• Nu S-Au Găsit Rezultate

View of On the existence and uniqueness of extensions of semi-Hölder real-valued functions

N/A
N/A
Protected

Academic year: 2022

Share "View of On the existence and uniqueness of extensions of semi-Hölder real-valued functions"

Copied!
7
0
0

Text complet

(1)

Rev. Anal. Num´er. Th´eor. Approx., vol. 39 (2010) no. 2, pp. 134–140 ictp.acad.ro/jnaat

ON THE EXISTENCE AND UNIQUENESS OF EXTENSIONS OF SEMI-H ¨OLDER REAL-VALUED FUNCTIONS

COSTIC ˘A MUST ˘AT¸ A

Abstract. Let (X, d) be a quasi-metric space,y0 X a fixed element andY a subset ofX such thaty0Y. Denote by (Λα,0(Y, d),k · |αY,d) the asymmetric normed cone of real-valued d-semi-H¨older functions defined on Y of exponent α (0,1], vanishing in y0, and by (Λα,0(Y,d),¯ k · |αY,d¯) the similar cone if d is replaced by conjugate ¯dofd.

One considers the following claims:

(a) For everyfin the linear space Λα,0(Y) = Λα,0(Y, d)∩Λα,0(Y,d) there exist¯ FΛα,0(X, d) such thatF|Y =f andkF|αY,d=kf|αY,d;

(b) For everyfΛα,0(Y) there exists ¯F Λα,0(X,d) such that ¯¯ F|Y =f and kF|¯αY,d¯=kf|αY,d¯;

(c) The extensionF in (a) is unique;

(d) The extension ¯F in (b) is unique;

(e) The annihilator Yd¯ of Y in Λα,0(X,d) is proximinal for the elements of¯ Λα,0(X) with respect to the distance generated byk · |αY,d;

(f) The annihilator Yd of Y in Λα,0(X, d) is proximinal for the elements of Λα,0(X) with respect to the distance generated byk · |αY,d¯;

(g) Yd¯in the claim (e) is Chebyshevian;

(h) Ydin the claim (f) is Chebyshevian.

Then the following equivalences hold:

(a)(e); (b)(f); (c)(g); (d)(h).

MSC 2000. 46A22, 41A50, 41A52.

Keywords. Extensions, semi-Lipschitz functions, semi-H¨older functions, best approximation, quasi-metric spaces.

1. INTRODUCTION

Let X be a nonempty set and d : X ×X → [0,∞) a function with the properties:

(QM1) d(x, y) =d(y, x) = 0 iffx=y, (QM2) d(x, y)≤d(x, z) +d(z, y), for all x, y, z∈X.

Then the function d is called a quasi-metric on X and the pair (X, d) is calledquasi-metric space ([13]).

“T. Popoviciu” Institute of Numerical Analysis, Cluj-Napoca, Romania, e-mail:

[email protected], [email protected].

(2)

Because, in general,d(x, y)6=d(y, x),forx, y∈Xone defines the conjugate quasi-metricdof d, by the equality ¯d(x, y) =d(y, x),for all x, y∈X.

LetY be a nonvoid subset of (X, d) and α∈(0,1] a fixed number.

Definition 1. a) A function f : Y → R is called d-semi-H¨older (of exponent α) if there exists a constant KY(f)≥0 such that

(1) f(x)−f(y)≤KY(f)dα(x, y), for allx, y∈Y.

b) f : Y → R is called d-semi-H¨older (of exponent α) if there exists a constant KY(f)≥0 such that

(2) f(x)−f(y)≤KY(f)·dα(y, x), for allx, y∈Y.

The smallest constantKY(f) in (1) is denoted bykf|αY,dand one shows that (3) kf|αY,d:= supn(f(x)−f(y))∨0

dα(x,y) :d(x, y)>0;x, y∈Yo . Analogously one defines kf|αY,d.

Observe that the function f is d-semi-H¨older onY iff −f is d-semi-H¨older on Y.Moreover

(4) kf|αY,d=k−f|αY,d.

Definition 2. ([14]). Let (X, d) be a quasi-metric space and Y ⊆ X a nonempty set. The function f :Y →Ris called≤d-increasing on Y if f(x)≤ f(y) whenever d(x, y) = 0, x, y∈Y.

The set of all ≤d-increasing functions on Y is denoted by RYd and it is a cone in the linear space RY of all real-valued functions on Y.

The set

(5) Λα(Y, d) :={f ∈RYd;f is d-semi-H¨older and kf|αY,d <∞}

is also a cone, called the cone ofd-semi-H¨older functions onY.

Ify0 ∈Y is arbitrary, but fixed, one considers the cone (6) Λα,0(Y, d) :={f ∈Λα(Y, d) :f(y0) = 0}.

Then the functional k |αY,d : Λα,0(Y, d) → [0,∞) is subadditive, positively homogeneous and the equalitykf|αY,d =k−f|αY,d= 0 impliesf ≡0.This means that k·|αY,d is an asymmetric norm (see [13], [14]), on the cone Λα,0(Y, d).

The pair

Λα,0(Y, d),k |αY,d

is called the asymmetric normed cone of d- semi-H¨older real-valued function onY (compare with [14]).

Analogously, one defines the asymmetric normed cone (Λα,0(Y, d),k · |α

Y,d).

of all d-semi-H¨older real-valued functions on Y, vanishing at the fixed point y0 ∈Y.

(3)

By the above definitions it follows that

f ∈(Λα,0(Y, d),k |αY,d) iff −f ∈(Λα,0(Y, d),k |αY,d) and, moreover, kf|αY,d=k−f|αY,d.

Defining Λα,0(Y) by

(7) Λα,0(Y) = Λα,0(Y, d)∩Λα,0(Y, d),

It follows that Λα,0(Y) is a linear subspace. The following, theorem holds.

Theorem 3. For every f ∈ Λα,0(Y) there exist at least one function F ∈ Λα,0(Y, d) and at least one function F ∈Λα,0(Y, d) such that

a) F|Y = F Y =f.

b) kF|αY,d =kf|αY,d and F

α

Y,d =kf|αY,d.

Proof. By Theorem 2 and Remark 3 in [11] it follows that the functions defined by the formulae:

F(f)(x) = inf

y∈Y{f(y) +kf|αY,ddα(x, y)}, x∈X, (8)

G(f)(x) = sup

y∈Y

{f(y)− kf|αY,ddα(y, x}, x∈X, are elements of Λα,0(X, d) and, respectively, the functions given by

F(f)(x) = inf

y∈Y{f(y) +kf|αY,ddα(y, x)}, x∈X, (9)

G(f)(x) = sup

y∈Y

{f(y)− kf|αY,ddα(x, y)}, x∈X are elements of Λα,0(X, d) such that

(10) F(f)|Y = G(f)|Y =f and kF(f)|αY,d=kG(f)|αY,d=kf|αY,d, respectively

(11) F(f)

Y = G(f)

Y =f and F(f)

α Y,d=

G(f)

α

Y,d=kf|αY,d.

Forf ∈Λα,0(Y) let us consider the following (nonempty) sets of extensions:

(12) Ed(f) :={H∈Λα,0(X, d) : H|Y =f and kH|αY,d =kf|αY,d} and

(13) Ed(f) :={H ∈ ∧α,0(X, d) : H

Y =f and H

α

Y,d=kf|αY,d}.

The sets Ed(f) andEd(f) are convex and

(14) F(f)(x)≥H(x)≥G(f)(x), x∈X for all H∈ Ed(f);

(15) F(f)(x)≥H(x)≥G(f)(x), x∈H, for all H∈ Ed(f).

(4)

Also, for F ∈Λα,0(X), F|Y ∈Λα,0(Y) and

F−H∈Λα,0(X, d), for all H∈ Ed(F|Y), F−H∈Λα,0(X, d) for all H∈ Ed(F|Y).

Let (X, d) be a quasi-metric space,y0∈X fixed andY ⊆Xsuch that y0∈Y.

Let

(16) Yd:={G∈Λα,0(X, d) : G|Y = 0}

and

(17) Yd:={G∈ ∧α,d(X, d) : G

Y = 0}.

Obviously, forF ∈Λα,0(X)

(18) F − Ed(F|Y)⊂Λα,0(X, d) and

(19) F − Ed(F|Y)⊂Λα,0(X, d).

In the sequel we prove a result of Phelps type ([1], [10], [12]) concerning the existence and uniqueness of the extensions preserving the smallest semi-H¨older constants and a problem of best approximation by elements of Yd and Yd, respectively.

Let (X,k |) be an asymmetric norm (see [13], [14]) and letM be a nonempty set ofX. The setM is calledproximinal for x∈X iff there exists at least one element m0 ∈M such that

kx−m0|= inf{kx−m|:m∈M}=ρ(x, M).

If M is proximinal for x, then the set PM(x) = {m0 ∈ M : kx−m0| = ρ(x, M)} is called the set of elements of best approximations for x in M. If cardPM(x) = 1 then the set M is called Chebyshevian forx.

The set M is called proximinal if M is proximinal for every x ∈ X, and Chebyshevian ifM is Chebyshevian for every x∈X.

Now, consider the following two problems of best approximation:

Pd(F).ForF ∈Λα,0(X) find G0∈Yd such that

(20) kF−G0|αY,d = inf{kF −G|αY,d :G∈Yd}=ρd(F, Yd) and

Pd(F).ForF ∈Λα,0(X) find G0∈Y

d such that

(21)

F −G0

α

Y,d= inf{

F −G

α

X,d:G∈Yd}=ρd(F, Yd).

Let

(22) PY

d

(F) :={G0 ∈Yd:

F−G0

α

X,dd(F, Yd)}

and

(23) PY

d (F) :={G0 ∈Yd:kF−G0|α

X,dd(F, Yd)}.

(5)

The following theorem holds.

Theorem 4. If F ∈Λα,0(X) then

(24) PY

d

(F) =F− Ed(F|Y),

(25) PY

d (F) =F− Ed(F|Y) and

(26) ρd(F, Yd) =kF|Y αY,d,

(27) ρd(F, Yd) =kF|Y

α Y,d . Proof. LetF ∈Λα,0(X)(= Λα,0(X, d)∩Λα,0(X, d)) Then, for every G∈Yd,

kF|Y

αY,d =kF|Y −G Y

αY,d ≤ F−G

α X,d. Taking the infimum with respect toG∈Y

d ,one obtainskF|Y |αY,d≤ρd(F, Y

d ).

On the other hand, for everyH ∈ Ed(F|Y), kF|Y

αY,d =kH|αX,d=kF −(F −H)|αX,d. Because F−H ∈Y

d ,it followskF|Y |αY,d≥ρd(F, Yd).

Consequently,Y

d is proximinal with respect to the distanceρdd-proximinal in short) and

ρd(F, Yd) =kF|Y αY,d. Now, letG0 ∈PY

d

(F).Then (F−G0)|Y = F|Y and

F−G0

α X,d= kF|Y |αY,d.This means thatF−G0∈ Ed(F|Y),i.e., G0∈F− Ed(F|Y).Conse- quently, G0∈PY

d

(F) impliesG0 ∈F − Ed(F|Y).

Taking into account the first part of the proof it follows PY

d

(F) = F − Ed(F|Y).

Analogously, one obtainsρd(F, Yd) =kF|Y |α

Y,dandPY

d (F) =F−Ed(F|Y).

By the equalities (22) and (23) it follows.

Corollary 5. Let F ∈Λα,0(X) andY ⊂X such that y0∈Y. Then a) cardEd(F|Y) = 1 iff Y

d is ρd-Chebyshevian;

b) cardEd(F|Y) = 1 iff Yd is ρd-Chebyshevian.

Remark6. Observe that the linear space Λα,0(X) = Λα,0(X, d)∩Λα,0(X, d) is a Banach space with respect to the norm

(28) kF|αX = max{kF|αX,d,kF|αX,d}.

(6)

In fact this space in the space of all real-valued Lipschitz functions defined on the quasi-metric space (X, dα),vanishing at a fixed pointy0 ∈X. Obviously, (29) kFkαX = sup

n|F(x)−F(y)|

dα(x,y) :d(x, y)>0; x, y∈X o

is a norm on Λα,0(X).

Corollary 7. For every element f in the space Λα,0(Y) = Λα,0(Y, d)∩ Λα,0(Y, d) there exists F ∈Λα,0(X) such that

F|Y =f and kFkαX =kfkαY .

The set of all extensions of f ∈Λα,0(Y) preserving the norm kfkαY (of the form (29), is denoted by E(f),i.e.,

(30) E(f) : ={F∈Λα,0(X) : F|Y =f and kFkαX =kfkαY}.

Denote by

(31) Y:={G∈Λα,0(X) : G|Y = 0}.

the annihilator of the set Y in Banach space Λα,0(X), and one considers the following problem of best approximation:

P.ForF ∈Λα,0(X) findG0 ∈Y such that

kF−G0kαX = inf{kF −GkαX :G∈Y}=ρ(F, Y).

Corollary 8. The subspace Yis proximinal in Λα,0(X) and the set of elements of best approximation for F ∈Λα,0(X) is

PY(F) =F− E(F|Y).

The distance of F toY is given by

ρ(F, Y) =kF|Y kαY .

The subspace Y is Chebyshevian for F iff cardE(F|Y) = 1.

Forf in the linear space Λα,0(Y),the equalitiesF(f)(x) =F(f)(x), x∈X and G(f)(x) = G(f)(x), x ∈ X are verified iff kf|αY,d = kf|αY,d. This means that kf|αY,d=k−f|αY,d and, consequently,

kfkαY = max{kf|αY,d;kf|αY,d}=kf|αY,d.

By Theorem 3 in [14], it follows that Λa,0(Y) is a Banach space and (Y, dα) is a metric space.

(7)

REFERENCES

[1] S. Cobzas¸,Phelps type duality reuslts in best approximation,Rev. Anal. Num´er. Th´eor.

Approx.,31, no. 1., pp. 29–43, 2002.

[2] J. Collins andJ. Zimmer, An asymmetric Arzel¯a-Ascoli Theorem, Topology Appl., 154, no. 11, pp. 2312–2322, 2007.

[3] P. FlectherandW.F. Lindgren,Quasi-Uniform Spaces, Marcel Dekker, New York, 1982.

[4] M.G. KreinandA.A. Nudel’man,The Markov Moment Problem and Extremum Prob- lems, Nauka, Moscow 1973 (in Russian), English translation: American Mathematical Society, Providence, R.I., 1977.

[5] E. Matouˇskova,Extensions of continuous and Lipschitz functions, Canad. Math. Bull., 43, no. 2, pp. 208–217, 2000.

[6] E.T. McShane,Extension of range of functions, Bull. Amer. Math. Soc.,40, pp. 837–

842, 1934.

[7] A. Mennucci, On asymmetric distances, Tehnical report, Scuola Normale Superiore, Pisa, 2004.

[8] C. Must˘at¸a,Best approximation and unique extension of Lipschitz functions, J. Ap- prox. Theory, 19, no. 3, pp. 222–230, 1977.

[9] C. Must˘at¸a,Extension of semi-Lipschitz functions on quasi-metric spaces, Rev. Anal.

Num´er. Th´eor. Approx.,30, no. 1, pp. 61–67, 2001.

[10] C. Must˘at¸a, A Phelps type theorem for spaces with asymmetric norms, Bul. S¸tiint¸.

Univ. Baia Mare, Ser. B. Matematic˘a-Informatic˘a,18, pp. 275–280, 2002.

[11] C. Must˘at¸a,Extensions of semi-H¨older real valued functions on a quasi-metric space, Rev. Anal. Num´er. Th´eor. Approx.,38, no. 2, pp. 164–169, 2009.

[12] R.R. Phelps, Uniqueness of Hahn-Banach extension and unique best approximation, Trans. Numer. Math. Soc.,95, pp. 238–255, 1960.

[13] S. Romagueraand M. Sanchis,Semi-Lipschitz functions and best approximation in quasi-metric spaces, J. Approx. Theory,103, pp. 292–301, 2000.

[14] S. RomagueraandM. Sanchis,Properties of the normed cone of semi-Lipschitz func- tions, Acta Math. Hungar,108, nos. 1–2, pp. 55–70, 2005.

[15] J.H. Wells and L.R. Williams, Embeddings and Extensions in Analysis, Springer- Verlag, Berlin, 1975.

Received by the editors: April 13, 2010.

Referințe

DOCUMENTE SIMILARE

An extension results for semi-Lipschitz functions, analogous to Mc Shane’s Extension Theorem [8] for real-valued Lipschitz functions defined on a subset of a metric space was proved

The papers [32, 33, 39] are concerned with best approximation in spaces of semi- Lipschitz functions defined on asymmetric metric spaces (called quasi-metric spaces) in connection

[9] Must˘ at ¸a, C., Extension of H¨ older functions and some related problems of best ap- proximation, “Babe¸s-Bolyai” University, Faculty of Mathematics, Research Seminar

[3] Halic , G., Local Approximation of Functions and Certain Extensions of Classical Anal- ysis, Section 4: Neutrices, Neutralized Integrals and Their Calculus. Research on Theory

[6] Marian, D., Generalized convex functions and mathematical analysis on networks, Re- search on Theory of Allure, Approximation, Convexity and Optimization, Cluj-Napoca, pp. ,

Several generalizations of the classical notions of convexity and quasicon- vexity of real-valued functions have been given for vector-valued functions and even for set-valued maps

In this note we shall present a procedure to obtain extremal elements of the unit ball of the quasi-normed semilinear space of real-valued semi-Lipschitz functions defined on

Lipschitz duals, meaning spaces of Lipschitz functions on a metric linear space, were used to study best approximation problems in such spaces (see [10]).. A good account on