Abstract
This chapter is devoted to some basic definitions and results that are needed in this book. We essentially focus on the Newton interpolation theory and the higher order convexity and concavity properties.
You have full access to this open access chapter, Download chapter PDF
This chapter is devoted to some basic definitions and results that are needed in this book. We essentially focus on the Newton interpolation theory and the higher order convexity and concavity properties.
Recall that, unless indicated otherwise, the symbol I always denotes an arbitrary real interval whose interior is nonempty.
2.1 Newton Interpolation Theory
In this first section, we recall some basic facts about Newton interpolation theory and divided differences. We also establish a result on the derivatives of interpolating polynomials. For background see, e.g., de Boor [32, Chapter 1], Gel’fond [39, Chapter 1], Quarteroni et al.[85, Section 8.2.2], and Stoer and Bulirsch [94, Section 2.1.3].
Let \(n\in \mathbb {N}\) and let x 0, x 1, …, x n be any (not necessarily distinct) points of I. Let also \(f\colon I\to \mathbb {R}\) be so that \(D^{m_i-1}f(x_i)\) exists for i = 0, …, n, where m i is the multiplicity of x i among the points x 0, x 1, …, x n.
We let
denote the divided difference of f at the points x 0, x 1, …, x n, and we let the map
denote the interpolating polynomial of f with nodes at x 0, x 1, …, x n, i.e., the unique polynomial P satisfying the equations
This polynomial has degree at most n.
Recall that f[x 0, x 1, …, x n] is precisely the coefficient of x n in the interpolating polynomial P n[f](x 0, x 1, …, x n;x). More precisely, the Newton interpolation formula states that
Moreover, the corresponding interpolation error at any x ∈ I can take the following form
Recall also that the map
is symmetric, i.e., invariant under any permutation of its arguments. Moreover, the divided differences of f can be computed via the following recurrence relation. For any k ∈{0, 1, …, n}, we have f[x k] = f(x k) and
When the points x 0, x 1, …, x n are pairwise distinct, we also have the following explicit expression
We now establish a proposition that shows how the derivative of an interpolating polynomial of a differentiable function f is related to the derivative of f.
Proposition 2.1
Suppose that I is an arbitrary nonempty open real interval. For any \(n\in \mathbb {N}^*\) , any system x 0 < x 1 < ⋯ < x n of n + 1 points in I, and any differentiable function \(f\colon I\to \mathbb {R}\) , there exist n points ξ 0, …, ξ n−1 in I such that, for i = 0, …, n − 1, we have x i < ξ i < x i+1 and
Moreover, we have
and
Proof
The function \(g\colon I\to \mathbb {R}\) defined by the equation
vanishes at the n + 1 points x 0, x 1, …, x n. The first part of the proposition then follows from applying Rolle’s theorem in each interval (x i, x i+1). Now, identity (2.6) immediately follows from (2.5) and the very definition of the interpolating polynomial. Identity (2.7) then follows by equating the coefficients of x n−1 in (2.6). □
2.2 Higher Order Convexity and Concavity
Let us recall the definitions of p-convex and p-concave functions and present some related results. For background see, e.g., Kuczma [58], Kuczma [61, Chapter 15], Popoviciu [84], and Roberts and Varberg [87, pp. 237–240].
Definition 2.2 (p-Convexity and p-Concavity)
A function \(f\colon I\to \mathbb {R}\) is said to be convex of order p (resp. concave of order p) or simply p-convex (resp. p-concave) for some integer p ≥−1 if for any system x 0 < x 1 < ⋯ < x p+1 of p + 2 points in I it holds that
Thus defined, a function \(f\colon I\to \mathbb {R}\) is 1-convex if it is an ordinary convex function; it is 0-convex if it is increasing (in the wide sense); it is (−1)-convex if it is nonnegative.
Let us now introduce a practical notation to denote the set of p-convex functions and the set of p-concave functions.
Definition 2.3
Let p ≥−1 be an integer.
-
We let \(\mathcal {K}^p_+(I)\) (resp. \(\mathcal {K}^p_-(I)\)) denote the set of functions \(f\colon I\to \mathbb {R}\) that are p-convex (resp. p-concave).
-
We let \(\mathcal {K}^p_+\) (resp. \(\mathcal {K}^p_-\)) denote the set of functions \(f\colon \mathbb {R}_+\to \mathbb {R}\) that are eventually p-convex (resp. eventually p-concave), i.e., p-convex (resp. p-concave) in a neighborhood of infinity.
We also set
The following proposition shows that both sets \(\mathcal {K}^p_+(I)\) and \(\mathcal {K}^p_-(I)\) are convex cones whose intersection is precisely the real linear space of all polynomials of degree less than or equal to p. A similar description of the sets \(\mathcal {K}^p_+\) and \(\mathcal {K}^p_-\) will be given in Corollary 4.6.
Proposition 2.4
For any \(p\in \mathbb {N}\) , the sets \(\mathcal {K}^p_+(I)\) and \(\mathcal {K}^p_-(I)\) are convex cones. These cones are opposite in the sense that f lies in \(\mathcal {K}^p_+(I)\) if and only if − f lies in \(\mathcal {K}^p_-(I)\) . Moreover, the intersection \(\mathcal {K}^p_+(I)\cap \mathcal {K}^p_-(I)\) is the real linear space of all polynomials of degree less than or equal to p.
Proof
That the sets \(\mathcal {K}^p_+(I)\) and \(\mathcal {K}^p_-(I)\) are convex cones is trivial; indeed, if f 1 and f 2 lie in \(\mathcal {K}^p_+(I)\) for instance, then so does c 1 f 1 + c 2 f 2 for any c 1, c 2 ≥ 0. By definition of \(\mathcal {K}^p_+(I)\) and \(\mathcal {K}^p_-(I)\), these cones are clearly opposite. Now, let f lie in \(\mathcal {K}^p_+(I)\cap \mathcal {K}^p_-(I)\) and let x 0 < ⋯ < x p be p + 1 points in I. By (2.2), for any x ∈ I we must have
which shows that f is a polynomial of degree at most p. Conversely, using (2.2) again, we can readily see that any such polynomial lies in \(\mathcal {K}^p_+(I)\cap \mathcal {K}^p_-(I)\). □
We now present an important lemma. It is interesting in its own right and will be very useful in the subsequent chapters. A variant of this result can be found in Kuczma [61, Lemma 15.7.2].
Recall first that for any \(f\colon I\to \mathbb {R}\), any \(p\in \mathbb {N}\), and any x ∈ I such that x + p ∈ I, we have
Lemma 2.5
Let \(p\in \mathbb {N}\) and let \(\mathcal {I}_{p+1}\) denote the set of tuples of I p+1 whose components are pairwise distinct. A function \(f\colon I\to \mathbb {R}\) lies in \(\mathcal {K}^p_+(I)\) (resp. \(\mathcal {K}^p_-(I)\) ) if and only if the restriction of the map
to \(\mathcal {I}_{p+1}\) is increasing (resp. decreasing) in each place. In particular, if I is not upper bounded, then for any function f lying in \(\mathcal {K}^p_+(I)\) (resp. \(\mathcal {K}^p_-(I)\) ), the function Δ p f is increasing (resp. decreasing) on I.
Proof
Using the definition of p-convexity and the standard recurrence relation (2.3) for divided differences, we can see that f lies in \(\mathcal {K}^p_+(I)\) if and only if, for any pairwise distinct x 0, …, x p+1 ∈ I, we have
Equivalently, for any pairwise distinct x 0, …, x p+1 ∈ I, we have
The latter condition exactly means that the map defined in the statement is increasing in the first place. Since this map is symmetric, it must be increasing in each place. The second part of the lemma follows from (2.8). □
We end this section with a second lemma, which provides some important connections between higher order convexity and higher order differentiability. In fact, these connections can be derived (sometimes tediously) from various results given in the references mentioned in the beginning of this section, especially the book by Kuczma [61, Chapter 15]. However, for the sake of self-containment we provide a detailed proof in Appendix A.
Lemma 2.6
Let I be an nonempty open real interval and let \(p\in \mathbb {N}\) . Then the following assertions hold.
-
(a)
We have \(\mathcal {K}^{p+1}(I)\subset \mathcal {C}^p(I)\).
-
(b)
Assume that I is not upper bounded. If \(f\in \mathcal {K}^p_+(I)\) , then \(\Delta ^jf\in \mathcal {K}^{p-j}_+(I)\) for every j ∈{0, …, p + 1}.
-
(c)
If \(f\in \mathcal {C}^{j}(I)\cap \mathcal {K}^p_+(I)\) for some j ∈{0, …, p + 1}, then \(f^{(j)}\in \mathcal {K}^{p-j}_+(I)\).
-
(d)
If \(f\in \mathcal {C}^1(I)\) and \(f'\in \mathcal {K}^{p-1}_+(I)\) , then \(f\in \mathcal {K}^p_+(I)\).
Proof
See Appendix A. □
2.3 A Key Lemma
Let \(p\in \mathbb {N}\), a > 0, and \(f\colon \mathbb {R}_+\to \mathbb {R}\). Combining Newton’s interpolation formula (2.1) with identity (2.8), we can readily see that the unique interpolating polynomial of f with nodes at the p points a, a + 1, …, a + p − 1 takes the form
If p = 0, then this polynomial is naturally the zero polynomial, which is assumed to have degree − 1. Moreover, using (2.2) we can immediately see that the corresponding interpolation error at any x > 0 is
Now, the right side of (2.10) is actually the remainder of the (p − 1)th degree Newton expansion of f(x) about x = a (see, e.g., Graham et al. [41, Section 5.3]). Note also that formula (2.10) is a pure identity in the sense that it is valid without any restriction on the form of f(x).
Using (2.9) and (2.10) we see that, for any a > 0, any x ≥ 0, any \(p\in \mathbb {N}\), and any \(g\colon \mathbb {R}_+\to \mathbb {R}\), the quantity \(\rho ^p_a[g](x)\) defined in (1.7) is precisely the interpolation error at a + x when considering the interpolating polynomial of g with nodes at a, a + 1, …, a + p − 1. We then immediately derive the following identities:
We note that identity (2.12) also extends to the case when x ∈{0, 1, …, p − 1}, even if g is not differentiable. Indeed, in this case we must have \(\rho ^p_a[g](x)=0\) by (2.11).
We now end this chapter with a key lemma that will be used repeatedly in this book. Although this lemma is rather technical, it is at the root of various fundamental convergence results of our theory. Recall first that, for any \(k\in \mathbb {N}\), the symbol ε k(x) stands for the sign of \(x^{ \underline {k}}\).
Lemma 2.7
Let \(p\in \mathbb {N}\), \(f\in \mathcal {K}^p\) , and a > 0 be so that f is p-convex or p-concave on [a, ∞). Then, for any x ≥ 0, we have
where ± stands for 1 or − 1 according to whether f lies in \(\mathcal {K}^p_+\) or \(\mathcal {K}^p_-\) . Moreover, if x ∈{0, 1, …, p} (i.e., ε p+1(x) = 0), then \(\rho _a^{p+1}[f](x)=0\).
Proof
If x ∈{0, 1, …, p}, then we have that \(\rho _a^{p+1}[f](x)=0\) by (2.11), and then the inequalities hold trivially. Let us now assume that x∉{0, 1, …, p}, which means that ε p+1(x) ≠ 0. Negating f if necessary, we may assume that it lies in \(\mathcal {K}^p_+\). By (2.12) we then have
Hence, using identities (2.3) and (2.8) and Lemma 2.5, we obtain
which proves the first two inequalities. The third one can be immediately proved using a telescoping sum. □
References
C. de Boor. A practical guide to splines. Revised edition. Applied Mathematical Sciences, 27. Springer-Verlag, New York, 2001.
A. O. Gel’fond. Calculus of finite differences. Translated from the Russian. International Monographs on Advanced Mathematics and Physics. Hindustan Publishing Corp., Delhi, 1971.
R. L. Graham, D. E. Knuth, and O. Patashnik. Concrete mathematics: a foundation for computer science. 2nd edition. Addison-Wesley Longman Publishing Co., Boston, MA, USA, 1994.
M. Kuczma. Sur une équation aux différences finies et une caractérisation fonctionnelle des polynômes. (French). Fund. Math., 55:77–86, 1964.
M. Kuczma. An introduction to the theory of functional equations and inequalities. Cauchy’s equation and Jensen’s inequality. 2nd edition. Birkhäuser Verlag AG, Basel, Switzerland, 2009.
T. Popoviciu. Sur quelques propriétés des fonctions d’une ou de deux variables réelles. Thèses de l’entre-deux-guerres, no. 146 (1933). 96 p. (http://www.numdam.org)
A. Quarteroni, R. Sacco, and F. Saleri. Numerical mathematics. 2nd Edition. TAM 37. Springer, Berlin, Germany, 2010.
A. W. Roberts and D. E. Varberg. Convex functions. Pure and Applied Mathematics, Vol. 57. Academic Press, New York-London, 1973.
J. Stoer and R. Bulirsch. Introduction to numerical analysis. 3rd Edition. TAM 12. Springer-Verlag, New York, USA, 2010.
Author information
Authors and Affiliations
Rights and permissions
Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.
The images or other third party material in this chapter are included in the chapter's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the chapter's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.
Copyright information
© 2022 The Author(s)
About this chapter
Cite this chapter
Marichal, JL., Zenaïdi, N. (2022). Preliminaries. In: A Generalization of Bohr-Mollerup's Theorem for Higher Order Convex Functions. Developments in Mathematics, vol 70. Springer, Cham. https://doi.org/10.1007/978-3-030-95088-0_2
Download citation
DOI: https://doi.org/10.1007/978-3-030-95088-0_2
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-95087-3
Online ISBN: 978-3-030-95088-0
eBook Packages: Mathematics and StatisticsMathematics and Statistics (R0)