J., Shiryaev, A.N.: Limit Theorems for Stochastic Processes, Volume 288 of Grundlehren der Mathematischen Wissenschaften [Fundamental Principles of Mathematical Sciences], 2nd edn. Springer, Berlin (2003)
Leão, D., Ohashi, A., Simas, A.B.: A weak version of path-dependent functional Itô calculus. Ann Probab. 46(6), 3399–3441 (2018)
Liang, G., Lyons, T., Qian, Zh.: Backward stochastic dynamics on a filtered probability space. Ann Probab. 39(4), 1422–1448 (2011)
Pardoux, É., Peng, S.: Adapted solution of a backward stochastic differential equation. Syst. Control Lett. 14(1), 55–61 (1990)
Pardoux, É. , Peng, S.: Backward stochastic differential equations and quasilinear parabolic partial differential equations. In: Stochastic Partial Differential Equations and Their Applications (Charlotte, NC, 1991), Volume 176 of Lecture Notes in Control and Inform. Sci., pp 200–217. Springer, Berlin (1992)
Peng, S., Wang, F.: BSDE, path-dependent PDE and nonlinear Feynman-Kac formula. Sci. Chin/ Math. 59(1), 19–36 (2016)
Acknowledgments
The authors are grateful to the anonymous Referee and Associated Editor for their stimulating comments on the first version of the paper. The research of the first named author was provided by a PhD fellowship (AMX) of the Ecole Polytechnique. The contribution of the second named author was partially supported by the grant 346300 for IMPAN from the Simons Foundation and the matching 2015-2019 Polish MNiSW fund.
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix
Appendix
1.1 A.1 Path-Dependent Martingale Additive Functionals
We here recall the notion of path-dependent Martingale Additive Functionals that we use in this paper. This was introduced in [6] and can be conceived as a path-dependent extension of the notion of non-homogeneous Martingale Additive Functionals of a Markov process developed in [2]. In this subsection, all results come from Section 4 in [6]. In this subsection we consider a progressive path-dependent canonical class \((\mathbbm {P}^{s,\eta })_{(s,\eta )\in \mathbbm {R}_+\times {\Omega }}\) satisfying Hypothesis 2.3 and the corresponding path-dependent system of projectors \((P_s)_{s\in \mathbbm {R}_+}\).
Definition A.1
On \(({\Omega },\mathcal {F})\), a path-dependent Martingale Additive Functional, in short path-dependent MAF will be a real-valued random-field M := (Mt, u)0≤t≤u verifying the two following conditions.
-
1.
For any 0 ≤ t ≤ u, Mt, u is \(\mathcal {F}^o_{u}\)-measurable;
-
2.
for any \((s,\eta )\in \mathbbm {R}_+\times {\Omega }\), there exists a real cadlag \((\mathbbm {P}^{s,\eta },\mathbbm {F}^{s,\eta })\)- martingale Ms, η (taken equal to zero on [0, s] by convention) such that for any η ∈Ω and s ≤ t ≤ u,
$$ M_{t,u} = M^{s,\eta}_{u}-M^{s,\eta}_{t} \text{ } \mathbbm{P}^{s,\eta}\text{ a.s.} $$
Ms, η will be called the cadlag version of M under\(\mathbbm {P}^{s,\eta }\).
A path-dependent MAF will be said to verify a certain property (being square integrable, having an absolutely continuous angular bracket) if under any \(\mathbbm {P}^{s,\eta }\) its cadlag version verifies it.
Proposition A.2
Let\((\mathcal {D}(A), A)\)bea weak generator of\((P_s)_{s\in \mathbbm {R}_+}\)and\((s,\eta ) \in {\mathbbm R}_+ \times {\Omega }\). Then for every\({\Phi }\in \mathcal {D}(A)\), \({\Phi }-{\int }_0^{\cdot }A({\Phi })_rdr\)admitsfor all (s, η) on\([s,+\infty [\)a\(\mathbbm {P}^{s,\eta }\)versionM[Φ]s, ηwhichis a\((\mathbbm {P}^{s,\eta },\mathbbm {F}^{s,\eta })\)-cadlagmartingale. In particular, the random field defined by\(M[{\Phi }]_{t,u}(\omega ):={\Phi }_u(\omega )-{\Phi }_t(\omega )-{\int }_t^uA{\Phi }_r(\omega )dr\)definesa MAF with cadlag versionM[Φ]s, ηunder\(\mathbbm {P}^{s,\eta }\).
Proposition A.3
LetM andN be two square integrable path-dependent MAFs and letMs, η(respectivelyNs, η)be the cadlag version ofM (respectivelyN) under a fixed\(\mathbbm {P}^{s,\eta }\). Assume thatN has an absolutely continuous angular bracket.
Then there exists an\(\mathbbm {F}^o\)-progressivelymeasurable process k such that for any\((s,\eta )\in \mathbbm {R}_+\times {\Omega }\),
Notation A.4
The processk whose existence is stated in Proposition A.3 will bedenoted\(\frac {d\langle M,N\rangle _t}{dt}\).
1.2 A.2 Proof of Proposition 3.10
In the sequel, we are in the framework of Section 3.
Lemma A.5
Let\(\tilde {f}\in \mathcal {L}^1_{uni}\). Then\(\begin {array}{rcl} (s,\eta )&\longmapsto &\mathbbm {E}^{s,\eta }[{\int }_s^{T}\tilde {f}_rdr]\\ \ [0,T]\times {\Omega }&\longrightarrow &\mathbbm {R} \end {array}\)is\(\mathbbm {F}^o\)-progressivelymeasurable.
Proof
We fix T0 ∈]0, T] and we will show that on [0, T0] × Ω, \((s,\eta )\longmapsto \mathbbm {E}^{s,\eta }[{\int }_s^T\tilde {f}_rdr]\) is \(\mathcal {B}([0,T_0])\otimes \mathcal {F}^o_{T_0}\)-measurable. We will start by showing that on [0, T0] × Ω × [0, T0], the function \(k^n:(s,\eta ,t)\mapsto \mathbbm {E}^{s,\eta }[{\int }_{t}^T((-n)\vee \tilde {f}_r\wedge n)dr]\) is \(\mathcal {B}([0,T_0])\otimes \mathcal {F}^o_{T_0}\otimes \mathcal {B}([0,T_0])\)-measurable, where \(n\in \mathbbm {N}\).
Let t ∈ [0, T0] be fixed. Then by Remark 2.4 \((s,\eta )\mapsto \mathbbm {E}^{s,\eta }[{{\int }_{t}^{T}}((-n)\vee \tilde {f}_r\wedge n)dr]\) is \(\mathcal {B}([0,T_0])\otimes \mathcal {F}^o_{T_0}\)-measurable.
Let (s, η) ∈ [0, T0] × Ω be fixed and \(t_m\underset {m\rightarrow \infty }{\longrightarrow } t\) be a converging sequence in [0, T0]. We then have
This sequence is uniformly bounded by nT, so by dominated convergence theorem, the convergence in Eq. A.1 also holds under the expectation, so that \(t\mapsto \mathbbm {E}^{s,\eta }[{\int }_{t}^T((-n)\vee \tilde {f}_r\wedge n)dr]\) is continuous. By Lemma 4.51 in [1], kn is therefore \(\mathcal {B}([0,T_0])\otimes \mathcal {F}^o_{T_0}\otimes \mathcal {B}([0,T_0])\)-measurable.
The composition of (s, η)↦(s, η, s) with the maps kn yields that, for any n ≥ 0, \(\tilde {k}^n:(s,\eta )\longmapsto \mathbbm {E}^{s,\eta }[{\int }_{s}^T((-n)\vee \tilde {f}_r\wedge n)dr]\) is (on [0, T0] × Ω) \(\mathcal {B}([0,T_0])\otimes \mathcal {F}^o_{T_0}\)-measurable. \(\tilde {k}^n\) therefore defines an \(\mathbbm {F}^o\)-progressively measurable process. Then by letting n tend to infinity, \(((-n)\vee \tilde {f}\wedge n)\) tends \(dt\otimes d\mathbbm {P}^{s,\eta }\) a.e. to \(\tilde {f}\) and since we assumed \(\mathbbm {E}^{s,\eta }[{\int }_s^T|\tilde {f}_r|dr]<\infty \), by dominated convergence, \(\mathbbm {E}^{s,\eta }[{\int }_{s}^T((-n)\vee \tilde {f}_r\wedge n)dr]\) tends to \(\mathbbm {E}^{s,\eta }[{\int }_s^T\tilde {f}_rdr]\). \((s,\eta )\longmapsto \mathbbm {E}^{s,\eta }[{\int }_s^T\tilde {f}(r,X_r)dr]\) is therefore an \(\mathbbm {F}^o\)-progressively measurable process as the pointwise limit of the \(\tilde {k}^n\) which are \(\mathbbm {F}^o\)-progressively measurable processes. □
We recall the following immediate consequence of Fubini’s Theorem which corresponds to Lemma 5.12 in [2].
Lemma A.6
Let ℙ bea probability measure on\(({\Omega },\mathcal {F})\)andϕ, ψbetwo measurable processes. Ifϕandψare ℙ-modificationsof each other, then they are equal\(dt\otimes d\mathbbm {P}\)a.e.
The proof of Proposition 3.10 goes through a linearization lemma.
Lemma A.7
Let\(\tilde {f}\in \mathcal {L}^2_{uni}\). Let, for every (s, η) ∈ [0, T] × Ω, \((\tilde Y^{s,\eta },\tilde M^{s,\eta })\)bethe unique solution of
in\(\left ({\Omega },\mathcal {F}^{s,\eta },\mathbbm {F}^{s,\eta },\mathbbm {P}^{s,\eta }\right )\). Then there exists a process\(\tilde Y\in \mathcal {L}^2_{uni}\), a square integrable path-dependent MAF\((\tilde M_{t,u})_{0\leq t\leq u}\)and\(\tilde Z^1,\cdots ,\tilde Z^d\in \mathcal {L}^2_{uni}\), such that for all (s, η) ∈ [0, T] × Ω the following holds.
-
1.
\(\tilde Y^{s,\eta }\)is on [s, T] a\(\mathbbm {P}^{s,\eta }\)-modification of\(\tilde Y\);
-
2.
\(\tilde M^{s,\eta }\)is the cadlag version of\(\tilde M\)under\(\mathbbm {P}^{s,\eta }\).
-
3.
For each integer 1 ≤ i ≤ d, \(\tilde Z^i=\frac {d\langle \tilde M^{s,\eta },N^{i,s,\eta }\rangle _t}{dt}\)\(dt\otimes d\mathbbm {P}^{s,\eta }\) a.e.
Remark A.8
The existence, for any (s, η), of a unique solution \((\tilde Y^{s,\eta },\tilde M^{s,\eta })\) of Eq. A.2 holds because ξ and \((t,\omega ,y,z)\mapsto \tilde {f}_t(\omega )\) trivially verify the hypothesis of Theorem 3.7.
Proof
We set \(\tilde Y:(s,\eta )\mapsto \mathbbm {E}^{s,\eta }\left [\xi + {\int }_s^T \tilde {f}_rdr\right ]\) which is \(\mathbbm {F}^o\)-progressively measurable by Remark 2.4 and Lemma A.5. Therefore, for a fixed t ∈ [s, T] we have \(\mathbbm {P}^{s,\eta }\)-a.s.
The second equality follows by Remark 2.4 and the third one uses (A.2). For every 0 ≤ t ≤ u and ω ∈Ω we set
For fixed (s, η), Eq. A.2 implies \(d\tilde Y^{s,\eta }_r=-\tilde {f}_rdr+d\tilde M^{s,\eta }_r\). On the other hand \({\int }_s^T|\tilde {f}|_rdr<+\infty \)\(\mathbbm {P}^{s,\eta }\) a.s.; so for any s ≤ t ≤ u we have \(\tilde M^{s,\eta }_u-\tilde M^{s,\eta }_t=\tilde M_{t,u}\)\(\mathbbm {P}^{s,\eta }\)- a.s. Taking into account that \(\tilde M^{s,\eta }\) is square integrable and the fact that previous equality holds for any (s, η) and t ≤ u, then \((\tilde M_{t,u})_{0\leq t\leq u}\) indeed defines a square integrable path-dependent MAF. Y belongs to \(\mathcal {L}^2_{uni}\) because the validity of the two following arguments hold for all (s, η). First Y is a \(\mathbbm {P}^{s,\eta }\)-modification of Ys, η on [s, T], so by Lemma A.6 Y = Ys, η\(dt\otimes d\mathbbm {P}^{s,\eta }\) a.e.; second \(Y^{s,\eta }\in \mathcal {L}^2(dt\otimes d\mathbbm {P}^{s,\eta })\). The existence of Z follows setting for all i, \(Z^i=\frac {d\langle \tilde M,N^i\rangle _t}{dt}\), see Notation A.4 and Proposition A.3. □
Notation A.9
For every fixed (s, η) ∈ [0, T] × Ω, we will denote by\((Y^{k,s,\eta },M^{k,s,\eta })_{k\in \mathbbm {N}}\)thePicard iterations associated toBSDEs, η(f, ξ) asdefined in Notation A.13 in [4] andZk, s, η := (Z1, k, s, η,⋯Zd, k, s, η) willdenote\(\frac {\langle M^{k,s,\eta },N^{s,\eta }\rangle _t}{dt}\).
This means that for all (s, η) ∈ [0, T] × Ω,(Y0, s, η, M0, s, η) ≡ (0, 0) andfor allk ≥ 1, we have on [s, T]
in the sense of\(\mathbbm {P}^{s,\eta }\)-indistinguishability, and that for all (s, η) ∈ [0, T] × Ω, k ≥ 0, Yk, s, η, Z1, k, s, η,⋯Zd, k, s, ηbelong to\(\mathcal {L}^2(dt\otimes d\mathbbm {P}^{s,\eta })\), see Notation A.13 and Lemma A.2 in [4].
A direct consequence of Proposition A.15 in [4] and the lines above it, is the following.
Proposition A.10
For every (s, η) ∈ [0, T] × Ω, each component of (Yk, s, η, Z1, k, s, η,⋯ , Zd, k, s, η) tendsto each component of (Ys, η, Z1, s, η,⋯ , Zd, s, η) in\(\mathcal {L}^2(dt\otimes d\mathbbm {P}^{s,\eta })\)and\(dt\otimes d\mathbbm {P}^{s,\eta }\)-a.e.when k tends to infinity.
Proposition A.11
For each\(k\in \mathbbm {N}\), there existprocesses\(Y^k\in \mathcal {L}^2_{uni}, Z^{k,1},\cdots , Z^{k,d}\in \mathcal {L}^2_{uni}\), a square integrablepath-dependent MAF\((M^k_{t,u})_{0\leq t\leq u}\)suchthat for all (s, η) ∈ [0, T] × Ω, we have the following.
-
1.
Yk, s, ηison [s, T] a\(\mathbbm {P}^{s,\eta }\)-modificationofYk;
-
2.
Mk, s, ηisthe cadlag version ofMkunder\(\mathbbm {P}^{s,\eta }\).
-
3.
For all (s, η) ∈ [0, T] × Ω andi ∈ ⟦1; d⟧, \( Z^{k,i}=\frac {d\langle M^{k,s,\eta },N^{i,s,\eta }\rangle _t}{dt}\)\(dt\otimes d\mathbbm {P}^{s,\eta }\)a.e.
Proof
We prove the statement by induction on k ≥ 0. It is clear that Y0 ≡ 0 and M0 ≡ 0 verify the assertion for k = 0.
Suppose the existence, for k ≥ 1, of a square integrable path-dependent MAF Mk− 1 and processes Yk− 1\( Z^{k-1,1},\cdots , Z^{k-1,d}\in \mathcal {L}^2_{uni}\) such that the statements 1. 2. 3. above hold replacing k with k − 1.
We fix (s, η) ∈ [0, T] × Ω. By Lemma A.6, (Yk− 1, s, η, Zk− 1, s, η) = (Yk− 1, Zk− 1) \(dt\otimes d\mathbbm {P}^{s,\eta }\) a.e. Therefore by Eq. A.4
According to Notation 3.8, the Eq. A.4 can be seen as a BSDE of the type \(BSDE^{s,\eta }(\tilde f,\xi )\) where \(\tilde f:(t,\omega )\longmapsto f(t,\omega ,Y^{k-1}_t(\omega ),Z^{k-1}_t(\omega ))\). We now verify that \(\tilde f\) verifies the conditions under which Lemma A.7 applies.
\(\tilde f\) is \(\mathbbm {F}^o\)-progressively measurable since Yk− 1, Zk− 1 are \(\mathbbm {F}^o\)-progressively measurable and f is \(\mathcal {P}ro^o\otimes \mathcal {B}(\mathbbm {R})\otimes \mathcal {B}(\mathbbm {R}^d)\)-measurable. Since
for all t, ω, with \(f(\cdot ,\cdot ,0,0),Y^{k-1},Z^{k-1,1},\cdots ,Z^{k-1,d}\in \mathcal {L}^2_{uni}\) by recurrence hypothesis, it is clear that \(\tilde f\in \mathcal {L}^2_{uni}\). Since (Yk, s, η, Mk, s, η) is a solution of \(BSDE^{s,\eta }(\tilde f,\xi )\), Lemma A.7 shows the existence of suitable Yk, Mk, Zk,1,⋯ , Zk, d verifying the statement for the integer k. □
Proof of Proposition 3.10
We define \(\bar {Y}\) and \(\bar Z^i, 1 \le i \le d\) by \({\bar Y}_s(\eta ):= \underset {k\in \mathbbm {N}}{\text {limsup }}Y^k_s(\eta )\) and \(\bar Z^i_s(\eta ):=\underset {k\in \mathbbm {N}}{\text {limsup }}Z^{k,i}_s(\eta ),\) for every (s, η) ∈ [0, T] × Ω. \(\bar {Y}\) and \(\bar Z:=(\bar Z^1,\cdots ,\bar Z^d)\) are \(\mathbbm {F}^o\)-progressively measurable. Combining Propositions A.11, A.10 and Lemma A.6 it follows that, for every (s, η) ∈ [0, T] × Ω,
Let us fix 1 ≤ i ≤ d and (s, η) ∈ [0, T] × Ω. There is a set As, η of full \(dt\otimes d\mathbbm {P}^{s,\eta }\) measure such that for all (t, ω) ∈ As, η we have
This implies
By Eqs. A.7 and ??, under every \(\mathbbm {P}^{s,\eta }\), we actually have
in the sense of \(\mathbbm {P}^{s,\eta }\)-indistinguishability, on the interval [s, T]. At this stage, in spite of Eq. A.7, \(\bar Y\) is not necessarily a modification of Ys, η. We will construct processes Y, Z fulfilling indeed the statement of Proposition 3.10. In particular Y fulfills item 1. that is a bit stronger than (A.7).
We set now \(\tilde f:(t,\omega )\mapsto f(t,\omega ,\bar {Y}_t(\omega ),\bar Z_t(\omega ))\); Eq. A.8 is now of the form (A.2) and we show that \(\tilde f\) so defined verifies the conditions under which Lemma A.7 applies. \(\tilde f\) is \(\mathbbm {F}^o\)-progressively measurable since f is \(\mathcal {P}ro^o\otimes \mathcal {B}(\mathbbm {R})\otimes \mathcal {B}(\mathbbm {R}^d)\)-measurable and \(\bar Y,\bar Z\) are \(\mathbbm {F}^o\)-progressively measurable.
Moreover, for any (s, η) ∈ [0, T] × Ω, Ys, η and Z1, s, η,⋯ , Zd, s, η belong to \(\mathcal {L}^2(dt\otimes d\mathbbm {P}^{s,\eta })\); therefore by Eq. A.6, so do \(\bar {Y}\) and \(\bar Z^1,\cdots ,\bar Z^d\).
Since this holds for all (s, η), then \(\bar {Y}\) and \(\bar Z^1,\cdots ,\bar Z^d\) belong to \(\mathcal {L}^2_{uni}\).
Finally, since \(|\tilde f(t,\omega )|=|f(t,\omega ,\bar Y_t(\omega ),\bar Z_t(\omega ))|\leq |f(t,\omega ,0,0)|+K(|\bar Y_t(\omega )|+\|\bar Z_t(\omega )\|)\) for all t, ω, with \(f(\cdot ,\cdot ,0,0),\bar Y,\bar Z^1,\cdots ,\bar Z^{d}\in \mathcal {L}^2_{uni}\), it is clear that \(\tilde f\in \mathcal {L}^2_{uni}\). Now Eq. A.8 can be considered as a BSDE where the driver does not depend on y and z of the form Eq. A.2. We can therefore apply Lemma A.7 to \(\tilde {f}\) and conclude on the existence of (Y, M, Z1,⋯ , Zd) verifying the three items of the proposition.
It remains to prove now the last assertion of Proposition 3.10. We fix some (s, η). The first item implies that \(Y_s= Y^{s,\eta }_s\)\(\mathbbm {P}^{s,\eta }\) a.s. But since Ys is \(\mathcal {F}^o_s\)-measurable and \(\mathbbm {P}^{s,\eta }(\omega ^s=\eta ^s)= 1\), it also yields that Ys is \(\mathbbm {P}^{s,\eta }\) a.s. equal to the deterministic value Ys(η) hence \(Y^{s,\eta }_s\) is \(\mathbbm {P}^{s,\eta }\) a.s. equal to the deterministic value Ys(η). This also proves that Y is unique because it is given by \(Y:(s,\eta )\longmapsto Y^{s,\eta }_s\). The uniqueness of Z up to zero potential sets is immediate by the third item of the proposition and Definition 3.2. □
Rights and permissions
About this article
Cite this article
Barrasso, A., Russo, F. Decoupled Mild Solutions of Path-Dependent PDEs and Integro PDEs Represented by BSDEs Driven by Cadlag Martingales. Potential Anal 53, 449–481 (2020). https://doi.org/10.1007/s11118-019-09775-x
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11118-019-09775-x