R E S E A R C H
Open Access
Interior fixed points of
unit-sphere-preserving Euclidean maps
Nirattaya Khamsemanan
1*, Robert F Brown
2, Catherine Lee
3and Sompong Dhompongsa
4*Correspondence:
1School of Information, Computer,
and Communication Technology, Sirindhorn International Institute of Technology (SIIT), Thammasat University, Prathum Thani, Thailand Full list of author information is available at the end of the article
Abstract
Schirmer proved that there is a class of smooth self-maps of the unit sphere in Euclideann-space with the property that any smooth self-map of the unit ball that extends a map of that class must have at least one fixed point in the interior of the ball. We generalize Schirmer’s result by proving that a smooth self-map of Euclidean n-space that extends a self-map of the unit sphere of that class must have at least one fixed point in the interior of the unit ball.
MSC: 55M20; 54C20
Keywords: interior fixed point theory
1 Introduction
For spacesX,Y and subsetsV⊆X,W⊆Y, a mapf:X→Y is anextensionof a map
φ: V →W iff(x) =φ(x) for allx∈V. We denote byBnthe unit ball inRn, bySn–its
boundary and byint(Bn) its interior.
Iff: B→Bis an extension ofφ: S→S={–, }andφhas no fixed points, thenf
must have aninteriorfixed point, that is, a fixed point inint(B). However, ifφhas a fixed
point, then there need not be any interior fixed points.
Ifn= , the situation is more complicated. Of course the Brouwer fixed point theorem implies that a mapf:B→Bmust have at least one interior fixed point if it is an extension
of a mapφ:S→Sthat has no fixed points. But it was proved in [] (see also []) that if
the extensionf is smooth, it may still be required to have interior fixed points for certain mapsφthat have many fixed points. Representing the points ofSby complex numbers,
letφ=φd:S→S, for an integerd, be thepower mapdefined byφd(z) =zd. Ifd≥ and
f:B→Bis a smooth extension ofφ
d, thenf has at least one interior fixed point. It is also
demonstrated in [] that interior fixed points of extensions need not exist ifd≤ or iff is not smooth. Schirmer generalized this interior fixed point result to smooth extensions f:Bn→Bnforn≥ to show in Example . of [] that iff is a smooth extension of a
‘sparse’ mapφ:Sn–→Sn–, a generalization ofφ
dthat is defined below, of degreedsuch
that (–)nd≥, thenf must have at least one interior fixed point.
Returning to the casen= , if we extend the mapφ:S→Swithout fixed points to a
mapf:R→R, there still must be a fixed point off inint(B). The reason for the interior
fixed points of the extensionf:B→Bof the map ofSwithout fixed points, namely that
(–, ) and (, –), lie in different components ofB×B\, where={(x,x) :x∈B},
ap-plies also to the extensionf:R→Rsince those points are also in different components
ofB×R\.
On the other hand, the reason for the presence of fixed points inint(Bn) for smooth
ex-tensions of certain maps ofSn–demonstrated in [] is considerably more subtle.
There-fore, it is reasonable to ask whether such fixed points would persist if, instead of smooth extensions f: Bn→Bn ofφ: Sn–→Sn–, we consider extensions that are smooth
Eu-clidean maps, that is, mapsf:Rn→Rn. Thus, we ask whether there still must be fixed
points off inint(Bn) if we allowf to map points ofint(Bn) outside ofBn.
We will prove that the interior fixed points do persist, even in this more general setting. As in the case of self-maps of balls, the interior fixed points of Euclidean maps are detected by means of a theorem that relates the index of a fixed point ofφ:Sn–→Sn–to its index
as a fixed point of an extension. We will therefore devote Section to a discussion of the properties of the fixed point index that we will use. In Section , we prove that a smooth extensionf:R→Rof a power mapφ
d:S→Sford≥ must have at least one
in-terior fixed point. Section then contains the proof that Schirmer’s result generalizes to smooth extensionsf:Rn→Rnof sparse mapsφ:Sn–→Sn–that satisfy the same degree restrictions.
2 The fixed point indices
Before extending the results of [, ] and [] to the case of a smooth Euclidean map f : (Rn,Sn–)→(Rn,Sn–) extendingφ:Sn–→Sn–, we need to define the relevant fixed
point indices. We will consider the restrictionf: (Bn,Sn–)→(Rn,Sn–). Since our goal is to establish conditions for the existence of fixed points on the interior ofBn, the behavior of the function outside ofBnis not relevant. Therefore, we will make use of the indices
i(Bn,f,p) andi(Sn–,φ,p) of an isolated fixed pointp∈Sn–. We do so by generalizing the
approach used in [] (see also []).
For an isolated fixed pointp∈Sn–, we can choose a small enough neighborhoodUso
that it contains only this fixed point and no other. We then may writef in this neighbor-hood ofpin terms of a local coordinate system in whichU∩Bnis contained in the upper half-space
Rn +=
(x,x, . . . ,xn)∈Rn|xn≥
in such a way thatpis the originin this setting andU∩Sn–is contained in the subspace
Rn–=(x
,x, . . . ,xn–, )∈Rn
.
In order to calculate the index ofpin each space, we consider the mapF:U→Rndefined by
F(x,x, . . . ,xn) =
⎧ ⎨ ⎩
(x,x, . . . ,xn) –f(x,x, . . . ,xn), ifxn≥,
(x,x, . . . ,xn) –f(x,x, . . . ,xn–, ), ifxn< .
Note thatFsends the origin, lower half-plane andRn–to itself respectively. Also,F(z)=z
forz= . The indexi(Sn–,φ,p) is equal toi(Rn–,F,) in this setting as in the traditional
definition of the index. Moreover,i(Bn,f,p) is identified withi(Rn,F,), which can be
3 Unit-circle-preserving maps of the plane
Brown, Greene and Schirmer proved
Theorem [, ] Let f :B→B be a smooth map with a finite number of fixed points
such that f(ζ) =ζkfor allζ ∈S for some k≥,where Bis the closed two-dimensional
ball with the boundary S.Ifπis a fixed point of f that lies in S,then either i(B,f,π) =
or i(B,f,π) = –.
The contractibility ofBimplies the following corollary.
Corollary [] Suppose f :B→Bis a smooth map such that f(ζ) =ζkfor allζ∈S,for
some k≥.Then there exists z∈int(B)such that f(z) =z.
We will extend Theorem to mapsf : (B,S)→(R,S) by modifying the proof of
The-orem in []. Corollary will then extend to mapsf: (R,S)→(R,S).
Theorem Let f : (B,S)→(R,S)be a smooth map with a finite number of fixed points
such that f(ζ) =ζkfor allζ ∈Sfor some k≥.Ifπis a fixed point of f that lies in S,then
either i(B,f,π) = or i(B,f,π) = –.
Proof Letπbe a fixed point offinS. We can write this fixed point in the polar coordinates
(r,θ) as (,θ). We will introduce new coordinates on a neighborhoodUofπas follows:
x=θ–θ, x= –r.
In the new coordinate setting, the fixed pointπis the origin andU∩Scorresponds to the
x-axis near , and the portion of the interior of the unit ball inUis contained in the upper
half-plane. Consider the following map (as described in Section ) in the new coordinate setting:
F(x,x) =
⎧ ⎨ ⎩
(x,x) –f(x,x), ifx≥,
(x,x) –f(x, ), ifx< .
Now writeF= (F,F) and defineg(x) =F(x, ). Sincef(ζ) =ζkforζ∈S, we have
g(x) =F(x, ) =x–kx= ( –k)x.
Since the mapf is defined to be smooth onB, the mapFis smooth on the upper
half-plane. LetF+denote the restriction ofFto the upper half-plane. We will see that
smooth-ness is only required in a neighborhood of the fixed point at . Since we are assuming that k≥, then
d
dxF(x, ) =g (x
) = –k<
and the smoothness ofF+implies that
∂F+ (x,x)
∂x
for (x,x) in an -neighborhood of the origin, for > sufficiently small and forx≥.
Letbe a circle of radius / about the origin.
Let+and–denote the half-circles above and below thex-axis respectively. Since
F takes the lower half-plane to itself, we know thatF maps– to the lower half-plane. Calculating the fixed point index off at the origin inRis equivalent to finding the winding number ofF() around . Thus, we need to understandF(+). Since+lies in the upper half-plane, we only considerF+. AssumingFhas only a finite number of fixed points, we
can choose small enough so that only one point onor its interior thatFmaps to the origin is the origin itself. Therefore, we can homotope the restriction ofF+to+inR\
to the restriction ofF+to the curve+
δ forδ> given by
+δ(t) =
(t– ),δ
– ( – t) ,
where ≤t≤.
We write the restriction ofF+to+
δ in coordinates as
F+δ+(t) =F+δ+(t),F++δ(t) =φδ(t),ψδ(t) .
The key idea of the proof is that forδsufficiently small, the smoothness ofF+and the fact
that
∂F+(x,x)
∂x
< implies that d
dtφδ(t) <
for allt. This tells us that thex-coordinate of the curveF+(+δ(t)) is a strictly monotone
function oft. In particular, the curveF+(+δ(t)) only crosses thex-axis once. This implies
the desired result since the winding number ofF() can then only be either or –. Notice that it is never specified thatf mapsBinto itself. In considering the mapF(z) =
z–f(z), although it is assumed thatFmaps the exterior of the disc to the exterior of the disc, the proof allows the image of the interior of the disc underFto lie anywhere inR.
Letf¯:B→Rbe the restriction off: (R,S)→(R,S) toB. SinceBis contractible,
the sum of the indices ofρf¯: B→Bequals one, and thereforeρ¯f(x) =xfor somex∈ int(B). But thenf(x) =xas well, sof: (B,S)→(R,S) has a fixed point in the interior
ofB. Therefore, we can extend Corollary as the following result.
Corollary Let f : (R,S)→(R,S)be a smooth map such that f(ζ) =ζkfor allζ ∈S
for some k≥.Then there exist z∈int(B)such that f(z) =z.
4 Interior fixed points of a mapf: (Rn,Sn–1)→(Rn,Sn–1)
Definition ([], p.) A smooth mapφ:Sn–→Sn–with finitely many fixed points is
transversely fixedifdφp–I:Tp(Sn–)→Tp(Sn–) is a nonsingular linear map for each fixed
pointp. ForF={p, . . . ,pr}a fixed point class ofφ, let
i(F) =
r
j=
Thetransverse Nielsen number N(φ) is defined by
N(φ) =
F∈F
i(F),
whereFis the set of fixed point classes ofφ.
A smooth mapφ:Sn–→Sn–issparseif it is transversely fixed and it hasN
(φ) fixed points.
In [], p. Schirmer obtained the following result.
Theorem Letφ:Sn–→Sn–be a sparse map of degree d and suppose f : (Bn,Sn–)→
(Bn,Sn–)is a smooth map extendingφ.If(–)nd≥,then f must have a fixed point in
int(Bn).
We will extend this result as a consequence of the following
Theorem Given a smooth map φ:Sn–→Sn– and a smooth map f : (Bn,Sn–)→
(Rn,Sn–)extendingφ,suppose that p∈Sn–is an isolated fixed point of f and that dφ p–I:
Tp(Sn–)→Tp(Sn–)is a nonsingular linear transformation.Then either i(Bn,f,p) = or
i(Bn,f,p) =i(Sn–,φ,p).
Proof The following proof is a modified version of Theorem . in []. We again writef in a small ball that containsp∈Sn–as described in Section and the mapF is also as
defined there. Moreover, forε> small enough, let
Dε(x, . . . ,xn) =
⎧ ⎨ ⎩
(εx,εx, . . . ,εxn) –f(εx,εx, . . . ,εxn), ifxn≥,
(εx,εx, . . . ,εxn) –f(εx,εx, . . . ,εxn–, ), ifxn< .
This then means that the indexi(Bn,f,p) =i(Bn,D
ε,) is the degree ofρ◦Dε:Sn–→Sn–,
whereρ(x) =x/|x|forx∈Rn\.
Note that
F(εx,εx, . . . ,εxn) –dFp(εx,εx, . . . ,εxn)≤o(ε)
becauseFis aCfunction. We also have
dFp(εx,εx, . . . ,εxn)≥Cε
for someC> independent ofεand (x,x, . . . ,xn) due to the fact thatdFp= –(dfp–I) is
nonsingular by hypothesis.
SincedFpis a nonsingular linear map, this last degree is easily seen to be , or±. But
the images ofDεof the upper and lower hemisphere are each contained entirely in either
the lower or upper half-space. This means thatDεis of degree orDε is homotopic in Rn\to the suspension ofD
ε|Sn–.
We can use Theorem to extend Theorem to the casef : (Rn,Sn–)→(Rn,Sn–). The
is solved in the previous section, the new material presented below extends the solution to all the cases.
Suppose we haveφ:Sn–→Sn–and a smooth mapf : (Rn,Sn–)→(Rn,Sn–) extend-ingφ. A fixed point classF off is called acommon fixed point classoff andφif there exists an essential fixed point classFofφwhich is contained inF.
We will again consider the restrictionf : (Bn,Sn–)→(Rn,Sn–). In the notation of [],
p., let
u(F) =max,iSn–,φ,F F⊂FandiSn–,φ,F > , l(F) =min
,iSn–,φ,F F⊂FandiSn–,φ,F < .
ThenFis atransversally common fixed point classoff andφif
l(F)≤i(F)≤u(F).
Definition Theboundary transversal Nielsen numberoff : (Bn,Sn–)→(Rn,Sn–) is
Nf;Bn,Sn– =N(φ) +N(f) –N(f,φ),
whereN(f,φ) is the number of essential and transversally common fixed point classes of f andφ.
Suppose thatφ:Sn–→Sn–has degreed. Thenf:Bn→Rnhas one fixed point classF
withi(F) = , and sol(F)≤ <i(F). Ifn= , thenφhas| –d|essential fixed point classes, each of the same index, andi(S,φ,F) =L(φ) = –d. Hence,i(F)≤u(F) if and only if
d≤, and
N(f,φ) =
⎧ ⎨ ⎩
ifd≤,
ifd> .
Ifn≥, thenφhas one fixed point classFwithi(Sn–,φ,F) = + (–)n–dand so
N(f,φ) = ⎧ ⎨ ⎩
if (–)n–d≥,
if (–)n–d< .
Note that this formula is still true for the casen= . Ifφ:Sn–→Sn–is a sparse map of degreedthenN
(φ) =| – (–)nd|for alldand all
n≥.
SinceN(f) = , for the case that (–)nd≤, the boundary transversal Nielsen number is
Nf;Bn,Sn– =N(φ) +N(f) –N(f,φ)
= – (–)nd+ –
As for the case that (–)nd> , the boundary transversal Nielsen number is
Nf;Bn,Sn– =N(φ) +N(f) –N(f,φ)
= – (–)nd+ –
= – (–)nd+ .
Hence, we have just proven the following
Proposition If f: (Bn,Sn–)→(Rn,Sn–) is a smooth extension of a sparse map φ : Sn–→Sn–of degree d,with n≥,then the boundary transversal Nielsen number is
Nf;Bn,Sn– = ⎧ ⎨ ⎩
| – (–)nd| if(–)nd≤, | – (–)nd|+ if(–)nd> .
As defined in [], p.,the extension Nielsen number N(f|φ) is a lower bound for the number of fixed points onSn–of continuous extensions of a continuous mapφ. It is equal
to the number of essential (classical) fixed point classesFoff withF∩Sn–=∅. A fixed point classFisrepresentableonSn–if there exists a subsetF⊂F∩Sn–withi(Bn,f,F) =
i(Sn–,φ,F). Thesmooth extension number N(f|φ) is the number of essential (classical)
fixed point classesFoff which are not representable onSn–. It is a lower bound for the
number of fixed points inSn–of a smooth extension of a smooth and transversally fixed
mapφ.
Proposition Ifφ:Sn–→Sn–is sparse,then
N(f|φ) =Nf;Bn,Sn– –N(φ), N(f|φ) =Nf;Bn,Sn– –N(φ).
Proof Our proof is modeled on the proofs of Proposition . and Corollary . in []. For any essential fixed point classFoff, sinceφis sparse,F∩Sn–containsu(F) fixed points psuch thati(Sn–,φ,p) = andl(F) fixed pointspsuch thati(Sn–,φ,p) = –. This means
thatFis representable onSn–if and only ifl(F)≤i(F)≤u(F). By the definitions of all the
Nielsen numbers involved, we have the result stated above forN(f|φ).
The result forN(f|φ) can be obtained in a similar manner by using Corollary . from [] along with the fact that all fixed point classes of a sparse map are essential.
By the definitions ofN(f,φ) in [] and the definition ofN(f|φ) defined early, we obtain
Proposition Ifφ:Sn–→Sn–is sparse,then the number of essential fixed point classes
of f which are common but not transversally common is
N(f|φ) –N(f|φ) =N(f,φ) –N(f,φ).
Theorem Let n≥,and letφ:Sn–→Sn–be a sparse map of degree d and suppose
f : (Bn,Sn–)→(Rn,Sn–)is a smooth map extendingφ.If(–)nd≥,then f must have a
fixed point inint(Bn).
Proof Sinceφhas degreedand it is sparse, by definition
N(f|φ) = ⎧ ⎨ ⎩
ifd= (–)n,
ifd= (–)n
and
N(f|φ) –N(f|φ) =Nf;Bn,Sn– –N(φ) –N(f|φ) (from Proposition ) =Nf;Bn,Sn– – – (–)nd–N(f|φ).
Applying Proposition , we have
N(f|φ) –N(f|φ) = ⎧ ⎨ ⎩
if (–)nd≤,
if (–)nd≥.
Thus, every smooth extension overBnof a sparse map ofSn–of degreedwith (–)nd≥
has a fixed point on the interior ofBn.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
All authors read and approved the final manuscript.
Author details
1School of Information, Computer, and Communication Technology, Sirindhorn International Institute of Technology
(SIIT), Thammasat University, Prathum Thani, Thailand. 2Department of Mathematics, University of California, Los Angeles, CA, USA.3Bard High School Early College, New York, NY, USA.4Department of Mathematics, Chiang Mai University, Chiang Mai, Thailand.
Acknowledgements
This work originated with Catherine Lee’s PhD dissertation under the supervision of Professor Robert F. Brown. The first author has been supported by the Thailand Research Fund grant number MRG5580232 under the mentorship of Professor Robert F. Brown of the Department of Mathematics, UCLA and Professor Sompong Dhompongsa of the Department of Mathematics, Chiang Mai University. She wishes to thank TRF as well. Last but not least, the authors would like to express their appreciation to both reviewers for their thorough reviews and many useful suggestions and comments.
Received: 29 June 2012 Accepted: 4 October 2012 Published: 17 October 2012
References
1. Brown, RF, Greene, RE, Schirmer, H: Fixed points of map extensions. In: Topological Fixed Point Theory and Applications, Tianjin, 1988. Lecture Notes in Math., vol. 1411, pp. 24-45. Springer, Berlin (1989)
2. Brown, RF, Greene, RE: An interior fixed point property of the disc. Am. Math. Mon.101(1), 39-47 (1994)
3. Schirmer, H: Nielsen theory of transversal fixed point sets. Fundam. Math.141(1), 31-59 (1992). With an appendix by Robert E. Greene
4. Dold, A: Fixed point index and fixed point theorem for Euclidean neighborhood retract. Topology4, 1-9 (1965) 5. Schirmer, H: A relative Nielsen number. Pac. J. Math.122, 459-473 (1986)
doi:10.1186/1687-1812-2012-183