Abstract
The present work is devoted to the construction of optimal quadrature formulas for the approximate calculation of the integrals in the Sobolev space . Here, is the Hilbert space of periodic and complex-valued functions whose m-th generalized derivatives are square-integrable. Here, firstly, in order to obtain an upper bound for the error of the quadrature formula, the norm of the error functional is calculated. For this, the extremal function of the considered quadrature formula is used. By minimizing the norm of the error functional with respect to the coefficients, an optimal quadrature formula is then obtained. Using the explicit form of the optimal coefficients, the norm of the error functional of the optimal quadrature formula is calculated. The convergence of the constructed optimal quadrature formula is investigated, and it is shown that the rate of convergence of the optimal quadrature formula is for and for . Finally, we present numerical results of comparison for absolute errors of the optimal quadrature formula with the weight in the case and the Midpoint formula. There, one can see the advantage of the optimal quadrature formulas.
Keywords:
Sobolev space of periodic functions; extremal function; error functional; optimal quadrature formulas; oscillating functions MSC:
65D30; 65D32
1. Introduction and Statement of the Problem
Approximate calculation of the integral
where is an integer and , plays an important role in computational mathematics. It is well-known that numerical calculation of such integrals encounters difficulties for large values of because the integrand oscillates strongly. Computation of the integral (1) are often done by the Filon method [1]. The Filon method reminds us of the Simpson quadrature formula. However, while in the Simpson method the entire integrand is replaced by a parabola, in the Filon method, only the function is replaced by a parabola. This way Filon obtained the quadrature formula with coefficients depending on [1].
Various aspects of the calculation of the integral (1) are discussed in a number of papers [2,3,4,5,6,7,8,9,10,11,12,13,14,15,16], where there are fairly complete bibliographies.
In the present work, combining S.L. Sobolev’s and I. Babuška’s methods for the approximate computation of the integral (1), an optimal quadrature formula is constructed, and the square of the norm for the error functional of the obtained optimal quadrature formula is evaluated in a certain Hilbert space.
Suppose is the Hilbert space of -periodic, complex-valued functions , , which are square-integrable with the mth generalized derivative. Every element of the space is a class of such functions that differ from each other by a constant term, and the inner product of the elements in this space is defined as
Here, the notation is the complex conjugate to g. The norm of a function in the space is defined by
For , we consider a quadrature formula of the form
where are coefficients of the quadrature formula, .
The error of the quadrature formula is given in the form
Here, is the Dirac delta-function,
ℓ is the periodic functional of the error for the quadrature formula, and is the value of the error functional at .
The challenge in the construction of an optimal quadrature formula for the approximate calculation of the integral (1) is calculating the quantity
that is, finding a function from for which the exact upper bound is attained, and the coefficients for which the exact lower bound are attained in (7).
In this case, the function is said to be the extremal for the given quadrature formula, values are called the the optimal coefficients for the quadrature Formula (4) in the space , and is the norm for the error functional of the optimal quadrature formula.
It should be noted that is the conjugate space to the space , and it consists of all periodic functionals that are orthogonal to the unity, i.e.,
The remaining part of the paper is organized as follows. In Section 2, we present the main results; i.e., we obtain the extremal function, the analytic expressions for the optimal coefficients, and the norm of the error functional for the optimal quadrature formula of the form (4). Further, in Section 3, Section 4, Section 5 and Section 6, we provide proofs of the main results.
2. Main Results
The central results of the present paper are the following.
Theorem 1.
Theorem 2.
Theorem 3.
In the space, the square of the norm of the error functionalfor the optimal quadrature formula of the form (4) has the following expression:
Hence, in particular, we obtain the following.
Remark 1.
For, we have the well-known optimal quadrature formula, which is the following rectangular formula
Moreover, for the norm of the error functiona, we obtain
whereis the Bernoulli number.
Remark 2.
If ω is a multiple for the number N of the nodes of the quadrature formula, i.e.,, then
and
Remark 3.
Calculations show that
wherevalues are the coefficients of the Euler–Frobenius polynomial of degree.
3. The Extremal Function to the Error Functional for the Quadrature Formula (4)
In order to find an explicit form of the norm for the error functional , we use an extremal function of the given functional, i.e., a function for which the following holds:
Proof of Theorem 1.
The idea of the proof is as follows. Using the Formula (2) and applying the Riesz theorem for the error functional ℓ, we find an explicit expression for ℓ by a new function , which is an element of the space . It is easy to see that the function is related to the error functional ℓ by the following differential equation:
Indeed, by the definition of the inner product, we have
Integrating by parts the right-hand side of (14) and taking into account that a function is infinitely differentiable and finite, i.e., , we obtain
On the other hand, by definition of the error functional , we have
Placing (16) on the left-hand side of (15), we find (13). Since the space of the infinity differentiable and finite functions is dense in the space , i.e., any function from can be approximated with arbitrary high accuracy by a sequence of the functions from the space , it follows that a periodic solution of Equation (13) is the extremal function of the error functional ℓ for the quadrature Formula (4) and .
Further, we find a periodic solution of Equation (13). In order to solve this equation, we use the Fourier transforms. For this, we provide some useful formulas:
Applying the Fourier transform to both sides of Equation (13), we obtain
where
By virtue of the Formula (20), we have
Now, using the definition of the Fourier transform, i.e., Formula (17), we directly obtain
Using the known properties of the delta-function, we have
and introducing the notation and using the Formula (21), we rewrite Equality (26) in the form
The right-hand side of (28) is equal to 0 at , since, in virtue of (8), the singularities
are mutually canceled.
Therefore, we can divide both sides of Equation (28) by . This division is not uniquely defined. For Equation (28), the function is defined up to the term of the form . Taking into account the aforementioned as well as the properties of the delta-function, we obtain
Applying the inverse Fourier transform to both sides of Equation (29) and using Formulas (18) and (19), after some calculations, we have
From (30), it follows that
Theorem 1 is completely proved. □
Now, we give an important theorem on zeros of the extremal function which is due to I. Babuška. This theorem was proved in the language of functional analysis (see, for example [17,18]).
Theorem 4
(I. Babuška). Suppose the error functional
- (a)
- is defined on the space—i.e., its value at constant is zero—and
- (b)
- is optimal; i.e., among all functionals of the formwith given system of nodes, it has the lowest norm in.
Then there exists a solutionof the equation
which is zero at pointsand belongs to.
4. The Square of the Error Functional of the Quadrature Formula (4)
It is easy to proof Babuška’s theorem also by algebraic way. For the sake of completeness, we give this proof here as well. A quadrature formula with the error functional in the space can be characterized by two manners. From one side, this quadrature formula is defined by coefficients , under the condition
From the other side, it is defined by the extremal function of the quadrature formula. The square of the norm for the error functional of the quadrature formula is expressed by the bilinear form with respect to coefficients of the formula and values of the extremal function. Indeed, since is a Hilbert space, we have
where is the extremal function of our quadrature formula. However,
so
Simple calculations show that
By virtue of
it follows that
Using Equalities (34) and (35) for the square of the norm for the error functional of the quadrature formula, we obtain the following analytical expression
For finding the minimum of the square for the error functional of the quadrature formula, we apply the method of indefinite Lagrange multipliers. For this we consider the following function:
Setting to 0 all partial derivatives by , , and of the function , we have
These give the following system of equations
It is not difficult to see that
Taking into account (37), it follows immediately that .
In order to find unknown coefficients , it is enough to solve the systems (38) and (39). The solution of this system, which we denote by , and , is a stationary point for the function . From the theory of the Lagrange method, it follows that values are searching values of the coefficients for the quadrature formula, . They give the conditional minimum for the square of the norm provided that (8) holds.
By virtue of condition (37), we see that the extremal function , defined by (9), vanishes at the nodes of the quadrature Formula (4), i.e., . This proves Babuška’s theorem.
Here, we assume that the systems (38) and (39) are solvable. Its solvability follows from the general theory of the Lagrange multipliers. However, as is shown in the calculations, the matrix of the systems (38) and (39) coincides with the matrix of the system considered in [10] in the construction of optimal cubature formulas in the Sobolev space of periodic functions. In [10], uniqueness of the set of the optimal coefficients was also proved. Hence, it immediately follows that the systems (38) and (39) have a unique solution.
5. Optimal Coefficients of the Quadrature Formula (4)
In the present section, we prove Theorem 2. In order to prove the theorem, it is enough to solve the systems (38) and (39) with respect to coefficients , . For a solution of this system, we search in the form
where is an unknown function. We are directly convinced that values, defined by Formula (40), , satisfy Equality (39). Substituting (40) into (38), we obtain
We introduce the following notation:
It is clear that
Therefore, denoting and then , we rewrite (42) in the form
From (41)–(43), it follows that
Hence, we have
From the uniqueness of the solution for the systems (38) and (39), we obtain
Hence, we directly obtain
We now refer to calculations of
As is known [11],
Taking the derivative of order with respect to z from both sides of (46) and taking (45) into account, we obtain
Now we calculate For this, we use Euler’s known formula
Denoting , we have
where , and .
Consequently,
Hence, as is known (see, for example, [18]), the Euler–Frobenius polynomial is defined by the formula
and
where
From (48), we have
By virtue of (49) and (47), Equality (45) takes the form
Using the formula
after some calculations, we obtain
By virtue of symmetry for the coefficients of Euler’s polynomial, we have
Accordingly, from (50), (44), and (45), we obtain
6. The Norm for the Error Functional of the Optimal Quadrature Formula
Here, we present the proof of Theorem 3. For this, we simplify the expression (36) as follows:
Hence, taking (38) into account for the square of the norm, we obtain
Then, using the analytic expressions of the optimal coefficients, i.e., by the Formula (52), we obtain
Provided that and , for the square of the norm, we finally obtain
Theorem 3 is completely proved.
Now, we give the square of the norm of the error functional for the several first values of m for .
If , then
Now let . It follows that
It is easy to see that, for and ,
In the case and , we have
From the last results, we can conclude that, for the functions from the space , the rate of convergence of the constructed optimal quadrature formula is for and for .
7. Numerical Results
In this section, we present numerical results of comparison for absolute errors of the optimal quadrature formula of the form (4) with the weight in the case and the Midpoint formula. We note that the rate of convergence for both of these formulas is . Here we use the Maple to obtain numerical results.
As an example, we consider calculation of the integral
with . Since for the functions we have the relation , we can consider it as a periodic function on the interval . For convenience, we denote the integrand as ; i.e., .
We approximately calculate the integral I by the Midpoint rule. Then approximate value for the integral (53) is then calculated as follows using the Midpoint rule
where , and .
Hence, for the function , the error of the Midpoint rule (54) is
In Table 1, we give the absolute values for the real part of the error (55) of the Midpoint rule for and .
Table 1.
The absolute values for the real part of the error (55) of the Midpoint rule for and .
It can be seen from the results given in Table 1 that the Midpoint rule converges for . In Figure 1, the process of this convergence is graphically shown when and for the real part of the function .
Figure 1.
The process of convergence for the Midpoint rule when and for the real part of the function .
In Figure 2 the graphs of numerical calculation of the integral (53) by the Midpoint rule for the case and are given. Here, we can see that the Midpoint rule does not converge when for the function .
Figure 2.
The Midpoint rule does not converge when for the function .
Now, we approximate the above integral (53) using the optimal quadrature formula of the form (4) with weight in the case . Then, taking into account (10), we have the approximate equality
for with the optimal coefficients
The approximate value for the integral (53) is calculated as follows, using the optimal quadrature formula:
Hence, for the function , the error of the optimal quadrature Formula (56) is
Thus, the numerical results of Table 2 show convergence of the optimal quadrature Formula (56) for and .
8. Conclusions
We obtained optimal quadrature formulas for the approximate calculation of the Fourier coefficients in Sobolev space . Firstly, in order to obtain an upper bound for the error of the quadrature formula, the norm of the error functional was calculated. Using the extremal function of the considered quadrature formula, we calculated the norm of the error functional. We found the explicit forms of the coefficients for the optimal quadrature formula, and they provide the minimum value to the norm of the error functional. Finally, we calculated the norm of the error functional of the optimal quadrature formulas. We show that, for the functions from the space , the rate of convergence of the constructed optimal quadrature formula is for and for . Finally, we presented numerical results of comparison for absolute errors of the constructed optimal quadrature formula in the case and the Midpoint formula, showing the advantage of the optimal quadrature formulas.
Author Contributions
The problem of the manuscript was stated by K.S. and A.H.; proofs of the main theoretical results were obtained by K.S., A.H. and B.A.; the numerical results were obtained by B.A. All authors have read and agreed to the published version of the manuscript.
Funding
This research received no external funding.
Institutional Review Board Statement
The study was conducted in accordance with the Scientific program of the Computational Mathematics Laboratory of V. I. Romanovskiy Institute of Mathematics and discussed in the Scientific seminar of the Laboratiry (protocol code 7, 8 July 2022).
Informed Consent Statement
Not applicable.
Data Availability Statement
Not applicable.
Acknowledgments
We are very thankful to the reviewers for valuable comments and remarks that have improved the quality of the paper.
Conflicts of Interest
The authors declare that they have no competing interests.
References
- Filon, L.N.G. On a quadrature formula for trigonometric integrals. Proc. Roy. Soc. Edinb. 1928, 49, 38–47. [Google Scholar] [CrossRef] [Scilit]
- Babuška, I. An optimal formula for computation of linear functionals. APL Mater. (Ger.) 1965, 10, 441–443. [Google Scholar]
- Boltaev, N.D.; Hayotov, A.R.; Shadimetov, K.M. Construction of optimal quadrature formulas for Fourier coefficients in Sobolev space (0,1). Numer. Algorithms 2017, 74, 307–336. [Google Scholar] [CrossRef] [Scilit]
- Boltaev, N.D.; Hayotov, A.R.; Milovanović, G.V.; Shadimetov, K.M. Optimal quadrature formulas for numerical evaluation of Fourier coefficients in . J. Appl. Anal. Comput. 2017, 7, 1233–1266. [Google Scholar]
- Filin, E.A. A modification of Filon’s method of numerical integration. J. Assoc. Comput. Mach. 1960, 7, 181–184. [Google Scholar] [CrossRef] [Scilit]
- Hayotov, A.R.; Jeon, S.; Lee, C.-O. On an optimal quadrature formula for approximation of Fourier integrals in the space . J. Comput. Appl. Math. 2020, 372, 112713. [Google Scholar] [CrossRef] [Scilit]
- Hayotov, A.R.; Jeon, S.; Lee, C.-O.; Shadimetov, K.M. Optimal Quadrature Formulas for Non-periodic Functions in Sobolev Space and Its Application to CT Image Reconstruction. Filomat 2021, 35, 4177–4195. [Google Scholar] [CrossRef] [Scilit]
- Hayotov, A.R.; Jeon, S.; Shadimetov, K.M. Application of optimal quadrature formulas for reconstruction of CT images. J. Comput. Appl. Math. 2021, 388, 113313. [Google Scholar] [CrossRef] [Scilit]
- Luke, Y.L. On the computation of oscillatory integrals. Proc. Camb. Phil. Soc. 1954, 50, 269–277. [Google Scholar] [CrossRef] [Scilit]
- Shadimetov, K.M. Optimal Lattice Quadrature and Cubature Formulas in the Sobolev Spaces; Fan va Texnologiya: Tashkent, Russia, 2019; 224p. [Google Scholar]
- Saks, S.; Zygmund, A. Analytic Functions; Monografie Matematyczne; Instytut Matematyczny Polskiej Akademii Nauk: Warsaw, Poland, 1952. [Google Scholar]
- Zadiraka, V.K. Theory of Computations of Fourier Transforms; Naukova Dumka: Kiev, Russia, 1988; 216p. [Google Scholar]
- Zhileikin, Y.M. The errors in the approximate computation of integrals of rapidly oscillating functions. Comput. Math. Math. Phys. 1971, 11, 344–348. [Google Scholar] [CrossRef] [Scilit]
- Zhileikin, Y.M.; Kukarkin, A.B. Optimal calculation of integrals of rapidly oscillating functions. Comput. Math. Math. Phys. 1978, 18, 15–21. [Google Scholar] [CrossRef] [Scilit]
- Bakhvalov, N.S. Numerical Methods; Nauka: Moscow, Russia, 1973; 632p. [Google Scholar]
- Babuška, I. The optimal computation of Fourier coefficients. APL Mater. (Ger.) 1966, 11, 113–123. [Google Scholar] [CrossRef] [Scilit]
- Sobolev, S.L. Introduction to the Theory of Cubature Formulas; Nauka: Moscow, Russia, 1974; 808p. [Google Scholar]
- Sobolev, S.L.; Vaskevich, V.L. The Theory of Cubature Formulas; Siberian Division of the Russia Academy of Sciences Novosibirsk: Novosibirsk, Russia, 1996; 484p. [Google Scholar]
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. |
© 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).

