Almost periodic functions in finite-dimensional space with the spectrum in a cone 7 0 Favorov S., Udodova O.1 0 2 n a Kharkiv National University J Department of Mechanics and Mathematics 8 md. Svobody 4, Kharkiv 61077, Ukraine 2 e-mail: [email protected], [email protected] ] V Abstract C We prove that an almost periodic function in finite-dimensional space extends to a holomorphic . h bounded function in a tube domain with a cone in the base if and only if the spectrum belongs to t the conjugate cone. We also prove that an almost periodic function in finite-dimensional space has the a boundedspectrum if and only if it extendsto an entirefunction of exponentialtype. m 2000 Mathematics Subject Classification 42A75, 30B50. [ 1 A continuous function f(z) on a strip v 0 S ={z =x+iy :x∈R, a≤y ≤b}, −∞≤a≤b≤+∞. a,b 2 8 is called almost periodic by Bohr on this strip, if for any ε>0 there exists l =l(ε) such that every interval 1 of the real axis of length l contains a number τ (ε-almost period for f(z)) with the property 0 7 sup |f(z+τ)−f(z)|<ε. (1) 0 z∈Sa,b / h In particular, when a=b=0 we obtain the class of almost periodic functions on the real axis. t a To each almost periodic function f(z) assign the Fourier series m ∞ v: an(y)eiλnx, λn ∈R, Xi nX=0 r where an(y) are continuous functions of the variable y ∈[a,b]. a In the case a = b = 0 all exponents λ are nonnegative if and only if the function f(x) extends to the n upper half-plane asa holomorphic boundedalmostperiodic function; the setof allexponents λ is bounded n if and only if f(x) extends to the plane C as an entire function of the exponential type σ =sup|λ |, which n n is almost periodic in every horizontal strip of finite width (see [1], [4]). A number of works connected with almost periodic functions of many variables on a tube set appeared recently (see [2], [6]-[9]). Recall that the set T ⊂Cm is a tube set if K T ={z =x+iy :x∈Rm,y ∈K}, K where K ⊂Rm is the base of the tube set. Definition. (See [6], [9]). A continuous function f(z), z ∈ T is called almost periodic by Bohr on T , if K K for any ε > 0 there exists l = l(ε), such that every m-dimensional cube on Rm with the side l contains at least one point τ ((T ,ε)-almost period from f(z)) with the property K sup |f(z+τ)−f(z)|<ε. (2) z∈TK 1TheworkwassupportedbytheINTAS-grant99-00089 1 Let f, g be locally integrable functions on every real plane {z =x+iy :x∈Rm}, y ∈K. 0 0 Definition. Stepanoff distance of the order p≥1 between functions f and g is the value 1 p S (f,g)= sup |f(z+u)−g(z+u)|pdu . p,TK z∈TK[0,Z1]m Using this definition, we can extend the concept of almost periodic functions by Stepanoff on a strip (see [4] p.197) to almost periodic functions on a tube set: Definition. A function f(z), z ∈ T , is called almost periodic by Stepanoff on T , if for any ε > 0 there K K exists l = l(ε) such that every m-dimensional cube with the side l contains at least one τ ((T ,ε,p)-almost K period by Stepanoff of the function f(z)) with the property S (f(z),f(z+τ))<ε. (3) p,TK The Fourier series for an almost periodic (by Bohr or by Stepanoff) function f(z) on a set T is the K series a(λ,y)eihx,λi, (4) λX∈Rm where hx,λi is the scalar product on Rm, and m a(λ,y)= lim 1 f(x+x′+iy)e−ihx+x′,λidx; (5) N→∞(cid:18)2N(cid:19) Z [−N,N]m this limit exists uniformly in the parameter x′ ∈Rm and does not depend on this parameter (see [6], [8]). A set of all vectors λ∈Rm such that a(λ,y)6≡0 is called the spectrum of f(z) and is denoted by spf; this set is at most countable, therefore the series (4) can be written in the form ∞ a (y)eihx,λni. n nX=0 Notethatpartialsumsoftheseries(4),generallyspeaking,donotconvergetothefunctionf(z). However the Bochner-Feyer sums 1 q−1 σq(z)= knqan(y)eihx,λni, 0≤knq <1, knq →1 as q →∞ nX=0 convergetothefunctionf(z)uniformlyforalmostperiodicfunctionsbyBohrandwithrespecttothemetric S foralmostperiodic functionsbyStepanoff; inparticular,iftwofunctions havethe sameFourierseries, p,TK then these functions coincide identically. For holomorphic almost periodic functions the series (4) can be written in the form ∞ ∞ ane−hy,λnieihx,λni = aneihz,λni, an ∈, (6) nX=0 nX=0 (see [8]). Any series of the form (6) is called Dirihlet series . By Γ we always denote a convex closed cone in Rm; by Γ we denote the conjugate cone to Γ: Γ={t∈Rm :ht,yi≥b0 ∀y ∈Γ}, ◦ b note that Γ=Γ. Also, Γ is the interior of a cone Γ. 1Forn=bb1see[4],forn>1see[8]. 2 Theorem 1. Let f(x) be an almost periodic function by Bohr on Rm with Fourier series ∞ aneihx,λni, (7) nX=0 where all the exponents λ belong to a cone Γ⊂Rm. Then f(x) continuously extends to the tube set T as n Γ an almost periodic by Bohr function F(z) with Fourier series (6). The function F(z) is holomorphic on the interior TΓ, and for any Γ′ ⊂⊂Γ uniformly w.r.t. z ∈TΓ′ b b b lim F(z)=a0, (8) kyk→∞ where a is the Fourier coefficient corresponding to the exponent λ = 0 (if 0 ∈/ sp f, put a = 0.) If 0 0 0 ◦ spf ⊂Γ, then (8) is true uniformly w.r.t. z ∈T . Γ Here the inclusion Γ′ ⊂⊂ Γ means that thebintersection of Γ′ with the unit sphere is contained in the interior of the intersection of Γ with this sphere. b To prove this theorem, we use the following lemmas. b Lemma 1. Supposethat aplurisubharmonic functionϕ(z)on Cm is boundedfrom aboveon asetT , where K K ⊂Rm is a convex set. Then the function ψ(y)= sup ϕ(x+iy) x∈Rm is convex on K. Proof of Lemma 1. Fix y , y ∈K. The plurisubharmonic on Cm function 1 2 ψ(y )−ψ(y ) 2 1 ϕ (z)=ϕ(z)− hImz,y −y i 1 ky −y k2 2 1 2 1 is boundedfromaboveonthe setT . Thereforethe subharmonicfunctionϕ (w)=ϕ ((y −y )w+iy ) [y1,y2] 2 1 2 1 1 is bounded on the strip {w =u+iv :u∈ R, 0 ≤ v ≤1}. Hence, the value of ϕ at any point of this strip 1 does not exceed max{supϕ (u),supϕ (u+i)}≤max{ sup ϕ (x+iy ), sup ϕ (x+iy )}. 2 2 1 1 1 2 u∈R u∈R x∈Rm x∈Rm Therefore, for any z =x+iy ∈T , [y1,y2] ψ(y )−ψ(y ) ψ(y )−ψ(y ) 2 1 2 1 ϕ (z)≤max{ψ(y )− hy ,y −y i,ψ(y )− hy ,y −y i} 1 1 ky −y k2 1 2 1 2 ky −y k2 2 2 1 2 1 2 1 ky k2−hy ,y i ky k2−hy ,y i 2 1 2 1 1 2 = ψ(y )+ ψ(y ). ky −y k2 1 ky −y k2 2 2 1 2 1 Hence for any y ∈[y ,y ] we have 1 2 ky k2−hy ,y i−hy,y −y i ky k2−hy ,y i+hy,y −y i 2 1 2 2 1 1 1 2 2 1 ψ(y)≤ ψ(y )+ ψ(y ). ky −y k2 1 ky −y k2 2 2 1 2 1 If y =λy +(1−λ)y , λ∈(0,1), then we obtain the inequality 1 2 ψ(λy +(1−λ)y )≤λψ(y )+(1−λ)ψ(y ). 1 2 1 2 Therefore, the function ψ(y) is convex on K. Lemma 2. Let ψ(y) be a convex bounded function on a cone Γ. Then ψ(y)≤ψ(0) for all y ∈Γ. 3 Proof of Lemma 2. Since ψ(y) is convex, we have 1 1 ψ(y)≤ 1− ψ(0)+ ψ(ty), t>1. (cid:18) t(cid:19) t Taking t→∞, we obtain ψ(y)≤ψ(0). Proof of Theorem 1. Let σ (x), q = 0,1,2,... be the Bochner-Feyer sums for the series (7). Obviously, q these functions are also defined for z ∈Cm. Assume that ϕ (z)=log(|σ (z)−σ (z)|). q,l q l For any fixed q and l, q >l, the function ϕ (z) is plurisubharmonic on Cm. Moreover,for z ∈T we have q,l Γ hy,λ i≥0 and n |σq(z)−σl(z)|≤|σq(z)|+|σl(z)|≤ b q−1 l−1 q−1 |an|e−hy,λni+ |an|e−hy,λni ≤2 |an|. nX=0 nX=0 nX=0 Consider the function ψ (y)= sup log(|σ (z)−σ (z)|). q,l q l x∈Rm Using lemma 1, we obtain that ψ (y) is convex in Γ. Therefore, by lemma 2, we have q,l b sup(|σ (z)−σ (z)|)≤ sup (|σ (x)−σ (x)|). (9) q l q l z∈T x∈Rm Γ Further, the function f(x) is almobst periodic, therefore the Bochner-Feyersums convergeuniformly on Rm, and for q,l≥N(ε) sup (|σ (x)−σ (x)|)≤ε. q l x∈Rm Hence, sup (|σ (z)−σ (z)|)≤ε for all z ∈T , q,l≥N(ε). q l x∈Rm Γ Thus the Bochner-Feyer sums uniformly converge on T , their limit is an almost periodic function by b Γ Bohr and holomorphic on the interior of T with the Dirihlet series (6). Γ Further, passing to the limit in (9) as q →∞, we get b b sup(|F(z)−σ (z)|)≤ sup (|f(x)−σ (x)|). l l z∈T x∈Rm Γ Choose l such that the right handbside of this inequality is less than ε. We have for Γ′ ⊂⊂Γ b sup |F(z)−a |≤ sup |F(z)−σ (z)|+ sup |σ (z)−a |≤ε+ sup |σ (z)−a |. 0 l l 0 l 0 z∈TΓ′ z∈TΓ′ z∈TΓ′ z∈TΓ′ Note that for any fixed λ ∈Γ\{0} the value hy,λ i tends to +∞ as kyk→∞, y ∈Γ′, therefore, we have n n l−1 l−1 |σl(z)−a0|=| kjlajeihx,λjie−hy,λji|≤ |aj|e−hy,λji →0 Xj=1 Xj=1 as kyk→∞ on Γ′. Hence, uniformly w.r.t. z ∈TΓ′ lim |F(z)−a |≤ε. (10) 0 kyk→∞ This is true for arbitraryε>0, then (8) follows. ◦ If spf ⊂ Γ, then for any λn ∈spf, hy,λni→+∞ as kyk→∞ uniformly w.r.t. y ∈Γ, therefore (10) is true uniformly w.r.t. z ∈T , and (8) is also true. The theorem has been proved. Γ b b 4 Theorem 2. Let f(x) be an almost periodic function by Stepanoff on Rm with the Fourier series (7). Let all the exponents λ belong to a cone Γ ⊂ Rm. Then there exists an almost periodic by Stepanoff function n F(z) in the tubeset T with the Fourier series (6) such that F(x)=f(x). The function F(z) is holomorphic Γ ◦ almost periodic by Bohr on any domain T , b∈Γ. Besides, for any cone Γ′ ⊂⊂Γ we have uniformly w.r.t. b Γ+b z ∈TΓ′ b b b lim F(z)=a0, (11) kyk→∞ ◦ where a0 is the Fourier coefficient for the exponent λ = 0. If spf ⊂Γ, then (11) is true uniformly w.r.t. ◦ z ∈T for any b∈Γ. Γ+b b Prboof. To prove the first part of the theorem, we need to replace ϕq,l(z) by 1 p ϕ (z)=log |σ (z+u)−σ (z+u)|pdu . q,l q l Z [0,1]m b Arguingasintheproofoftheorem1,weobtainthattheBochner-Feyersumsσ (z)convergeintheStepanoff q metric uniformly w.r.t. z ∈T to an almost periodic function by Stepanoff F(z) with Fourier series (6). Γ ◦ Let b ∈Γ. The module of the function σ (z)−σ (z) is estimated from above by the mean value on the b q l corresponding ball contained in T . Using the Ho¨lder inequality, we have b Γ 1 b p sup |σ (x+bi)−σ (x+bi)|≤C sup |σ (z+u)−σ (z+u)|pdu , q l q l x∈Rm z∈T Z Γ[0,1]m b where the constant C depends only on b and Γ. Applying lemmas 1 and 2 to the functions b ψ˜ (y)= sup log|σ (z+bi)−σ (z+bi)|, q,l,b q l x∈Rm wegetthatthe Bochner-Fouriersums convergeuniformly onT to F(z),thusF(z)isholomorphicalmost Γ+b ◦ periodic by Bohr in T for any b∈Γ. b Γ+b Then the other statements of the theorem follow from theorem 1. b b Now we prove the inverse statements to theorems 1 and 2. Theorem 3. Suppose that an almost periodic by Bohr function f(x) continuously extends to the interior of TΓ as a holomorphic function F(z). If F(z) is bounded on any set TΓ′, Γ′ being a the cone in Rm, Γ′ ⊂⊂Γ, then F(z) is an almost periodic function by Bohr on T and the spectrum of F(z) is contained in Γ. Γ ◦ b Proof. Take λ∈/ Γ. Then there exists y0 ∈ Γ such that hy0,λi<0. ◦ ChooseaneighboburhoodU ⊂Γofy0 suchthathy,λi≤ 12hy0,λiforally ∈U.LetAbeanynondegenerate operator in Rm such that A maps all the vectors e =(1,0,...,0),...,e =(0,0,...,1) into U. 1 m The function F(Aζ) is holomorphic and bounded on the set {ζ =ξ+iη ∈Cm :ξ ∈Rm, ηj >0,j =1,...,m} because A{η :ηj ≥0}⊂Γ. If for each coordinates ζ1,...,ζm we change the integration over the segments −N ≤ξj ≤N, ηj =0 to the integration over the half-circles ζj =Neiθj, 0≤θj ≤π, j =1,...,m we obtain the equality m m m 1 F(Aξ)e−ihAξ,λidξ = i F(ANeiθ) eiθj−iNeiθjhAej,λidθ, (12) (cid:18)2N(cid:19) Z (cid:18)2(cid:19) Z jY=1 [−N,N]m [0,π]m 5 where θ =(θ1,...,θm), eiθ =(eiθ1,...,eiθm). Since hAe ,λi<0 for j =1,...,m, we see that the integrand in the right-hand side of (12) is uniformly j bounded for all N >1. By Lebesgue theorem (12) tends to zero as N →∞. Thus 1 lim f(x)e−ihx,λidx=0. (13) N→∞(2N)m Z A([−N,N]m) CoverthesetA([−N2,N2]m)bycubesL =x′ +[−N,N]m suchthattheinteriorsofthesecubesarenotin- j j tersected. WemayassumethatthenumberofthecubesintersectingtheboundaryofthesetA([−N2,N2]m) is O(Nm−1) as N → ∞. Taking into account boundedness of the function f(x)e−ihx,λi on Rm and equal- ity (13), we have 1 1 f(x)e−ihx,λidx= f(x)e−ihx,λidx+O(N2m−1)=o(1) (14) (2N)2m Z (2N)2m Z ∪Lj A([−N2,N2]m) Since (5) we see that uniformly w.r.t. x′ ∈Rm as N →∞ 1 f(x)e−ihx,λidx=a(λ,f)+o(1). (15) (2N)m Z x′+[−N,N]m On the other hand, the number of the cubes L equals O(Nm) as N →∞, then the equality a(λ,f)=0 j follows from (14) and (15). This yields the inclusion spf ⊂ Γ. Using theorem 1 we complete the proof of our theorem. b Theorem 4. If F(z) is bounded on each set TΓ′, Γ′ ⊂⊂ Γ, and the nontangential limit value of F(z) as y → 0 is an almost periodic function by Stepanoff on Rm, then F(z) extends to T as an almost periodic Γ function by Stepanoff and the spectrum of F(z) is contained in Γ. Proof. The proof of this theorem is the same as of theoremb3, but we have to use theorem 2 instead of theorem 1. To formulate further results we need the concept of P-indicator. ( See, for example, [5], p. 275.) Definition. P-indicator of an entire function F(z) on Cm is the function 1 h (y)= sup lim log|F(x+iry)|. F x∈Rmr→∞r Theorem 5. (For m = 1 see [3], [4].) Let f(x), x ∈ Rm be an almost periodic function by Stepanoff with the Fourier series (7), and let kλ k ≤ C < ∞ for all n. Then f(x) extends to Cm as an entire function n F(z) of exponential type, which is almost periodic by Bohr on any tube domain in Cm with bounded base; F(z) has the Dirichlet series (6), and P-indicator h (y) satisfies the equation h (y) = H (−y), where F F spf H (µ):= sup hx,µi is the support function of the set spf. spf x∈spf Proof. Take µ∈Rm such that kµk=1. Put fµ(x)=f(x)e−i[Hspf(µ)+ε]hx,µi. ∞ The Fourier series aneihx,λn−(Hspf(µ)+ε)µi corresponds to the function fµ(x), hence nP=0 spf ⊂{x∈Rm :hx,µi ≤−ε}. µ Since spf is bounded, we obtain for some δ >0 µ spf (x)⊂Γ ={λ∈Rm :hλ,−µi≥δkλk}. µ δ,−µ 6 Theorem 2 yields that f (x) extends to the interior of the domain T , where µ Γδ,−µ Γ ={y :hy,−µi≥ 1−δ2kybk} δ,−µ p is the conjugate cone to Γ , as anbalmost periodic function by Bohr F (z). This function is holomorphic δ,−µ µ ◦ on any domain T , b∈Γ with the Dirichlet series Γδ,−µ+b δ,−µ b b ∞ aneihz,λn−[Hspf(µ)+ε]µi, nX=0 and Fµ(z)→0 as kyk→∞ uniformly w.r.t. z ∈TΓ′ for any cone Γ′ ⊂⊂ Γδ,−µ. Using (5) we get b ane−hy,λn−[Hspf(µ)+ε]µi ≤ sup |Fµ(x+iy)|, y ∈Γ′. (16) (cid:12) (cid:12) x∈Rm (cid:12) (cid:12) (cid:12) (cid:12) Put F(z):=Fµ(z)ei[Hspf(µ)+ε]hz,µi. F(z) is almost periodic on TΓ′ with Dirichlet series (6). Therefore it follows from (16) that |an|≤ sup |F(x+iy)|ehy,λni. (17) x∈Rm On the other hand, the function Fµ(z) is bounded on TΓ′, hence |F(z)|≤C(Γ′)e−[Hspf(µ)+ε]hy,µi, z ∈TΓ′ (18) Coverthe space Rm by the interiors of a finite number of cones Γ′,...,Γ′ . There exist holomorphic on 1 N the interior of Γ′ almost periodic functions F (z), k = 1,...,N with identical Dirichlet series (6). Using k k the uniqueness theorem, we obtain that these functions coincide on the intersections of the cones and thus defineaholomorphicfunctionF(z)onCm\Rm. TheBochner-FeyersumsforF(z)convergetothisfunction uniformly on any set {z =x+iy :x∈Rm, kyk=r >0}. Hence, these sums converge on the tube domain T . Thus F(z) extends to Cm as the holomorphic {kyk<r} function, which is almost periodic on any tube set with a bounded base. Owing to the uniqueness of expansion into Fourier series, we have F(x)=f(x). Let us prove that h (y)=H (−y). From inequality (18) with µ=−y it follows that F spf 1 h (y)≤ lim [H (−y)+ε]hry,yi=H (−y)+ε. F spf spf r→∞r The functions h (y) and H (y) are positively homogenious, hence the inequality F spf h (y)≤H (−y) F spf is true for all y ∈Rm. Further, fix x,y ∈ Rm. The holomorphic on C function ϕ(w) = F(x+wy) is bounded on the axis Imw =0. Then the estimate |ϕ(w)|≤Cea|Imw| for some a>0 and all w∈C follows from (18). Using the definition of P-indicator, we get 1 lim log|ϕ(iv)|≤h (y). F v→+∞v Therefore the function ϕ(w)ei(hF(y)+ε)w is bounded on the positive part of the imaginary axis. Applying the Fragmen-Lindelof principle to the quadrants Rew ≥ 0, Imw ≥ 0 and Rew ≤ 0, Imw ≥ 0, we get boundedness of this function on the upper half-plane. Applying the Fragmen-Lindelof principle to the half-plane Imw ≥0, we get the inequality |ϕ(w)|≤ sup |ϕ(w)| ehF(y)Imw (Imw>0). (cid:18) (cid:19) Imw=0 7 Hence, for all z ∈Cm, we have |F(z)|≤ sup |F(x)|ehF(y). x∈Rm Now using formula (17) for coefficients of the Dirichlet series of the function F(z), we get the estimate |an|≤ sup |f(x)|ehF(y)+hy,λni. (19) x∈Rm Suppose hy ,λ i+h (y ) < 0 for some y ∈Rm. Put y =ty in (19) and let t→∞. We obtain a =0. 0 n F 0 0 0 n This is impossible because λ ∈spf. n Thus for all y ∈Rm and λ ∈spf we have h (y)+hy,λ i≥0, hence n F n H (−y)= sup h−y,λ i≤h (y). spf n F λn∈spf This completes the proof of the theorem. The following theorem is inverse to the previous one. Theorem 6. (For m = 1 see [3], [4].) Let F(z) be an entire function on Cm, |F(z)| ≤ Cebkzk, let F(x), x ∈ Rm be an almost periodic function by Stepanoff with the Fourier series (7). Then F(z) is an almost periodic function by Bohr on any tube domain T ⊂ Cm with the bounded base, F(z) has the Dirichlet D series (6), and spF ⊂ {λ:kλk≤b}. Proof. It follows from theorem 5, that it suffices to prove the inclusion spF ⊂{λ:kλk≤b}. Let the function F(x) be bounded on Rm. Arguing as in theorem 5, we see that for all z ∈Cm |F(z)|≤ sup |F(x)|ehF(y), x∈Rm where h (y) is P-indicator for F(z). Further, for all x∈Rm we have F 1 1 h (y)= sup lim log|F(x+iry)|≤ sup lim (logC+bkx+iryk)≤bkyk, F x∈Rmr→∞r x∈Rmr→∞r therefore for all z ∈Cm |F(z)|≤Cebkyk. Take ε>0, µ∈Rm, kµk=1. Consider the function F (z)=F(z)e−ihz,µi(b+ε). µ Since |F (z)| ≤ Cebkyk+(b+ε)hy,µi uniformly w.r.t. x ∈ Rm, then F (z) is uniformly bounded for z ∈ T , µ µ Γ−µ where Γ is the cone {y : hy,−µi ≥ (1− ε )kyk}. Using theorem 4, we obtain that the spectrum F is −µ b+ε µ contained in Γ and −µ spF =spF +(b+ε)µ⊂Γ +(b+ε)µ. b µ −µ Finally, using the inclusion b (Γ +(b+ε)µ)⊂{λ:kλk≤b+ε} −µ µ:k\µk=1 b andthe arbitrarityofchoiceofε we getthe assertionofthe theoreminthe case ofboundedonRm function F(z). Now let the function F(z) be unbounded on Rm. Put for some N >0 1 g(z)= F(z+t)dt. Nm Z [0,N]m The function g(z) satisfies the estimate on Cm |g(z)|≤CebmNebkzk. 8 Asinthecasem=1(see [4]),wecanprovethatg(x)isanalmostperiodicfunctionbyBohrandisbounded on Rm. The function g(x) has the Fourier series ∞ a eiλ1nN −1...eiλmnN −1eihx,λni, n Nλ1 Nλm nX=0 n n where λj are coordinatesof the vectorλ (if λj =0, the correspondingmultiplier shouldbe replacedby 1). n n n Using countabilityofspF,we canchooseN in sucha waythatnone ofthe numbersλjN coincideswith n 2πk, k ∈ Z\{0}. In this case spg = spF. Applying the proved above statement to the function g(z), we obtain the inclusion spF ⊂{λ:kλk≤b}. REFERENCES [1] BohrH. Zur Theorie der fastperiodischen Funktionen. III Teil: Dirichletentwicklung analytischerFunk- tionen.//Acta math. – 47. – 1926. – P.237-281. [2] FavorovS.Yu.,RashkovskiiA.Yu.andRonkinL.I.Almostperiodiccurrentsandholomorphicchains.// C. R. Acad. Sci. Paris.– Serie I. – 327. – 1998. – P.302-307. [3] Levin B.Ya. Almost periodic functions with the bounded spectrum.//Sb. Actual’nyje voprosy matem- aticheskogo analisa. – Published by Rostov university. – 1978. – P.112-124.(Russian) [4] Levitan B.M. Almost periodic functions.– M.: Gostehizdat. – 1953. – 396 p. (Russian) [5] Ronkin L.I. Introduction to the theory of entire functions of several variables. – M.: ”Nayka”. – 1971. – 430 p. (Russian) [6] Ronkin L.I. CR-functions and holomorphic almost periodic functions with an integer base.//at. Fiz. Anal. Geom. – .4. – 1997.– P.472-490.(Russian) [7] Ronkin L.I. On a certain class of holomorphic almost periodic functions.//Sb.Mat. – 33. – 1992. – P.135-141.(Russian) [8] RonkinL.I.Almostperiodicdistributionsintube domains.–Zap.Nauchn.Sem.POMI.–247.–1997. – P.210-236.(Russian) [9] RonkinL.I.Jessentheoremforholomorphicalmost-periodicfunctionsintubedomains.//Sb.Mat–28. – 1987. – P.199-204.(Russian) 9