Abstract
We calculate the rate of convergence of the double rational Fourier series for regular, bounded, measurable, and two-variable functions. The rectangular oscillation of the two-variable function is used to quantify this rate. Additionally, we give an approximation of convergence rate of the double rational Fourier series for continuous functions with generalized bounded variation.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
Bojanic and Waterman [4, 5, 27] examined the quantitative version of the Dirichlet-Jordan test, concentrating on a class of functions with generalized bounded variation. In [27], Waterman presented an estimate of the Fourier series convergence rate for functions that are closer to the class of harmonic bounded variation. Móricz contributed a quantitative version of the Dirichlet-Jordan test for double Fourier series [14], considering functions with bounded variation.
This study extends and broadens the findings to include double rational Fourier series applied to continuous two-variable functions, which are close to the harmonic bounded variation class, building on the foundation established in [27] and [14]. The study of (double) rational Fourier series is important from a variety of perspectives, including approximation theory and Fourier analysis [10, 12, 16, 21, 22, 24] as well as signal processing, system identification, and control theory [3, 15, 17,18,19, 25].
During the 1920 s, two independent definitions of the rational orthogonal system emerged, attributed to Malmquist and Takenaka. Consequently, this orthogonal system is alternatively known as the Takenaka-Malmquist (or Malmquist-Takenaka) system. M. M. Džrbažyan [8] conducted research on rational Fourier series, employing the rational orthogonal system as his foundation. The definition of the rational orthogonal system is outlined as follows:
and \(\phi _{-n}(e^{ix})=\overline{\phi _n(e^{ix})},\ \forall n\in {\mathbb {N}}\). Here, \(\{\alpha _n\}_{n\in {\mathbb {N}}}\) is complex sequence such that \(\alpha _k\)’s are in open unit disk \({\mathbb {D}}\). Achieser [2, p. 244] (or see [6, p. 151]) observed that the system in (1) is complete in \(L^2[0,2\pi ]\) if and only if \(\sum _{n=1}^{\infty }(1-\mid \alpha _k\mid )=\infty .\) The simplest way to satisfy the previous completeness condition is to assume
In the sequel, the above condition (2) is assumed to be satisfied.
In 2009, Fu and Li [9] worked on convergence of the Cesàro mean of rational Fourier series. In 2012, Tan and Zhou [24] studied convergence behaviour of rational Fourier series and proved Jordan Dirichlet test for it. Later on, in 2013, Tan and Qian [22] gave quantitative version of Jordan Dirichlet test and proved convergence of rational Fourier series for functions of harmonic bounded variation. Recently, the quantitative version of Jordan Dirichlet test of rational Fourier series was generalized for certain functions of \(\Lambda -\) bounded variation [12]. In 2013, Tan and Zhou [23] worked on order of magnitude of rational Fourier coefficients and recently those results were further extended for multiple rational Fourier coefficients [11]. In 2022, Mnatsakanyan [13] proved \(L^p\) bounds for the maximal partial sum operator of the rational Fourier series. Recently, Abdullayev and Savchuk [1] gave analogous version of Fejer theorem for rational Fourier series.
If an integrable function f is \(2\pi \) periodic in both the variables, then double rational Fourier series of f is defined as
where \(\hat{f}(m,n)\) is the \((m,n)^{th}\) double rational Fourier coefficient of f, given by
Note that, if \(\alpha _k=0,\ \forall k\in {\mathbb {N}}\) in (1) then \(\{\phi _n(e^{ix})\}_{n=-\infty }^{\infty }\) reduces to \(\{e^{inx}\}_{n=-\infty }^{\infty }\) and therefore double rational Fourier series reduces to classical double Fourier series.
2 Preliminaries and Notations
The notations and definitions mentioned in this section will be used in the sequel.
-
1.
Waterman [26] defined the concept of \(\Lambda -\) bounded variation as follows.
Let \(\{\lambda _n\}\) be a non decreasing sequence of positive numbers such that \(\sum \frac{1}{\lambda _n}\) diverges then a real valued function f is said to be of \(\Lambda -\) bounded variation on [a, b] (i.e. \(f\in \Lambda BV[a,b])\) if
$$\begin{aligned} V_1(f,[a,b])=\sup \sum _{k=1}^n\frac{\mid f(b_k)-f(a_k)\mid }{\lambda _k}<\infty , \end{aligned}$$for every sequence of non-overlapping intervals \([a_k,b_k],k=1,...,n,\) which is contained in [a, b].
The notion of bounded variation is extended for two variables in various ways like in the sense of Arzelà, Vitali, Hardy and others as can be seen in [7]. Here, we will need the definition of \(\Lambda \) bounded variation in the sense of Hardy, as given by Sablin [20]. A measurable function f defined on a rectangle \(I\times J:=[a,b]\times [c,d]\) is said to be of \(\Lambda -\) bounded variation in the sense of Hardy (that is, \(f\in \Lambda BV(R^2)\)) if the marginal functions \(f(.,c)\in \Lambda _1 BV([a,b]),\ f(a,.)\in \Lambda _2 BV([c,d])\) and
$$\begin{aligned}V_2(f,R^2)=\sup _{J_1,J_2}\left\{ \sum _i\sum _j \frac{\mid f(I_i\times K_j)\mid }{\lambda _i\mu _j}\right\} <\infty . \end{aligned}$$Here, \(\Lambda =(\Lambda _1,\Lambda _2)\) where \(\Lambda _1=\{\lambda _i\}_{i=1}^{\infty }\) and \(\Lambda _2=\{\mu _j\}_{j=1}^{\infty }\); \(J_1\) and \(J_2\) are finite collections of non-overlapping subintervals \(\{I_i\}\) and \(\{K_j\}\) in [a, b] and [c, d] respectively; and \(f(I\times J)=\Delta f_{(a,c)}^{(b,d)}=f(b,d)-f(a,d)-f(b,c)+f(a,c).\)
-
2.
As mentioned in [27], we suppose that \(\frac{\lambda _{\mid j\mid }}{\mid j\mid }\) is non increasing and if m is fixed then H(t) is a continuously non increasing function on \([-\pi ,0)\) and \((0,\pi ]\) such that
$$\begin{aligned} H(t)=\frac{\lambda _{\mid j\mid }}{t};\ t=\frac{j\pi }{m+1} \text{ and } j=\pm 1,\pm 2,...,\pm (m+1). \end{aligned}$$Similarly, we suppose that \(\frac{\mu _{\mid k\mid }}{\mid k\mid }\) is non increasing and if n is fixed then G(t) is a continuously non increasing function on \([-\pi ,0)\) and \((0,\pi ]\) such that
$$\begin{aligned} G(t)=\frac{\mu _{\mid k\mid }}{t};\ t=\frac{k\pi }{n+1} \text{ and } k=\pm 1,\pm 2,...,\pm (n+1). \end{aligned}$$ -
3.
The oscillation of a function \(g:[a,b]\rightarrow {\mathbb {C}}\) over a subinterval \([a_1,b_1]\) of [a, b] is defined as
$$\begin{aligned}osc_1(g,[a_1,b_1]) =\sup _{t,y\in [a_1,b_1]}\mid g(t)-g(y)\mid .\end{aligned}$$ -
4.
The oscillation of a function \(h:[a,b]\times [c,d]\rightarrow {\mathbb {C}}\) over a sub-rectangle \([a_1,b_1]\times [c_1,d_1]\) of \([a,b]\times [c,d]\) is defined as
$$\begin{aligned} osc_2&(h,[a_1,b_1]\times [c_1,d_1])\\&=\sup _{\begin{array}{c} u_1,u_2\in [a_1,b_1];\\ v_1,v_2\in [c_1,d_1] \end{array}}\mid h(u_1,v_1) -h(u_2,v_1)-h(u_1,v_2)+h(u_2,v_2)\mid . \end{aligned}$$ -
5.
For \(m\in {\mathbb {N}}\cup \{0\}\), we define
$$\begin{aligned}\eta _{km}=\frac{k\pi }{m+1},\ \forall k=0,1,2,...,m;\\ I_{km}^{+}=[\eta _{km},\eta _{(k+1)m}]\end{aligned}$$and
$$\begin{aligned}I_{km}^{-}=[-\eta _{(k+1)m},-\eta _{km}].\end{aligned}$$ -
6.
For a function \(f\in \mathbb {\overline{T}}^2:=[-\pi ,\pi ]^2,\) we define
$$\begin{aligned} S(f;x,y)&:=\frac{1}{4}\{f(x+0,y+0)+f(x-0,y+0)+f(x+0,y-0)\\&\quad +f(x-0,y-0)\} \end{aligned}$$and \( \psi (u,v):=\psi _{xy}(u,v)\)
$$\begin{aligned}:= {\left\{ \begin{array}{ll} S(f;x,y)-f(x-u,y-v) &{}\quad \hbox {if}\ u,v\ne 0 \\ S(f;x,y)-\frac{f(x-0,y-v)+f(x+0,y-v)}{2} &{} \quad \text {if } u=0,\ v\ne 0 \\ S(f;x,y)-\frac{f(x-u,y+0)+f(x-u,y-0)}{2} &{} \quad \text {if } u\ne 0,\ v=0 \\ 0 &{} \quad \hbox {if}\ u=v=0. \end{array}\right. } \end{aligned}$$Note that, here \(f(x+0,y+0):=\lim \{f(x+u,y+v):u,v\rightarrow 0\text { and }u,v>0\}\) and similary other limits like \(f(x-0,y+0),\ f(x+0,y-0)\) and \(f(x-0,y-0)\) are defined.
-
7.
In view of [24, Lemma 2.1] and [22, Lemma 2.1], for \(\alpha _k=\mid \alpha _k\mid e^{ia_k},\ n\in {\mathbb {N}}\) and \(x\in [-\pi ,\pi ]\), the partial sums of rational Fourier series of integrable function g(x) is given by
$$\begin{aligned} S_ng(x)=\frac{1}{\pi }\int _{-\pi }^{\pi }g(x-t)D_n(x-t,x)dt,\end{aligned}$$where rational Dirchlet kernel is given by
$$\begin{aligned} D_n(t,x)=\frac{1}{2}\sum _{k=-n}^{n}\overline{\phi _k(e^{it})} \phi _k(e^{ix})=\frac{\sin \left[ \frac{x-t}{2}+\theta _n(t,x) \right] }{2\sin \left( \frac{x-t}{2}\right) } \end{aligned}$$(5)and
$$\begin{aligned}\theta _n(t,x)=\int _t^x\sum _{k=1}^n\frac{1-\mid \alpha _k\mid ^2}{1-2\mid \alpha _k\mid \cos (y-a_k)+\mid \alpha _k\mid ^2}dy.\end{aligned}$$Here,
$$\begin{aligned} \int _{-\pi }^{\pi }D_n(t,x)du=\pi \end{aligned}$$(6) -
8.
For \(m,n\in {\mathbb {N}}\), the partial sum of double rational Fourier series of an integrable function f(x, y) is given by
$$\begin{aligned}S_{mn}(f;x,y)=\sum _{j=-m}^{m}\sum _{k=-n}^{n}\hat{f}(j,k)\phi _j(e^{ix})\phi _k(e^{iy}).\end{aligned}$$Using (4), (5) and the fact that f is \(2\pi \) periodic in both variables, we have
$$\begin{aligned} S_{mn}(&f;x,y)\nonumber \\&=\frac{1}{4\pi ^2}\sum _{j=-m}^{m}\sum _{k=-n}^{n}\left\{ \int \int _{\mathbb {\overline{T}}^2}f(u,v) \overline{\phi _j(e^{iu})\phi _k(e^{iv})} \phi _j(e^{ix})\phi _k(e^{iy})dudv\right\} \nonumber \\&=\frac{1}{\pi ^2}\int \int _{\mathbb {\overline{T}}^2}f(u,v)D_m(u,x)D_n(v,y)du dv\nonumber \\&=\frac{1}{\pi ^2}\int \int _{\mathbb {\overline{T}}^2}f(x-u,y-v)D_m(x-u,x)D_n(y-v,y)du dv. \end{aligned}$$(7) -
9.
A two variable function f(x, y) is said to be regular in \(\mathbb {\overline{T}}^2\) if \(f(x\pm 0, y\pm 0),\ f(x\pm 0,.)\) and \(f(.,y\pm 0)\) exist for \((x,y)\in \mathbb {\overline{T}}^2.\)
3 Results
Understanding the rate of convergence is essential because it provides insights into how many terms are needed for a rational Fourier series to approximate a given function accurately. Analyzing and characterizing the rate of convergence is a crucial step in utilizing rational Fourier series effectively in a wide range of applications. The upcoming result, Theorem 1, gives rate of convergence of double rational Fourier series for a bounded, measurable and regular two variable function in terms of osciallations. Later on, the convergence is given in terms for two variable functions of generalized bounded variation in Theorem 2 and Corollary 1.
Theorem 1
If f is bounded, measurable and regular function on \(\mathbb {\overline{T}}^2\) and \(2\pi \) periodic in each variable, then for any \(m,n\in {\mathbb {N}}\cup \{0\},\)
Proof
Let \(h(u,v)=\psi (u,v)-\psi (u,0)-\psi (0,v).\) Therefore,
We will mainly use the following inequalities
and
Using (9), we have
Similarly,
Let
Now, using summation by parts, we get
In view of [22, Lemma 2.3], we have for \(0<t<\pi \),
Thus,
Similarly,
Using (10), we get
Using summation by parts,
Similarly,
Using double summation by parts,
Using (11), we get
Now from all the inequalities, it is easy to deduce
Now, let
In view of [22, Lemma 2.3], we have for \(0<t<\pi \),
Thus,
Now using (9),(10),(12) and following similar steps as before, we get
and
In view of [12, Theorem 1], we get
and
Thus, the result is proved. \(\square \)
Remark 1
Theorem 1 is analogous result of [14, Theorem 2] for double rational Fourier series and it is extension of [12, Theorem 1] for two variable functions.
Theorem 2
If \(f\in \Lambda BV ([0,\pi ]\times [0,\pi ]);\ f\) is continuous in \([0,\pi ]\times [0,\pi ];\) for \(m,n\in {\mathbb {N}};\) \(\frac{\pi }{m+1}=a_m<a_{m-1}<...<a_0=\pi \) and \(\frac{\pi }{n+1}=b_n<b_{n-1}<...<b_0=\pi ;\) then
Proof
We will follow similar technique as in [5, 14, 27]. Let
and
Now for \(\ j=0,1,\dots , m-1\) and \(k=0,1,\dots n-1,\) define M(u, v) on the rectangle \([\pi /(m+1),\pi )\times [\pi /(n+1),\pi )\) and N(u) and R(v) on the intervals \([\pi /(m+1),\pi )\) and \([\pi /(n+1),\pi )\) respectively as follows:
and
Therefore,
and
Now, applying double summation by parts, we have
Note that, we have \(-H(u)\) and \(-G(v)\) as non decreasing and continuous function on \((0,\pi ]\). Thus, by properties of two dimensional Riemann-Stieltjes integrals we can estimate \(A_1,\ A_2\) and \(A_3\) in the following manner.
Now, consider \(A_2\).
Similarly,
Since f is continuous in \([0,\pi ]\times [0,\pi ]\),we get,
and
Thus,
and
Consider,
Thus, we get,
Similarly, we get,
By substituting (15), (17), (18) and (16) in (14), we get the result. \(\square \)
Remark 2
Theorem 2 is extension of [27, Theorem on p. 52] for two variable continuous functions.
Corollary 1
If \(f\in \Lambda BV ([-\pi ,\pi ]\times [-\pi ,\pi ]);\ f\) is continuous in \([-\pi ,\pi ]\times [-\pi ,\pi ];\) for \(m,n\in {\mathbb {N}};\) \(\frac{\pi }{m+1}=a_m^{(1)}<a_{m-1}^{(1)}<...<a_0^{(1)}=\pi ,\ \frac{-\pi }{m+1}=a_m^{(2)}>a_{m-1}^{(2)}>...>a_0^{(2)}=-\pi ,\ \frac{\pi }{n+1}=b_n^{(1)}<b_{n-1}^{(1)}<...<b_0^{(1)}=\pi \) and \(\frac{-\pi }{n+1}=b_n^{(2)}>b_{n-1}^{(2)}>...>b_0^{(2)}=-\pi ,\) then
where
for \(a>0, \ T(a)=[0,a]\) and for \(a<0, \ T(a)=[a,0].\)
Proof
In view of Theorem 1, [27, Theorem on p. 52] and by proceeding in the similar manner as in the proof of Theorem 2, the result is proved. \(\square \)
Remark 3
In the above corollary, by setting \(\lambda _k=\mu _k=1,\) for all \(k\in {\mathbb {N}};\) \(a_i^{(1)}=\frac{\pi }{i+1},\ a_i^{(2)}=\frac{-\pi }{i+1},\) for \(i=0,1,...,m;\ b_j^{(1)}=\frac{\pi }{j+1},\ b_j^{(2)}=\frac{-\pi }{j+1}\) for \(j=0,1,...,n;\) and \(H(t)=G(t)=1/t;\) we get analogous result of [14, Theorem 3] for double rational Fourier series of continuous functions of bounded variation in two variables.
Data availability
Not applicable
References
Abdullayev, F.G., Savchuk, V.V.: Fejér-type positive operator based on Takenaka-Malmquist system on unit circle. J. Math. Anal. Appl. 529(2), 127298 (2024)
Achieser, N.I.: Theory of approximation Translated from the Russian. In: ed Charles J, Frederick Ungar Publishing Co, Hyman. New York. p 1956 (1947)
Akçay, H.: On the uniform approximation of discrete-time systems by generalized Fourier series. IEEE Trans. Signal Process. 49(7), 1461–1467 (2001)
Bojanić, R.: An estimate of the rate of convergence for Fourier series of functions of bounded variation. Publ. Inst. Math. (Beograd) 26(40), 57–60 (1979)
Ranko, B., Waterman, D.: On the rate of convergence of Fourier series of functions of generalized bounded variation. Akad. Nauka Umjet. Bosne Hercegov. Rad. Odjelj. Prirod. Mat. Nauka 22, 5–11 (1983)
Bultheel, A., González-Vera, P., Hendriksen, E., Njastad, O., et al.: Cambridge Monographs on Applied and Computational Mathematics, p. 5. Cambridge University Press, Cambridge (1999)
Clarkson, J.A., Adams, C.R.: On definitions of bounded variation for functions of two variables. Trans. Am. Math. Soc. 35(4), 824–854 (1933)
Džrbašyan, M.M.: On the theory of series of Fourier in terms of rational functions. Akad. Nauk Armyan. SSR. Izv. Fiz.-Mat Estest. Tehn. Nauki 9, 3–28 (1956)
Fu, Y.X., Li, L.Q.: Convergence of the Cesàro mean for rational orthonormal bases. Acta Math. Sin. Engl. Ser. 25(4), 581–592 (2009)
Kazlouskaya, N.Y., Rovba, Y.A.: Approximation of the function \(|{\rm SIN}\, x|^s\) by the partial sums of the trigonmometric rational Fourier series. Dokl. Nats. Akad. Nauk Belarusi 65(1), 11–17 (2021)
Khachar, H.J., Vyas, R.G.: A note on multiple rational Fourier series. Period. Math. Hung. 85(2), 264–274 (2022)
Khachar, Hardeepbhai J., Vyas, Rajendra G.: Rate of convergence for rational and conjugate rational fourier series of functions of generalized bounded variation. Acta Comment. Univ. Tartu. Math. 26(2), 233–241 (2022)
Mnatsakanyan, G.: On almost-everywhere convergence of Malmquist-Takenaka series. J. Funct. Anal. 282(12), 109461–109533 (2022)
Móricz, F.: A quantitative version of the Dirichlet-Jordan test for double Fourier series. J. Approx. Theory 71(3), 344–358 (1992)
Ninness, Brett, Gustafsson, Fredrik: A unifying construction of orthonormal bases for system identification. IEEE Trans. Automat. Control 42(4), 515–521 (1997)
Pap, Margit, Schipp, Ferenc: Equilibrium conditions for the Malmquist-Takenaka systems. Acta Sci. Math. (Szeged) 81(3–4), 469–482 (2015)
Qian, Tao: Two-dimensional adaptive Fourier decomposition. Math. Methods Appl. Sci. 39(10), 2431–2448 (2016)
Qian, Tao, Li, Hong, Stessin, Michael: Comparison of adaptive mono-component decompositions. Nonlinear Anal. Real World Appl. 14(2), 1055–1074 (2013)
Tao, Q., Xiaoyin, W., Liming, Z.: MIMO frequency domain system identification using matrix-valued orthonormal functions. Automatica J. IFAC 133, 109882 (2021)
Sablin, A.I.: \(\Lambda \)-variation and Fourier series. Izv. Vyssh. Uchebn. Zaved. Mat. 10, 66–68 (1987)
Savchuk, V.V.: Best approximations of the Cauchy-Szegö kernel in the mean on the unit circle. Ukr. Math. J. 70(5), 817–825 (2018)
Tan, L., Qian, T.: On convergence of rational Fourier series of functions of bounded variations (in Chinese). Sci. Sin. Math. 43(6), 541–550 (2013). https://doi.org/10.1360/012013-127
Tan, L., Zhou, C.: On the order of rational Fourier coefficients of various bounded variations. Compl. Var. Elliptic Equ. 58(12), 1737–1743 (2013)
Tan, Lihui, Zhou, Chaoying: The point-wise convergence of general rational Fourier series. Math. Methods Appl. Sci. 35(17), 2067–2074 (2012)
Z. Wang, Chi Man W., Agostinho R., Tao Q., and Feng W.: Adaptive Fourier decomposition for multi-channel signal analysis. IEEE Trans. Signal Process. 70, 903–918 (2022)
Waterman, Daniel: On \(L\)-bounded variation. Studia Math. 57(1), 33–45 (1976)
Waterman, D.: Estimating functions by partial sums of their Fourier series. J. Math. Anal. Appl. 87(1), 51–57 (1982)
Acknowledgements
The authors are thankful to the referee for their valuable suggestions in improving quality of the paper.
Author information
Authors and Affiliations
Contributions
Both the authors contributed equally.
Corresponding author
Ethics declarations
Funding
The first author is thankful to Council of Scientific and Industrial Research, India for providing financial support through SRF (File no.: 09/0114(11228)/2021-EMR-I).
Conflict of interest
There is no conflict of interest.
Additional information
Communicated by Tao Qian.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Khachar, H.J., Vyas, R.G. Rate of Convergence for Double Rational Fourier Series. Complex Anal. Oper. Theory 18, 99 (2024). https://doi.org/10.1007/s11785-023-01479-w
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s11785-023-01479-w
Keywords
- Rational Fourier series
- Double rational Fourier series
- Rate of convergence
- Generalized bounded variation