On a surjectivity theorem

Abstract

Authors

Costica Mustata
“Tiberiu Popoviciu” Institute of Numerical Analysis, Romanian Academy, Romania

Keywords

Paper coordinates

C. Mustăţa, On a surjectivity theorem, Seminar of Functional Analysis and Numerical Methods, Preprint Nr. 1, (1985), 73-84.

PDF

About this paper

Journal
Publisher Name
DOI
Print ISSN
Online ISSN

MR # 832505

google scholar link

[1] Aronsson, G., A note on the extension of Lipschitz and Holder functions (preprint, 5 p.)
[2] Aronsson, G., Extension of functions satisfying Lipschitz conditions, Arkiv for Mathematik 6, 28 (1967), 551-561.
[3] Gutier, S., Isac, G., Penot, J.P., Surjectivity of multifunctions under generalized differentiability assumptions, Bull. Austral. Math. Soc., 28 (1983), 13-21.
[4] Mustata, C., Best Approximation and Unique Exctension of Lipschitz Functions, Journal of Approx. Theory 19, 3 (1977), 222-230.
[5] Mustata, C., Extension of bounded Lipschitz functions “Babes-Bolyai” University, Faculty of Mathematics, Research Seminaries, Seminar of Funcitonal Analysis and Numerical Methods, Preprint nr.1, 1980, 1-20.
[6] Mc Shane, E.J., Extension of range of functions, Bull. Amer. Math. Soc. 40 (1934), 837-842.
[7] Singer, I., Cea mai bună aproximare în spații vectoriale normate prin elemente din subspații vectoriale, Editura Academiei R.S. România, București, 1067.

Paper (preprint) in HTML form

1985-Mustata-UBB-Seminar-On-a-subjectivity-theorem

ON A SURJECTIVITY THEOREM

Costică Mustăţa

It was proved in[3]that if X X XXX and Y Y YYY are two Banach spaces and F: X Y X Y X rarr YX \rightarrow YXY is a multifunction such that F ( X ) F ( X ) F(X)F(X)F(X) is a closed sub- set of Y Y YYY ,then the multifunction F F FFF is surjective if and only if for overy b P ( X ) b P ( X ) b in P(X)b \in P(X)bP(X) the tangent cone in b b bbb to the set F ( X ) F ( X ) F(X)F(X)F(X) is Y Y YYY
In this note we shall give necesarry and sufficient conditions for the surjectivity of a multifunction F : X X F : X X F:X rarr XF: X \rightarrow XF:XX ,when X X XXX and Y Y YYY are arbitrary motric spaces.
For a metric space Z Z ZZZ and a fixed point z 0 Z z 0 Z z_(0)in Zz_{0} \in Zz0Z denote by Ifp ( Z , R ) ( Z , R ) (Z,R)(\mathrm{Z}, \mathrm{R})(Z,R) the aet of all real-valued Lipschitz functions on Z which vanish at g 0 , 1.0 g 0 , 1.0 g_(0),1.0g_{0}, 1.0g0,1.0
(1) Lip 0 ( z 0 , z ) = { f 8 , z R , f Lip 0 z 0 , z = f 8 , z R , f Lip_(0)(z_(0),z)={f8,z rarr R,f:}\operatorname{Lip}_{0}\left(z_{0}, z\right)=\left\{f 8, z \rightarrow R, f\right.Lip0(z0,z)={f8,zR,f is Lipschitz and f ( z 0 ) = 0 } f z 0 = 0 {:f(z_(0))=0}\left.f\left(z_{0}\right)=0\right\}f(z0)=0}
Recall that a function is Z R Z R Z rarr RZ \rightarrow RZR is called Lipschitz if there ucists 1 0 1 0 1 >= 01 \geqslant 010 suobs that
(2) | x ( z 1 ) = f ( z 2 ) | M a ( z 1 , z 2 ) x z 1 = f z 2 M a z 1 , z 2 quad|x(z_(1))=f(z_(2))| <= Ma(z_(1),z_(2))\quad\left|x\left(z_{1}\right)=f\left(z_{2}\right)\right| \leq M a\left(z_{1}, z_{2}\right)|x(z1)=f(z2)|Ma(z1,z2)
for all z 1 , z 2 z z 1 , z 2 z z_(1),z_(2)in zz_{1}, z_{2} \in zz1,z2z ,where d d ddd denotes the metric of Z Z ZZZ .The smal- lest number 1 for which(2)holds is called the Lipschitz norm of 1 and 18 denoted by | f | | f | |f||f||f| .The norm is given by :
(3)|2||=⿱⿴囗十力 { 2 ( g 1 ) 1 ( g 2 / A ( z 1 , z 2 ) : z 1 , z 2 z , z 1 z 2 } 2 g 1 1 g 2 / A z 1 , z 2 : z 1 , z 2 z , z 1 z 2 {∣2(g_(1))-1(g_(2)∣//A(z_(1),z_(2)):z_(1),z_(2)in z,z_(1)!=z_(2)}:}\left\{\mid 2\left(g_{1}\right)-1\left(g_{2} \mid / A\left(z_{1}, z_{2}\right): z_{1}, z_{2} \in z, z_{1} \neq z_{2}\right\}\right.{2(g1)1(g2/A(z1,z2):z1,z2z,z1z2}

- 80 -

The space Lip 0 ( Z , R ) Lip 0 ( Z , R ) Lip_(0)(Z,R)\operatorname{Lip}_{0}(Z, R)Lip0(Z,R) equipped with the norm (3) is a Banach space :
If Z 1 Z Z 1 Z Z_(1)sub ZZ_{1} \subset ZZ1Z is such that Z 0 Z 1 Z 0 Z 1 Z_(0)inZ_(1)Z_{0} \in Z_{1}Z0Z1 and f Lip 0 ( Z 1 , R ) f Lip 0 Z 1 , R f inLip_(0)(Z_(1),R)f \in \operatorname{Lip}_{0}\left(Z_{1}, R\right)fLip0(Z1,R) then a function f ¯ : Z R f ¯ : Z R bar(f):Z rarr R\bar{f}: Z \rightarrow Rf¯:ZR, f ¯ Lip 0 ( Z , R ) f ¯ Lip 0 ( Z , R ) bar(f)inLip_(0)(Z,R)\bar{f} \in \operatorname{Lip}_{0}(Z, R)f¯Lip0(Z,R) is called an extension of f f fff if
(4) I | Z 1 = I I ¯ Z 1 = I quad bar(I)|_(Z_(1))=I\left.\quad \overline{\mathrm{I}}\right|_{\mathrm{Z}_{1}}=\mathrm{I}I|Z1=I and I = I I ¯ = I || bar(I)||=||I||\|\overline{\mathrm{I}}\|=\|\mathrm{I}\|I=I.
In general, every f I i p 0 ( Z 1 , R ) f I i p 0 Z 1 , R f in Iip_(0)(Z_(1),R)f \in I i p_{0}\left(Z_{1}, R\right)fIip0(Z1,R) has a least one extension f ¯ Lip o ( Z , R ) f ¯ Lip o ( Z , R ) bar(f)inLip_(o)(Z,R)\bar{f} \in \operatorname{Lip}_{o}(Z, R)f¯Lipo(Z,R). Concerning the properties of the set of all extensions of f f fff see [4], [5].
A subset A A AAA of the metric space ( z , d ) ( z , d ) (z,d)(z, d)(z,d) is called uniformly discrete if there exists δ > 0 δ > 0 delta > 0\delta>0δ>0 such that a ( x , y ) δ a ( x , y ) δ a(x,y) >= deltaa(x, y) \geqslant \deltaa(x,y)δ for all x , y A , x y x , y A , x y x,y in A quad,x!in yx, y \in A \quad, x \notin yx,yA,xy.
For Z 1 Z Z 1 Z Z_(1)sub ZZ_{1} \subset ZZ1Z denote by Z 1 Z 1 Z_(1)^(_|_)Z_{1}^{\perp}Z1 the subspace of Lip 0 ( Z , R ) Lip 0 Z , R Lip_(0)(Z_(,)R)\operatorname{Lip}_{0}\left(Z_{,} R\right)Lip0(Z,R) defined by
(5) z 1 = { e Iip 0 ( z , B ) , z 1 = e Iip 0 ( z , B ) , quadz_(1)^(_|_)={e inIip_(0)(z,B),quad:}\quad z_{1}^{\perp}=\left\{e \in \operatorname{Iip}_{0}(z, B), \quad\right.z1={eIip0(z,B), I z 1 = 0 } z 1 = 0 {:z_(1)=0}\left.z_{1}=0\right\}z1=0}.
Obviously, Z 1 Z 1 Z_(1)^(_|_)Z_{1}^{\perp}Z1 is a closed subspace of the Banach space Lip 0 ( Z , R ) Lip 0 ( Z , R ) Lip_(0)(Z,R)\operatorname{Lip}_{0}(Z, R)Lip0(Z,R).
A subset Y Y YYY of a normed space X X XXX is called proximinal if every x X x X x in Xx \in XxX has a best approximation element in Y Y YYY, 1.a. Ior every., x X x X x in Xx \in XxX there exists y 0 Y y 0 Y y_(0)in Yy_{0} \in Yy0Y such that
(6) x y 0 = inf { x y : y Y } (6) x y 0 = inf { x y : y Y } {:(6)||x-y_(0)||=i n f{||x-y||:y in Y}:}\begin{equation*} \left\|x-y_{0}\right\|=\inf \{\|x-y\|: y \in Y\} \tag{6} \end{equation*}(6)xy0=inf{xy:yY}
If every x X x X x in Xx \in XxX, has exactly one element of best approximation in Y Y YYY, then the set Y Y YYY is called a Cebrsextan subset of X X XXX.
Let now X X XXX and Y Y YYY be a a aaa metric space and I & X Y I & X Y I&X rarr YI \& X \rightarrow YI&XY a, multifunction. It was proved in [4] that F ( X ) F ( X ) F(X)^(_|_)F(X)^{\perp}F(X) is always a proximinal subset of Y Y YYY and that F ( X ) F ( X ) F(X)^(_|_)F(X)^{\perp}F(X) is Cebyserian if and only if

81 81 -81--81-81

I ( I ) = { 0 } I ( I ) = { 0 } I(I)^(_|_)={0}I(I)^{\perp}=\{0\}I(I)={0} (Ieaman 1 and Cocollary I I III in [4] )
The matr wesult of this Jote is the following theorem: InTrorrer 1. Ial X , Y X , Y X,YX, YX,Y be two matric spaces ; let F s X Y F s X Y FsX rarr YF s X \rightarrow YFsXY be e millitunction guch that Γ ( X ) Γ ( X ) Gamma(X)\Gamma(X)Γ(X) is closed and non-uniformly discrete and let y 0 P ( x ) y 0 P ( x ) y_(0)in P(x)y_{0} \in P(x)y0P(x) be fixed . Then the following conditroves area equilaring:
(a) I Is aughestive ( ( ((( ioe. D ( X ) = I ) D ( X ) = I ) D(X)=I)D(X)=I)D(X)=I);
(b) Evaxy f Irg 0 ( H ( X ) , F ) f Irg 0 ( H ( X ) , F ) f inIrg_(0)(H(X),F)f \in \operatorname{Irg}_{0}(H(X), F)fIrg0(H(X),F) has exactly one extension I ¯ I t p 0 ( I , E ) I ¯ I t ¯ p 0 ( I , E ) bar(I)in bar(It)p_(0)(I,E)\bar{I} \in \overline{I t} p_{0}(I, E)I¯Itp0(I,E);
(c) The gubspace I ( X ) I ( X ) I(X)^(_|_)I(X)^{\perp}I(X) is Cebysevian in Lip 0 ( Y 1 R ) Lip 0 Y 1 R Lip_(0)(Y_(1)R)\operatorname{Lip}_{0}\left(Y_{1} R\right)Lip0(Y1R).
Proof. II I Is dirfective then P ( X ) = Y P ( X ) = Y P(X)=YP(\mathrm{X})=YP(X)=Y so that Itp 0 ( I ( X ) , I ) = Itg 0 ( Y 0 , I ) Itp 0 ( I ( X ) , I ) = Itg 0 Y 0 , I Itp_(0)(I(X),I)=Itg_(0)(Y_(0),I)\operatorname{Itp}_{0}(I(X), I)=\operatorname{Itg}_{0}\left(Y_{0}, I\right)Itp0(I(X),I)=Itg0(Y0,I) and I ( X ) = I = { 0 } I ( X ) = I = { 0 } I(X)^(_|_)=I^(_|_)={0}I(X)^{\perp}=I^{\perp}=\{0\}I(X)=I={0} which show that the Implicationa ( a ) ( b ) ( a ) ( b ) (a)=>(b)(a) \Rightarrow(b)(a)(b) and ( a ) ( c ) ( a ) ( c ) (a)=>(c)(a) \Rightarrow(c)(a)(c) hold true.
suppose now that condition (b) is verified then, by [4, Th.2] It follows that F ( X ) = Y F ( X ) ¯ = Y bar(F(X))=Y\overline{F(X)}=YF(X)=Y and since F ( X ) F ( X ) F(X)F(X)F(X) is closed in Y Y YYY it follows that I ( X ) = I , i 0 . e 0 ( b ) ( a ) I ( X ) = I , i 0 . e 0 ( b ) ( a ) I(X)=I,i_(0).e_(0)(b)=>(a)I(X)=I, i_{0} . e_{0}(b) \Rightarrow(a)I(X)=I,i0.e0(b)(a).
The equivalence ( b ) ( c ) ( b ) ( c ) (b)<=>(c)(b) \Leftrightarrow(c)(b)(c) follows by Lemma 1 and Corollary 1 from [4]. Theorem 1 is completely proved.
If X X XXX and Y Y YYY are Banach spaces then condition (b) is equivalent also to the condition :
(d) For every y F ( X ) y F ( X ) y in F(X)y \in F(X)yF(X) the tangent cone T y F ( X ) T y F ( X ) T_(y)F(X)T_{y} F(X)TyF(X) to F ( X ) F ( X ) F(X)F(X)F(X) in y y yyy is I , ( [ 3 ] I , ( [ 3 ] I,([3]I,([3]I,([3], Theorem 3 ) ) ))).
COBOTTIARY 1. Tet X , Y X , Y X,YX, YX,Y be two Banach spaces and let F : X Y F : X Y F:X rarr YF: X \rightarrow YF:XY be a multifunction such that F ( X ) F ( X ) F(X)F(X)F(X) is closed. Then
( a ) ( d ) ; ( b ) ( c ) and ( d ) ( b ) ; ( a ) ( d ) ; ( b ) ( c )  and  ( d ) ( b ) (a)<=>(d)quad;quad(b)<=>(c)" and "(d)=>(b)"; "(a) \Leftrightarrow(d) \quad ; \quad(b) \Leftrightarrow(c) \text { and }(d) \Rightarrow(b) \text {; }(a)(d);(b)(c) and (d)(b)
If. furthermore F ( X ) F ( X ) F(X)F(X)F(X) is not uniformly discrete then ( b ) ( d ) ( b ) ( d ) (b)=>(d)(b) \Rightarrow(d)(b)(d) too.
Example. The hypothesis that F ( X ) F ( X ) F(X)F(X)F(X) is not uniformly alscrete is essential in Theorem 1.
Let X = Y = [ 0 , 1 ] X = Y = [ 0 , 1 ] X=Y=[0,1]X=Y=[0,1]X=Y=[0,1] and F : X Y F : X Y F:X rarr YF: X \rightarrow YF:XY defined by
F ( x ) = { 0 , x [ 0 , 1 / 2 ] I , x ( 1 / 2 , 1 ] F ( x ) = 0 ,      x [ 0 , 1 / 2 ] I ,      x ( 1 / 2 , 1 ] F(x)={[0",",x in[0","1//2]],[I",",x in(1//2","1]]:}F(x)= \begin{cases}0, & x \in[0,1 / 2] \\ I, & x \in(1 / 2,1]\end{cases}F(x)={0,x[0,1/2]I,x(1/2,1]
Then F ( X ) = { 0 , 1 } F ( X ) = { 0 , 1 } F(X)={0,1}F(X)=\{0,1\}F(X)={0,1} is closed and uniformly discrete. Let Lip 0 ( [ 0 , 1 ] , R ) = { f : [ 0 , 1 ] R , f Lip 0 ( [ 0 , 1 ] , R ) = { f : [ 0 , 1 ] R , f Lip_(0)([0,1],R)={f:[0,1]rarr R,f\operatorname{Lip}_{0}([0,1], R)=\{f:[0,1] \rightarrow R, fLip0([0,1],R)={f:[0,1]R,f is Iipschitz and f ( 0 ) = 0 } f ( 0 ) = 0 } f(0)=0}f(0)=0\}f(0)=0} Every function f Lip ( { Q , I } , R ) f Lip ( { Q , I } , R ) f in Lip({Q,I},R)f \in \operatorname{Lip}(\{Q, I\}, R)fLip({Q,I},R) has a unique extension f ¯ f ¯ bar(f)\bar{f}f¯ defined by f ( x ) = f ( I ) x , x [ 0 , x ] f ¯ ( x ) = f ( I ) x , x [ 0 , x ] bar(f)(x)=f(I)x,x in[0,x]\overline{\mathrm{f}}(\mathrm{x})=\mathrm{f}(\mathrm{I}) \mathrm{x}, x \in[0, x]f(x)=f(I)x,x[0,x]. Indeed
I ¯ | { 0 , 1 } = 1 and I ¯ = P = | I ( 1 ) | I ¯ { 0 , 1 } = 1  and  I ¯ = P = | I ( 1 ) | ( bar(I))|_({0,1})=1" and "|| bar(I)||=||P||=|I(1)|\left.\bar{I}\right|_{\{0,1\}}=1 \text { and }\|\bar{I}\|=\|P\|=|I(1)|I¯|{0,1}=1 and I¯=P=|I(1)|
Therefore (b) does not imply (a).
Also
F ( { 0 , 1 } ) = { f Ii p 0 ( [ 0 , I ] , R ) , f ( 1 ) = 0 } F ( { 0 , 1 } ) = f Ii p 0 ( [ 0 , I ] , R ) , f ( 1 ) = 0 F({0,1})^(_|_)={f in Iip_(0)([0,I],R),f(1)=0}F(\{0,1\})^{\perp}=\left\{f \in \operatorname{Ii} p_{0}([0, I], R), f(1)=0\right\}F({0,1})={fIip0([0,I],R),f(1)=0}
and for every f Lip 0 ( [ 0 , 1 ] , R ) f Lip 0 ( [ 0 , 1 ] , R ) f inLip_(0)([0,1],R)f \in \operatorname{Lip}_{0}([0,1], R)fLip0([0,1],R) the element g 0 g 0 g_(0)g_{0}g0 of best approximation for f f f^(')f^{\prime}f in P ( { 0 , 1 } ) P ( { 0 , 1 } ) P({0,1})^(_|_)P(\{0,1\})^{\perp}P({0,1}) is
g 0 ( x ) = f ( x ) f ( 1 ) x , x [ 0 , 1 ] . g 0 ( x ) = f ( x ) f ( 1 ) x , x [ 0 , 1 ] . g_(0)(x)=f(x)-f(1)x quad,quad x in[0,1].g_{0}(x)=f(x)-f(1) x \quad, \quad x \in[0,1] .g0(x)=f(x)f(1)x,x[0,1].
This element is unique since g 0 ( x ) = f ( x ) f ( x ) g 0 ( x ) = f ( x ) f ( x ) g_(0)(x)=f(x)-f(x)g_{0}(x)=f(x)-f(x)g0(x)=f(x)f(x) and f ¯ f ¯ bar(f)\bar{f}f¯ is the only extension of f f fff (see [4], Lemma 1 ). Therefore (c) does not imply (a) .

REIPERINNCES

I. ARONSSON , G. , A note on the extension of Lipschitz and Hölder functions (preprint, 5 p.)
2. ARONSSON , G., Extension of functions satisfying Lipschitz conditions , Arkiv för Mathematik 6, 28 (1967), 551-561.
3. GATIXI, S., IS40, G, PMOT J. P, Gurfectivity of maltifunctiona under genaralized differentiability assumptionell. Bull. Mastral. Math. Soc., 28 (1983), 13-21.
4. Musrimp, Go. Best Approximation and Untque Bxtension of Higchity Tunctions, Journal of Approx. Theory 19, 3 (1977), 222-230.
5. GUSTITA, C n C n C_(n)C_{n}Cn, Batension of bounded Iipschitz functions, "Babeq-Bolyay" University. Faculty of Mathematics, Research Seminaries, Seminar of Functional Analysis and Numerlcal Methods, Preprint Nr.1, 1980, 1-20.
6. Me SHANE, E.J., Extansion of range of functions, Bull. Amer. Hath. Soc. 40 (1934), 837-842.
7. STIVGER, Io, Gea, mai tună aproximare in spatii vectoriale normate prin elemente din subspatii vectoriale, Faltura Academiei R.S. România, Bucureşti 1967.
1985

Related Posts