You are on page 1of 10

MATHEMATICAL

COMPUTER
MODELLING
PERGAMON Mathematical and Computer Modelling 32 (2000) 1367-1376
www.elsevier.nl/locate/mcm

Some Elementary Examples


in Nonlinear Operator Theory
J. APPELL
Department of Mathematics, University of Wiirzburg
Am Hubland, D-97074 Wiirzburg, Germany
appell@mathematik.uni-wuerzburg.de
P. P. ZABREJKO
Faculty of Mathematics, Belgosuniversity
pr. Skoriny 4, BY-220050 Minsk, Belorussia
and
National Academy of Sciences of Belorussia
Institute of Mathematics, ul. Surganova 11
BY-220072 Minsk, Belorussia.
zabreiko@mmf.bsu.unibel.by

Abstract-This note is concerned with a comparison between various metric and topological
properties of nonlinear maps between Banach spaces, such ss properness, coercivity, invertibility,
Lipschitz continuity, and compactness. A series of (mostly elementary) examples which illustrate the
abstract results is given as well. @ 2000 Elsevier Science Ltd. All rights reserved.

Keywords-Nonlinear operator, Global invertibility, Local invertibility, Bay-invertibility, Proper-


ness, Bay-properness, Coercivity, Bay-coercivity, Lipschitz condition, Measure of noncompactness,
cr-Lipschitz condition.

1. PROPER AND RAY-PROPER MAPS


Throughout the following, X and Y are real Banach spaces, and F : X --+ Y is a continuous (in
general, nonlinear) map. Recall that F is called proper if the preimage F-(C) = {X : 3: f X,
F(z) E C} c X of any compact set C C Y is compact, and ray-proper [l] if the preimage
F-([6, y]) c X of the “ray” [0, y] = {tg : 0 5 t < 1) is compact for any y E Y. Moreover, F is
called closed if F maps closed subsets of X onto closed subsets of Y. Some simple interconnections
between these properties are collected in the following proposition.

PROPOSITION 1. With X, Y, and F as above, the following statements are true.


(a) Every proper map F is ray-proper, but not vice versa.
(b) F is proper if and only if F is closed and ray-proper.
(c) If F is linear, the properness of F is equivalent to the existence and boundedness of F-l
on F(X).

The above properties are of particular interest in invertibility results for nonlinear maps. As
a matter of fact, each of these properties is “missing” if a map is only locally invertible, but

089571?7/00/$ - see front matter @ 2000 Elsevier Science Ltd. All rights reserved. Typeset by A@-TBX
PII: SO895-7177(00)00210-7
1368 J. APPALLAND P. P. ZABRESKO

not globally invertible {see Proposition 2 below). Results of this type are not only of theoretical
interest, but also important in view of applications. Thus, to prove the global invertibility of a
proper map simply reduces to proving its local invertibility which may often be achieved by quite
elementary means (e.g., by the inverse function theorem and its various generalizations).
Before stating a corresponding theorem, we have to recall another notion. Following [2], we
call a map F : X --+ Y ray-invertible if the following holds: fur each y E Y there exists sume
continuous map y : [O, I] -+ X such that F(y(7)) = my for 0 5 T < 1. Observe that a ray-
invertible map is always onto, since y = F(y(1)) E F(X) for a suitably chosen path y.

PROPOSITION 2. A map F : X + Y is a global homeomorphism if and on& if any one of the


following four conditions is satisfied:
(a) F is a local borneomorphism and proper (See [3j.)
(b) F is a local homeomorphism and ray-proper (See f-21.)
(c) F is a local homeomorphism and closed (‘See (41.)
(d) F is a local homeomorphism and ray-invertible (See [2j.)

The standard first-year calculus example of a local homeomorphism which is not globally
invertible is F : R2 -+ R2 defined by

F(zl, ~2) = (eZ1 COSZ~,e”’ sinx2). (1)

Of course, the map (1) is not ray-proper since F-( [8, y]) = (--oo,O) x (2k7r : k E Z} for y = (1, 0),
and thus, not proper either. F is not closed since F(R x (0)) = (0, W) x (0). Finally, F is not
ray-invertible since it is not onto.
Let us now make a comparison of the various conditions on F arising in Propositions 1 and 2. A
careful analysis of the implications of these theorems shows that there are only ten combinations
which do not lead to a contradiction; we collect them in Table 1.

Table 1.

’ GIobal N Local N Proper Ray-Proper Ray-Invertible 1 Closed

Example 1 Yes Yes Y- Yes Yes Yes

I Example 2 no I IlO I yes I 3-s I 3-s 1 Yes

I Example 3 I no I ye= I no no I no I no

r--
Example 4 1 no I no I no 1 yes I yes l-l0

f Example 5 no ll0 no no Yes Yes

Example 6 no no no no Yes no

Example 7 no no no Yes no no

Example 8 no no no no no Yes

Example 9 no no Ye Yes no Yes

Example 10 no ll0 no no I30 IlO

To show that these possibilities actually occur, we give a series of ten examples. The only
example which is not straightforward is Example 4.
EXAMPLE 1. Let X = Y = Iw and F(z) = x. Obviously, F has all six properties.

EXAMPLE 2. Let X = IEs3,Y = II!‘, and F(z~,xz,z~) = (rcost,rsint)(? = XT + 2; +x$, t =


nxl/r). Clearly, F is not locally invertible. On the other hand, F is proper, since ]]F(s)II = 11~11
for all z E Iw3. It is also clear that F is closed.
NonlinearOperator Theory 1369

Observe that Example 2 is not possible for Y = W, since any continuous surjective map
F : X -+ R, with dimX 2 2, has the property that F-((y)) is unbounded for any y E R.
EXAMPLE 3. Let X = Y = W and F(z) = arctanz. Then, F is locally invertible, but has none
of the remaining properties.
EXAMPLE 4. Let X = Y be a real separable Hilbert space with orthonormal basis {el, e2, es,, . . }.
For 0 < 6 < l/a, let

Obviously, B, n B, = 8 for m # n. We define F by

I x, ifzcX\B,
F(x) = 1_ IIZ--e,Il e + lb-enlIz,
6 > n
if Ic E B,.
6
Then, F is a continuous map with F(B,) = T,, where T, is the “ice cream cone”

T,= co ({ie,,}~B~), n=l,2,3 ,....

It is not hard to see that F is neither globally nor locally invertible. Furthermore, F is not proper,
since the set F-({el, (1/2)ez, (1/3)es,. . . } U (0)) contains the noncompact set {el, e2, es,. . . }.
Similarly, F is not closed, since the set F({el,ez,e3,. . .}) = {el, (1/2)ez, (1/3)es,. . .} is not
closed. We claim that F is both ray-proper and ray-invertible.
Fix y E Y. If the ray [0, y] does not meet any of the sets T,, we have F-([8, y]) = [0, y], and
there is nothing to prove. In the opposite case, we have [8, y] n T, # 0 for just one n. Since T,
is closed and convex, we get then [6, y] rl T, = {ty : t_ I t 5 t+} for some scalars t_, t+ 2 0.
If y = ve, for some q > 0, then all points zc E F-([8, y]) belong to the ray {te, : 0 5 t < co}.
By definition, F maps the generic point z = te, of this ray to the point F(te,) = m(t)en,
where T,, is the piecewise linear function

;+
f & ifOIt<l-6,

1
‘-nn6fnb(t-1), ifl-s<t<l,
m(t) =
;+ n+~~M1(t-l), ifl<t<1+6,

l 6 ifl+JIl<co.

This shows that, for every q > 0, the preimage F-( (f3,qe,]) is some interval containing 6. On the
other hand, if y is not of the form y = ve, for some q~> 0, then F-([8, y]) is contained in the plane
II = span {e,, y}. The restriction of F to this plane is the identity outside B, f~ II; in particular,
the boundary C of B,, n II is kept fixed. To understand the action of F on the interior of the ball
B, n II, observe that the two segments joining e, with the two points, where the tangent rays
starting from (l/n)e, hit the boundary C, are mapped by F to the segments joining (l/n)e,
with these boundary points. (Geometrically, F maps the “night-cap” with peak e, continuously
to the “night-cap” with peak (l/n)e,, keeping the boundary C fixed.) This geometric reasoning
shows that the preimage of the portion of [13,y] inside B, n II is a continuous curve in B, n II
which is mapped l-l onto [0, y] n B, n Il. Altogether, we conclude that F is both ray-proper and
ray-invertible.
EXAMPLE 5. Let X = W2, Y = R, and F( q,~) = ~1. Then, F is closed and ray-invertible but
has none of the remaining properties.
1370 3. APPELL AND P. P. ZABREJKO

EXAMPLE 6. Let X = R2, Y = R, and F(zl, ~2) = ~1~0~x2. Then, F is ray-invertible, but has
none of the remaining properties.
EXAMPLE 7. (See [5].) Let X = Y = W2, G( 21,221 = (1~1l,arctanz2),H(yl,y2) = (@I COSYZ,
eglsin yz), and F = H o G. It is not hard to see that F is ray-proper. On the other hand, F is
not proper since
F- ({(YI,Yz) : pi: +Y; = 1)) = (0) x R

is not compact. Similarly, F is not closed since the set

FW) xW = {(YI,YZ) : Y: +Y; = 1, YI > O}

is not closed in R2. It is clear that F is neither ray-invertible nor globally invertible. By
Proposition l(b), F cannot be locally invertible either.
EXAMPLE 8. Let X = Y = R and F(z) E 0. Obviously, F is closed, but has none of the
remaining properties.
EXAMPLE 9. Let X = Y = I[$ and F(z) = (x1. Obviously, F is proper and closed, but not
ray-invertible.
EXAMPLE 10. Let X = Y = W and F(z) = sinz Obviously, F is neither ray-proper nor
ray-invertible. To see that F is not closed, consider the discrete set (27~ + l/n : n E II).
It is interesting to compare Example 7 with the following (linear!) map between infinite di-
mensional spaces.
EXAMPLE 7'. (See [6].) Let X = Y = CIO, 11, equipped with the usual max-norm, and let F be
the map defined by
t

F(x)(t) = z(s) ds. (2)


s0

The range of F is the subspace CA[O, l] = {y : y E C’[O, 11, y(0) = 0) of X, and F is invertible
on this subspace with inverse F-‘(y) = y’. However, F-’ is not bounded, since we consider Y
equipped with the C-norm, and thus, F is not proper, by Proposition l(c). It is easy to check
that
if CAP,
9 E
otherwise,
11,

and hence, F is ray-proper. On the other hand, F is not closed since Ci[O, l] is not closed in Y.
Finally, F is not ray-invertible either, since F is not onto. Thus, in this case, Proposition 2(a),
(c), (d) do not apply, but Proposition 2(b) does. In fact, F cannot be a local homeomorphism
by Proposition 2(b); this may, of course, also be verified directly.
Now, take X = CIO, l] and Y = CA[O, l] with norm llylly = Ily’llx, and let F be defined again
by (2). In this case, the map F is not any more compact, but a global homeomorphism with
continuous F-‘(y) = y’. Consequently, F is not only ray-proper now, but even proper, closed,
and ray-invertible.

2. COERCIVE AND RAY-COERCIVE MAPS

Recall that a map F : X --) Y is called coercive if the preimage F-(B) of any bounded set
B c Y is bounded in X, and ray-coercive [5] if the preimage F-([e, y]) of [0, y] is bounded in X
for any y E Y. It is easy to see that the coercivity of F is equivalent to the condition

,,$ym IIF = 00. (3)


Nonlinear Operator Theory 1371

PROPOSITION 3. With X, Y, and F as above, the following statements are true.


(a) Every coercive map F is ray-coercive, but not vice versa.
(b) Every ray-proper map F is ray-coercive; the converse is true in case X = R?.
(c) Every coercive map F : EP + Y is proper; the converse is true in case Y = JR?.
(d) If F is linear, the coercivity of F is equivalent to the existence and boundedness of F-’
on F(X).

Again, the implications of Proposition 3 show that there are precisely seven combinations which
do not lead to a contradiction; we collect them in Table 2.

Table 2.

I Proper Ray-Proper Coercive Ray-Coercive

Example 1 Ye Yes yes

Example 11 yes no yes

Example 12 no Yes yes

Example 7 no no Yes

Example 13 no no Yes yes

Example 14 no no no yes

I Example 8 1 no 1 no 1 no 1 no

We make some remarks on Table 2. The map F from Example 1 is trivially coercive. The
map F from Example 7 is ray-coercive by Proposition 3(b), but not coercive by Proposition 3(c).
The fact that F from Example 8 is not ray-coercive is again trivial. The following four nontrivial
examples fit in the remaining rows in Table 2.
EXAMPLE 11. Let X = Y be as in Example 4, and put now, for 0 < S < l/a,

B, = {z : Ic E X, 115- ne,II 2 ns}, B= i=jBn,


n=l

and

K, = co({e,) U B,), K= ij K,.


n=l

Obviously, the sets K, are all closed and satisfy D(Km, K,) 2 l/&! - 6 > 0 for m # n, where
D(A, B) denotes the distance of the sets A and B. Now, we define F by

ifxEX\B,
F(x) = 1 _ 115- n4I e, + 112- %II 2, if x E B,.
nd > n6

Then, F is a continuous map with F(B,) = K,,. Since F(ne,) = e,, F cannot be coercive.
However, F is proper. In fact, let C c Y be compact. Then, C meets only a finite number of the
sets K,; therefore, it suffices to prove that the sets F-(C \ K) and F-(C n K,) are compact.
By definition of F, we have F-(C \ K) = C \ K. M oreover, for fixed n, we have F- (C n Kn) c
C u B,, since F(B,) = K,,. Thus, we only have to show that the set [F-(C n Kn)] n B, is
compact.
For fixed n, consider the sequence (yk)k defined by

Yk = 1 _ lh - n4I e, + llzk - nedl zk


( 726 > nS
1372 J. APPELLAND P. P. ZABREJKO

in B,. Suppose that this sequence converges to some y* E Y. Without loss of generality, we
may assume that the real sequence (&)k with & = l]xk - nen]] converges to some c E [0, n6] as
k 4 CQ. If < = 0, then (sk)k converges to ne, as k -+ cm. If 5 > 0, then & > 0 for large k, and
hence,

r*=Z(Tlt- [1-$]eJ -t2.=;(y*- [I--$Je,),

as k ---f co. In any case, (zk)k contains a convergent subsequence, and thus, the set [F-(C II
Km)] n B, is compact as claimed. The ray-properness and ray-coercivity of F follows from
Propositions 1 (b) and 3(b).
EXAMPLE 12. Let X = Y be as in Example 4, and let, again for 0 < 6 < l/a,

Again, Km n K, = 0 for m # n. Define F by

( x7 ifa:EX\K,
1 n - 1 ll4144l x

I n+n if x E K,, and ]]xl] 5 1,


F(x) =
cwl > ’
IIx/i- 1 + (2 - 11x]]) if x E K, and 1 < )(x]l 5 2,

I x, if x E K, and ]]x]] 2 2.

Then, F is continuous on X. Since ]/F(x)]] = 11x11for ]lx]] 2 2, F is certainly coercive. On the


other hand, F cannot be proper, because F(e,) = (l/n)e,. However, F is ray-proper. In fact,
for any y E Y, the preimage F-([B, y]) is b ounded by the coercivity of F and contained in the
ray {ty : 0 5 t < 00).
EXAMPLE 13. Let X = Y be an arbitrary infinite dimensional Banach space, e E X fixed with
]]e]] = 1, and F(x) = IIxlle. B ein g compact, F cannot be proper. On the other hand, F is
coercive, since llF(x)II = IIxlI for all x E X. A trivial calculation shows that

{x : x E x, llxll 5 T), if y = re for some r > 0,


F-(]e, Yl) = (@,
otherwise.

Consequently, F is ray-coercive but not ray-proper since X is infinite dimensional.


EXAMPLE 14. Let X = Y be as in Example 4, and let F(x) = f( Ilxll), where f : [0, co) -+ X is
the continuous vector function given by

(n - t)e, + (t - n + l)e,+l
f(t) = Il(n - t)e, + (t - n + l)e,+llI ’
n-llt<n, n = 1,2,3 ,***’

Since llF(x)II = 1 for all x E X, F cannot be coercive. It is not difficult to check that the
function f is l-l. Consequently, for any y E Y, the preimage F-((y)) is either empty or
coincides with the sphere {x : z E X, ~~x~~ = f-l(y)}. Th is s h ows that F cannot be ray-proper,
let alone proper.
Finally, for any y E Y, the preimage F- ([l-l,y]) is again either empty or some sphere, and
hence, F is ray-coercive.

3. LIPSCHITZ AND a-LIPSCHITZ MAPS


As before, let F : X 4 Y be a continuous map between two Banach spaces X and Y. Given
such a map, put
[FILip = sup ]]F(x) - F(‘)]]
+#5 lb - fll ’ (4
NonlinearOperator Theory 1373

and

(5)
As usual, F is called Lipschitz continuous if [F]Lip < co. In case of a bounded linear operator F,
we have of course [F]Lip = llF[l. A detailed d iscussion of various properties of Lipschitz continuous
nonlinear maps may be found in [7].
Let X be a Banach space and M C X a bounded subset. The (Hausdorff) measure of non-
compactness of M is defined by

o(M) = inf{& : e > 0, M has a finite c-net in X},

where by a finite s-net we mean, as usual, a covering by finitely many balls with radius c. A
continuous map F : X -+ Y is called a-Lipschitz if there exists some k > 0 such that

W’(W) I MM), (6)


for any bounded subset M c X. In this case, we put

(F]A = inf{k : k > 0, (6) holds} = sup *$$)),


C%(M)>0

and call the characteristic [F],J the a-Lipschitz norm (or measure of noncompactness) of F. In
particular, in case [F]A < 1, the map F is called condensing. Intuitively speaking, the condition
[F]A < 1 means that the image F(M) of any bounded set M c X is “more compact” than M
itself.
Parallel to (6) and (7), we will also be interested in the “reverse” condition

W(M)) L k4W, (8)


for any bounded subset M c X, and in the “lower” characteristic

[F& = sup{k : k > 0, (8) holds} = inf “zg)).


a(M)>0

Measures of noncompactness and condensing maps are nowadays a classical tool in both the
theory and application of nonlinear problems involving noncompact operators. For example,
there is a famous fixed-point theorem by Darbo f8] which states that a condensing map which
leaves a nonempty convex bounded closed subset of a Banach space invariant has a fixed point
in this set. This fixed-point theorem applies to integral equations, boundary value problems
for both ordinary and partial differential equations, functional-differential equations, imbedding
theorems, and many other problems. A detailed account of the state-of-the-art may be found
in [9].
We collect some information on the characteristics introduced above in the following proposi-
tion.

PROPOSITION 4. The characteristics (41, (5), (71, and (9) have the following properties.

(4 [F]L~* = 0 and only if F is constant.


(b) [F]li, > 0 implies that F is 1-1 and closed.
(cl [F-l]~i,, = [F&j if F is a homeomorpbism.
(4 [F],J = 0 if and only if F is compact.
(4 [Flo > 0 implies that F is coercive.
(0 [F-~]A = [F];l if F is a homeomorphism.
(fd [F]A I [F]Lip-
U-4[F]lipI [F]a ifx is infinite dimensional.
1374 J.APPELLAND P. P. ZABREJKO

The proof of most of these properties is straightforward. Let us just prove the assertion (e),
since it provides a link to our discussion of the previous two sections. If (3) is false, we may find a
bounded set B c Y such that F-(B) C X contains an unbounded sequence (CC,),. If ]]g]] 5 c for
all y E B, by the continuity of F, we may choose a sequence (6n)n of positive numbers such that
IIF(F(s9II 5 c for z E Bn:={x: 2 E X, I(z-x2,1) Id,}. Thesets M;:=Bl, iid2:=BlUBz,
. . . . Mn:=B1uB2u~~~uB, ,... are then bounded and satisfy 0 < cx(A&) --+ co as n -+ 00. On
the other hand, a(F(M,)) < 2c; consequently, [Fla = 0.
We remark that the converse of Proposition 4(e) is not true; i.e., the coercivity of F does not
imply that [Fla > 0. For example, the operator F : X 4 X defined by F(z) = ljz11z is a coercive
homeomorphism with continuous inverse F-‘(y) = y/m (F-l(d) = 0). On the other hand,
the sphere Sri, = {z : 2 E X, ]]z]l = l/n} is mapped onto the sphere Srjnz under F, and hence,
[Fla = 0.
Let us point out that the properties (d) and (g) are particularly interesting from the viewpoint,
of fixed-point theory. In fact, (d) shows that every compact operator is condensing, and (g) shows
that every contraction (i.e., [Fj~i~ < 1) is condensing as well. In other words, the above-mentioned
fixed-point theorem by Darbo is a common generalization of both Schauder’s fixed-point principle
and (a special version of) Banach’s contraction mapping principle. Many operators arising in
applications, however, satisfy the strict inequality 0 < [F]A < 1 5 [F]Lip which means that
they are natural examples of condensing operators which are neither compact nor contractive.
Examples of this kind which are related to the essential spectrum of an elliptic differential operator
over unbounded domains may be found in [lo].
If X is an infinite dimensional space, there are, in principle, eight possibilities for equality and
strict inequality occuring in the chain of estimates

[F]li, < [F], I [F]A I [F]Lip,

which is a combination of Proposition 4(g) and (h). The following eight elementary examples
correspond to these possibilities. In all these examples, we may choose X = l,, the linear space
of all bounded real sequences equipped with the supremum norm, and surprisingly, F as a linear
map.
EXAMPLE 15. Let F(zi,2s,zs,z4,2s ,...) = (21,22,53,24,55,,..); then

[F]li, = [F]a = [F]A = [F]Lip = 1.

Of course, the identity on any infinite dimensional space will do the same job.
EXAMPLE 16. Let F(zl, ~2, z3,24,z5,. . .) = (3% 23,24,25,&3,. . .); then

[F]np= 0, [F], = [FIA = [FJLip= 1.


For the first equality, just consider z = (1, 0, 0, . . 4) and z? = (0, 0, 0, . . . ).
EXAMPLE 17. Let F(q, z2,23,24,q,,. . .) = (x1,0, x3,0, ~5,. . .); then

[F]li, = [J’la= 0, [F]A = [J’]Lip = 1.

To see that [FIa = 0, observe that the set M = (z : 5 = (0,~,0, z4,0,. . . ), ]]z]] = 1) satisfies
cr(iV) = 1.
EXAMPLE 18. Let F(z~,x~,z~,x~,x~,. ..) = (x1,0,0,0,0,...); then

[F]lip
= [F]a= [F]A ~0, [F]Lip= 1.

EXAMPLE 19. LetF(zl,z2,23,x4,25,...)=(252,23,224,55,22s,...);then

[F]lip
= 0, [F]a= 1, [F]A = [F]Lip = 2.
Nonlinear Operator Theory 1375

EXAMPLE 20. Let F(zl, ~2, ~3, ~4, ~5,. . . ) = (‘&, x3,24, zg, ~6,. . . ); then

[J’jli, = 0, [% = [F]A = 1, [F]Lip = 2.

EXAMPLE 21. Let F(21,22,23,24,25 ,...) = (221,0,2s,O,25 ,... ); then

[Flli, = [J’lcz= 0, [F]A = 1, [Fluid= 2.

EXAMPLE 22. Let F(q,z2, x3, 24, x5,. . . ) = (3z2, x3, 224,x5, 2x6,. . . ); then

[F]e, = 0, [Fla = 1, [F]A = 2, [F]Lip = 3.

To obtain a general view of these examples, we collect them in Table 3.


Finally, let us illustrate the assertions of Propositions l-3 in the scheme presented in Figure 1
(courtesy of Miiuschen) .

Table 3.

Example 15 [F]lip = [I$ = [F]A = [F]Lip

Example 16 [qlleP < [Flo. = [F]A = [PILip


Example 17 [F]lid = [PIP < [F]A = (F]Lip
Example 18 [F]li, = [F], = [F]A < [F]Lip

Example 19 [p]billip
< [& < [F]A = [FjLip
Example 20 [PIlip < [PIa = [F]A < [F]Li,
Example 21 [Fllip = [Fla < [F]A < [F]Lip

Example 22 [F]lip < [& < [F]A < [F]Lip


t

-1. piaiq-* Iraycoercivc[,_._i


!
I ! 7-l

__) between any spaces X and Y

.___. _) under re.strictions on dim X or dim Y

Figure 1.
1376 J. A~~ELL AND P. P. ZABREJKO

REFERENCES
1. M.Z. Nashed and J. Hernlndez, Global invertibility in nonlinear functional analysis, In Proc. Z”d Intern.
Conf. Fixed Point Theory Appl., World Scientific, Singapore, (1991).
2. M. DGrfner, On a theorem of HernBndez and Nashed, Zeitschr. Anal. Anw. 1’7 (2), 267-270, (1998).
3. S. Bansch and S. Mazur, ijber mehrdeutige stetige Abbildungen, Studia Math. 5, 174-178, (1934).
4. F.E. Browder, Covering spaces, fiber spaces and local homeomorphisms, Duke Math. J. 21, 329-336, (1954).
5. J. HernBndez, Global invertibility in smooth and nonsmooth analysis, Ph.D. Thesis, Univ. of Delaware (1991).
6. M. Darfner, Beitriige zur Spektraltheorie nichtlinearer Operatoren, Ph.D. Thesis, Univ. of Wiirzburg (1997).
7. R.H. Martin, Nonlinear Operator and Differential Equations in Banach Spaces, Wiley, New York, (1976).
8. G. Darbo, Punti uniti in trssformazioni a codominio non compatto, Rend. Sem. Mat. Univ. Padova 24,
84-92, (1955).
9. R.R. Akhmerov, M.I. Kamenskij, A.S. Potapov, A.E. Flodkina and B.N. Sadovskij, Measures of Noncompact-
ness and Condensing Operators, (in Russian), Nauka, Novosibirsk, (1986); English translation Birkhliuser,
Basel, (1992).
10. J. Appell, Darbo’s fixed point principle after 30 years, In Proc. NATO Advanced Study Conf. Nonlin. Fun&.
Anal. Appl., Reidel, Dordrecht, (1986).

You might also like