This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Multisummability of formal solutions for a family of generalized singularly perturbed moment differential equations

Alberto Lastra Departamento de Física y Matemáticas
University of Alcalá
Ap. de Correos 20, E-28871 Alcalá de Henares (Madrid), Spain
[email protected]
Sławomir Michalik Faculty of Mathematics and Natural Sciences, College of Science
Cardinal Stefan Wyszyński University
Wóycickiego 1/3, 01-938 Warszawa, Poland
[email protected] http://www.impan.pl/~slawek
 and  Maria Suwińska Faculty of Mathematics and Natural Sciences, College of Science
Cardinal Stefan Wyszyński University
Wóycickiego 1/3, 01-938 Warszawa, Poland
[email protected]
Abstract.

The notion of moment differentiation is extended to the set of generalized multisums of formal power series via an appropriate integral representation and accurate estimates of the moment derivatives.

The main result is applied to characterize generalized multisummability of the formal solution to a family of singularly perturbed moment differential equations in the complex domain, broadening widely the range of singularly perturbed functional equations to be considered in practice, such as singularly perturbed differential equations and singularly perturbed fractional differential equations.

Key words and phrases:
multisummability, formal solution, moment estimates, moment derivatives, moment differential equations, singular perturbation
2020 Mathematics Subject Classification:
34K26, 34K41, 34E10, 34M30

1. Introduction

The main aim of this work is to give a step forward in the theory of summability of formal solutions to functional equations in the complex domain. More precisely, we deal with the so-called moment (partial) differential equations, in which the operators known as moment derivatives act on the unknown function.

The main advances in the present study are twofold: on the one hand, we provide an integral representation of the moment derivatives of generalized sums of formal power series which can be extended to an infinite sector satisfying some generalized exponential growth. In addition to this, we describe the dependence of such moment derivatives with respect to three elements, namely the moment sequence, the sequence involved in the asymptotic representation of the sum, and also the generalized exponential growth at infinity (Theorem 2). Consequently, a novel definition of moment derivatives acting on generalized multisums of formal power series makes sense, and induces many possible applications in the theory, as the following one. On the other hand, we apply the previous result to achieve the main result of this research, namely the multisummability of the formal solutions to certain family of singularly perturbed moment differential equations (Theorem 3). More precisely, we prove that the formal solution to

(1) {εka(z)m2,zpω(z,ε)ω(z,ε)=f^(z,ε)m2,zjω(0,ε)=ψ^j(ε),j=0,,p1,\left\{\begin{aligned} \varepsilon^{k}a(z)\partial_{m_{2},z}^{p}\omega(z,\varepsilon)-\omega(z,\varepsilon)&=\hat{f}(z,\varepsilon)\\ \partial_{m_{2},z}^{j}\omega(0,\varepsilon)&=\hat{\psi}_{j}(\varepsilon),\qquad j=0,\ldots,p-1,\end{aligned}\right.

is multisummable along a certain appropriate multidirection (d1,d2)2(d_{1},d_{2})\in\mathbb{R}^{2} with respect to the perturbation parameter ε\varepsilon if and only if the forcing term f^(z,ε)[[z,ε]]\hat{f}(z,\varepsilon)\in\mathbb{C}[[z,\varepsilon]] and the initial conditions ψ^j(ε)[[ε]]\hat{\psi}_{j}(\varepsilon)\in\mathbb{C}[[\varepsilon]], for 0jp10\leq j\leq p-1 are multisummable along the same multidirection. Here, ε\varepsilon stands for a small complex parameter, a(z)a(z) is a holomorphic function near the origin, m2m_{2} is a sequence of moments, and k,pk,p are positive integers with k<pk<p. The precise description of the elements involved in the problem is given in Section 5.

The study of moment differential equations is motivated by the versatility of the moment derivative operator, whose definition was initially put forward by W. Balser and M. Yoshino in [2] for formal power series. Let m=(m(p))p0m=(m(p))_{p\geq 0} be a sequence of positive real numbers. Then, the moment derivative m,z:[[z]][[z]]\partial_{m,z}:\mathbb{C}[[z]]\to\mathbb{C}[[z]] is defined by m,z(zp)=m(p)m(p1)zp1\partial_{m,z}(z^{p})=\frac{m(p)}{m(p-1)}z^{p-1} for every positive integer pp and m,z(1)=0\partial_{m,z}(1)=0, defining moment derivatives for elements in [[z]]\mathbb{C}[[z]] by linearity. It is natural to extend the previous definition to a holomorphic function on some disc at the origin by identifying the function with its Taylor series at z=0z=0. In addition to this, the authors proved that one can extend the definition of moment differentiation to the generalized sums in a direction of a formal power series, as the sum of the moment derivative of that formal power series (see Corollary 1 and Definition 10, [13]). In this work, we further extend it to the generalized multisums of a formal power series (Corollary 1 and Definition 10).

The most suitable choice for mm is to be a moment sequence of certain Laplace-like operator. It is clear that the sequence m1=(p!)p0m_{1}=(p!)_{p\geq 0} gives rise to the usual derivative when considering the moment operator m1,z\partial_{m_{1},z}. Apart from that choice for mm, many other derivations which appear in concrete applications can be represented as a moment derivative. For example, for every k>0k>0, the sequence m1/k=(Γ(1+p/k))p0m_{1/k}=(\Gamma(1+p/k))_{p\geq 0} is associated with the Caputo fractional derivative z1/k\partial_{z}^{1/k} via the relation (m1/k,zf^)(z1/k)=z1/k(f^(z1/k))(\partial_{m_{1/k},z}\hat{f})(z^{1/k})=\partial_{z}^{1/k}(\hat{f}(z^{1/k})), valid for every f^(z)[[z]]\hat{f}(z)\in\mathbb{C}[[z]] (see [16], Remark 3, for further details). Recently, many applications of Caputo derivatives appear in the literature such as [3, 8], also in the study of the asymptotic periodic solutions of evolution equations [21], numerical studies, etc. Let q(0,1)q\in(0,1). The sequence mq=([p]q!)p0m_{q}=([p]_{q}!)_{p\geq 0}, with [p]!q=[1]q[2]q[p]q[p]!_{q}=[1]_{q}[2]_{q}\cdots[p]_{q} is known as the sequence of qq-factorials, where []q=j=01qj[\ell]_{q}=\sum_{j=0}^{\ell-1}q^{j}. It determines the moment differentiation which coincides with the qq-derivative Dq,zD_{q,z} given by Dq,zzp=[p]qzp1D_{q,z}z^{p}=[p]_{q}z^{p-1} for every pp\in\mathbb{N}. This moment differentiation is quite related to the dilation operator, appearing in the study of qq-difference equations which is of great interest in the scientific community with interesting advances in the knowledge of the asymptotic behavior of the solutions of qq-difference equations (see [9, 15] among others, and the references therein).

The interest of moment functional equations has increased in the last decade, and recent achievements have been reached in this concern. A first step was given in the seminal work [2], where the authors study the formal solutions and Gevrey estimates of their coefficients of linear moment partial differential equations with constant coefficients. The development of a more general theory through the construction of kernels for generalized summability by J. Sanz in [22] allows to enlarge the class of moment sequences considered, in the framework of strongly regular sequences. The sequence m1/km_{1/k} above belongs to such a family for every k>0k>0, whereas mqm_{q} does not. The so-called 1+1+ level, appearing in the asymptotic study of difference equations is also related to a strongly regular sequence [4, 5].

After the seminal work [2], the second author gave answer to the problem of analyticity of such problems [16], via splitting of the characteristic equation with respect to one of its variables. We also refer to [17] for a further study in the homogeneous situation, while dropping the condition of the convergence of the initial data. Further knowledge on the solutions to moment partial differential equations with constant coefficients is given in [18] to study the growth properties and summability of the formal solutions. It is also worth mentioning the family of partial differential equations studied in [10], where the coefficients of the equation under study belong to certain functional spaces associated with functions whose derivatives are uniformly bounded in terms of some strongly regular sequence.

The last advances in the theory of the asymptotic behavior of solutions to moment functional equations have been obtained recently regarding the summability of certain families of moment integro-differential equations [12], and also Maillet-type theorems [11, 24]. We mention the recent works by P. Remy in the study of partial differential equations [19] and integro-differential equations [20] of a similar nature as those considered in these works.

Recent results on generalized multisummability of formal power series concerning different (comparable and nonequivalent) levels associated with ultraholomorphic classes achieved in [6] are applied in the present study to achieve asymptotic properties of the solutions to a singularly perturbed moment differential equation. That concept of multisummability as long as previous results achieved by the authors in [13] have been the key points used to describe generalized multisummability of the formal solution of the main equation (1). As mentioned above, novel integral representations and accurate estimates of the moment derivatives of generalized sums of formal power series are needed, arriving to the coherent definition of moment derivation of the generalized multisum of a formal power series. More precisely, the first main result of the present study, Theorem 2, resorts to an appropriate deformation path which is split in order to provide upper bounds for the moment derivative of a sectorial holomorphic function, quite related to the multisummability process. As a consequence, Corollary 1 and Definition 10 encompass the notion of moment derivation of the multisum of a formal power series along some multidirection. As an application of these results, Section 5 characterizes multisummability of the formal solution to (1) in terms of that of the initial data and the forcing term (Theorem 3). The proof of this last result is based on the properties of formal Borel transform which transform the problem into an auxiliary moment partial differential equation, which is easier to handle.

The paper is structured as follows. We fix notation in Section 2, followed by Section 3, where we recall the definition and main results on strongly regular sequences, generalized summability and multisummability of formal power series. In that section, some technical results needed in the sequel are also proved. The main purpose of Section 4 is to state Theorem 2, leading to a coherent definition of moment derivative of the generalized multisum of a formal power series. Section 5 describes an application (Theorem 3) of the previous results in the framework of singularly perturbed moment differential equations.

2. Notation

By \mathbb{N} we shall denote the set of all positive integers, i.e., {1,2,}\{1,2,\dots\} and 0={0}\mathbb{N}_{0}=\mathbb{N}\cup\{0\}.

\mathcal{R} stands for the Riemann surface of the logarithm.

For all r>0r>0 and z0z_{0}\in\mathbb{C}, D(z0,r)D(z_{0},r) stands for the open disc in the complex plane centered at z0z_{0} and with radius rr. For any fixed θ>0\theta>0 and dd\in\mathbb{R} a subset of \mathcal{R} defined as

Sd(θ)={z:|argzd|<θ2}S_{d}(\theta)=\left\{z\in\mathcal{R}:\ |\arg z-d|<\frac{\theta}{2}\right\}

is an open infinite sector with vertex at the origin, bisecting direction dd and opening θ\theta. In cases where the opening is not specified, we simply write SdS_{d}. For every r>0r>0, we write Sd(θ;r):=Sd(θ)D(0,r)S_{d}(\theta;r):=S_{d}(\theta)\cap D(0,r). A sectorial region Gd(θ)G_{d}(\theta) is a subset of \mathcal{R} such that there exists r>0r>0 for which Gd(θ)Sd(θ;r)G_{d}(\theta)\subset S_{d}(\theta;r) and for any 0<θ<θ0<\theta^{\prime}<\theta there exists 0<r<r0<r^{\prime}<r such that Sd(θ;r)Gd(θ)S_{d}(\theta^{\prime};r^{\prime})\subset G_{d}(\theta).

We put S^d(θ;r):=Sd(θ)D(0,r)\hat{S}_{d}(\theta;r):=S_{d}(\theta)\cup D(0,r). Analogously, we write S^d(θ)\hat{S}_{d}(\theta) (resp. S^d\hat{S}_{d}) whenever the radius r>0r>0 (resp. the radius and the opening r,θ>0r,\theta>0) can be omitted. We write SSd(θ)S\prec S_{d}(\theta) if SS is an infinite sector with vertex at the origin such that S¯Sd(θ)\bar{S}\subset S_{d}(\theta), where the closure is considered with respect to \mathcal{R}. Similarly, for two sectorial regions Gd(θ)G_{d}(\theta) and Gd(θ)G_{d^{\prime}}(\theta^{\prime}) we write Gd(θ)Gd(θ)G_{d}(\theta)\prec G_{d^{\prime}}(\theta^{\prime}) whenever Gd(θ)Gd(θ)G_{d}(\theta)\subset G_{d^{\prime}}(\theta^{\prime}) and relation \prec holds for the sectors appearing in the definitions of both sectorial regions.

If (𝔼,𝔼)(\mathbb{E},\|\cdot\|_{\mathbb{E}}) is a complex Banach space, by 𝒪(U,𝔼)\mathcal{O}(U,\mathbb{E}) we denote the set of all functions holomorphic on the open set UU\subset\mathbb{C} with values from 𝔼\mathbb{E}. For 𝔼=\mathbb{E}=\mathbb{C} we simply write 𝒪(U)\mathcal{O}(U). The set of all formal power series in tt with coefficients in 𝔼\mathbb{E} is denoted by 𝔼[[z]]\mathbb{E}[[z]].

3. Preliminary results and definitions

The purpose of this section is to recall the main facts on the theory of generalized summability and also one of the equivalent notions of generalized multisummability of a formal power series along certain multidirection, developed in [6]. We first remind the main elements regarding the theory of strongly regular sequences and related properties together with the theory of generalized summability, for the sake of completeness. These definitions and the detailed constructions can be found in [23] and the references therein.

3.1. Strongly regular sequences

The concept of strongly regular sequences was put forward by V. Thilliez in [25].

Definition 1.

Let 𝕄=(Mp)p0\mathbb{M}=(M_{p})_{p\geq 0} be a sequence of positive real numbers with M0=1M_{0}=1. 𝕄\mathbb{M} is a strongly regular sequence if the following statements hold:

  • (lc)

    Mp2Mp1Mp+1M_{p}^{2}\leq M_{p-1}M_{p+1}, p1p\geq 1 (𝕄\mathbb{M} is logarithmically convex).

  • (mg)

    there exists A1>0A_{1}>0 such that Mp+qA1p+qMpMqM_{p+q}\leq A_{1}^{p+q}M_{p}M_{q}, for all p,q0p,q\geq 0 (𝕄\mathbb{M} is of moderate growth).

  • (snq)

    there exists A2>0A_{2}>0 such that qpMq(q+1)Mq+1A2MpMp+1\sum_{q\geq p}\frac{M_{q}}{(q+1)M_{q+1}}\leq A_{2}\frac{M_{p}}{M_{p+1}}, for all p0p\geq 0 (𝕄\mathbb{M} satisfies the strong non-quasianalyticity condition).

The previous notion generalizes that of Gevrey sequences of order α>0\alpha>0, (p!α)p0(p!^{\alpha})_{p\geq 0}, which widely appear in the theory of summability of formal solutions to functional equations. In association with a strongly regular sequence 𝕄\mathbb{M}, one can define the function

(2) M(t):={supp0log(tpMp) for t>00 for t=0M(t):=\left\{\begin{aligned} \sup_{p\geq 0}\log\left(\frac{t^{p}}{M_{p}}\right)&\quad\textrm{ for }t>0\\ \\ 0\qquad\quad&\quad\textrm{ for }t=0\end{aligned}\right.

It turns out that MM is a non-decreasing and continuous function in [0,)[0,\infty), with limtM(t)=+\lim_{t\to\infty}M(t)=+\infty. We also consider the positive real number

ω(𝕄):=(limsuprmax{0,log(M(r))log(r)})1,\omega(\mathbb{M}):=\left(\lim\sup_{r\to\infty}\max\left\{0,\frac{\log(M(r))}{\log(r)}\right\}\right)^{-1},

which determines the limit opening for a sector to admit nontrivial flat ultraholomorphic functions defined on them. We refer to [7] for a deeper study in this direction.

Following [22, 23, 25], one has the next results.

Lemma 1 ((17), [22]).

For every H>0H>0, there exist C,D>0C,D>0 such that for all p0p\geq 0 one has

0tp1exp(M(t/H))𝑑tCDpMp.\int_{0}^{\infty}t^{p-1}\exp(-M(t/H))dt\leq CD^{p}M_{p}.
Lemma 2.

Let 𝕄\mathbb{M} be a strongly regular sequence, and let s1s\geq 1. There exists ρ(s)1\rho(s)\geq 1 (only depending on 𝕄\mathbb{M} and ss) such that

exp(M(t))exp(sM(t/ρ(s))),t0.\exp(-M(t))\leq\exp(-sM(t/\rho(s))),\qquad t\geq 0.
Lemma 3.

Let 𝕄=(Mp)p0\mathbb{M}=(M_{p})_{p\geq 0} be a strongly regular sequence. The sequence 𝕄s:=(Mps)p0\mathbb{M}^{s}:=(M_{p}^{s})_{p\geq 0} defines a strongly regular sequence for every s>0s>0. Moreover, ω(𝕄s)=sω(𝕄)\omega(\mathbb{M}^{s})=s\omega(\mathbb{M}).

Lemma 4.

Let 𝕄\mathbb{M} be a sequence satisfying (lc) property. Then

  • (Mp1/p)p0(M_{p}^{1/p})_{p\geq 0} is nondecreasing.

  • MpMqMp+qM_{p}M_{q}\leq M_{p+q} for all p,q0p,q\in\mathbb{N}_{0}.

3.2. Generalized summability

In this subsection, (𝔼,𝔼)(\mathbb{E},\left\|\cdot\right\|_{\mathbb{E}}) stands for a complex Banach space.

The classical summability theory of formal power series related to Gevrey sequences (see for example [1, 14]) has recently been adapted to the more general settings involving strongly regular sequences (see [22, 23]). This notion leans on the approximation of holomorphic functions in sectors of the complex plane by formal power series whenever the approximation is given in terms of a given strongly regular sequence.

Definition 2.

Let 𝕄=(Mp)p0\mathbb{M}=(M_{p})_{p\geq 0} be a sequence of positive real numbers, and let Gd(θ)G_{d}(\theta)\subseteq\mathcal{R} be a sectorial region, for some θ>0\theta>0 and dd\in\mathbb{R}. A function f𝒪(Gd(θ),𝔼)f\in\mathcal{O}(G_{d}(\theta),\mathbb{E}) admits the formal power series f^(z)=p0fpzp𝔼[[z]]\hat{f}(z)=\sum_{p\geq 0}f_{p}z^{p}\in\mathbb{E}[[z]] as its 𝕄\mathbb{M}-asymptotic expansion in Gd(θ)G_{d}(\theta) if for every 0<θ<θ0<\theta^{\prime}<\theta and r>0r>0 with Sd(θ;r)Gd(θ)S_{d}(\theta^{\prime};r)\subseteq G_{d}(\theta) and all integer N1N\geq 1, there exist C,A>0C,A>0 with

f(z)p=0N1fpzp𝔼CANMN|z|N,\left\|f(z)-\sum_{p=0}^{N-1}f_{p}z^{p}\right\|_{\mathbb{E}}\leq CA^{N}M_{N}|z|^{N},

for all zSd(θ;r)z\in S_{d}(\theta^{\prime};r).

Further details on the following result can be found in Section 3 [23].

Lemma 5.

In the situation of the previous definition, there exist C~,A~>0\tilde{C},\tilde{A}>0 such that

fp𝔼C~A~pMp\left\|f_{p}\right\|_{\mathbb{E}}\leq\tilde{C}\tilde{A}^{p}M_{p}

for every p0p\geq 0.

The exponential growth in sectors of the complex plane is extended in terms of the function M()M(\cdot) as follows.

Definition 3.

Let θ>0\theta>0 and dd\in\mathbb{R}, and suppose that 𝕄\mathbb{M} is a fixed sequence of positive real numbers. We define the set 𝒪𝕄(Sd(θ),𝔼)\mathcal{O}^{\mathbb{M}}(S_{d}(\theta),\mathbb{E}) as consisting of all functions f𝒪(Sd(θ),𝔼)f\in\mathcal{O}(S_{d}(\theta),\mathbb{E}) such that for every 0<θ<θ0<\theta^{\prime}<\theta there exist constants C,K>0C,K>0 satisfying

(3) f(z)𝔼Cexp(M(|z|K)) for every zSd(θ).\|f(z)\|_{\mathbb{E}}\leq C\exp\left(M\left(\frac{|z|}{K}\right)\right)\textrm{ for every }z\in S_{d}(\theta^{\prime}).

The construction of operators involved in the summability process leans on the existence of kernel functions for generalized summability, related to a given strongly regular sequence.

Definition 4.

Let 𝕄\mathbb{M} be a strongly regular sequence with ω(𝕄)<2\omega(\mathbb{M})<2 and with function M()M(\cdot) defined as in (2). Two complex functions e,Ee,E are strong kernel functions for 𝕄\mathbb{M}-summability if the following properties hold:

  • e𝒪(S0(ω(𝕄)π))e\in\mathcal{O}(S_{0}(\omega(\mathbb{M})\pi)). There exists α>0\alpha>0 such that for all bounded proper subsectors TT of S0(ω(𝕄π))S_{0}(\omega(\mathbb{M}\pi)), there exists C>0C>0 with

    (4) |e(z)|C|z|α,zT.|e(z)|\leq C|z|^{\alpha},\qquad z\in T.

    Furthermore, ee is a flat function in every infinite subsector of S0(ω(𝕄))S_{0}(\omega(\mathbb{M})). More precisely, for every ε>0\varepsilon>0 there exist C,K>0C,K>0 such that

    (5) |e(z)|Cexp(M(|z|K)) for every zS0(ω(𝕄)πε).|e(z)|\leq C\exp\left(-M\left(\frac{|z|}{K}\right)\right)\quad\hbox{ for every }z\in S_{0}(\omega(\mathbb{M})\pi-\varepsilon).

    We also assume that e(x)e(x)\in\mathbb{R} for every real x>0x>0.

  • E𝒪()E\in\mathcal{O}(\mathbb{C}) with generalized exponential growth at infinity

    (6) |E(z)|c~exp(M(|z|k~))for every z,|E(z)|\leq\tilde{c}\exp\left(M\left(\frac{|z|}{\tilde{k}}\right)\right)\quad\textrm{for every }z\in\mathbb{C},

    for some c~,k~>0\tilde{c},\tilde{k}>0. There also exists β>0\beta>0 such that for all 0<θ~<2πω(𝕄)π0<\tilde{\theta}<2\pi-\omega(\mathbb{M})\pi and ME>0M_{E}>0, there exists c~2>0\tilde{c}_{2}>0 with

    (7) |E(z)|c~2|z|β,zSπ(θ~)D(0,ME).|E(z)|\leq\frac{\tilde{c}_{2}}{|z|^{\beta}},\quad z\in S_{\pi}(\tilde{\theta})\setminus D(0,M_{E}).
  • Functions ee and EE are connected by the moment function associated with ee defined by

    (8) me(z):=0tz1e(t)𝑑t.m_{e}(z):=\int_{0}^{\infty}t^{z-1}e(t)dt.

    mem_{e} is a holomorphic function on {z:Re(z)>0}\{z\in\mathbb{C}:\hbox{Re}(z)>0\}, continuous up to its boundary. Indeed, EE is determined from ee via the sequence of moments associated with ee, (me(p))p0(m_{e}(p))_{p\geq 0}, by

    (9) E(z)=p0zpme(p),z.E(z)=\sum_{p\geq 0}\frac{z^{p}}{m_{e}(p)},\quad z\in\mathbb{C}.
Lemma 6 (Proposition 5.7, [22]).

Given a kernel function ee for 𝕄\mathbb{M}-summability, associated with some strongly regular sequence 𝕄\mathbb{M}, the sequence of moments (me(p))p0(m_{e}(p))_{p\geq 0} and 𝕄\mathbb{M} are equivalent, i.e., there exist C,D,C~,D~>0C,D,\tilde{C},\tilde{D}>0 such that

CDpme(p)MpC~D~pme(p),p0.CD^{p}m_{e}(p)\leq M_{p}\leq\tilde{C}\tilde{D}^{p}m_{e}(p),\qquad p\geq 0.

As a matter of fact, the classical kernels for summability involved in the Gevrey theory satisfy weaker properties (see [1]). These more restrictive conditions are justified (see Section 4.2, [6]) regarding their applicability and adaptability to practical situations. Indeed, given a strongly regular sequence, the existence of a pair of kernels for 𝕄\mathbb{M}-summability is guaranteed whenever 𝕄\mathbb{M} admits a nonzero proximate order. We refer to Section 2.3 [6] for a brief review on sequences admitting a nonzero proximate order. This property will turn into an assumption for every strongly regular sequence under consideration hereinafter.

The classical formal Borel transform can also be adapted to this framework.

Definition 5.

Let (me(p))p0(m_{e}(p))_{p\geq 0} be a sequence of moments. Then the formal mem_{e}-Borel moment transform ^me,z:𝔼[[z]]𝔼[[z]]\hat{\mathcal{B}}_{m_{e},z}\colon\mathbb{E}[[z]]\to\mathbb{E}[[z]] is given by

^me,t(p0upzp)=p0upme(p)zp.\hat{\mathcal{B}}_{m_{e},t}\left(\sum_{p\geq 0}u_{p}z^{p}\right)=\sum_{p\geq 0}\frac{u_{p}}{m_{e}(p)}z^{p}.

An 𝕄\mathbb{M}-analog of Laplace transform is also available (see Section 6, [23]).

Proposition 1.

Let dd\in\mathbb{R}. Given a strongly regular sequence 𝕄\mathbb{M} which admits a nonzero proximate order, and a pair of kernel functions for 𝕄\mathbb{M}-summability associated, say ee and EE, we define for every f𝒪𝕄(Sd,𝔼)f\in\mathcal{O}^{\mathbb{M}}(S_{d},\mathbb{E}) the ee-Laplace transform of ff along direction τarg(Sd)\tau\in\hbox{arg}(S_{d}) by

(Te,τf)(z)=0(τ)e(u/z)f(u)duu,(T_{e,\tau}f)(z)=\int_{0}^{\infty(\tau)}e(u/z)f(u)\frac{du}{u},

for all |arg(z)τ|<ω(𝕄)π/2|\hbox{arg}(z)-\tau|<\omega(\mathbb{M})\pi/2, and small enough |z||z|. The variation of τ\tau among the arguments of SdS_{d} determines a holomorphic function, denoted Te,dfT_{e,d}f, defined in a sectorial region of bisecting direction dd and opening larger than ω(𝕄)π\omega(\mathbb{M})\pi.

As a matter of fact, there exists a generalization to the classical Borel-Laplace procedure for the effective summation of a given formal power series.

Definition 6.

Let 𝕄\mathbb{M} be a strongly regular sequence which admits a nonzero proximate order and let mem_{e} denote a sequence of moments associated with 𝕄\mathbb{M}. The series u^𝔼[[z]]\hat{u}\in\mathbb{E}[[z]] is 𝕄\mathbb{M}-summable along direction dd\in\mathbb{R} if ^me,z(u^(z))\hat{\mathcal{B}}_{m_{e},z}(\hat{u}(z)) is a series with a positive radius of convergence, and the analytic function defining such series, say u(z)u(z), can be extended to an infinite sector of bisecting direction dd, say S^d\hat{S}_{d}, with u(z)𝒪𝕄(S^d,𝔼)u(z)\in\mathcal{O}^{\mathbb{M}}(\hat{S}_{d},\mathbb{E}).

Proposition 2.

In the situation of the previous definition, the function v(z)=(Te,du)(z)v(z)=(T_{e,d}u)(z) is holomorphic on a bounded sector of bisecting direction dd and opening larger than ω(𝕄)π\omega(\mathbb{M})\pi.

Definition 6 does not depend on the kernel functions for 𝕄\mathbb{M}-summability (and therefore on the moment sequence) considered. In addition to this, the procedure described there provides us with the only function (due to Watson’s Lemma, see Corollary 3.16, [7]) admitting the initial formal power series as its 𝕄\mathbb{M}-asymptotic expansion in a wide enough sector of bisecting direction dd, known as the 𝕄\mathbb{M}-sum of the formal power series along direction dd.

Definition 7.

The function vv in Proposition 2 is known as the 𝕄\mathbb{M}-sum of u^\hat{u} along direction dd\in\mathbb{R}, and is denoted by 𝒮𝕄,d(u^)\mathcal{S}_{\mathbb{M},d}(\hat{u}). The set of formal power series with coefficients in 𝔼\mathbb{E} which are 𝕄\mathbb{M}-summable along direction dd is denoted by 𝔼{z}𝕄,d\mathbb{E}\{z\}_{\mathbb{M},d}.

Lemma 7.

Let f^(z)\hat{f}(z) be a formal power series, and let k0k\in\mathbb{N}_{0}. We define the formal power series g^(z):=zkf^(z)\hat{g}(z):=z^{k}\hat{f}(z).

Let 𝕄\mathbb{M} be a strongly regular sequence admitting a nonzero proximate order. Let dd\in\mathbb{R}. Then, the following statements are equivalent:

  • The formal power series f^\hat{f} is 𝕄\mathbb{M}-summable in the direction dd.

  • The formal power series g^\hat{g} is 𝕄\mathbb{M}-summable in the direction dd.

If one of the previous equivalent statements holds, then

(10) 𝒮𝕄,d(g^)=zk𝒮𝕄,d(f^).\mathcal{S}_{\mathbb{M},d}(\hat{g})=z^{k}\mathcal{S}_{\mathbb{M},d}(\hat{f}).
Proof.

It is straightforward that f^\hat{f} being 𝕄\mathbb{M}-summable in the direction dd yields g^\hat{g} being 𝕄\mathbb{M}-summable in the same direction, as the set of 𝕄\mathbb{M} summable functions in a direction is an algebra. The second part of the equivalence can be proved following an analogous argument as that for Exercise 3, Section 4.5, in [1]. More precisely, an iterative argument allows us to assume that k=1k=1. Let g^(z)=p1gpzp.\hat{g}(z)=\sum_{p\geq 1}g_{p}z^{p}. Then, there exists a bounded sector with bisecting direction dd and opening larger than ω(𝕄)π\omega(\mathbb{M})\pi such that for any subsector TT there exist C,A>0C,A>0 with

𝒮𝕄,d(g^)(z)p=1n1gpzpCAnMn|z|n,\left\|\mathcal{S}_{\mathbb{M},d}(\hat{g})(z)-\sum_{p=1}^{n-1}g_{p}z^{p}\right\|\leq CA^{n}M_{n}|z|^{n},

valid for every zTz\in T and n2n\geq 2. Therefore, one has that

z1𝒮𝕄,d(g^)(z)p=0n2gp+1zpCAnMn|z|n1CAA1M1(AA1)n1Mn1|z|n1,\left\|z^{-1}\mathcal{S}_{\mathbb{M},d}(\hat{g})(z)-\sum_{p=0}^{n-2}g_{p+1}z^{p}\right\|\leq CA^{n}M_{n}|z|^{n-1}\leq CAA_{1}M_{1}(AA_{1})^{n-1}M_{n-1}|z|^{n-1},

regarding property (mg)(mg) of 𝕄\mathbb{M}. Observe that f^(z)=p0gp+1zp\hat{f}(z)=\sum_{p\geq 0}g_{p+1}z^{p}, which concludes the proof. ∎

3.3. Generalized multisummability

As in the classical theory, the procedure of Borel-Laplace summation does not succeed when dealing with the formal solutions to some functional equations. As a matter of fact, a more general approach called multisummability is needed in the study of formal solutions to ordinary differential equations. A generalized theory of multisummability can be considered in this framework from different points of view.

The theory of generalized multisummability deals with summability processes with respect to sequences obtained by algebraic actions on the initial strongly regular sequences handled. More precisely, given two sequences of positive real numbers 𝕄=(Mp)p0\mathbb{M}=(M_{p})_{p\geq 0} and 𝕃=(Lp)p0\mathbb{L}=(L_{p})_{p\geq 0}, we denote 𝕄/𝕃:=(Mp/Lp)p0\mathbb{M}/\mathbb{L}:=(M_{p}/L_{p})_{p\geq 0}. The comparison of sequences and their properties is studied in Sections 3.1 and 3.2 [6] in a more general framework. In the present work, we focus on the case where both 𝕄\mathbb{M} and 𝕃\mathbb{L} are powers of some strongly regular sequence admitting a nonzero proximate order, the sequence of quotients being a positive power of the initial sequence, which turns out to be a strongly regular sequence (see Lemma 3) admitting a nonzero proximate order (see Remark 4.8 (i), [22]).

The iterated procedure approach to multisummability in the more general context of strongly regular sequences reads as follows.

Definition 8 (Definition 4.22, [6]).

Let 𝕄j\mathbb{M}_{j}, where j=1,2j=1,2, be two strongly regular sequences admitting nonzero proximate orders. We assume that ω(𝕄1)<ω(𝕄2)<2\omega(\mathbb{M}_{1})<\omega(\mathbb{M}_{2})<2. For j=1,2j=1,2, we consider a strong kernel eje_{j} of 𝕄j\mathbb{M}_{j}-summability and its associated sequence of moments mjm_{j}. The formal power series f^=p0apzp𝔼[[z]]\hat{f}=\sum_{p\geq 0}a_{p}z^{p}\in\mathbb{E}[[z]] is (𝕄1,𝕄2)(\mathbb{M}_{1},\mathbb{M}_{2})-summable in the multidirection (d1,d2)2(d_{1},d_{2})\in\mathbb{R}^{2} with |d1d2|<π(ω(𝕄2)ω(𝕄1))/2|d_{1}-d_{2}|<\pi(\omega(\mathbb{M}_{2})-\omega(\mathbb{M}_{1}))/2 if:

  • (i)

    g^=^m1,z(f^(z))\hat{g}=\hat{\mathcal{B}}_{m_{1},z}(\hat{f}(z)) is 𝕄2/𝕄1\mathbb{M}_{2}/\mathbb{M}_{1}-summable along direction d2d_{2}. Let gg denote such 𝕄2/𝕄1\mathbb{M}_{2}/\mathbb{M}_{1}-sum.

  • (ii)

    gg admits analytic continuation g1g_{1} in an infinite sector Sd1S_{d_{1}} of bisecting direction d1d_{1} with g1𝒪𝕄1(Sd1,𝔼)g_{1}\in\mathcal{O}^{\mathbb{M}_{1}}(S_{d_{1}},\mathbb{E}).

The (𝕄1,𝕄2)(\mathbb{M}_{1},\mathbb{M}_{2})-sum of f^\hat{f} in the multidirection (d1,d2)(d_{1},d_{2}) is given by Te1,d1g1T_{e_{1},d_{1}}g_{1}, which determines a holomorphic function on a bounded sector of bisecting direction dd and opening slightly larger than ω(𝕄1)π\omega(\mathbb{M}_{1})\pi (see Proposition 2). We denote the (𝕄1,𝕄2)(\mathbb{M}_{1},\mathbb{M}_{2})-sum of f^\hat{f} in the multidirection (d1,d2)(d_{1},d_{2}) by 𝒮(𝕄1,𝕄2),(d1,d2)(f^)\mathcal{S}_{(\mathbb{M}_{1},\mathbb{M}_{2}),(d_{1},d_{2})}(\hat{f}), and 𝔼{z}(𝕄1,𝕄2),(d1,d2)\mathbb{E}\{z\}_{(\mathbb{M}_{1},\mathbb{M}_{2}),(d_{1},d_{2})} stands for the set of all formal power series f^(z)𝔼[[z]]\hat{f}(z)\in\mathbb{E}[[z]] which are (𝕄1,𝕄2)(\mathbb{M}_{1},\mathbb{M}_{2})-multisummable along the multidirection (d1,d2)(d_{1},d_{2}).

It is worth remarking that 𝕄\mathbb{M}-summability along a direction is stated in [6] in terms of weight sequences satisfying less restrictive conditions than strongly regular sequences. However, any weight sequence admitting a nonzero proximate order is indeed a strongly regular sequence. According to Theorem 4.23 [6], we recall that the previous construction does not depend on the kernels for 𝕄j\mathbb{M}_{j}-summability considered in the process, j=1,2j=1,2. Indeed, an equivalent definition of multisummability is the following.

Proposition 3 (Definition 4.1, [6]).

In the situation of Definition 8, the formal power series f^\hat{f} is (𝕄1,𝕄2)(\mathbb{M}_{1},\mathbb{M}_{2})-summable in the multidirection (d1,d2)(d_{1},d_{2}) if there exist a formal power series f^1\hat{f}_{1} which is 𝕄1\mathbb{M}_{1}-summable in d1d_{1} and a formal power series f^2\hat{f}_{2} which is 𝕄2\mathbb{M}_{2}-summable in d2d_{2} such that f^=f^1+f^2\hat{f}=\hat{f}_{1}+\hat{f}_{2}. Moreover, the (𝕄1,𝕄2)(\mathbb{M}_{1},\mathbb{M}_{2})-sum of f^\hat{f} in the multidirection (d1,d2)(d_{1},d_{2}) is given by 𝒮𝕄1,d1(f^1)+𝒮𝕄2,d2(f^2)\mathcal{S}_{\mathbb{M}_{1},d_{1}}(\hat{f}_{1})+\mathcal{S}_{\mathbb{M}_{2},d_{2}}(\hat{f}_{2}).

The splitting of f^\hat{f} into a sum in Proposition 3 is essentially unique (see Proposition 4.2, [6]). This equivalent definition of multisummability allows to give a direct proof of the following result.

Lemma 8.

Let f^(z)\hat{f}(z) be a formal power series, and let k0k\in\mathbb{N}_{0}. We define the formal power series g^(z):=zkf^(z)\hat{g}(z):=z^{k}\hat{f}(z).

Let 𝕄j\mathbb{M}_{j}, where j=1,2j=1,2, be two strongly regular sequences admitting nonzero proximate orders. We assume that ω(𝕄1)<ω(𝕄2)<2\omega(\mathbb{M}_{1})<\omega(\mathbb{M}_{2})<2. We choose (d1,d2)2(d_{1},d_{2})\in\mathbb{R}^{2} with |d1d2|<π(ω(𝕄2)ω(𝕄1))/2|d_{1}-d_{2}|<\pi(\omega(\mathbb{M}_{2})-\omega(\mathbb{M}_{1}))/2. Then, the following statements are equivalent:

  • The formal power series f^\hat{f} is (𝕄1,𝕄2)(\mathbb{M}_{1},\mathbb{M}_{2})-summable in the multidirection (d1,d2)(d_{1},d_{2}).

  • The formal power series g^\hat{g} is (𝕄1,𝕄2)(\mathbb{M}_{1},\mathbb{M}_{2})-summable in the multidirection (d1,d2)(d_{1},d_{2}).

If one of the previous equivalent statements hold, then one has that

(11) 𝒮(𝕄1,𝕄2),(d1,d2)(g^)=zk𝒮(𝕄1,𝕄2),(d1,d2)(f^).\mathcal{S}_{(\mathbb{M}_{1},\mathbb{M}_{2}),(d_{1},d_{2})}(\hat{g})=z^{k}\mathcal{S}_{(\mathbb{M}_{1},\mathbb{M}_{2}),(d_{1},d_{2})}(\hat{f}).
Proof.

If f^\hat{f} is (𝕄1,𝕄2)(\mathbb{M}_{1},\mathbb{M}_{2})-summable in the multidirection (d1,d2)(d_{1},d_{2}), then f^=f^1+f^2\hat{f}=\hat{f}_{1}+\hat{f}_{2}, with f^1\hat{f}_{1} being 𝕄1\mathbb{M}_{1}-summable in d1d_{1} and f^2\hat{f}_{2} being 𝕄2\mathbb{M}_{2}-summable in d2d_{2}. Then, zkf^1z^{k}\hat{f}_{1} is 𝕄1\mathbb{M}_{1}-summable in d1d_{1} and zkf^2z^{k}\hat{f}_{2} is 𝕄2\mathbb{M}_{2}-summable in d2d_{2}. Then, g^=zkf^=zkf^1+zkf^2\hat{g}=z^{k}\hat{f}=z^{k}\hat{f}_{1}+z^{k}\hat{f}_{2}, which entails that g^\hat{g} is (𝕄1,𝕄2)(\mathbb{M}_{1},\mathbb{M}_{2})-summable in the multidirection (d1,d2)(d_{1},d_{2}).

On the other hand, if g^\hat{g} is (𝕄1,𝕄2)(\mathbb{M}_{1},\mathbb{M}_{2})-summable in the multidirection (d1,d2)(d_{1},d_{2}), then g^=g^1+g^2\hat{g}=\hat{g}_{1}+\hat{g}_{2}, with g^1\hat{g}_{1} being 𝕄1\mathbb{M}_{1}-summable in d1d_{1} and g^2\hat{g}_{2} being 𝕄2\mathbb{M}_{2}-summable in d2d_{2}. As the splitting is essentially unique, and g^(z):=zkh^(z)\hat{g}(z):=z^{k}\hat{h}(z), one can choose a splitting in which g^j=zkh^j\hat{g}_{j}=z^{k}\hat{h}_{j}, for j=1,2j=1,2. Lemma 7 guarantees that the formal power series g^j\hat{g}_{j} is 𝕄j\mathbb{M}_{j}-summable along direction djd_{j} iff h^j\hat{h}_{j} is 𝕄j\mathbb{M}_{j}-summable along direction djd_{j}. Then, f^\hat{f} can be written in the form f^=h^1+h^2\hat{f}=\hat{h}_{1}+\hat{h}_{2}, where h^j\hat{h}_{j} is 𝕄j\mathbb{M}_{j}-summable along direction djd_{j}, for j=1,2j=1,2, leading to multisummability of f^\hat{f}.

Regarding the construction of the sums above, one also arrives at (11).

4. On moment differentiation

In this section, we focus our attention on the concept of a moment derivative and recall some properties associated with this notion. We also state some new results to be applied in the work.

The notion of a generalized derivative operator allows to consider functional problems under greater generality. More precisely, we deal with the following formal operator.

Definition 9.

Let (𝔼,𝔼)(\mathbb{E},\left\|\cdot\right\|_{\mathbb{E}}) be a complex Banach space. For any fixed sequence of moments (me(p))p0(m_{e}(p))_{p\geq 0} we define the mm-differential operator me,t:𝔼[[t]]𝔼[[t]]\partial_{m_{e},t}\colon\mathbb{E}[[t]]\to\mathbb{E}[[t]] by the formula:

me,z(p0apme(p)zp):=p0ap+1me(p)zp.\partial_{m_{e},z}\left(\sum_{p\geq 0}\frac{a_{p}}{m_{e}(p)}z^{p}\right):=\sum_{p\geq 0}\frac{a_{p+1}}{m_{e}(p)}z^{p}.

Usual derivatives are recovered when considering the moment sequence (p!)p0(p!)_{p\geq 0}, which is associated with the kernel function e(z)=zexp(z)e(z)=z\exp(-z). In addition to this, moment derivatives can be read in terms of Caputo α\alpha-fractional derivatives zα\partial_{z}^{\alpha} as follows. Let α\alpha be a positive rational number. The moment sequence mα:=(Γ(1+αp))p0m_{\alpha}:=(\Gamma(1+\alpha p))_{p\geq 0} is associated with the kernel function e(z)=1αz1αexp(z1α)e(z)=\frac{1}{\alpha}z^{\frac{1}{\alpha}}\exp(-z^{\frac{1}{\alpha}}) and the fractional derivative of order α\alpha is defined on formal power series in zαz^{\alpha} by

zα(p0apΓ(1+αp)zαp)=p0ap+1Γ(1+αp)zαp.\partial_{z}^{\alpha}\left(\sum_{p\geq 0}\frac{a_{p}}{\Gamma(1+\alpha p)}z^{\alpha p}\right)=\sum_{p\geq 0}\frac{a_{p+1}}{\Gamma(1+\alpha p)}z^{\alpha p}.

Therefore, one has

(mαf^)(zα)=zα(f^(zα)),(\partial_{m_{\alpha}}\hat{f})(z^{\alpha})=\partial_{z}^{\alpha}(\hat{f}(z^{\alpha})),

for every f^𝔼[[z]]\hat{f}\in\mathbb{E}[[z]].

We may also observe that qq-derivatives defined by

Dq,zf(z)=f(qz)f(z)qzzD_{q,z}f(z)=\frac{f(qz)-f(z)}{qz-z}

for some fixed q(0,1)q\in(0,1) can be interpreted in terms of the moment derivatives associated with the moment sequence ([p]q!)p0([p]_{q}!)_{p\geq 0}, with [p]q!=[1]q[2]q[p]q[p]_{q}!=[1]_{q}[2]_{q}\cdots[p]_{q} and [j]q=h=0j1qh[j]_{q}=\sum_{h=0}^{j-1}q^{h}.

We also have the following result.

Lemma 9 (Lemma 3, [13]).

Let mj=(mj(p))p0m_{j}=(m_{j}(p))_{p\geq 0} for j=1,2j=1,2 be two sequences of moments. Then,

  • -

    The sequence m1m2=(m1(p)m2(p))p0m_{1}m_{2}=(m_{1}(p)m_{2}(p))_{p\geq 0} is a sequence of moments.

  • -

    m1,zm2,zm1m2,z^m1,z\mathcal{B}_{m_{1},z}\circ\partial_{m_{2},z}\equiv\partial_{m_{1}m_{2},z}\circ\hat{\mathcal{B}}_{m_{1},z} as operators defined in 𝔼[[z]]\mathbb{E}[[z]].

Moment differentiation can be naturally extended to holomorphic functions on some neighborhood of the origin by identifying the function with its Taylor series at the origin. However, this formal differentiation does not preserve convergence unless some regularity property is assumed for the sequence of moments. However, if one departs from a strongly regular sequence 𝕄\mathbb{M} which admits a nonzero proximate order, and considers a pair of kernel functions associated with it and then constructs the corresponding sequence of moments mm, then it holds that 𝕄\mathbb{M} and mm generate the same ultraholomorphic space of functions (i.e., they are equivalent sequences) and mm is indeed a strongly regular sequence (see Remark 3.8, [10]). In [13], the definition of moment differentiation was also extended to the generalized sum along a direction of a formal power series as the generalized sum along that same direction of the formal moment derivative of the initial formal power series. For that purpose, the moment derivative was also provided for any function defined on some neighborhood of the origin with holomorphic extension to an infinite sector and with certain generalized exponential growth at infinity (as described in Definition 3). Indeed, the first part of Theorem 3 [13] reads as follows.

Theorem 1.

Let me=(me(p))p0m_{e}=(m_{e}(p))_{p\geq 0} be a sequence of moments. We also fix d,θ,rd,\theta,r\in\mathbb{R} with θ,r>0\theta,r>0 and φ𝒪(S^d(θ;r),𝔼)\varphi\in\mathcal{O}(\hat{S}_{d}(\theta;r),\mathbb{E}). Then, there exists 0<r~<r0<\tilde{r}<r such that for all 0<θ1<θ0<\theta_{1}<\theta, all zS^d(θ1;r~)z\in\hat{S}_{d}(\theta_{1};\tilde{r}) and all n0n\in\mathbb{N}_{0}, one has that

(12) me,znφ(z)=12πiΓzφ(ω)0(τ)ξnE(zξ)e(ωξ)ωξ𝑑ξ𝑑ω,\partial_{m_{e},z}^{n}\varphi(z)=\frac{1}{2\pi i}\oint_{\Gamma_{z}}\varphi(\omega)\int_{0}^{\infty(\tau)}\xi^{n}E(z\xi)\frac{e(\omega\xi)}{\omega\xi}d\xi d\omega,

with τ=τ(ω)(arg(ω)ω(me)π2,arg(ω)+ω(me)π2)\tau=\tau(\omega)\in(-\arg(\omega)-\frac{\omega(m_{e})\pi}{2},-\arg(\omega)+\frac{\omega(m_{e})\pi}{2}). The integration path Γz\Gamma_{z} is a deformation of the circle {|ω|=r1}\{|\omega|=r_{1}\}, for any choice of 0<r1<r0<r_{1}<r, which depends on zz. More precisely, such deformation consists of substituting some arc of the circle contained in S^d(θ;r)\hat{S}_{d}(\theta;r) by a simple path which attains an adequate sufficient distance to the origin while it remains inside Sd(θ)S_{d}(\theta).

At this point, we give a step forward in order to define the moment derivatives on functions defined on some sectorial region, and which can be extended under certain generalized exponential growth to infinity.

Theorem 2.

Let me=(me(p))p0m_{e}=(m_{e}(p))_{p\geq 0} be a sequence of moments, and 𝕄=(Mp)p0\mathbb{M}=(M_{p})_{p\geq 0} be a strongly regular sequence admitting a nonzero proximate order. Let d1,d2d_{1},d_{2}\in\mathbb{R} satisfying |d1d2|<aω(𝕄)π2|d_{1}-d_{2}|<a\frac{\omega(\mathbb{M})\pi}{2} for some a>0a>0. We choose u^𝔼{z}𝕄a,d2\hat{u}\in\mathbb{E}\{z\}_{\mathbb{M}^{a},d_{2}} and write u=S𝕄a,d2(u^)𝒪(G,𝔼)u=S_{\mathbb{M}^{a},d_{2}}(\hat{u})\in\mathcal{O}(G,\mathbb{E}), for some sectorial region G=Gd2(θ)G=G_{d_{2}}(\theta) with θ>aπω(𝕄)\theta>a\pi\omega(\mathbb{M}), as seen in Figure 1. Assume moreover that uu can be extended (the extension is also denoted by uu) to an infinite sector of bisecting direction d1d_{1}, with u𝒪𝕄b(Sd1,𝔼)u\in\mathcal{O}^{\mathbb{M}^{b}}(S_{d_{1}},\mathbb{E}). Then, the following statements hold:

  • (a)

    There exists r~>0\tilde{r}>0 such that for every SSd1S^{\prime}\prec S_{d_{1}} and GGD(0,r~)G^{\prime}\prec G\cap D(0,\tilde{r}) and all zS~:=SGz\in\tilde{S}:=S^{\prime}\cup G^{\prime} and n0n\in\mathbb{N}_{0}

    me,znu(z)=12πiΛzu(w)0(τ)ξnE(zξ)e(wξ)wξ𝑑ξ𝑑w,\partial_{m_{e},z}^{n}u(z)=\frac{1}{2\pi i}\oint_{\Lambda_{z}}u(w)\int_{0}^{\infty(\tau)}\xi^{n}E(z\xi)\frac{e(w\xi)}{w\xi}d\xi dw,

    where τ=τ(ω)(arg(w)ω(me)π2,arg(w)+ω(me)π2)\tau=\tau(\omega)\in(-\arg(w)-\frac{\omega(m_{e})\pi}{2},-\arg(w)+\frac{\omega(m_{e})\pi}{2}). The path Λz\Lambda_{z} depends on zz.

  • (b)

    There exist C4,C5,C6>0C_{4},C_{5},C_{6}>0 such that

    (13) me,znu(z)𝔼C4C5nme(n)Mnaexp(Mb(C6|z|)) for all n0 and zS~,\|\partial_{m_{e},z}^{n}u(z)\|_{\mathbb{E}}\leq C_{4}C_{5}^{n}m_{e}(n)M^{a}_{n}\exp\left(M^{b}(C_{6}|z|)\right)\quad\textrm{ for all }n\in\mathbb{N}_{0}\textrm{ and }z\in\tilde{S},

    where Mb(t)M^{b}(t) is the function defined in (2), corresponding to the strongly regular sequence 𝕄b\mathbb{M}^{b}.

Refer to caption
Figure 1. Example of configuration of the sets Sd1S_{d_{1}} and GG
Proof.

The proof of the previous result is based on that of Theorem 3 and Proposition 2, [13]. We provide a complete proof for the sake of completeness and focus on distinctive points with respect to the results in that previous work.

Let SSd1S^{\prime}\prec S_{d_{1}} and G′′GG^{\prime\prime}\prec G. Let r1>0r_{1}>0 and θ>θ>0\theta>\theta^{\prime}>0 such that Sd2(θ;2r1)GS_{d_{2}}(\theta;2r_{1})\subseteq G and SG′′Sd2(θ)S^{\prime}\cup G^{\prime\prime}\subset S_{d_{2}}(\theta^{\prime}). We take r~:=k~r1Kρ(2)\tilde{r}:=\frac{\tilde{k}r_{1}}{K\rho(2)}, where k~\tilde{k} is given in (6), KK is defined in (5), and ρ()\rho(\cdot) is as in Lemma 2. Let G:=G′′D(0,r~)G^{\prime}:=G^{\prime\prime}\cap D(0,\tilde{r}) and S~:=SG\tilde{S}:=S^{\prime}\cup G^{\prime}. Choose zS~z\in\tilde{S}. The path Λz\Lambda_{z} is constructed as follows. Let r1eiθ1,r1eiθ2r_{1}e^{i\theta_{1}},r_{1}e^{i\theta_{2}} be the points in {z:|z|=r1}Sd\{z\in\mathbb{C}:|z|=r_{1}\}\cap S_{d} and let P~=r1eiθ~1,Q~=r1eiθ~2\tilde{P}=r_{1}e^{i\tilde{\theta}_{1}},\tilde{Q}=r_{1}e^{i\tilde{\theta}_{2}} be the points in {z:|z|=r1}S\{z\in\mathbb{C}:|z|=r_{1}\}\cap S^{\prime}. We assume θ1<θ~1<θ~2<θ2\theta_{1}<\tilde{\theta}_{1}<\tilde{\theta}_{2}<\theta_{2}. We define Λ1:=[0,r1]ei(d2θ/2)\Lambda_{1}:=[0,r_{1}]e^{i(d_{2}-\theta/2)}, Λ2\Lambda_{2} is the arc of the circle of radius r1r_{1} from r1ei(d2θ/2)r_{1}e^{i(d_{2}-\theta/2)} to r1ei(θ1+θ~1)/2r_{1}e^{i(\theta_{1}+\tilde{\theta}_{1})/2}. We also put Λ3:=[r1ei(θ1+θ~1)/2,Rei(θ1+θ~1)/2]\Lambda_{3}:=[r_{1}e^{i(\theta_{1}+\tilde{\theta}_{1})/2},Re^{i(\theta_{1}+\tilde{\theta}_{1})/2}], with R=R(z)>0R=R(z)>0 to be determined. Λ4\Lambda_{4} is the arc of circle from Rei(θ1+θ~1)/2Re^{i(\theta_{1}+\tilde{\theta}_{1})/2} to Rei(θ2+θ~2)/2Re^{i(\theta_{2}+\tilde{\theta}_{2})/2}, Λ5:=[r1ei(θ2+θ~2)/2,Rei(θ2+θ~2)/2]\Lambda_{5}:=[r_{1}e^{i(\theta_{2}+\tilde{\theta}_{2})/2},Re^{i(\theta_{2}+\tilde{\theta}_{2})/2}], Λ6\Lambda_{6} is the arc of the circle of radius r1r_{1} from r1ei(θ2+θ~2)/2r_{1}e^{i(\theta_{2}+\tilde{\theta}_{2})/2} to r1ei(d2+θ/2)r_{1}e^{i(d_{2}+\theta/2)} and Λ7:=[0,r1]ei(d2+θ/2)\Lambda_{7}:=[0,r_{1}]e^{i(d_{2}+\theta/2)}. We finally define the integration path

Λz:=Λ1+Λ2+Λ3+Λ4Λ5+Λ6Λ7,\Lambda_{z}:=\Lambda_{1}+\Lambda_{2}+\Lambda_{3}+\Lambda_{4}-\Lambda_{5}+\Lambda_{6}-\Lambda_{7},

see Figure 2. In case that |z|<r~=k~r1Kρ(2)|z|<\tilde{r}=\frac{\tilde{k}r_{1}}{K\rho(2)}, where k~\tilde{k} is given in (6), KK is defined in (5), and ρ()\rho(\cdot) is as in Lemma 2, then one can choose P~=Q~\tilde{P}=\tilde{Q} and remove Λ3\Lambda_{3}, Λ4\Lambda_{4} and Λ5\Lambda_{5} from the concatenation. Otherwise, R:=ρ(2)Kk~|z|R:=\frac{\rho(2)K}{\tilde{k}}|z|.

Refer to caption
Figure 2. Integration path Λz\Lambda_{z}

For the first part of the proof, we observe from (35) in [23] that

0(τ)E(zξ)e(wξ)wξ𝑑ξ=1wz,\int_{0}^{\infty(\tau)}E(z\xi)\frac{e(w\xi)}{w\xi}d\xi=\frac{1}{w-z},

for every pair of complex numbers (z,w)(z,w) in which both sides of the previous expression are defined. Hence, if u𝒪(GSd1)u\in\mathcal{O}(G\cup S_{d_{1}}) then one may replace the contour Γz\Gamma_{z} in (12) by Λz\Lambda_{z}.

We now provide the estimates in (13). We first observe (see (17) in [13]) that

(14) |0(τ)ξnE(zξ)e(ωξ)ωξ𝑑ξ|A0B0nme(n),n0,\left|\int_{0}^{\infty(\tau)}\xi^{n}E(z\xi)\frac{e(\omega\xi)}{\omega\xi}d\xi\right|\leq A_{0}B_{0}^{n}m_{e}(n),\qquad n\geq 0,

valid for all zz\in\mathbb{C} with |z|r~|z|\leq\tilde{r} and τ(arg(ω)ω(me)π2,arg(ω)+ω(me)π2)\tau\in\left(-\arg(\omega)-\frac{\omega(m_{e})\pi}{2},-\arg(\omega)+\frac{\omega(m_{e})\pi}{2}\right), for ω\omega\in\mathbb{C} with |ω|=r1|\omega|=r_{1}. This entails there exist A11,B11>0A_{11},B_{11}>0 such that

12πiΛju(ω)0(τ)ξnE(zξ)e(ωξ)ωξ𝑑ξ𝑑ω𝔼(sup|ω|=r1,ωSd2(θ;2r1)u(ω))A11B11nme(n),\left\|\frac{1}{2\pi i}\int_{\Lambda_{j}}u(\omega)\int_{0}^{\infty(\tau)}\xi^{n}E(z\xi)\frac{e(\omega\xi)}{\omega\xi}d\xi d\omega\right\|_{\mathbb{E}}\leq\left(\sup_{|\omega|=r_{1},\omega\in S_{d_{2}}(\theta;2r_{1})}\left\|u(\omega)\right\|\right)A_{11}B_{11}^{n}m_{e}(n),

for j=2,6j=2,6, valid for every n0n\in\mathbb{N}_{0}. We recall that in this case Λ3,Λ4\Lambda_{3},\Lambda_{4} and Λ5\Lambda_{5} do not appear in the integration path. It only remains to give upper bounds regarding the paths Λ1\Lambda_{1} and Λ7\Lambda_{7}, which are postponed.

If |z|>r~|z|>\tilde{r}, for every ωΛ2Λ6\omega\in\Lambda_{2}\cup\Lambda_{6} one can choose τ\tau with

τ(arg(ω)ω(me)π2,arg(ω)+ω(me)π2)(arg(z)+ω(me)π2,arg(z)+2πω(me)π2).\tau\in\left(-\arg(\omega)-\frac{\omega(m_{e})\pi}{2},-\arg(\omega)+\frac{\omega(m_{e})\pi}{2}\right)\cap\left(-\arg(z)+\frac{\omega(m_{e})\pi}{2},-\arg(z)+2\pi-\frac{\omega(m_{e})\pi}{2}\right).

This entails (see (22)–(26) in the proof of Theorem 3 [13]) the existence of A12,B12>0A_{12},B_{12}>0 with

12πiΛju(ω)0(τ)ξnE(zξ)e(ωξ)ωξ𝑑ξ𝑑ω𝔼(sup|ω|=r1,ωSd2(θ;2r1)u(ω))A12B12nme(n),\left\|\frac{1}{2\pi i}\int_{\Lambda_{j}}u(\omega)\int_{0}^{\infty(\tau)}\xi^{n}E(z\xi)\frac{e(\omega\xi)}{\omega\xi}d\xi d\omega\right\|_{\mathbb{E}}\leq\left(\sup_{|\omega|=r_{1},\omega\in S_{d_{2}}(\theta;2r_{1})}\left\|u(\omega)\right\|\right)A_{12}B_{12}^{n}m_{e}(n),

for j=2,6j=2,6, and all n0n\in\mathbb{N}_{0}. The previous expression can be bounded from above by

A12B12nme(n)exp(Mb(C12|z|))A_{12}B_{12}^{n}m_{e}(n)\exp(M^{b}(C_{12}|z|))

for some C12>0C_{12}>0 taking into account the growth at infinity determined by vv and the choice of the integration paths. On the other hand, parametrizing the integration path Λ3\Lambda_{3} and analogous estimates as in the previous expression yield

12πiΛ3u(ω)0(τ)ξnE(zξ)e(ωξ)ωξ𝑑ξ𝑑ω𝔼A13B13nme(n)exp(Mb(C13|z|)),\left\|\frac{1}{2\pi i}\int_{\Lambda_{3}}u(\omega)\int_{0}^{\infty(\tau)}\xi^{n}E(z\xi)\frac{e(\omega\xi)}{\omega\xi}d\xi d\omega\right\|_{\mathbb{E}}\leq A_{13}B_{13}^{n}m_{e}(n)\exp(M^{b}(C_{13}|z|)),

for some A13,B13,C13>0A_{13},B_{13},C_{13}>0 (see (28) in [13]). These same upper bounds hold for the integration along Λ5\Lambda_{5}. Again, the parametrization of Λ4\Lambda_{4} and usual estimates yield

12πiΛ4u(ω)0(τ)ξnE(zξ)e(ωξ)ωξ𝑑ξ𝑑ω𝔼A14B14nme(n)exp(Mb(C14|z|)),\left\|\frac{1}{2\pi i}\int_{\Lambda_{4}}u(\omega)\int_{0}^{\infty(\tau)}\xi^{n}E(z\xi)\frac{e(\omega\xi)}{\omega\xi}d\xi d\omega\right\|_{\mathbb{E}}\leq A_{14}B_{14}^{n}m_{e}(n)\exp(M^{b}(C_{14}|z|)),

for some A14,B14,C14>0A_{14},B_{14},C_{14}>0 (see (31) in [13]).

At this point, it only remains to provide upper bounds for

(15) 12πiΛju(ω)0(τ)ξnE(zξ)e(ωξ)ωξ𝑑ξ𝑑ω𝔼\left\|\frac{1}{2\pi i}\int_{\Lambda_{j}}u(\omega)\int_{0}^{\infty(\tau)}\xi^{n}E(z\xi)\frac{e(\omega\xi)}{\omega\xi}d\xi d\omega\right\|_{\mathbb{E}}

for j=1,7j=1,7 for all zS~z\in\tilde{S}. Let n0n\in\mathbb{N}_{0}. We write u^(z)=p0upzp\hat{u}(z)=\sum_{p\geq 0}u_{p}z^{p} and define

v(z):=1zn(u(z)p=0n1upzp).v(z):=\frac{1}{z^{n}}\left(u(z)-\sum_{p=0}^{n-1}u_{p}z^{p}\right).
Lemma 10.

The function vv belongs to 𝒪𝕄b(Sd1,𝔼)\mathcal{O}^{\mathbb{M}^{b}}(S_{d_{1}},\mathbb{E}). In addition to this, for every SSd1S^{\prime}\prec S_{d_{1}} there exist C~,B~,A~>0\tilde{C},\tilde{B},\tilde{A}>0, which do not depend on nn, such that

v(z)𝔼C~A~nMnaexp(Mb(|z|B~)),\left\|v(z)\right\|_{\mathbb{E}}\leq\tilde{C}\tilde{A}^{n}M_{n}^{a}\exp\left(M^{b}\left(\frac{|z|}{\tilde{B}}\right)\right),

for all zSz\in S^{\prime}.

Proof.

It is clear that v𝒪𝕄b(Sd1,𝔼)v\in\mathcal{O}^{\mathbb{M}^{b}}(S_{d_{1}},\mathbb{E}). We have that for all SSd1S^{\prime}\prec S_{d_{1}} one has that (3) holds for some C,K>0C,K>0.

Bearing in mind that uu admits u^\hat{u} as its asymptotic expansion, there exist K1,C~1,A~1>0K_{1},\tilde{C}_{1},\tilde{A}_{1}>0 such that

(16) v(z)𝔼C~1A~1nMna,\left\|v(z)\right\|_{\mathbb{E}}\leq\tilde{C}_{1}\tilde{A}_{1}^{n}M^{a}_{n},

for all zSz\in S^{\prime} with |z|K1|z|\leq K_{1}, provided that the opening of Sd1S_{d_{1}} is small enough. On the other hand, for every zSz\in S^{\prime} with |z|K1|z|\geq K_{1} and all 0pn10\leq p\leq n-1, one can apply Lemma 5 and usual estimates to arrive at

upzpn𝔼1K1npC~1A~1pMpaC~2A~2nMna.\left\|u_{p}z^{p-n}\right\|_{\mathbb{E}}\leq\frac{1}{K_{1}^{n-p}}\tilde{C}_{1}\tilde{A}_{1}^{p}M_{p}^{a}\leq\tilde{C}_{2}\tilde{A}_{2}^{n}M_{n}^{a}.

Also,

(17) u(z)𝔼|z|n1K1nCexp(Mb(|z|/K))\frac{\left\|u(z)\right\|_{\mathbb{E}}}{|z|^{n}}\leq\frac{1}{K_{1}^{n}}C\exp(M^{b}(|z|/K))

for all zSz\in S^{\prime} with |z|K1|z|\geq K_{1}. The result follows from here. ∎

Observe that for every n0n\in\mathbb{N}_{0} one has that

(18) me,znu(z)=me,zn(u(z)p=0n1upzp)=me,zn(znv(z)).\partial_{m_{e},z}^{n}u(z)=\partial_{m_{e},z}^{n}\left(u(z)-\sum_{p=0}^{n-1}u_{p}z^{p}\right)=\partial_{m_{e},z}^{n}(z^{n}v(z)).

In view of (18) one may substitute the study of upper estimates of (15) by those of

(19) Ij:=12πiΛjv(ω)0(τ)ξnωnE(zξ)e(ωξ)ωξ𝑑ξ𝑑ω𝔼.I_{j}:=\left\|\frac{1}{2\pi i}\int_{\Lambda_{j}}v(\omega)\int_{0}^{\infty(\tau)}\xi^{n}\omega^{n}E(z\xi)\frac{e(\omega\xi)}{\omega\xi}d\xi d\omega\right\|_{\mathbb{E}}.

We provide upper bounds for (19) for j=1j=1, which remain valid for the case j=7j=7.

In view of Lemma 10, one derives

(20) v(ω)𝔼C~1A~1nMnaexp(Mb(|ω|B~))C~1exp(Mb(r1B~))A~1nMna.\left\|v(\omega)\right\|_{\mathbb{E}}\leq\tilde{C}_{1}\tilde{A}_{1}^{n}M_{n}^{a}\exp\left(M^{b}\left(\frac{|\omega|}{\tilde{B}}\right)\right)\leq\tilde{C}_{1}\exp\left(M^{b}\left(\frac{r_{1}}{\tilde{B}}\right)\right)\tilde{A}_{1}^{n}M_{n}^{a}.

By Proposition 2, [13] there exists ε>0\varepsilon>0, such that the estimation (13) holds for zS~D(0,ε)z\in\tilde{S}\cap D(0,\varepsilon). Hence, we may assume that |z|ε|z|\geq\varepsilon. Let MEM_{E} be a positive constant given in Definition 4.

We split the path of integration in the inner integral in (19) into the segment [0,eiτME/|z|][0,e^{i\tau}M_{E}/|z|] and the ray [eiτME/|z|,(τ)][e^{i\tau}M_{E}/|z|,\infty(\tau)], and we interchange the order of integration in (19). Next, we observe that arg(ω)=d2θ/2\hbox{arg}(\omega)=d_{2}-\theta/2 and arg(ξ)=τ(arg(ω)πω(me)2,arg(ω)+πω(me)2)\hbox{arg}(\xi)=\tau\in(-\hbox{arg}(\omega)-\frac{\pi\omega(m_{e})}{2},-\hbox{arg}(\omega)+\frac{\pi\omega(m_{e})}{2}). Hence

(21) I112πsupωΛ1v(ω)𝔼(0ME/|z||E(zseiτ)|0r1snwn|e(wseiθ~)|dwwdss+ME/|z||E(zseiτ)|0r1snwn|e(wseiθ~)|dwwdss)=:12πsupωΛ1v(ω)𝔼(I11+I12),I_{1}\leq\frac{1}{2\pi}\sup_{\omega\in\Lambda_{1}}\|v(\omega)\|_{\mathbb{E}}\bigg{(}\int_{0}^{M_{E}/|z|}|E(zse^{i\tau})|\int_{0}^{r_{1}}s^{n}w^{n}|e(wse^{i\tilde{\theta}})|\frac{dw}{w}\frac{ds}{s}\\ +\int_{M_{E}/|z|}^{\infty}|E(zse^{i\tau})|\int_{0}^{r_{1}}s^{n}w^{n}|e(wse^{i\tilde{\theta}})|\frac{dw}{w}\frac{ds}{s}\bigg{)}=:\frac{1}{2\pi}\sup_{\omega\in\Lambda_{1}}\|v(\omega)\|_{\mathbb{E}}(I_{11}+I_{12}),

for θ~:=d2θ/2+τ(πω(me)/2,πω(me)/2)\tilde{\theta}:=d_{2}-\theta/2+\tau\in(-\pi\omega(m_{e})/2,\pi\omega(m_{e})/2).

Observe that

(22) 0r1snwn|e(wseiθ~)|dww0sr1tn1|e(teiθ~)|𝑑t.\int_{0}^{r_{1}}s^{n}w^{n}|e(wse^{i\tilde{\theta}})|\frac{dw}{w}\leq\int_{0}^{sr_{1}}t^{n-1}|e(te^{i\tilde{\theta}})|dt.

If s<ME/|z|s<M_{E}/|z| then using (4) we continue the estimation (22) by

(23) 0r1stn1|e(teiθ~)|𝑑tC0r1stn1tα𝑑tCr1n+αn+αsn+αC~2A~2nsn+α,\int_{0}^{r_{1}s}t^{n-1}|e(te^{i\tilde{\theta}})|dt\leq C\int_{0}^{r_{1}s}t^{n-1}t^{\alpha}dt\leq\frac{Cr_{1}^{n+\alpha}}{n+\alpha}s^{n+\alpha}\leq\tilde{C}_{2}\tilde{A}_{2}^{n}s^{n+\alpha},

for some C~2,A~2>0\tilde{C}_{2},\tilde{A}_{2}>0 and α>0\alpha>0.

By (23) we get

(24) I11sup|ζ|ME|E(ζ)|0ME/εC~2A~2nsn+α1𝑑sC~3A~3n,I_{11}\leq\sup_{|\zeta|\leq M_{E}}|E(\zeta)|\int_{0}^{M_{E}/\varepsilon}\tilde{C}_{2}\tilde{A}_{2}^{n}s^{n+\alpha-1}ds\leq\tilde{C}_{3}\tilde{A}_{3}^{n},

for some C~3,A~3>0\tilde{C}_{3},\tilde{A}_{3}>0.

On the opposite case sME/|z|s\geq M_{E}/|z|, using Lemma 1 and Lemma 6 and (5) we estimate (22) by

(25) 0r1stn1|e(teiθ~)|𝑑tC~4A~4nme(n),\int_{0}^{r_{1}s}t^{n-1}|e(te^{i\tilde{\theta}})|dt\leq\tilde{C}_{4}\tilde{A}_{4}^{n}m_{e}(n),

for some constants C~4,A~4>0\tilde{C}_{4},\tilde{A}_{4}>0.

Since θ<θ\theta^{\prime}<\theta and arg(z)(d2θ/2,d2+θ/2)\hbox{arg}(z)\in(d_{2}-\theta^{\prime}/2,d_{2}+\theta^{\prime}/2), there exists δ>0\delta>0 such that for every zS~z\in\tilde{S} there exists τ(d2+θ/2πω(me)/2,d2+θ/2+πω(me)/2)\tau\in(-d_{2}+\theta/2-\pi\omega(m_{e})/2,-d_{2}+\theta/2+\pi\omega(m_{e})/2) satisfying arg(z)+τ(πω(me)/2δ,πω(me)/2+δ)\hbox{arg}(z)+\tau\not\in(-\pi\omega(m_{e})/2-\delta,\pi\omega(m_{e})/2+\delta). Hence, using (7) and (25) we estimate

(26) I12C~4A~4nme(n)c~2|z|βMe/|z|1sβ+1𝑑sC~5A~5nme(n),I_{12}\leq\tilde{C}_{4}\tilde{A}_{4}^{n}m_{e}(n)\frac{\tilde{c}_{2}}{|z|^{\beta}}\int_{M_{e}/|z|}^{\infty}\frac{1}{s^{\beta+1}}ds\leq\tilde{C}_{5}\tilde{A}_{5}^{n}m_{e}(n),

for some C~5,A~5>0\tilde{C}_{5},\tilde{A}_{5}>0 and β>0\beta>0.

Taking into account (20), (21), (24) and (26), upper estimates as above yield

I1C~6C~7nme(n)Mnaexp(Mb(C~8|z|)),I_{1}\leq\tilde{C}_{6}\tilde{C}_{7}^{n}m_{e}(n)M_{n}^{a}\exp\left(M^{b}(\tilde{C}_{8}|z|)\right),

for some C~j>0\tilde{C}_{j}>0, 6j86\leq j\leq 8.

This concludes the proof of (13). ∎

In the last part of this section, we describe compatibility conditions regarding asymptotic expansions and moment derivation, which allows to provide a differential structure to the set of multisummable formal power series.

Lemma 11.

In the situation of Theorem 2, one has that me,zu(z)\partial_{m_{e},z}u(z) is the 𝕄a\mathbb{M}^{a}-sum of me,zu^(z)\partial_{m_{e},z}\hat{u}(z) in GG.

Proof.

Observe that Eme(zξ)=n0(zξ)nme(n)E_{m_{e}}(z\xi)=\sum_{n\geq 0}\frac{(z\xi)^{n}}{m_{e}(n)}. Therefore, for every p0p\geq 0 one has

limz0me,zpEme(zξ)=ξpme(0).\lim_{z\to 0}\partial_{m_{e},z}^{p}E_{m_{e}}(z\xi)=\frac{\xi^{p}}{m_{e}(0)}.

Analogously,

limz0zpEme(zξ)=p!me(p)ξp,\lim_{z\to 0}\partial_{z}^{p}E_{m_{e}}(z\xi)=\frac{p!}{m_{e}(p)}\xi^{p},

which yields

limz0me,zpEme(zξ)=me(p)me(0)p!limz0zpEme(zξ).\lim_{z\to 0}\partial_{m_{e},z}^{p}E_{m_{e}}(z\xi)=\frac{m_{e}(p)}{m_{e}(0)p!}\lim_{z\to 0}\partial_{z}^{p}E_{m_{e}}(z\xi).

In view of (12), this means that given uu which admits u^(z)=n0unzn\hat{u}(z)=\sum_{n\geq 0}u_{n}z^{n} as its asymptotic expansion at the origin in GG, then

limz0,zGme,zpu(z)\displaystyle\lim_{z\to 0,z\in G}\partial_{m_{e},z}^{p}u(z) =me(p)me(0)p!limz0,zGzpu(z)\displaystyle=\frac{m_{e}(p)}{m_{e}(0)p!}\lim_{z\to 0,z\in G}\partial_{z}^{p}u(z)
=me(p)me(0)p!zpu^(z)|z=0=me(p)me(0)p!p!up=me(p)me(0)up.\displaystyle=\frac{m_{e}(p)}{m_{e}(0)p!}\partial_{z}^{p}\hat{u}(z)\left.\right|_{z=0}=\frac{m_{e}(p)}{m_{e}(0)p!}p!u_{p}=\frac{m_{e}(p)}{m_{e}(0)}u_{p}.

On the other hand,

me,zpu^(z)=n0un+pme(n+p)me(n)zn,\partial_{m_{e},z}^{p}\hat{u}(z)=\sum_{n\geq 0}\frac{u_{n+p}m_{e}(n+p)}{m_{e}(n)}z^{n},

which entails that

me,zpu^(z)|z=0=upme(p)me(0).\partial_{m_{e},z}^{p}\hat{u}(z)\left.\right|_{z=0}=u_{p}\frac{m_{e}(p)}{m_{e}(0)}.

We conclude that

limz0,zGme,zpu(z)=me,zpu^(z)|z=0.\lim_{z\to 0,z\in G}\partial_{m_{e},z}^{p}u(z)=\partial_{m_{e},z}^{p}\hat{u}(z)\left.\right|_{z=0}.

This is an equivalent condition for me,zu(z)\partial_{m_{e},z}u(z) to admit me,zu^(z)\partial_{m_{e},z}\hat{u}(z) as its asymptotic expansion in GG. ∎

Corollary 1.

Let me=(me(p))p0m_{e}=(m_{e}(p))_{p\geq 0} be a sequence of moments, and let 𝕄\mathbb{M} be a strongly regular sequence which admits a nonzero proximate order. We also choose positive numbers a,ba,b such that ω(𝕄)<2/(a+b)\omega(\mathbb{M})<2/(a+b). Given any multidirection (d1,d2)2(d_{1},d_{2})\in\mathbb{R}^{2} such that |d1d2|<aω(𝕄)π2|d_{1}-d_{2}|<a\frac{\omega(\mathbb{M})\pi}{2}, the space 𝔼{z}(𝕄b,𝕄a+b),(d1,d2)\mathbb{E}\{z\}_{(\mathbb{M}^{b},\mathbb{M}^{a+b}),(d_{1},d_{2})} is closed under mem_{e}-differentiation.

Proof.

In view of Lemma 3, we have that 𝕄b\mathbb{M}^{b} is a strongly regular sequence which admits a nonzero proximate order (see Remark 4.8 (i), [22]). This guarantees the existence of a moment sequence meb:=(me(p))p0m_{e^{b}}:=(m_{e}(p))_{p\geq 0} associated with some kernel function ebe^{b} for 𝕄b\mathbb{M}^{b}-summability.

Let u^𝔼{z}(𝕄b,𝕄a+b),(d1,d2)\hat{u}\in\mathbb{E}\{z\}_{(\mathbb{M}^{b},\mathbb{M}^{a+b}),(d_{1},d_{2})}. It holds that ^mebu^\hat{\mathcal{B}}_{m_{e^{b}}}\hat{u} is 𝕄a\mathbb{M}^{a}-summable in direction d2d_{2}. Equivalently, there exists U𝒪(G,𝔼)U\in\mathcal{O}(G,\mathbb{E}), for some sectorial region GG of bisecting direction d2d_{2} and opening larger than aπω(𝕄)a\pi\omega(\mathbb{M}) such that UU admits ^mebu^\hat{\mathcal{B}}_{m_{e^{b}}}\hat{u} as its 𝕄a\mathbb{M}^{a}-asymptotic expansion in GG. Lemma 11 guarantees that memeb,zU(z)\partial_{m_{e}\cdot m_{e^{b}},z}U(z) is the 𝕄a\mathbb{M}^{a}-sum of memeb,z(^mebu^)\partial_{m_{e}\cdot m_{e^{b}},z}(\hat{\mathcal{B}}_{m_{e^{b}}}\hat{u}) in GG.

We recall from Lemma 9 that

memeb,z(^mebu^)=^meb(me,zu^(z)).\partial_{m_{e}\cdot m_{e^{b}},z}(\hat{\mathcal{B}}_{m_{e^{b}}}\hat{u})=\hat{\mathcal{B}}_{m_{e^{b}}}\left(\partial_{m_{e},z}\hat{u}(z)\right).

This entails that the formal power series ^meb(me,zu^(z))\hat{\mathcal{B}}_{m_{e^{b}}}\left(\partial_{m_{e},z}\hat{u}(z)\right) is 𝕄a\mathbb{M}^{a}-summable along direction d2d_{2}. Moreover, its sum can be extended to an infinite sector Sd1S_{d_{1}} of bisecting direction d1d_{1}, which belongs to 𝒪𝕄b(Sd1,𝔼)\mathcal{O}^{\mathbb{M}^{b}}(S_{d_{1}},\mathbb{E}) due to the fact that assumptions in Theorem 2 hold, and regarding (13). This allows to conclude that me,zu^(z)𝔼{z}(𝕄b,𝕄a+b),(d1,d2)\partial_{m_{e},z}\hat{u}(z)\in\mathbb{E}\{z\}_{(\mathbb{M}^{b},\mathbb{M}^{a+b}),(d_{1},d_{2})}. ∎

As a consequence, a definition of moment derivative for multisums of formal power series can be stated.

Definition 10.

Let me=(me(p))p0m_{e}=(m_{e}(p))_{p\geq 0} be a sequence of moments, and let 𝕄\mathbb{M} be a strongly regular sequence which admits a nonzero proximate order. We also choose positive numbers a,ba,b such that ω(𝕄)<2/(a+b)\omega(\mathbb{M})<2/(a+b). Given any multidirection (d1,d2)2(d_{1},d_{2})\in\mathbb{R}^{2} such that |d1d2|<aω(𝕄)π2|d_{1}-d_{2}|<a\frac{\omega(\mathbb{M})\pi}{2}. For every f^𝔼{z}(𝕄b,𝕄a+b),(d1,d2)\hat{f}\in\mathbb{E}\{z\}_{(\mathbb{M}^{b},\mathbb{M}^{a+b}),(d_{1},d_{2})} we define

me,z(𝒮(𝕄b,𝕄a+b),(d1,d2)(f^)):=𝒮(𝕄b,𝕄a+b),(d1,d2)(me,zf^).\partial_{m_{e},z}(\mathcal{S}_{(\mathbb{M}^{b},\mathbb{M}^{a+b}),(d_{1},d_{2})}(\hat{f})):=\mathcal{S}_{(\mathbb{M}^{b},\mathbb{M}^{a+b}),(d_{1},d_{2})}(\partial_{m_{e},z}\hat{f}).

We conclude with some properties of multisummable series which will appear in the next section.

Lemma 12.

Let 𝔼\mathbb{E} be the Banach space of holomorphic functions on some nonempty closed neighborhood of the origin, D¯\overline{D}, endowed with the supremum norm. Let 𝕄j\mathbb{M}_{j} for j=1,2j=1,2 be two strongly regular sequences admitting nonzero proximate orders, with ω(𝕄1)<ω(𝕄2)<2\omega(\mathbb{M}_{1})<\omega(\mathbb{M}_{2})<2. Let (d1,d2)2(d_{1},d_{2})\in\mathbb{R}^{2} such that |d1d2|<π(ω(𝕄2)ω(𝕄1))|d_{1}-d_{2}|<\pi(\omega(\mathbb{M}_{2})-\omega(\mathbb{M}_{1})). We also consider a moment sequence mm and a(z)𝒪(D¯)a(z)\in\mathcal{O}(\overline{D}).

For every f^(t,z)𝔼[[t]]\hat{f}(t,z)\in\mathbb{E}[[t]] with f^𝔼{t}(𝕄1,𝕄2),(d1,d2)\hat{f}\in\mathbb{E}\{t\}_{(\mathbb{M}_{1},\mathbb{M}_{2}),(d_{1},d_{2})}, one has:

  • a(z)f^(t,z)𝔼{t}(𝕄1,𝕄2),(d1,d2)a(z)\hat{f}(t,z)\in\mathbb{E}\{t\}_{(\mathbb{M}_{1},\mathbb{M}_{2}),(d_{1},d_{2})}.

  • m,zf^𝔼{t}(𝕄1,𝕄2),(d1,d2)\partial_{m,z}\hat{f}\in\mathbb{E^{\prime}}\{t\}_{(\mathbb{M}_{1},\mathbb{M}_{2}),(d_{1},d_{2})}, where 𝔼\mathbb{E}^{\prime} is the Banach space of holomorphic functions on some D¯D\overline{D}^{\prime}\subseteq D endowed with the supremum norm.

Proof.

We write 𝕄j:=(Mj,p)p0\mathbb{M}_{j}:=(M_{j,p})_{p\geq 0} for j=1,2j=1,2. Regarding Proposition 3, one can write f^=f^1+f^2\hat{f}=\hat{f}_{1}+\hat{f}_{2}, with f^j\hat{f}_{j} being 𝕄j\mathbb{M}_{j}-summable along direction djd_{j}, j=1,2j=1,2. We write f^j(t,z)=p0aj,p(z)tp𝔼[[t]]\hat{f}_{j}(t,z)=\sum_{p\geq 0}a_{j,p}(z)t^{p}\in\mathbb{E}[[t]] for j=1,2j=1,2. We observe that

a(z)f^(t,z)=a(z)f^1(t,z)+a(z)f^2(t,z).a(z)\hat{f}(t,z)=a(z)\hat{f}_{1}(t,z)+a(z)\hat{f}_{2}(t,z).

For j=1,2j=1,2, there exists a sectorial region GjG_{j} of bisecting direction djd_{j} and opening larger than ω(𝕄j)π\omega(\mathbb{M}_{j})\pi, a holomorphic function fj𝒪(G×D¯)f_{j}\in\mathcal{O}(G\times\overline{D}), and positive constants C,AC,A such that for all tGGt\in G^{\prime}\prec G and all nn\in\mathbb{N} one has

(27) fj(t,z)p=0n1aj,p(z)tp𝔼CAnMj,n|t|n,\left\|f_{j}(t,z)-\sum_{p=0}^{n-1}a_{j,p}(z)t^{p}\right\|_{\mathbb{E}}\leq CA^{n}M_{j,n}|t|^{n},

for all (t,z)G×D¯(t,z)\in G^{\prime}\times\overline{D}. Therefore, the series a(z)f^j(t,z)𝔼[[t]]a(z)\hat{f}_{j}(t,z)\in\mathbb{E}[[t]] is such that

a(z)fj(t,z)p=0n1a(z)aj,p(z)tp𝔼C(supzD¯|a(z)|)AnMj,n|t|n,\left\|a(z)f_{j}(t,z)-\sum_{p=0}^{n-1}a(z)a_{j,p}(z)t^{p}\right\|_{\mathbb{E}}\leq C\left(\sup_{z\in\overline{D}}|a(z)|\right)A^{n}M_{j,n}|t|^{n},

for all (t,z)G×D¯(t,z)\in G^{\prime}\times\overline{D}. This entails that afjaf_{j} is the 𝕄j\mathbb{M}_{j}-sum of af^ja\hat{f}_{j} along direction djd_{j}. This concludes that a(z)f^(t,z)𝔼{t}(𝕄1,𝕄2),(d1,d2)a(z)\hat{f}(t,z)\in\mathbb{E}\{t\}_{(\mathbb{M}_{1},\mathbb{M}_{2}),(d_{1},d_{2})} again in view of Proposition 3.

For the second part of the proof, it only remains to check that for all j=1,2j=1,2 the formal power series m,zf^j(t,z)𝔼[[t]]\partial_{m,z}\hat{f}_{j}(t,z)\in\mathbb{E}^{\prime}[[t]] is 𝕄j\mathbb{M}_{j}-summable along direction djd_{j}, for j=1,2j=1,2. We first observe that the coefficients of m,zf^j(t,z)\partial_{m,z}\hat{f}_{j}(t,z) as a formal power series in tt are holomorphic and bounded functions on some common neighborhood of the origin DD^{\prime}, due to the properties of mm. Let

gn(t,z)=tn(fj(t,z)p=0n1aj,p(z)tp)g_{n}(t,z)=t^{-n}\left(f_{j}(t,z)-\sum_{p=0}^{n-1}a_{j,p}(z)t^{p}\right)

and define 𝔼′′\mathbb{E}^{\prime\prime} as the Banach space of holomorphic and bounded functions in GG^{\prime} with the sup. norm. We observe from (27), which is valid for all zD¯z\in\overline{D^{\prime}}, that

tn(fj(t,z)p=0n1aj,p(z)tp)𝔼CAnMj,n,\left\|t^{-n}\left(f_{j}(t,z)-\sum_{p=0}^{n-1}a_{j,p}(z)t^{p}\right)\right\|_{\mathbb{E}^{\prime}}\leq CA^{n}M_{j,n},

for all nn\in\mathbb{N}. We have obtained that

(28) gn(t,z)𝔼′′CAnMj,n.\left\|g_{n}(t,z)\right\|_{\mathbb{E}^{\prime\prime}}\leq CA^{n}M_{j,n}.

We write

tn(m,zfj(t,z)p=0n1(m,zaj,p(z))tp)𝔼=m,z(tn(fj(t,z)p=0n1aj,p(z)tp))𝔼.\left\|t^{-n}\left(\partial_{m,z}f_{j}(t,z)-\sum_{p=0}^{n-1}(\partial_{m,z}a_{j,p}(z))t^{p}\right)\right\|_{\mathbb{E}^{\prime}}=\left\|\partial_{m,z}\left(t^{-n}\left(f_{j}(t,z)-\sum_{p=0}^{n-1}a_{j,p}(z)t^{p}\right)\right)\right\|_{\mathbb{E}^{\prime}}.

An analogous argument as that of Theorem 2 yields

m,zgn(t,z)=12πi|ω|=r~jgn(t,ω)0(τ)E(zξ)e(ωξ)ω𝑑ξ𝑑ω,\partial_{m,z}g_{n}(t,z)=\frac{1}{2\pi i}\int_{|\omega|=\tilde{r}_{j}}g_{n}(t,\omega)\int_{0}^{\infty(\tau)}E(z\xi)\frac{e(\omega\xi)}{\omega}d\xi d\omega,

for some r~j<r\tilde{r}_{j}<r, which can be bounded from above taking into account (14) and (28). This way one arrives at

m,zgn(t,z)𝔼′′CA0B0r~jm(1)AnMj,n,\left\|\partial_{m,z}g_{n}(t,z)\right\|_{\mathbb{E}^{\prime\prime}}\leq CA_{0}B_{0}\tilde{r}_{j}m(1)A^{n}M_{j,n},

or equivalently

|m,zfj(t,z)p=0n1(m,zaj,p(z))tp|CA0B0r~jm(1)AnMj,n|t|n,\left|\partial_{m,z}f_{j}(t,z)-\sum_{p=0}^{n-1}(\partial_{m,z}a_{j,p}(z))t^{p}\right|\leq CA_{0}B_{0}\tilde{r}_{j}m(1)A^{n}M_{j,n}|t|^{n},

valid for all (t,z)G×D¯(t,z)\in G^{\prime}\times\overline{D}^{\prime}. This yields that m,zfj(t,z)\partial_{m,z}f_{j}(t,z) is the 𝕄j\mathbb{M}_{j}-sum of p0(m,zaj,p(z))tp\sum_{p\geq 0}(\partial_{m,z}a_{j,p}(z))t^{p} along direction djd_{j}.

5. Application. Multisummability of formal solutions to singularly perturbed moment differential equations

In this section, we state the main result of the present work, namely the multisummability properties of the formal solutions to certain families of singularly perturbed moment differential equations.

Let 𝕄=(Mp)p0\mathbb{M}=(M_{p})_{p\geq 0} be a strongly regular sequence which admits a nonzero proximate order.

Let k,pk,p\in\mathbb{N} with 1k<p1\leq k<p. Let m1=(m1(p))p0m_{1}=(m_{1}(p))_{p\geq 0} and m2=(m2(p))p0m_{2}=(m_{2}(p))_{p\geq 0} be two sequences of moments associated with two strongly regular sequences admitting nonzero proximate order, and assume that m1=(m1(p))p0m_{1}=(m_{1}(p))_{p\geq 0} is the moment sequence associated with 𝕄s1\mathbb{M}^{s_{1}} for some s1>0s_{1}>0. Moreover, we assume that m2=(m2(p))p0m_{2}=(m_{2}(p))_{p\geq 0} is an 𝕄\mathbb{M}-sequence of order s2>0s_{2}>0, i.e., there exist positive constants c~1,c~2\tilde{c}_{1},\tilde{c}_{2} such that

c~1p(Mp)s2m2(p)c~2p(Mp)s2,p0.\tilde{c}_{1}^{p}(M_{p})^{s_{2}}\leq m_{2}(p)\leq\tilde{c}_{2}^{p}(M_{p})^{s_{2}},\qquad p\geq 0.

We also assume that s2p>s1ks_{2}p>s_{1}k, and that ω(𝕄)s2pk<2\omega(\mathbb{M})\frac{s_{2}p}{k}<2.

Let a(z)a(z) be a holomorphic function on some closed disc centered at the origin, say D¯\overline{D}, such that a(z)1𝒪(D¯)a(z)^{-1}\in\mathcal{O}(\overline{D}). Moreover, f^(z,ε)[[z,ε]]\hat{f}(z,\varepsilon)\in\mathbb{C}[[z,\varepsilon]] and ψ^j(ε)[[ε]]\hat{\psi}_{j}(\varepsilon)\in\mathbb{C}[[\varepsilon]] are formal power series.

We consider a singularly perturbed moment differential equation of the form

(29) {εka(z)m2,zpω(z,ε)ω(z,ε)=f^(z,ε)m2,zjω(0,ε)=ψ^j(ε),j=0,,p1,\left\{\begin{aligned} \varepsilon^{k}a(z)\partial_{m_{2},z}^{p}\omega(z,\varepsilon)-\omega(z,\varepsilon)&=\hat{f}(z,\varepsilon)\\ \partial_{m_{2},z}^{j}\omega(0,\varepsilon)&=\hat{\psi}_{j}(\varepsilon),\qquad j=0,\ldots,p-1,\end{aligned}\right.

where ε\varepsilon is a small complex parameter.

The main result of the present work reads as follows.

Theorem 3.
  • (i)

    There exists a unique formal solution ω^(z,ε)[[z,ε]]\hat{\omega}(z,\varepsilon)\in\mathbb{C}[[z,\varepsilon]] of (29), which belongs to 𝒪(D¯)[[ε]]\mathcal{O}(\overline{D})[[\varepsilon]] if f^𝒪(D¯)[[ε]]\hat{f}\in\mathcal{O}(\overline{D})[[\varepsilon]].

  • (ii)

    Let s1(0,s2pk)s_{1}\in(0,\frac{s_{2}p}{k}) and choose (d1,d2)2(d_{1},d_{2})\in\mathbb{R}^{2} with

    |d1d2|<πω(𝕄)2(s2pks1).|d_{1}-d_{2}|<\frac{\pi\omega(\mathbb{M})}{2}\left(\frac{s_{2}p}{k}-s_{1}\right).

    The following statements are equivalent:

    • (ii.1)

      ω^(z,ε)\hat{\omega}(z,\varepsilon) is (𝕄s1,𝕄s2pk)(\mathbb{M}^{s_{1}},\mathbb{M}^{\frac{s_{2}p}{k}})-summable in the multidirection (d1,d2)2(d_{1},d_{2})\in\mathbb{R}^{2}.

    • (ii.2)

      f^(z,ε)\hat{f}(z,\varepsilon) and m2,zjω^(0,ε)\partial_{m_{2},z}^{j}\hat{\omega}(0,\varepsilon), j=0,1,,p1j=0,1,\ldots,p-1, are (𝕄s1,𝕄s2pk)(\mathbb{M}^{s_{1}},\mathbb{M}^{\frac{s_{2}p}{k}})-summable in the multidirection (d1,d2)(d_{1},d_{2}).

The proof of the main result is left to the end of the work, and is preceded by some auxiliary results modifying the shape of the main problem or stating asymptotic results of related problems.

Let u^=εkw^\hat{u}=\varepsilon^{k}\hat{w}. Then equation (29) can be rewritten in the form

(30) a(z)m2,zpu^(z,ε)εku^(z,ε)=f^(z,ε).a(z)\partial_{m_{2},z}^{p}\hat{u}(z,\varepsilon)-\varepsilon^{-k}\hat{u}(z,\varepsilon)=\hat{f}(z,\varepsilon).

Let us write u^(z,ε)=nkun(z)εnm1(n)\hat{u}(z,\varepsilon)=\sum_{n\geq k}u_{n}(z)\frac{\varepsilon^{n}}{m_{1}(n)}. Then the following lemma holds for the formal operator ^m1,ε\hat{\mathcal{B}}_{m_{1},\varepsilon}:

Lemma 13.

If u^εk[[z,ε]]\hat{u}\in\varepsilon^{k}\mathbb{C}[[z,\varepsilon]] then ^m1,ε(εku^)=m1,εk^m1,εu^\hat{\mathcal{B}}_{m_{1},\varepsilon}(\varepsilon^{-k}\hat{u})=\partial_{m_{1},\varepsilon}^{k}\hat{\mathcal{B}}_{m_{1},\varepsilon}\hat{u}.

Proof.

Notice that

^m1,ε(εku^(z,ε))=n0un+k(z)εnm1(n)m1(n+k)=m1,εkn0un+k(z)εn+k(m1(n+k))2=m1,εk^m1,εu^(z,ε).\hat{\mathcal{B}}_{m_{1},\varepsilon}(\varepsilon^{-k}\hat{u}(z,\varepsilon))=\sum_{n\geq 0}u_{n+k}(z)\frac{\varepsilon^{n}}{m_{1}(n)m_{1}(n+k)}\\ =\partial_{m_{1},\varepsilon}^{k}\sum_{n\geq 0}u_{n+k}(z)\frac{\varepsilon^{n+k}}{(m_{1}(n+k))^{2}}=\partial_{m_{1},\varepsilon}^{k}\hat{\mathcal{B}}_{m_{1},\varepsilon}\hat{u}(z,\varepsilon).

After applying the formal Borel transform ^m1,ε\hat{\mathcal{B}}_{m_{1},\varepsilon} to both sides of (30), from Lemma 13 we receive

(31) a(z)m2,zpU^(z,ε)m1,εkU^(z,ε)=F^(z,ε),a(z)\partial_{m_{2},z}^{p}\hat{U}(z,\varepsilon)-\partial_{m_{1},\varepsilon}^{k}\hat{U}(z,\varepsilon)=\hat{F}(z,\varepsilon),

where U^(z,ε)=^m1,εu^(z,ε)\hat{U}(z,\varepsilon)=\hat{\mathcal{B}}_{m_{1},\varepsilon}\hat{u}(z,\varepsilon) and F^(z,ε)=^m1,εf^(z,ε)\hat{F}(z,\varepsilon)=\hat{\mathcal{B}}_{m_{1},\varepsilon}\hat{f}(z,\varepsilon).

We fix s1,s2>0s_{1},s_{2}>0 such that s2p>s1ks_{2}p>s_{1}k. Let 𝕄=(Mp)p0\mathbb{M}=(M_{p})_{p\geq 0} be a strongly regular sequence which admits nonzero proximate order.

The following result is a direct consequence of Lemma 8

Lemma 14.

The formal power series ω^(z,ε)\hat{\omega}(z,\varepsilon) is (𝕄s1,𝕄s2pk)(\mathbb{M}^{s_{1}},\mathbb{M}^{\frac{s_{2}p}{k}})-multisummable (with respect to ε\varepsilon) in the multidirection (d1,d2)(d_{1},d_{2}) if and only if u^(z,ε)\hat{u}(z,\varepsilon) is (𝕄s1,𝕄s2pk)(\mathbb{M}^{s_{1}},\mathbb{M}^{\frac{s_{2}p}{k}})-multisummable (with respect to ε\varepsilon) in the multidirection (d1,d2)(d_{1},d_{2}). Let ω\omega and uu be the corresponding (𝕄s1,𝕄s2pk)(\mathbb{M}^{s_{1}},\mathbb{M}^{\frac{s_{2}p}{k}})-sums. Then, it holds that u=εkωu=\varepsilon^{k}\omega.

The next result is a direct consequence of Lemma 14 and Definition 8.

Proposition 4.

The formal power series w^(z,ε)\hat{w}(z,\varepsilon) is (𝕄s1,𝕄s2pk)(\mathbb{M}^{s_{1}},\mathbb{M}^{\frac{s_{2}p}{k}})-multisummable in the admissible multidirection (d1,d2)(d_{1},d_{2}) if and only if the following conditions are met

  1. (1)

    U^\hat{U} is 𝕄s2pks1\mathbb{M}^{\frac{s_{2}p}{k}-s_{1}}-summable in the direction d2d_{2},

  2. (2)

    the sum of U^\hat{U}, denoted by UU, is analytically continued to Sd1S_{d_{1}} and U𝒪𝕄s1(Sd1,𝔼)U\in\mathcal{O}^{\mathbb{M}^{s_{1}}}(S_{d_{1}},\mathbb{E}).

By Theorem 4 from [13] the formal power series U^(z,ε)\hat{U}(z,\varepsilon) is 𝕄s2pks1\mathbb{M}^{\frac{s_{2}p}{k}-s_{1}}-summable in the direction d2d_{2} if and only if F^(z,ε)\hat{F}(z,\varepsilon) and m2,zjw^(0,ε)\partial_{m_{2},z}^{j}\hat{w}(0,\varepsilon), j=0,1,,k1j=0,1,\ldots,k-1, are 𝕄s2pks1\mathbb{M}^{\frac{s_{2}p}{k}-s_{1}}-summable in the same direction d2d_{2}.

The next result extends Lemma 4 and Theorem 4, [13].

Theorem 4 (Compare Lemma 4 and Theorem 4, [13]).

Let us consider the Cauchy problem

(32) {(m1,tka(z)m2,zp)u(t,z)=f^(t,z)[[t,z]]m1,tju(0,z)=φj𝒪(D¯),\left\{\begin{aligned} (\partial_{m_{1},t}^{k}-a(z)\partial_{m_{2},z}^{p})u(t,z)&=\hat{f}(t,z)\in\mathbb{C}[[t,z]]\\ \partial_{m_{1},t}^{j}u(0,z)&=\varphi_{j}\in\mathcal{O}(\overline{D}),\end{aligned}\right.

with DD being a fixed neighborhood of the origin.

  • 1.

    There exists a unique formal solution u^(t,z)[[t,z]]\hat{u}(t,z)\in\mathbb{C}[[t,z]] of (32), which belongs to 𝒪(D¯)[[t]]\mathcal{O}(\overline{D})[[t]] if f^𝒪(D¯)[[t]]\hat{f}\in\mathcal{O}(\overline{D})[[t]].

  • 2.

    Let 𝔼\mathbb{E} denote the Banach space of holomorphic functions in D¯\overline{D} with the norm of the supremum. Assume that f^𝒪(D¯)[[t]]\hat{f}\in\mathcal{O}(\overline{D})[[t]]. The following statements are equivalent:

    1. 2.1.

      u^(t,z)\hat{u}(t,z) is 𝕄s2pks1\mathbb{M}^{\frac{s_{2}p}{k}-s_{1}}-summable in direction d2d_{2} (seen as a formal power series in tt with coefficients in 𝔼\mathbb{E}) with sum u(t,z)u(t,z) being an analytic solution of (32), and moreover u(t,z)𝒪𝕄s1(Sd1,𝔼)u(t,z)\in\mathcal{O}^{\mathbb{M}^{s_{1}}}(S_{d_{1}},\mathbb{E}^{\prime}), where 𝔼\mathbb{E}^{\prime} stands for the Banach space of holomorphic and bounded functions defined on D(0,r)D(0,r^{\prime}) for some 0<r<r0<r^{\prime}<r, endowed with the supremum norm.

    2. 2.2.

      f^(t,z)\hat{f}(t,z) and m2,zju^(t,0)\partial_{m_{2},z}^{j}\hat{u}(t,0) are 𝕄s2pks1\mathbb{M}^{\frac{s_{2}p}{k}-s_{1}}-summable in direction d2d_{2} with sums f(t,z)f(t,z) and m2,zju(t,0)\partial_{m_{2},z}^{j}u(t,0), respectively. Moreover, f(t,z)𝒪𝕄s1(Sd1,𝔼)f(t,z)\in\mathcal{O}^{\mathbb{M}^{s_{1}}}(S_{d_{1}},\mathbb{E}) and m2,zju(t,0)𝒪𝕄s1(Sd1)\partial_{m_{2},z}^{j}u(t,0)\in\mathcal{O}^{\mathbb{M}^{s_{1}}}(S_{d_{1}}).

Proof.

The proof heavily rests on that of Lemma 4 and Theorem 4, [13]. We only give details at the points in which it differs from the proof of those previous results.

The existence of a unique formal solution u^(t,z)\hat{u}(t,z) follows from the recursion satisfied by their coefficients, written as a formal power series in tt. Holomorphy of the coefficients is also guaranteed from that recursion formula.

For the second statement, we first observe that the implication (2.1.2.2.)2.1.\Rightarrow 2.2.) follows from the fact that 𝕄s2p/ks1\mathbb{M}^{s_{2}p/k-s_{1}}-summable formal power series along any fixed direction are compatible with respect to sums, product, and also moment derivation (see Corollary 1, [13]), we obtain that f^(t,z)\hat{f}(t,z) and m2,zju^(t,0)\partial_{m_{2},z}^{j}\hat{u}(t,0) are 𝕄s2p/ks1\mathbb{M}^{s_{2}p/k-s_{1}}-summable along the same direction. In addition to this, f(t,z)𝒪𝕄s1(Sd1,𝔼)f(t,z)\in\mathcal{O}^{\mathbb{M}^{s_{1}}}(S_{d_{1}},\mathbb{E}) due to Theorem 2. It is easy to check that m2,zju(t,0)𝒪𝕄s1(Sd1)\partial_{m_{2},z}^{j}u(t,0)\in\mathcal{O}^{\mathbb{M}^{s_{1}}}(S_{d_{1}}).

We proceed to prove the implication (2.2.2.1.)2.2.\Rightarrow 2.1.). Let D:=D(0,r)D:=D(0,r). We denote by (𝔼,𝔼)(\mathbb{E},\left\|\cdot\right\|_{\mathbb{E}}) the Banach space of holomorphic and bounded functions on D¯\overline{D}, where 𝔼\left\|\cdot\right\|_{\mathbb{E}} is the norm defined by

f(z)r:=p0|fp|rp,\left\|f(z)\right\|_{r}:=\sum_{p\geq 0}|f_{p}|r^{p},

for f𝒪(D¯)f\in\mathcal{O}(\overline{D}), with f(z)=p0fpzpf(z)=\sum_{p\geq 0}f_{p}z^{p} for zD¯z\in\overline{D}. Denoting ω^(t,z):=m2,zpu^(t,z)\hat{\omega}(t,z):=\partial_{m_{2},z}^{p}\hat{u}(t,z), one has that ω^(t,z)\hat{\omega}(t,z) is a formal solution of

(11a(z)m1,tkm2,zp)ω^(t,z)=g^(t,z),\left(1-\frac{1}{a(z)}\partial_{m_{1},t}^{k}\partial_{m_{2},z}^{-p}\right)\hat{\omega}(t,z)=\hat{g}(t,z),

where g^(t,z)=a(z)1m1,tk(ψ^0(t)+zψ^1(t)++zp1ψ^p1(t))a(z)1f^(t,z)\hat{g}(t,z)=a(z)^{-1}\partial_{m_{1},t}^{k}(\hat{\psi}_{0}(t)+z\hat{\psi}_{1}(t)+\ldots+z^{p-1}\hat{\psi}_{p-1}(t))-a(z)^{-1}\hat{f}(t,z), and with ψ^0(t)=u^(t,0)\hat{\psi}_{0}(t)=\hat{u}(t,0), ψ^j(t)=m2(0)m2(j)m2,zju^(t,0)\hat{\psi}_{j}(t)=\frac{m_{2}(0)}{m_{2}(j)}\partial_{m_{2},z}^{j}\hat{u}(t,0). We also define

ω^(t,z)=q0ω^q(t,z),\hat{\omega}(t,z)=\sum_{q\geq 0}\hat{\omega}_{q}(t,z),

with ω^0(t,z)=g^(t,z)\hat{\omega}_{0}(t,z)=\hat{g}(t,z) and ω^q(t,z)=a(z)1m1,tkm2,zpω^q1(t,z)\hat{\omega}_{q}(t,z)=a(z)^{-1}\partial_{m_{1},t}^{k}\partial_{m_{2},z}^{-p}\hat{\omega}_{q-1}(t,z) for q1q\geq 1. From the hypotheses made one has that ω^0(t,z)𝔼[[t]]\hat{\omega}_{0}(t,z)\in\mathbb{E}[[t]] is 𝕄s2p/ks1\mathbb{M}^{s_{2}p/k-s_{1}}-summable in direction d2d_{2}, with sum ω0(t,z)𝒪(G×D¯)\omega_{0}(t,z)\in\mathcal{O}(G\times\overline{D}), for some sectorial region GG of opening larger than π(s2p/ks1)ω(𝕄)\pi(s_{2}p/k-s_{1})\omega(\mathbb{M}) and bisecting direction d2d_{2}. By Theorem 4, [13] and by Theorem 2 we obtain that for every GGG^{\prime}\prec G and every SSS^{\prime}\prec S there exist positive constants C4,C5,C6C_{4},C_{5},C_{6} such that

m1,tnω0(t,z)rC4C5nm1(n)Mns2pks1exp(Ms1(C6|t|))C~1C~2nMns2pkexp(Ms1(C6|t|)),\left\|\partial_{m_{1},t}^{n}\omega_{0}(t,z)\right\|_{r}\leq C_{4}C_{5}^{n}m_{1}(n)M_{n}^{\frac{s_{2}p}{k}-s_{1}}\exp\left(M^{s_{1}}(C_{6}|t|)\right)\leq\tilde{C}_{1}\tilde{C}_{2}^{n}M_{n}^{\frac{s_{2}p}{k}}\exp\left(M^{s_{1}}(C_{6}|t|)\right),

for all tGSt\in G^{\prime}\cup S^{\prime} and n0n\in\mathbb{N}_{0}, and some C~1,C~2>0\tilde{C}_{1},\tilde{C}_{2}>0. An induction argument and the application of Lemma 5 [13] (see the proof of Theorem 4, [13]) yield that ω^q(t,z)𝔼[[t]]\hat{\omega}_{q}(t,z)\in\mathbb{E}[[t]] is 𝕄s2p/ks1\mathbb{M}^{s_{2}p/k-s_{1}}-summable in direction d2d_{2}, and

m1,tnωq(t,z)r~C~1CqC~2qk+nC5nMqk+ns2pk|z|pqm2(pq)exp(Ms1(C6|t|)),\left\|\partial_{m_{1},t}^{n}\omega_{q}(t,z)\right\|_{\tilde{r}}\leq\tilde{C}_{1}C^{q}\tilde{C}_{2}^{qk+n}C_{5}^{n}M_{qk+n}^{\frac{s_{2}p}{k}}\frac{|z|^{pq}}{m_{2}(pq)}\exp\left(M^{s_{1}}(C_{6}|t|)\right),

for all tGSt\in G^{\prime}\cup S^{\prime}, zD¯z\in\overline{D} with r~=|z|\tilde{r}=|z| and with C=a(z)1rC=\left\|a(z)^{-1}\right\|_{r}. It is direct to check that

q0m1,tnωq(t,z)r~C~1C~3nMns2pkexp(Ms1(C6|t|)),\sum_{q\geq 0}\left\|\partial_{m_{1},t}^{n}\omega_{q}(t,z)\right\|_{\tilde{r}}\leq\tilde{C}_{1}\tilde{C}_{3}^{n}M_{n}^{\frac{s_{2}p}{k}}\exp\left(M^{s_{1}}(C_{6}|t|)\right),

valid for some C~3>0\tilde{C}_{3}>0 and all tGGt\in G^{\prime}\prec G, all zD(0,r)z\in D(0,r^{\prime}), for some r>0r^{\prime}>0. This entails that ω(t,z):=q0ωq(t,z)\omega(t,z):=\sum_{q\geq 0}\omega_{q}(t,z) is a holomorphic function on (GSd1)×D(0,r)(G\cup S_{d_{1}})\times D(0,r^{\prime}) which is the 𝕄s2pk\mathbb{M}^{\frac{s_{2}p}{k}}-sum of ω^(t,z)=q0ω^q(t,z)𝔼[[t]]\hat{\omega}(t,z)=\sum_{q\geq 0}\hat{\omega}_{q}(t,z)\in\mathbb{E}[[t]] along direction d2d_{2}. A direct application of Watson’s lemma (Corollary 4.12, [22]) allows us to conclude that the sum of u^(t,z)𝔼[[t]]\hat{u}(t,z)\in\mathbb{E}[[t]], denoted by u(t,z)u(t,z) is an analytic solution of (32) with u(t,z)𝒪𝕄s1(Sd1,𝔼)u(t,z)\in\mathcal{O}^{\mathbb{M}^{s_{1}}}(S_{d_{1}},\mathbb{E}^{\prime}), where 𝔼\mathbb{E}^{\prime} stands for the Banach space of holomorphic and bounded functions defined on D(0,r)D(0,r^{\prime}). ∎

All the previous arguments allow us to conclude with the proof of the main result of the present work.

Proof of Theorem 3.

It is straightforward to check that (29) admits a unique formal solution which is obtained from the unique solution U^(t,z)\hat{U}(t,z) of (32) obtained in Theorem 4 by reversing the relations u^=εkω^\hat{u}=\varepsilon^{k}\hat{\omega} and U^(z,ε)=^m1,εu^(z,ε)\hat{U}(z,\varepsilon)=\hat{\mathcal{B}}_{m_{1},\varepsilon}\hat{u}(z,\varepsilon).

If ω^(z,ε)𝔼[[ε]]\hat{\omega}(z,\varepsilon)\in\mathbb{E}[[\varepsilon]] is (𝕄s1,𝕄s2pk)(\mathbb{M}^{s_{1}},\mathbb{M}^{\frac{s_{2}p}{k}})-summable in (d1,d2)(d_{1},d_{2}), it is clear that a(z)m2,zpω^(z,ε)𝔼[[ε]]a(z)\partial_{m_{2},z}^{p}\hat{\omega}(z,\varepsilon)\in\mathbb{E}[[\varepsilon]] is also (𝕄s1,𝕄s2pk)(\mathbb{M}^{s_{1}},\mathbb{M}^{\frac{s_{2}p}{k}})-summable in (d1,d2)(d_{1},d_{2}), and from Lemma 8 and Lemma 12 we have that εka(z)m2,zpω^(z,ε)\varepsilon^{k}a(z)\partial_{m_{2},z}^{p}\hat{\omega}(z,\varepsilon) also belongs to 𝔼{ε}(𝕄s1,𝕄s2pk),(d1,d2)\mathbb{E}\{\varepsilon\}_{(\mathbb{M}^{s_{1}},\mathbb{M}^{\frac{s_{2}p}{k}}),(d_{1},d_{2})}. The sum of two elements in 𝔼{ε}(𝕄s1,𝕄s2pk),(d1,d2)\mathbb{E}\{\varepsilon\}_{(\mathbb{M}^{s_{1}},\mathbb{M}^{\frac{s_{2}p}{k}}),(d_{1},d_{2})} remains in that space, which in view of (29) entails that

f^(z,ε)𝔼{ε}(𝕄s1,𝕄s2pk),(d1,d2).\hat{f}(z,\varepsilon)\in\mathbb{E}\{\varepsilon\}_{(\mathbb{M}^{s_{1}},\mathbb{M}^{\frac{s_{2}p}{k}}),(d_{1},d_{2})}.

It is clear that ω^𝔼{ε}(𝕄s1,𝕄s2pk),(d1,d2)\hat{\omega}\in\mathbb{E}\{\varepsilon\}_{(\mathbb{M}^{s_{1}},\mathbb{M}^{\frac{s_{2}p}{k}}),(d_{1},d_{2})} implies m2,zjω^(0,ε){ε}(𝕄s1,𝕄s2pk),(d1,d2)\partial_{m_{2},z}^{j}\hat{\omega}(0,\varepsilon)\in\mathbb{C}\{\varepsilon\}_{(\mathbb{M}^{s_{1}},\mathbb{M}^{\frac{s_{2}p}{k}}),(d_{1},d_{2})}.

We proceed to give a proof for the implication (ii.2)(ii.1)(ii.2)\Rightarrow(ii.1). In view of Proposition 4 one only has to check that U^𝔼[[ε]]\hat{U}\in\mathbb{E}[[\varepsilon]] is 𝕄s2pks1\mathbb{M}^{\frac{s_{2}p}{k}-s_{1}}-summable in direction d2d_{2} and its sum can be extended to an infinite sector of bisecting direction d1d_{1}, say Sd1S_{d_{1}}, being that extension in the space 𝒪𝕄s1(Sd1,𝔼)\mathcal{O}^{\mathbb{M}^{s_{1}}}(S_{d_{1}},\mathbb{E}). U^\hat{U} turns out to be a formal solution of (31).

We observe that F^(z,ε)𝔼[[ε]]\hat{F}(z,\varepsilon)\in\mathbb{E}[[\varepsilon]] and m1,εjU^(z,0)\partial_{m_{1},\varepsilon}^{j}\hat{U}(z,0) are both 𝕄s2pks1\mathbb{M}^{\frac{s_{2}p}{k}-s_{1}}-summable in direction d2d_{2}. On the one hand, F^(z,ε)=^f^(z,ε)\hat{F}(z,\varepsilon)=\hat{\mathcal{B}}\hat{f}(z,\varepsilon) and f^𝔼[[ε]]\hat{f}\in\mathbb{E}[[\varepsilon]] is (𝕄s1,𝕄s2pk)(\mathbb{M}^{s_{1}},\mathbb{M}^{\frac{s_{2}p}{k}})-multisummable in (d1,d2)(d_{1},d_{2}). On the other hand, we have

^m1,ε(m2,zjω^(0,ε))=m2,zj(^m1,εω^(0,ε))=m2,zj(m1,εkU^)(0,ε),\hat{\mathcal{B}}_{m_{1},\varepsilon}(\partial_{m_{2},z}^{j}\hat{\omega}(0,\varepsilon))=\partial_{m_{2},z}^{j}(\hat{\mathcal{B}}_{m_{1},\varepsilon}\hat{\omega}(0,\varepsilon))=\partial_{m_{2},z}^{j}(\partial_{m_{1},\varepsilon}^{k}\hat{U})(0,\varepsilon),

which is 𝕄s2pks1\mathbb{M}^{\frac{s_{2}p}{k}-s_{1}}-summable in d2d_{2}.

We are now in conditions to apply Theorem 4 to arrive at u^(z,ε)𝔼[[ε]]\hat{u}(z,\varepsilon)\in\mathbb{E}[[\varepsilon]] being (𝕄s1,𝕄s2pk)(\mathbb{M}^{s_{1}},\mathbb{M}^{\frac{s_{2}p}{k}})-multisummable in (d1,d2)(d_{1},d_{2}). We conclude after applying Lemma 8 that ω^(z,ε)𝔼[[ε]]\hat{\omega}(z,\varepsilon)\in\mathbb{E}[[\varepsilon]] is (𝕄s1,𝕄s2pk)(\mathbb{M}^{s_{1}},\mathbb{M}^{\frac{s_{2}p}{k}})-multisummable in (d1,d2)(d_{1},d_{2}). ∎

According to our best knowledge, the main result of the paper is new even in the case of singularly perturbed differential equations and classical multisummability (for the classical approach to multisummability see Section 10 [1] or Section 7 [14]). Namely, putting 𝕄=m2=(p!)p0\mathbb{M}=m_{2}=(p!)_{p\geq 0} and s2=1s_{2}=1 in Theorem 3 we conclude that

Corollary 2.
  • (i)

    Let a(z)𝒪(D¯)a(z)\in\mathcal{O}(\overline{D}) be such that also a(z)1𝒪(D¯)a(z)^{-1}\in\mathcal{O}(\overline{D}). Moreover, f^(z,ε)[[z,ε]]\hat{f}(z,\varepsilon)\in\mathbb{C}[[z,\varepsilon]] and ψ^j(ε)[[ε]]\hat{\psi}_{j}(\varepsilon)\in\mathbb{C}[[\varepsilon]] are formal power series. Then there exists a unique formal solution ω^(z,ε)[[z,ε]]\hat{\omega}(z,\varepsilon)\in\mathbb{C}[[z,\varepsilon]] of the singularly perturbed differential equation of the form

    {εka(z)zpω(z,ε)ω(z,ε)=f^(z,ε)zjω(0,ε)=ψ^j(ε),j=0,,p1,\left\{\begin{aligned} \varepsilon^{k}a(z)\partial_{z}^{p}\omega(z,\varepsilon)-\omega(z,\varepsilon)&=\hat{f}(z,\varepsilon)\\ \partial_{z}^{j}\omega(0,\varepsilon)&=\hat{\psi}_{j}(\varepsilon),\quad j=0,\ldots,p-1,\end{aligned}\right.

    where ε\varepsilon is a small complex parameter. If additionally f^𝒪(D¯)[[ε]]\hat{f}\in\mathcal{O}(\overline{D})[[\varepsilon]] then also ω^(z,ε)𝒪(D¯)[[ε]]\hat{\omega}(z,\varepsilon)\in\mathcal{O}(\overline{D})[[\varepsilon]].

  • (ii)

    Let s(0,pk)s\in(0,\frac{p}{k}) and choose (d1,d2)2(d_{1},d_{2})\in\mathbb{R}^{2} with |d1d2|<π2(pks)|d_{1}-d_{2}|<\frac{\pi}{2}(\frac{p}{k}-s). The following statements are equivalent:

    • (ii.1)

      ω^(z,ε)\hat{\omega}(z,\varepsilon) is (kp,1s)(\frac{k}{p},\frac{1}{s})-summable in the multidirection (d1,d2)(d_{1},d_{2}).

    • (ii.2)

      f^(z,ε)\hat{f}(z,\varepsilon) and zjω^(0,ε)\partial_{z}^{j}\hat{\omega}(0,\varepsilon), j=0,1,,p1j=0,1,\ldots,p-1, are (kp,1s)(\frac{k}{p},\frac{1}{s})-summable in the multidirection (d1,d2)(d_{1},d_{2}).

References

  • [1] W. Balser, Formal power series and linear systems of meromorphic ordinary differential equations, Universitext, Springer-Verlag, New York, 2000.
  • [2] W. Balser, M. Yoshino, Gevrey order of formal power series solutions of inhomogeneous partial differential equations with constant coefficients, Funkcial. Ekvac. 53 (2010), 411–434.
  • [3] M.I. Gomoyunov, On representation formulas for solutions of linear differential equations with Caputo fractional derivatives, Fract. Calc. Appl. Anal. 23, No 4 (2020), 1141–1160.
  • [4] G.K. Immink, Exact asymptotics of nonlinear difference equations with levels 1 and 1+, Ann. Fac. Sci. Toulouse T.XVII (2) (2008), 309–356.
  • [5] G. K. Immink, Accelero-summation of the formal solutions of nonlinear difference equations, Ann. Inst. Fourier (Grenoble) 61 (2011), No 1, 1–51.
  • [6] J. Jiménez-Garrido, S. Kamimoto, A. Lastra, J. Sanz, Multisummability in Carleman ultraholomorphic classes by means of nonzero proximate orders, J. Math. Anal. Appl. 472 (2019), No. 1, 627–686.
  • [7] J. Jiménez-Garrido, J. Sanz, G. Schindl, Injectivity and surjectivity of the asymptotic Borel map in Carleman ultraholomorphic classes, J. Math. Anal. Appl. 469 (2019), 136–168.
  • [8] A. Kilbas, H. Srivastava, J. Trujillo, Theory and applications of fractional differential equations, North-Holland Math. Stud. 204, Elsevier, Amsterdam, 2006.
  • [9] A. Lastra, S. Malek, On multiscale Gevrey and qq-Gevrey asymptotics for some linear qq-difference-differential initial value Cauchy problems, J. Difference Equ. Appl. 23 (2017), No. 8, 1397–1457.
  • [10] A. Lastra, S. Malek, J. Sanz, Summability in general Carleman ultraholomorphic classes, J. Math. Anal. Appl. 430 (2015), 1175–1206.
  • [11] A. Lastra, S. Michalik, M. Suwińska, Estimates of formal solutions for some generalized moment partial differential equations, J. Math. Anal. Appl. 500 (2021), no. 1, Paper No. 125094, 18 pp.
  • [12] A. Lastra, S. Michalik, M. Suwińska, Summability of formal solutions for a family of generalized moment integro-differential equations, Fract. Calc. Appl. Anal. 24 (2021), no. 5, 1445–1476.
  • [13] A. Lastra, S. Michalik, M. Suwińska, Summability of formal solutions for some generalized moment partial differential equations, Results Math. 76, 22 (2021), 27 pp.
  • [14] M. Loday-Richaud, Divergent series, summability and resurgence II, Simple and multiple summability, Lecture Notes in Math. 2154. Springer, 2016.
  • [15] S. Malek, Asymptotics and confluence for some linear qq-differencedifferential Cauchy problem, J. Geom. Anal 32, 93 (2022).
  • [16] S. Michalik, Analytic solutions of moment partial differential equations with constant coefficients, Funkcial. Ekvac. 56 (2013), 19–50.
  • [17] S. Michalik, Summability of formal solutions of linear partial differential equations with divergent initial data, J. Math. Anal. Appl. 406 (2013), no. 1, 243–260.
  • [18] S. Michalik, B. Tkacz, The Stokes phenomenon for some moment partial differential equations, J. Dyn. Control Syst. 25 (2019), no. 4, 573–598.
  • [19] P. Remy, Summability of the formal power series solutions of a certain class of inhomogeneous nonlinear partial differential equations with a single level, J. Differ. Equations 313 (2022), 450–502.
  • [20] P. Remy, Gevrey order and summability of formal series solutions of certain classes of inhomogeneous linear integro-differential equations with variable coefficients, J. Dyn. Control Syst. 23 (2017), 853–878.
  • [21] L. Ren, J. Wang, M. Fečkan, Asymptotically periodic solutions for Caputo type fractional evolution equations, Fract. Calc. Appl. Anal. 21, No 5 (2019), 1294–1312.
  • [22] J. Sanz, Flat functions in Carleman ultraholomorphic classes via proximate orders, J. Math. Anal. Appl. 415 (2014), no. 2, 623–643.
  • [23] J. Sanz, Asymptotic analysis and summability of formal power series, Analytic, Algebraic and Geometric Aspects of Differential Equations, Trends Math., Birkhäuser/Springer, Cham (2017), 199–262.
  • [24] M. Suwińska, Gevrey estimates of formal solutions for certain moment partial differential equations with variable coefficients, J. Dyn. Control Syst. 27 (2021), no. 2, 355–370.
  • [25] V. Thilliez, Division by flat ultradifferentiable functions and sectorial extensions, Results Math. 44 (2003), 169–188.