Abstract
Positive recurrence of a d-dimensional diffusion with an additive Wiener process, with switching and with one recurrent and one transient regimes and variable switching intensities is established under suitable conditions. The approach is based on embedded Markov chains.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
Let us consider the process \((X_t,Z_t)\) with a continuous component X and discrete one Z described by the stochastic differential equation in \({\mathbb {R}}^d\)
for the component X, while \(Z_t\) is a continuous-time conditionally Markov process given X on the state space \(S= \{0,1\}\) with positive intensities of respective transitions \( \lambda _{01}(x) =: \lambda _0(x), \, \& \, \lambda _{10}(x) =: \lambda _1(x)\); here the variable x signifies a certain (arbitrary Borel measurable) dependence on the component X; the trajectories of Z are assumed to be càdlàg; the probabilities of jumps for Z are conditionally independent given the trajectory of the component X (see the precise description in what follows). Denote
It is assumed that
These conditions along with the boundedness of the function b in x suffice for the process \((X_t,Z_t)\) to be well-defined. A rigorous construction of the system (X, Z) of this type may be given by the SDE system
where \(\pi ^i_t, \, i=0,1,\) are two Poisson processes with intensities \(\lambda _i(X_t), \, i=0,1\), respectively. More precisely,
where \({\bar{\pi }}^i_t, i=0,1\), are, in turn, two standard Poisson processes with a constant intensity one, independent of the Wiener process \((W_t)\) and of each other, and the time changes
are applied to each of them, respectively.
By virtue of the assumption (2) the equation between the jumps only concerns the diffusion part of the SDE (3), for which it is well-known since [13] that the equation has a pathwise unique strong solution. The jump moments are stopping times with respect to the filtration \((\mathcal{F}_t = \mathcal{F}_t^{W,\pi ^0,\pi ^1}, \, t\ge 0)\), and the position of the system after any jump \((X_\tau ,Z_\tau )\) is uniquely determined by the left limiting values \((X_{\tau -},Z_{\tau -})\):
After any such jump, the diffusion part of the SDE is solved starting from the position \(X_{\tau }\) until the next jump, say, \(\tau '\), of the component Z, and the moment of this next jump is determined by the trajectories of \(\pi ^0_t\) and (or) of \(\pi ^1_t\) and by the intensity \(\lambda _{Z_s}(X_s), s < \tau '\). Since there might be only a finite numbers of jumps on any bounded interval of time, then pathwise (and, hence, also weak) uniqueness follows on \([0,\infty )\). Therefore, the process (X, Z) exists and is markovian. Its (quasi-) generator has a form
where \({\bar{z}} : = 1(z=0)\) (that is, \({\bar{z}}\) is not a z, the other state from \(\{0,1\}\)).
For any \(t>0\) fixed let us define the function
The vector-function \(v(s,x) = (v(s,x,0),v(s,x,1))\) satisfies the system of PDEs
where
Due to the results of [12, Theorem 5.5] (see also [7]), its solution is continuous in the variable s for any bounded and continuous f. Hence, the process is Feller’s (that is, \(\mathbb {E}_{x,z} h(X_t)\) is continuous in x and, of course, bounded for any \(h\in C_b\) and any \(t>0\)). Since the process is Markov and càdlàg, then it is also strong Markov according to the well-known sufficient condition.
The SDE solution is assumed ergodic under the regime \(Z=0\) and transient under \(Z=1\). We are looking for sufficient conditions for positive recurrence of the strong Markov process \((X_t,Z_t)\). Such a problem was considered in [2] for the exponentially recurrent case; for other references see [1, 5, 8, 11], and the references therein. Under weak ergodic and transient conditions the setting was earlier investigated in [15] for the case of the constant intensities \(\lambda _{0}, \lambda _{1}\) (i.e., not depending on x). In [9] and [10] other interesting results about the transience and recurrence for diffusions with switching were established; in particular, examples were given of conditions under which the solutions of the SDEs on the half line with reflection with two transient “pure” regimes are recurrent, and, vice versa, where the solutions of the SDEs with two recurrent “pure” regimes are transient. Here we tackle the case of a combination of one transient and one recurrent regime. In the case of [15] the lengths of intervals between successive jumps of the discrete component were all independent and independent of the Wiener process. In the general case under the consideration in the present paper they are not independent of W via the component X, and, hence, not independent of each other. This difficulty will be overcome with the help of certain comparison arguments.
The paper consists of the sections: Introduction, Main result, Auxiliaries, and Proof of main result.
2 Main result
Theorem 1
Let the drift \(b = (b_+,b_-)\) be bounded and Borel measurable, and let there exist \(r_-, r_+,M>0\) such that
and
Then the process (X, Z) is positive recurrent; moreover, there exists \(C>0\) such that for all \(M_1\) large enough and all \(x \in {\mathbb {R}}^d\) and for \(z=0,1\)
where
Remark 1
On the basis of the theorem 1 it may be proved that the process \((X_t,Z_t)\) has a unique invariant measure (see [4, Sect. 4.4]), and for each nonrandom initial condition x, z there is a convergence to this measure in total variation when \(t\rightarrow \infty \). We leave a rigorous presentation of this claim till next publications for a wider class of models.
3 Auxiliaries
Denote \(\Vert b\Vert = \sup _{x,z}|b(x,z)|\). Let \(M_1 \gg M\) (the value \(M_1\) will be specified later). Let
and
where \(T_n\) for each \(n\ge 1\) is defined by induction as
Let
To prove the theorem it suffices to evaluate from above the value \({\mathbb {E}}_{x,z}\tau \) because \(\tau _{M_1} \le \tau \). Let \(\epsilon >0, q<1\) be positive values satisfying the equality
(see (6)). In the proof of the theorem it suffices to assume \(|x| > M\).
Lemma 1
Under the assumptions of the theorem for any \(\delta >0\) there exists \(M_1\) such that
Let us denote by \(X^i_t, \, i=0,1\) the solutions of the equations
Proof of lemma 1
Let \(Z_0=0\); then \(T_0=0\). We have,
due to the uniqueness of solutions of the SDEs (1) (or (3)) and (10) and because of the property of stochastic integrals [6, Theorem 2.8.2] to coincide almost surely (a.s.) on the set where the integrands are equal. Therefore, we estimate for any \(|x| > M\) with \(z=0\):
Let us fix some \(t_0\), so that
Then
Now, with this \(t_0\) already fixed, by virtue of the boundedness of b there exists \(M_1>M\) such that for any \(|x|\ge M_1\) we get
Similarly, the bound for the second term in (9) follows if we replace the process \(X^0\) by \(X^1\) and the intensity \(\lambda _0\) by \(\lambda _1\). \(\square \)
Lemma 2
If \(M_1\) is large enough, then under the assumptions of the theorem for any \(|x|>M_1\) for any \(k=0,1,\ldots \)
Corollary 1
If \(M_1\) is large enough, then under the assumptions of the theorem for any \(|x|>M_1\) for any \(k=0,1,\ldots \)
and
Proof of lemma 2
1. Recall that \(T_0=0\) under the condition \(Z_0=0\). We have,
In other words, the moment \(T_{2k+1}\) may be treated as “\(T_{1}\) after \(T_{2k}\)”. Under \(Z_0=0\) the process \(X_t\) coincides with \(X^0_t\) until the moment \(T_1\). Hence, we have on \(t\in [0,T_1]\) by Itô’s formula
on the set \((|X_t|> M)\) due to the assumptions (5). Further, since \(1(|X_t| > M) = 1 - 1(|X_t| \le M)\), we obtain
Thus, always for \(|x|>M_1\),
For a fixed \(\epsilon >0\) let us choose \(\delta = {\bar{\lambda }}_0^{-1}\epsilon / (2M \Vert b\Vert + 2r_-) \). Then, since \(|x|>M_1\) implies \(T_1\wedge \tau = T_1\) on \((Z_0=0)\), and since
we get with \(z=0\)
Substituting here x by \(X_{T_{2k}}\) and writing \(\mathbb {E}_{x,z}(\cdot |{{\mathcal {F}}}_{T_{2k}})\) instead of \(\mathbb {E}_{x,z}(\cdot )\), and multiplying by \(1(\tau > T_{2k})\), we obtain the bounds (11) and (12), as required.
Note that the bound (15) follows straightforwardly from
and similarly
2. The condition \(Z_0=1\) implies the inequality \(T_0>0\). We have,
In other words, the moment \(T_{2k+2}\) may be treated as “\(T_{0}\) after \(T_{2k+1}\)”. Under the condition \(Z_0=1\) the process \(X_t\) coincides with \(X^1_t\) until the moment \(T_0\). Hence, we have on \([0,T_0]\) by Itô’s formula
on the set \((|X_t|> M)\) due to the assumptions (5). Further, since \(1(|X_t| > M) = 1 - 1(|X_t| \le M)\), we obtain
Thus, for \(|x|>M_1\) and with \(z=1\) we have,
For a fixed \(\epsilon >0\) let us choose \(\delta = {\underline{\lambda }}_1^{-1}\epsilon / (2M \Vert b\Vert ) \). Then, since \(|x|>M_1\) implies \(T_0\wedge \tau = T_0\) on the set \((Z_0=1)\), we get (recall that \(z=1\))
Substituting here \(X_{T_{2k+1}}\) instead of x and writing \(\mathbb E_{x,z}(\cdot |{{\mathcal {F}}}_{T_{2k+1}})\) instead of \(\mathbb E_{x,z}(\cdot )\), and multiplying by \(1(\tau > T_{2k+1})\), we obtain the bounds (13) and (14), as required. Lemma 2 is proved. \(\square \)
Proof of corollary 1
is straightforward by taking expectations.
Lemma 3
If \(M_1\) is large enough, then under the assumptions of the theorem for any \(k=0,1,\ldots \)
and
Corollary 2
If \(M_1\) is large enough, then under the assumptions of the theorem for any \(k=0,1,\ldots \)
and
Proof of lemma 3
Let \(Z_0=0\); recall that it implies \(T_0=0\). If \(\tau \le T_{2k+1}\), then (16) is trivial. Let \(\tau > T_{2k+1}\). We will substitute x instead of \(X_{T_{2k}}\) for a while, and will be using the solution \(X^1_t\) of the equation
For \(M_1\) large enough, since \(|x|\wedge |X_{T_1}|>M_1\) implies \(T_2\le \tau \), and due to the assumptions (5) the double bound
is guaranteed in the same way as the bounds (13) and (14) in the previous lemma. In particular, it follows that for \(|x|>M_1\)
since \(|X_{T_1}|\le M_1\) implies \(\tau \le T_1\) and \(\mathbb E_{X_{T_1},1} X_{T_2\wedge \tau }^2 - X_{T_1\wedge \tau }^2=0\). So, on the set \(|x|>M_1\) we have with \(z=0\)
Now substituting back \(X_{T_{2k}}\) in place of x and multiplying by \(1(\tau > T_{2k+1}) \), we obtain the inequalities (16) and (17), as required.
For \(Z_0=1\) we have \(T_0>0\), and the bounds (18) and (19) follow in a similar way. Lemma 3 is proved. \(\square \)
Proof of corollary 2
is straightforward by taking expectations. \(\square \)
Lemma 4
Under the assumptions of the theorem for any \(k=0,1,\ldots \)
and
Proof of lemma 4
On the set \(\tau >T_{2k+1}\) we have,
Similarly, on the set \(\tau >T_{2k}\)
On the sets \(\tau \le T_{2k+1}\) and \(\tau \le T_{2k}\), respectively, both sides of the required inequalities equal zero. Lemma 4 follows. \(\square \)
4 Proof of theorem 1
Consider the case \(Z_0=0\) where \(T_0=0\). Since the identity
we have,
Due to the convergence \(T_n\uparrow \infty \), we get by the monotone convergence theorem
By virtue of the corollary 1, we have
Therefore,
By virtue of Fatou’s lemma we get
Note that \(1(\tau> T_{2k+1}) \le 1(\tau > T_{2k})\). So, \(\mathbb P_{x,0}(\tau> T_{2k+1}) \le {\mathbb {P}}_{x,0}(\tau > T_{2k})\). Hence,
Thus,
Therefore, we estimate
So, (21) implies that
Denoting \(\displaystyle c:= \min \left( \frac{1-q}{2q}\, ((2r_++d)+ \epsilon ), \frac{1-q}{2}\, ((2r_--d)- \epsilon )\right) \), we conclude that
So, as \(m\uparrow \infty \), by the monotone convergence theorem we get the inequality
Due to (20), it implies that
as required. Recall that this bound is established for \(|x|>M_1\), while in the case of \(|x|\le M_1\) the left hand side in this inequality is just zero.
In the case of \(Z_0=1\) (and, hence, \(T_0>0\)), we have to add the value \({\mathbb {E}}_{x,z} T_0\) satisfying the bound \({\mathbb {E}}_{x,1} T_0\le {{\underline{\lambda }}}_1^{-1}\) to the right hand side of (22), which leads to the bound (7), as required. Theorem 1 is proved. \(\square \)
Remark 2
The results of the paper may be extended to the equation
with a Borel measurable \(\sigma \) under the assumptions of the existence of a strong solutions, or of a weak solution which is weakly unique (because the strong Markov property is needed), in addition to the standing balance type conditions replacing (5) and (6) (while (4) is still valid): \(a(x,z)=\sigma \sigma ^*(x,z)\) and
with some \(R_-, R_+ >0\), and
where the definitions of \({\underline{\lambda }}_1\) and \({\overline{\lambda }}_0 \) do not change. The proofs will now involve the diffusion coefficient and will use the assumptions (24) and (25), but otherwise will remain the same as in the case of the unit diffusion matrix.
Data availability
Data sharing is not applicable to this article as no datasets were generated or analysed during the current study.
References
Anulova, S., Veretennikov, A.: Exponential Convergence of degenerate Hybrid Stochastic Systems with full Dependence. In: Modern Stochastics and Applications, Springer, Optimization and Its Applications 90, Springer International Publishing Switzerland, 159-174 (2014). https://doi.org/10.1007/978-3-319-03512-3_10
Cloez, B., Hairer, M.: Exponential ergodicity for Markov processes with random switching. Bernoulli 21(1), 505–536 (2015). https://doi.org/10.3150/13-BEJ577
Gikhman, I.I., Skorokhod, A.V.: Stochastic Differential Equations. Springer, Berlin et al (1972)
Khasminskii, R.Z.: Stochastic stability of differential equations, 2nd edn. Springer, Berlin (2012). https://doi.org/10.1134/S0032946012030064
Khasminskii, R.Z.: Stability of regime-switching stochastic differential equations. Probl. Inf. Transm. 48, 259–270 (2012). https://doi.org/10.1134/S0032946012030064
Krylov, N.V.: Introduction to the Theory of Random Processes, AMS (2002)
Krylov, N.V., Safonov, M.V.: A certain property of solutions of parabolic equations with measurable coefficients. Math. USSR-Izv. 16(1), 151–164 (1981)
Mao, X., Yin, G., Yuan, C.: Stabilization and destabilization of hybrid systems of stochastic differential equations. Automatica 43, 264–273 (2007). https://doi.org/10.1016/j.automatica.2006.09.006
Pinsky, R., Scheutzow, M.: Some remarks and examples concerning the transience and recurrence of random diffusions. Annales de l’I.H.P. Probabilités et statistiques 28.4: 519-536 (1992) http://eudml.org/doc/77443
Pinsky, M., Pinsky, R.G.: Transience/Recurrence and Central Limit Theorem Behavior for Diffusions in Random Temporal Environments. Ann. Probab. 21(1), 433–452 (1993). https://doi.org/10.1214/aop/1176989410
Shao, J., Yuan, C.: Stability of regime-switching processes under perturbation of transition rate matrices. Nonlinear Anal. Hybrid Syst 33, 211–226 (2019). https://doi.org/10.1016/j.nahs.2019.02.009
Solonnikov, V.A.: On boundary value problems for linear parabolic systems of differential equations of general form. Proc. Steklov Inst. Math. 83, 1–184 (1965)
Veretennikov, A. Yu.: On strong solutions and explicit formulas for solutions of stochastic integral equations. Math. USSR-Sb. 39(3), 387–403 (1981). https://doi.org/10.1070/SM1981v039n03ABEH001522
Veretennikov, A. Yu.: On polynomial mixing and convergence rate for stochastic difference and differential equations. Theory Probab. Appl. 44(2), 361–374 (2000). https://doi.org/10.1137/S0040585X97977550
Veretennikov, A.Yu.: On Positive Recurrence of One-Dimensional Diffusions with Independent Switching. In: Shiryaev A.N., Samouylov K.E., Kozyrev D.V. (eds) Recent Developments in Stochastic Methods and Applications. ICSM-5 2020. Springer Proc. in Mathem. & Stat., Springer, Cham, vol 371, pp. 242 - 252 (2021). https://doi.org/10.1007/978-3-030-83266-7_18
Author information
Authors and Affiliations
Corresponding author
Additional information
This article is dedicated to István Gyöngy on the occasion of his 70th birthday.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This research was funded by Russian Foundation for Basic Research grant 20-01-00575a.
Rights and permissions
Springer Nature or its licensor holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Veretennikov, A. Positive recurrence of a solution of an SDE with variable switching intensities. Stoch PDE: Anal Comp 10, 1165–1179 (2022). https://doi.org/10.1007/s40072-022-00265-7
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s40072-022-00265-7