A NNALES DE LA FACULTÉ DES SCIENCES DE T OULOUSE
V ALERIU A NISIU
Porosity and continuous, nowhere differentiable functions
Annales de la faculté des sciences de Toulouse 6
esérie, tome 2, n
o1 (1993), p. 5-14
<http://www.numdam.org/item?id=AFST_1993_6_2_1_5_0>
© Université Paul Sabatier, 1993, tous droits réservés.
L’accès aux archives de la revue « Annales de la faculté des sciences de Toulouse » (http://picard.ups-tlse.fr/~annales/) implique l’accord avec les conditions générales d’utilisation (http://www.numdam.org/conditions).
Toute utilisation commerciale ou impression systématique est constitu- tive d’une infraction pénale. Toute copie ou impression de ce fichier doit contenir la présente mention de copyright.
Article numérisé dans le cadre du programme Numérisation de documents anciens mathématiques
http://www.numdam.org/
- 5 -
Porosity and continuous,
nowhere differentiable functions(*)
VALERIU
ANISIU(1)
Annales de la Faculte des Sciences de Toulouse Vol. II, nO 1, 1993
Dans l’espace de Banach
C(
0 ,1 ~
des fonctions contenues sur( 0 ,
1],
les fonctions nulle part differentiables sont typiques non seulement dans le sens de la catégorie de Baire, mais aussi dans le sens plus restrictif, en utilisant la porosité. Les resultats sont ensuite generalises,en remplacant la différentiabilité par la diflerentiabilite approximative.
ABSTRACT. - In the Banach space
C[ 0, 1 ~ ] of
all continuous functionson
( 0 ,
1~,
the nowhere differentiable functions are typical not only in the Baire category sense but also in a stronger sense, using porosity.The results also hold when differentiability is replaced with approximate differentiability.
1. Introduction
S. Banach
[1]
and S. Mazurkiewicz[6]
showed that in the Banach spaceC[ 0 1]
of all continuous functions on[0, 1 ~,
endowed with the sup-norm, the set of nowhere differentiable functions is a residual( i. e.
acomplement
of a meager
set).
The firstexample
of a nowhere differentiable function wasconstructed
by
Weierstrass in1873,
but evenrecently,
such functions with furtherproperties
are exhibited(M.
Hata[4], [5]).
The notion of a set of
r-porosity,
definedby
L.Zajicek
in thegeneral
context of metric spaces, allows sometimes to
improve
resultsconcerning
meager sets.
(*) Reçu le 5 j anvier 1993
(1) Faculty of Mathematics, University of Cluj-Napoca, Str. Kogalniceanu 1, 3400 Cluj-Napoca, Romania
Let
(X, d)
be a metric space, A a subset of X and xE X.
For R > 0 wedenote
.S(c, R)
the open ball with center a~ and radius R and define:We say that A is porous at x if
p(x, A)
> 0. It iseasily
seen that A is porousat x if and
only
if there exists p > 0 such that for each c >0,
there existR, 0 R ~ ~ and z ~ X with
The set A is said to be porous if it is porous at all its
points. Finally,
Ais said to be r-porous if it is a union of
countably
many porous sets.It is
proved
in[7]
that any u-porous set is of the firstcategory (meager)
and if X is an euclidean space, it has
Lebesgue
measure zero, but in any Banach space there exist meager sets which are not r-porous.Let be a
property
for thepoints
of X.We say that P is
generic
withrespect
toporosity (or p-generic)
if the setof the
points
that do not possess thisproperty
is ~-porous. In this case, we shall also say that ap-typical
element of X has theproperty
P.2. Nowhere differentiable continuous functions We need the
following simple lemma,
similar with that of Evans[2].
LEMMA 2.1.2014 For each a > 0 there ezists a
function
ud : ~ --~IR, 1/ a-Lipschitzian
andhaving
theperiod
2a such that:a) 0 ud(t)
1 = EIR;
b) for
each t EIR,
there ezists an interval C( a/8 , a ~, having
thelengtha/8
so thatfor
h EIa(t):
Proof
. DefineThen ua is
evidently 2a-periodic, 1/a-Lipschitzian
and satisfiesa).
It issufhcient to check
b)
for tE ~ [0, 2a).
1ft E
~ 0 , 3a/4~ U ~
a,7a/4)
take =~ a/8 , a/4 ] and
if t E~ 3a/4 , a~ U [ 7a/4 2a)
takeI a ( t)
=~ 7a/8 , a ~
.The conclusion follows after some
simple computations. 0
THEOREM 2.2. - In the Ba.nach space
C~ 0 , 1 ~ of
all real continuousfunctions
on~ 0 , 1 ~,
, endowed with the sup-norm, ap-typical function
has atno
point finite
one-sided derivatives.Moreover, for
such afunction
~, onehas , ,
Proof.
- For n EIN ,
n >2,
denoteIt is clear that
U°° 2 An
contains all the functions x EC[0 , 1 ~ ] for
whichthere exists
[0,1)
such thatIn
particular,
if ac has a finiteright
derivative at apoint t 6 [0,1)
thenx E
An.
The case of the left derivatives can be handled in a similar manner, or can be reduced to theprevious
oneusing
themapping t
-; 1- t.So,
it is sufficient to prove that for afixed n 2: 2,
the setAn
is porous.Let x be an
arbitrary element
inAn
and e > 0.Using
the uniformcontinuity of z,
there exists 5 > 0 such that fort’, t~~
E~ 0 , 1 ~, ~t~
-we have
For this a >
0,
take ~othe function constructed in lemma
2.1,
restrictedto [0, 1].
We have then ud E
C [0 , 1], ~ua~ ( =
1 andwhere
Ia(t)
C~ a/8 , a ~, length Ia(t)
=a/8.
Denote u = 75 ~ua, z = x + u. We shall show that
In
fact,
for y EB (z, E},
t E~ 0 ,
1 -1 l n and
h EIa(t),
we have:Hence y ~ An
and(*)
holds.Because ~x - z~
= 75 ~, we havewhence
An
isporous. ~
Remark 2.3. - There are many different notions of
porosity ( cf.
L. Za-jicek [12]).
Forexample,
T. Zamfirescu[13]
uses aslightly
different one: theset A is said to be a-porous
(0
a1)
if for each ae E X and ~ > 0 there exists z EE)
such thatIt follows from
(**)
thatAn
is1/76-porous.
Let us mention
that,
the value of a in oursetting
is notsignificant
fora 03C3-porous
set,
because a theorem ofZajicek [11]
states that there existsa countable
decomposition
which has a uniformporosity arbitrarly
closedto
1/2.
Note also
that,
as the refereepointed out,
theproof
of the theorem 2.2also
gives
that theexceptional
set is 03C3-very porous(in terminology
of[12]).
The theorem 2.2 says
nothing
about thepossibility
that thetypical
function x has an
infinite,
evenbilateral,
derivative at somepoints.
Totreat this
aspect,
we need thefollowing
lemma.LEMMA 2.4. - For each a >
0,
there ezists afunction
ud : ~ -~!R, 1/a-Lipschitzian
andhaving
theperiod
7a such that:a) i 1 = I , t
EIR;
b)
For each t EIR,
there exists and intervalIa(t)
C[a, 6a], having
thelength a/10
such thatfor
h EIa(t):
Proof
. DefineThen
a)
is obvious and to checkb)
it is suficient to consider tE ~ [ 0, 7a)
andtake: i , ~ _ u _ _ ~ ,~ . r .. v r ~ . v
THEOREM 2.5. - In the Banach space
C[ 0 , 1 ~,
ap-typical function
hasat no
point
afinite
orinfinite
two-sided derivative.Moreover, for
such afunction lY,
on has:Proof
. For n EIN,
n >3,
denoteThe set U
Bn)
contains all the functions x EC~ 0 , ’ ~
for whichthere exists t E
(0,1)
such thatIt is sufficient to prove that for a
fixed n 2: 3,
the setAn
is porous(the
caseof
Bn being similar).
Let x be an
arbitrary
element inAn
obtained from the uniformcontinuity
of a*, like in theproof
of the theorem2.2,
choosea > 0 such that
Take ua
the function constructed in the lemma2.4,
restrictedto [0,
1],
We shall show that:
In
fact,
for y EB (z, ~),
t E[ 1 /n , 1 -
and t~ EIa(t),
we have:Hence y
~ An
and because =400 c,
we obtaintherefore
An
isporous. 0
Remark 2.6. - The
question
if ap-typical
function inC[ 0 1 ~
has at nopoint
a finite or infinite one-sidedderivative,
has anegative
answer. Infact,
even if such functions
exist,
the firstexample
was constructedby
Besicovitch(see
also[3], [4]), they
do not form a residual inC~ 0 , 1 ~, cf.
S. Saks[8],
and a
fortiori they
are notp-typical.
Note however that J.
Maly (5~
constructed a nonvoidcompact
subset inC[ 0, 1 ~ where
the functions with no(finite
orinfinite)
one-sided derivatives form a residual.Maly’s
construction does not use the existence of theBesicovitch-type
functions.3. The case of
approximate
derivativesThe
preceding
theorems can beimproved, replacing
thedifferentiability,
with
approximate differentiability,
as did Evans[2]
in the case of Bairecategory
results.Recall that
if
denotes theLebesgue
measure in R and A ç IR ismeasurable then the
right
lowerdensity
of A at t E IR is defined(see [9], [10])
as:If x is a measurable function then:
Similarly
theapproximate right
lower limit is defined. Then theapproxi-
mate
right
limit is the common value of these two extreme limits shouldthey
be the same. The
approximate
left limit and theapproximate
derivative aredefined in the standard manner.
THEOREM 3.1.2014 In the Banach space
C[ 0 , 1 ],
, ap-typical function
hasat no
point
afinite
one-sidedapprozimate
derivative.Moreover, for
such afunction
x, one has:Proof
. For n EIN,
n >2,
denote for7/8
c1,
then x E
An
It is then sufficient to prove that for
each n 2:
2(fixed),
the setAn
isporous.
Let x be an
arbitrary
element inAn
and c > 0.Taking
a, ua, u, z as in theproof
of the theorem2.2,
we have for y EB(z, ~),
t E~ 0 ,
1 -1/n ~,
s E
Ia(t):
Therefore
and we obtain
and so,
y ~ An.
Hence,
like in theproof
of the theorem 2.2:Remarks 3.2
a) Using
a similarmethod,
the theorem 2.5 may beimproved, substituing
the derivative with the
approximate
derivative and the extreme limits with theapproximate
extreme limits.b)
M. J. Evans[2] proved
that inC[0, 1]
the functions for whichare
typical.
This result can also be extended to ap-typical
one,using
the lemma in
[2]
and the above method.Note that in
[2]
a continuous functionsatisfying (***)
is first cons-tructed.
- 14 - References
[1]
BANACH(S.)
.2014 Uber die Baire’sche Kategorie gewisser Funktionenmengen,Studia. Math. 3
(1931),
pp. 174-179.[2]
EVANS(M. J.).
2014 Approximate smoothness of continuous functions, Colloq.Math. 54
(1987),
2, pp. 307-313.[3]
HATA(M.) .
2014 Singularities of the Weierstrass type functions, J. Analyse Math.51
(1988),
pp. 62-90.[4]
HATA(M.) .
2014 On continuous functions with no unilateral derivatives, Ann. Inst.Fourier Grenoble 38
(1988),
2, pp. 43-62.[5]
MALY(J.).
2014 Where the continuous functions without unilateral derivatives aretypical, Trans. Amer. Math. Soc. 283
(1984),
1, pp. 169-175.[6]
MAZURKIEWICZ(S.).
2014 Sur les fonctions non dérivables, Studia Math. 3(1931),
pp. 92-94.
[7]
PREISS(D.)
and ZAJICEK(L.).
2014 Fréchet differentiation of convex functions ina Banach space with a separable dual, Proc. Amer. Math. Soc. 91
(1984),
2,pp. 202-204.
[8]
SAKS(S.) .
2014 On the functions of Besicovitch in the space of continuous functions, Fund. Math. 19( 1932),
pp. 211-219.[9]
SAKS(S.).
2014 Theory of the integral, Warsaw, 1937.[10]
THOMSON(B. S.)
. 2014 Real functions, Lecture Notes in Math. 1170, Springer1985.
[11]
ZAJICEK(L.) .
2014 Sets of 03C3-porosity and sets of 03C3-porosity(q),
Casopis Pest. Mat.101
(1976),
pp. 350-359.[12]
ZAJICEK(L.).
2014 Porosity and 03C3-porosity, Real Anal. Exchange 13(1987-88),
pp. 314-350.