Almost periodic solution in distribution for stochastic differential equations with Stepanov almost periodic coefficients Fazia Bedouhene, Nouredine Challali, Omar Mellah, Paul Raynaud de Fitte, Mannal Smaali To cite this version: Fazia Bedouhene, Nouredine Challali, Omar Mellah, Paul Raynaud de Fitte, Mannal Smaali. Almost periodic solution in distribution for stochastic differential equations with Stepanov almost periodic coefficients. 2017. ￿hal-01478199v3￿ HAL Id: hal-01478199 https://hal.archives-ouvertes.fr/hal-01478199v3 Preprint submitted on 24 Oct 2017 HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. Almost periodic solution in distribution for stochastic differential equations with Stepanov almost periodic coefficients Fazia Bedouhene∗, Nouredine Challali†, Omar Mellah ‡, Paul Raynaud de Fitte §, and Mannal Smaali¶, October 24, 2017 Abstract This paper deals with the existence and uniqueness of (µ-pseudo) almost periodic mild solution to some evolution equations with Stepanov (µ-pseudo) almost periodic coefficients, in both determinist and stochastic cases. After revisiting some known conceptsandpropertiesofStepanov(µ-pseudo)almostperiodicityincompletemetric space, we consider a semilinear stochastic evolution equation on a Hilbert separable space with Stepanov (µ-pseudo) almost periodic coefficients. We show existence and uniqueness of the mild solution which is (µ-pseudo) almostperiodic in 2-distribution. We also generalize a result by Andres and Pennequin, according to which there is no purely Stepanov almost periodic solutions to differential equations with Stepanov almost periodic coefficients. Keywords : Weighted pseudo almost periodic; Stepanov almost periodic; Stochastic evolution equations; Pseudo almost periodic in 2-UI distribution 1 Introduction The concept of Stepanov almost periodicity, which is the central issue in this paper, was first introduced in the literature by Stepanov [56], and is a natural generalization of the concept of almost periodicity in Bohr’s sense. Important contributions upon such concept wheresubsequently made by N. Wiener [65], P. Franklin [38], A. S. Besicovitch [11], B. M. Levitan and V. V. Zhikov [45, 46], Amerio and Prouse [1], S.Zaidman [67], A. S. Rao [54], C. Corduneanu [22], L. I. Danilov [26, 27], S. Stoin´ski [58, 57], J. Andres, A. M. Bersani, G. Grande, K. Lesniak [3, 4, 2, 5]. Outside the field of harmonic analysis, a substantial application of Stepanov almost periodicity lies in the theory of differential equations [50]. ∗Mouloud Mammeri University of Tizi-Ouzou, Laboratoire de Math´ematiques Pures et Appliqu´ees, Tizi-Ouzou, Algeria E-Mail: [email protected] ORCID ID 0000-0002-2664-2445 †Mouloud Mammeri University of Tizi-Ouzou, Laboratoire de Math´ematiques Pures et Appliqu´ees, Tizi-Ouzou, Algeria E-Mail: [email protected] ORCID ID 0000-0002-6965-5086 ‡Mouloud Mammeri University of Tizi-Ouzou, Laboratoire de Math´ematiques Pures et Appliqu´ees, Tizi-Ouzou, Algeria E-Mail: [email protected] ORCIDID 0000-0002-3042-9719 §Normandie Univ, Laboratoire Rapha¨el Salem, UMR CNRS 6085, Rouen, France E-Mail: prf@univ- rouen.fr ORCID ID 0000-0001-5527-9393 ¶Mouloud Mammeri University of Tizi-Ouzou, Laboratoire de Math´ematiques Pures et Appliqu´ees, Tizi-Ouzou, Algeria E-Mail: smaali [email protected] ORCID ID 0000-0002-4556-8930 1 In this context, the theory of dynamical systems is pertinent and, in particular, in the study of various kinds and extensions of almost periodic and (or) almost automorphic motions. This is due mainly to their importance and applications in physical sciences. One can mention e.g. T. Diagana [29, 32, 33, 34], J. Blot et al. [17, 15, 16, 13], G. M. N’Gu´er´ekata et al. [51, 52], J. Andres and D. Pennequin [7, 6], Z. Hu and A. B. Mingarelli [40, 41, 39]. Though there has been a significant attention devoted to the theory of Stepanov al- most periodicity in the deterministic case, there are few works related to the notion of Stepanovalmostperiodicityforstochasticprocesses. Toourknowledge, thefirstworkded- icated to Stepanov almost periodically correlated (APC) processes is due to L. H. Hurd and A. Russek [42], where Gladyshev’s characterization of APC correlation functions was extended to Stepanov APC processes. In the framework of Stochastic differential equa- tion, Bezandry and Diagana [12] introduced the concept of Stepanov almost periodicity in mean-square. Their aim was to prove, under some conditions, existence and unique- ness of Stepanov (quadratic-mean) almost periodic solution for a class of nonautonomous stochastic evolution equations on a separable real Hilbertspace. Thispaper was thestart- ing point of other works on stochastic differential equations with Stepanov-like (µ-pseudo) almost periodic (automorphic) coefficients (see, e.g. [21, 20, 66, 60]). Unfortunately, the claimed results are erroneous [9, 49]. The motivation of this paper has two sources. The first one comes from the papers by Andres and Pennequin [7, 6], who show the nonexistence of purely Stepanov-almost periodic solutions of ordinary differential equations in uniformly convex Banach spaces. The second one comes from our paper [9, Example 3.1], where, with the simple counterex- ample of Ornstein-Uhlenbeck process, we have shown that even a one-dimensional linear equation with constant coefficients has no nontrivial solution which is Stepanov almost periodic in mean-square. In this paper, we revisit the question of existence and uniqueness of Stepanov almost periodic solutions in both deterministic and stochastic cases. More precisely, we con- sider two semilinear stochastic evolution equations in a Hilbert space. The first one has Stepanov almost periodic coefficients, and the second one has Stepanov µ-pseudo almost periodic coefficients. We show that each equation has a unique mild solution which is almost periodicin 2-UI distribution in the firstcase, and µ-pseudoalmost periodicin 2-UI distribution in the second case. Our results generalize and complete those of Da Prato and Tudor [23], and those obtained recently by Kamenskii et al. [43], and Bedouhene et al. [9]. We also show, by mean of a new superposition theorem in the deterministic case, the nonexistence of purely Stepanov µ-pseudo almost periodic solutions to some evolution equations, generalizing a result of Andres and Pennequin [7]. The rest of the paper is organized as follows. In Section 2, we investigate several notions of Stepanov almost periodicity in Lebesgue measure, and Stepanov (µ-pseudo) almostperiodicityformetric-valuedfunctions. Weseeinparticularthatalmostperiodicity inStepanov sensedependsontheuniformstructureofthestate space. Specialattention is paidtosuperpositionoperatorsbetweenthespacesofStepanov(µ-pseudo)almostperiodic metric-valued functions. Section 3 is the main part of this paper. Therein, we study existence and uniqueness of bounded mild solutions to the abstract semilinear stochastic evolution equation on a Hilbert separable space dX(t) = AX(t)dt+F(t,X(t))dt+G(t,X(t))dW(t), where F and G are Stepanov (µ-pseudo) almost periodic, satisfying Lipschitz and growth 2 conditions. In the case of uniqueness, the solution can be(µ-pseudo) almost periodic in 2- UI distribution. Our approach is inspired from Kamenskii et al. [43], Da Prato and Tudor [23], and Bedouhene et al. [9]. The major difficulty is the treatment of the limits F∞ and G∞ provided by the Bochner criterion in Stepanov sense applied to F and G respectively. Thanks to an application of Komlo´s’s theorem [44], this difficulty dissipates by showing that F∞ and G∞ inherit the same properties as F and G respectively. Finally, Section 4 is devoted to some remarks and conclusions about the problem of existence of purely Stepanov almost periodic solutions. We show by a simple example in a one-dimensional setting, that one can obtain bounded purely Stepanov almost periodic solutions when the forcing term is purely Stepanov almost periodic in Lebesgue measure. 2 Stepanov Almost periodicity and its variants in metric space In this section, we present the concept of Stepanov (µ-pseudo) almost periodic function andrelated concepts like almost periodicity in Lebesguemeasure. Moreover, wealso recall some useful and key results. We begin with some notations. 2.1 Notations In what follows, (E,d) is a complete metric space. Unless otherwise stated, we keep the notation d to designate the metric of any metric space X. When X is a Banach space, its norm induced by d will be denoted by . . k k LetX andYbetwo complete metric spaces, we denotesomeclassical spaces as follows: C(X,Y), the space of continuous functions from X to Y; • C (X,Y), the space C(X,Y) endowed with the topology of uniform convergence on k • compact subsets of X; BC(X,Y), the space of bounded continuous functions from X to Y; • If Y is Banach space, we denote by CUB R,Y the space BC(R,Y) endowed with • the topology of uniform convergence on R whose norm is noted by . . (cid:0) (cid:1) k k∞ 2.2 Stepanov and Bohr almost periodicity Letusrecallsomedefinitionsof(Stepanov)almostperiodicfunctionsandsomekeyresults. Almost periodicity Recall that a set A R is relatively dense if there exists a real ⊂ numberℓ > 0, suchthatA [a,a+ℓ]= , for alla inR. We say thatacontinuous function ∩ 6 ∅ f :R E is Bohr almost periodic (or simply almost periodic) if for all ε > 0, the set → T(f,ε) := τ R, supd(f(t),f(t+τ)) < ε (cid:26) ∈ t∈R (cid:27) is relatively dense [1, 11, 22]. The numbers τ T(f,ε) are called ε-almost periods. We ∈ denote by AP(R,E) the space of Bohr almost periodic functions. We have the follow- ing criteria for Bohr almost periodicity, established by Bochner [19] for complex-valued functions, see also [23] for metric-valued functions: 3 Theorem 2.1 ([19, 23]) Let f : R E be a continuous function. The following condi- → tions are equivalent: 1. f AP(R,E). ∈ 2. f satisfies Bochner criterion, namely, the set f(t+.), t R is relatively compact { ∈ } in BC(R,E) with respect to the uniform metric. 3. For every pair of sequences (α′ ) R and (β′) R, one can extract common n ⊂ n ⊂ subsequences (α ) (α′ ) and (β ) (β′) such that n ⊂ n n ⊂ n lim lim f(t+α +β ) = lim f(t+α +β ) (1) n m n n n→∞m→∞ n→∞ pointwise. If condition 3 holds, we say that f satisfies Bochner’s double sequence criterion. This criterion turns out to be a very powerful tool in applications to differential equations, see for instance [23]. Stepanov almost periodicity In all the sequel, unless stated otherwise, p denotes a realnumber,withp 1. Following [25,28],letM(R,E)bethesetofmeasurablefunctions ≥ from R to E (we do not distinguish between functions that coincide almost everywhere for Lebesgue’s measure). We fix a point x in E. We denote by Lp(R,E), the subset of 0 M(R,E) of locally p-integrable functions, that is, Lp(R,E) = f M(R,E), for any a,b R; dp(f(t),x )dt < + . 0 ∈ ∈ ∞ n [aZ,b] o Define Lp(0,1;E) as the set Lp(0,1;E) = f M(R,E), dp(f(t),x )dt < + 0 ∈ ∞ n [0Z,1] o which is a complete metric space, when it is endowed with the metric 1/p Lp(f,g) =  dp(f(t),g(t))dt . D Z [0,1]     We denote by L∞(R,E) the space of all E-valued essentially bounded functions, endowed with essential supremum metric. Obviously, all the previous spaces do not depend on the choice of the point x E. 0 ∈ We say that a locally p-integrable function f : R E is Stepanov almost periodic of → order p or Sp-almost periodic, if, for all ε > 0, the set SpT(f,ε) := τ R, Dd (f(.+τ),f (.)) ε Sp ∈ ≤ n o is relatively dense, where, for any locally p-integrable functions f,g : R E, → x+1 1/p Dd (f,g) = sup dp(f(t),g(t))dt . Sp x∈R(cid:18)Zx (cid:19) 4 ThespaceofSp-almostperiodicE-valuedfunctionsisdenotedbySpAP(R,E). LetS∞AP(R,E) be the space of functions f L∞(R,E) such that for any ε > 0, there exists a relatively ∈ dense Td (f,ε) such that L∞ d (f(.+τ),f (.)) ε, for allτ Td (f,ε). DL∞ ≤ ∈ L∞ The relation lim Dd (f,g) = Dd (f,g) holds for any functions f,g M(R,E), see [7] p→∞ Sp L∞ ∈ for the proof. As in the case of Bohr-almost periodic functions, we have similar characterizations of Stepanov almost periodic functions. More precisely, let f Lp(R,E), p [1,+ [. Then, the following state- ∈ ∈ ∞ ments are equivalent (compare with [41, Theorem 1 and Proposition 6]): f is Sp-almost periodic. • f is Sp-almost periodic in Bochner sense, that is, from every real sequence (α′ ) R • n ⊂ one can extract a subsequence (α ) of (α′ ) and there exists a function g Lp(R,E) n n ∈ such that lim Dd (f(.+α ),g(.)) = 0. Sp n n→+∞ f satisfies Bochner’s type double sequence criterion in Stepanov sense, that is, for • every pair of sequences α′ R and β′ R, there exist common subsequences { n} ⊂ { n} ⊂ (α ) (α′ ) and (β ) (β′) such that, for every t R, the limits n ⊂ n n ⊂ n ∈ lim lim f(t+α +β ) and lim f(t+α +β ), (2) n m n n n→∞m→∞ n→∞ exist and are equal, in the sense of the Lp-metric 1/p d (h(t+.),g(t+.)) = dp(h(t+s),g(t+s))ds DLp Z[0,1] ! for h,g Lp(R,E). ∈ The equivalence between these statements was originally established in [41] in the context of Banach spaces. However, onecanprovideasimplerproofof theequivalence ofthethree previous items based on the following observation that the concept of Stepanov almost periodicity can be seen as Bohr almost periodicity of some function with values in the Lebesgue space Lp(0,1;E). More precisely, let fb denote the Bochner transform [18] of a function f Lp(R,E): ∈ R E[0,1] fb : → t f(t+.). (cid:26) 7→ Then f SpAP(R,E) if, and only if, fb AP(R,Lp(0,1;E), and the previous equivalences ∈ ∈ become a simple consequence of the fact that f f if and only if fb fb (see e.g. n → n → [1, 5, 8, 46]). Since functions in SpAP(R,E) are bounded with respect to the Stepanov metric, one denotes by Sp(R,E) (or Sp(R) when E = R, and S(R,E) when p = 1) the set of all Dd -bounded functions, that is, for some (or any) fixed x E, Sp 0 ∈ Sp(R,E) = f M(R,E);Dd (f,x ) < + . Sp 0 { ∈ ∞} 5 So, from now on, the space SpAP(R,E) will be seen as a (closed) subset of the complete metric space (Sp(R,E),Dd ). We have the following inclusions: Sp AP(R,E) S∞AP(R,E) SpAP(R,E) SqAP(R,E) S1AP(R,E) S(R,E) ⊂ ⊂ ⊂ ⊂ ⊂ for p q 1 and AP(R,E) = SpAP(R,E) (R,E), where (R,E) denotes the set of u u ≥ ≥ ∩C C E-valued uniformly continuous functions on R. FormorepropertiesanddetailsaboutrealandBanach-valuedStepanovalmostperiodic functions, we refer the reader for instance to the papers and monographs [1, 2, 8, 11, 22, 37, 45, 46]. Beside the previous characterizations of the class SpAP(R,E), there is an other one based on the concept of Stepanov almost periodicity in (Lebesgue) measure, invented by Stepanov [56]. This concept plays a significant role in the proof of our superposition theorem in SpAP(R,E) (Theorem 2.10). Stepanov almost periodicity in Lebesgue measure : the space S0AP For any measurable set A R, let ⊂ κ(A) = supmeas([ξ,ξ+1] A), ξ∈R ∩ where meas is the Lebesgue measure. A measurable function f : R E is said to be → Stepanov almost periodic in Lebesgue measure or S0-almost periodic if for any ε, δ > 0, the set Tκ(f,ε,δ) := τ R, supmeas t [ξ,ξ +1],d f(t+τ),f(t) ε < δ ( ∈ ξ∈R ∈ ≥ ) (cid:8) (cid:0) (cid:1) (cid:9) is relatively dense. It should be mentioned that this almost periodicity was introduced as µ-almost periodicity [57]. We denote by S0AP(R,E) (S0AP(R) when E = R) the space of such fonctions. This space was studied in depth by several authors (in both normed and metric spaces). One can quote Stoinski’s works [57, 59], where an approximation property and some compactness criterion are given. Danilov [25, 26, 28] has explored this class in the framework of almost periodicmeasure-valued functions. Therecently publishedpaper [53], that we discovered at the time of writing this paper, completes the previous ones. The authors of this paper investigate some other properties, in particular, they show that in general the mean value of S0-almost periodic functions may not exist, furthermore, S0-almost periodic functions are generally not Stepanov-bounded. As pointed out by Danilov [26], S0-almost periodicity coincides with classical Stepanov almost periodicity when replacing the metric d by d′ = min(d,1). In other words, we have the following characterization (see [26, 27]) S0AP(R,E) = S1AP R,(E,d′) = SpAP R,(E,d′) , p > 0. (3) ∀ More generally, Stepanov almost pe(cid:0)riodicity c(cid:1)an be see(cid:0)n as S0-al(cid:1)most periodicity under a uniformintegrability condition inStepanov sense(see[25,38,53,56]). Tobemoreprecise, let M′(R,E) be the set of Dd -bounded functions such that p Sp lim sup sup dp f(t),x dt = 0. (4) 0 δ→0+ξ∈RT⊂[ξ,ξ+1]ZT measT≤δ (cid:0) (cid:1) 6 The space M′(R,E), p 1, is a closed subset of (Sp(R,E),Dd ). In [26, page 1420], p ≥ Sp Danilov gives an elegant characterization of Stepanov almost periodic functions in terms of M′(R,E) and S0AP(R,E), more precisely: p SpAP(R,E) = S0AP(R,E) M′(R,E). (5) ∩ p A rather interesting result about the space S0AP(R,E) is reported in the following theorem[27,Theorem3],whichgivesauniformapproximationofStepanovalmostperiodic functions by Bohr almost periodic functions, in the context of normed space E. Before, let us denote by (R) the collection of measurable sets T R such that the indicator S ⊂ function of T, 1l , belongs to S1AP(R), and by Tc the complementary set of T. T Theorem 2.2 (Danilov [27]) Let f S0AP(R,E), then for any δ > 0, there exist a set ∈ T (R) and a Bohr almost periodic function F such that κ(Tc) < δ and f(t) = F (t) δ ∈ S δ δ δ for all t T . δ ∈ As consequence, we have the following corollary: Corollary 2.3 Let f S0AP(R,E). Then, for all ε > 0, there exist a measurable set ∈ T (R) and a compact subset K of E such that κ(Tc)< ε and f(t) K , t T . ε ∈ S ε ε ∈ ε ∀ ∈ ε Danilov has shown that this property remains valid even in the metric framework [28]. Remark 2.4 1. Unlike almost periodicity in Bohr sense and almost periodicity in Lebesque measure for function with values in a metric space (E,d), which depend only on thetopological structureof Eand noton its metric (see e.g., [10]and [26]re- spectively), Stepanov almost periodicity is a metric property. In fact, as the metrics dandd′ = min(d,1) aretopologically equivalent on E,weonlyneed toshowthatthe inclusion S1AP R,(E,d) S1AP R,(E,d′) is strict, since in view of (3), we have ⊂ S1AP R,(E,d′) = S0AP(R,E). Consider the example given in [7, Remark 3.3]. As (cid:0) (cid:1) (cid:0) (cid:1) shown by the authors, the function g = exp +∞ g , where g is the 4n-periodic (cid:0) (cid:1) n=2 n n function given by (cid:0)P (cid:1) 2 g (t) = β 1 t n 1l (t), t [ 2n,2n], n n(cid:18) − αn | − |(cid:19) [n−α2n,n+α2n] ∈ − with α = 1/n5 and β = n3, is not in S1AP(R). Using Danilov’s Corollary [26], we n n get that g belongs to S0AP(R), as a superposition of a continuous function and a periodic, continuous and bounded function. 2. Still in the spirit of the link between the spaces S1AP(R) and S0AP(R), an inter- esting property established by Stoin´ski says that the inverse of any trigonometric polynomial with constant sign is S0-almost periodic. Inparticular, the Levitan func- 1 tion f : R R given by f(t) = is S0-almost periodic but not → 2+cos(t)+cos(2t) Stepanov almost periodic (see Example 4.1 and [53] for the second statement). 3. Uniform integrability in Stepanov sense is a metric property, that is, the space M′(R,(E,d)) depends on the metric d. p 7 Bohr and Stepanov almost periodic functions depending on a parameter Here- inafter, some definitions of Bohr and Stepanov almost periodicity for metric-valued para- metric functions are presented. Such definitions are simple adaptation of the well-known ones in the literature, see in particular [9, 17, 31, 35, 47]. 1. Wesaythataparametricfunctionf : R X Yisalmost periodic withrespecttothe × → first variable, uniformly with respect to the second variable in bounded subsets of X (respectively in compact subsets of X) if, for every bounded (respectively compact) subset B of X, the mapping f : R C(B,Y) is almost periodic. We denote by → APU (R X,Y) and APU (R X,Y) respectively the spaces of such functions. b c × × 2. We say that a function f : R X Y is Sp-almost periodic if, for every x X, the × → ∈ Y-valued function f(.,x) is Sp-almost periodic. We denote by SpAP(R X,Y) the × space of such functions. 3. Letf SpAP(R X,Y). Wesaythatf isSp-almostperiodicuniformlywithrespect ∈ × to the second variable in compact (resp. bounded) subsets of X if f(.,x) is Sp- almost periodic uniformly with respect to x K for any compact (resp. bounded) ∈ subset K of X. The space of such functions is denoted by SpAPU (R X,Y) (resp. c × SpAPU (R X,Y)). b × Clearly, we have the following inclusions: SpAPU (R X,Y) SpAPU (R X,Y) SpAP(R X,Y) Lp(R X,Y), b c × ⊂ × ⊂ × ⊂ × where Lp(R X,Y) denotes the set of measurable functions f :R X Y such that, for × × → all x X; f(.,x) Lp(R,Y). ∈ ∈ The following proposition will be very useful in the sequel. Proposition 2.5 Let Y be a complete metric space, and let X be a complete separable metric space. Let f SpAP(R X,Y) satisfying the following Lipschitz condition: ∈ × d(f(t,x),f(t,y)) K(t)d(x,y), t R, x,y X, (6) ≤ ∀ ∈ ∈ for some positive function K(.) in Sp(R). Then for every real sequence (α′ ), there exist n a subsequence (α ) (α′ ) (independent of x) and a function f∞ SpAP(R X,Y) such n ⊂ n ∈ × that for every t R and x X, we have ∈ ∈ t+1 lim d(f(s+α ,x),f∞(s,x))pds = 0. (7) n n Zt Proof Firstly, let us show that f∞ is Lipschitz with respect to the second variable in the Stepanov metric sense. Let x,y X. We consider a real sequence (α′ ) R. Since f SpAP(R X,Y), for every x X,∈we can find a subsequence (α ) (αn′ )⊂(depending ∈ × ∈ n ⊂ n on x) such that limDd (f(.+α ,x) f∞(.,x)) = 0. (8) Sp n n − For the same reason, for every y X, there exists a subsequence of (α ), (depending on n ∈ both x and y still noted (α ) for simplicity) such that n limDd (f(.+α ,y),f∞(.,y)) = 0. (9) Sp n n 8 Then, by (6), (8) and (9), we get Dd (f∞(.,x),f∞(.,y)) limDd (f∞(.,x),f(.+α ,x))+limDd (f(.+α ,x),f(.+α ,y)) Sp Sp n Sp n n ≤ +limDd (f(.+α ,y),f∞(.,y)) Sp n K Spd(x,y). (10) ≤ k k Secondly, let us show (7). Let (α′ ) be a real sequence. Since X is separable, let D n be a dense countable subset of X. Using (8) and a diagonal procedure, we can find a subsequence (α ) of (α′ ) such that for every t R and x D, we have n n ∈ ∈ t+1 lim (d(f(s+α ,x),f∞(s,x)))pds = 0 (11) n n Zt Let x X, there exists a sequence (x ) D such that lim d(x ,x) = 0. From (6), we k k k ∈ ⊂ deduce t+1 lim (d(f(s+α ,x ),f(s+α ,x)))pds = 0, (12) n k n k Zt uniformly with respect to n N. Now from (11), we obtain, for every t R and k N, ∈ ∈ ∈ t+1 lim (d(f(s+α ,x ),f∞(s,x )))pds = 0. (13) n k k n Zt Using (12), (13) and by a classical result on interchange of limits, we deduce limf(s+α ,x) = limf∞(s,x ) = f∞(s,x) n k n k in Stepanov metric. The last equality follows from (10). 2.3 Bohr and Stepanov weighted pseudo almost periodic functions with values in metric space The notions of Stepanov-like weighted pseudo almost periodicity and Stepanov pseudo almost periodicity of functions, with values in Banach space X, were introduced by T. Diagana [29, 30, 31] as natural generalizations of the pseudo almost periodicity invented by Zhang [68, 69]. Here we give the definitions of these different notions for functions with values in a complete metric space E. C. and M. Tudor [62] have proposed an elegant definition of pseudo almost periodicity in the context of metric spaces, which is slightly restrictive, since it requires compactness of the range of the function instead of its boundedness. A more general definition of weighted pseudo almost periodicity (automorphic) has been introduced in [9], where it is shown that there is no need to assume that E is a vector space, nor a metric space, and these notions depend only on the topological structure of E. The definition we propose here (see Definition 2.6) is an intermediate between that of C. and M. Tudor and that in the wide sense [9, Proposition 2.5 (i)]. It coincides with the one existing in the literature when E is a normed space [15, 16]. We begin by recalling thedefinition of µ-ergodicity for vector-valued functions[15,16]. Let (X; . ) be a Banach space. Let µ be a Borel measure on R satisfying k k µ(R) = and µ(I)< for every bounded interval I. (14) ∞ ∞ 9