C OMPOSITIO M ATHEMATICA
F U C HENG H SIANG
On the convergence of a lacunary trigonometrical series
Compositio Mathematica, tome 15 (1962-1964), p. 28-33
<http://www.numdam.org/item?id=CM_1962-1964__15__28_0>
© Foundation Compositio Mathematica, 1962-1964, tous droits réser- vés.
L’accès aux archives de la revue « Compositio Mathematica » (http:
//http://www.compositio.nl/) implique l’accord avec les conditions gé- nérales d’utilisation (http://www.numdam.org/conditions). Toute utili- sation commerciale ou impression systématique est constitutive d’une infraction pénale. Toute copie ou impression de ce fichier doit conte- nir la présente mention de copyright.
Article numérisé dans le cadre du programme Numérisation de documents anciens mathématiques
http://www.numdam.org/
Series
by
Fu
Cheng Hsiang
1. A
lacunary trigonometrical
series is a series for which theterms different from zero are very sparse. This kind of
trigonomet-
rical series may be
put
in the form:We here assume, for
simplicity,
that the constant term of theseries also vanishes. A series S Ci is said to possess a gap
(u, v)
if c, = 0 for u i v. It is known that
[3,
p.251, § 10.81]
i f
a series ~Ci
possessesinlinitely
many gaps(Mkl m’ k )
such thatM’k/Mk > U
> 1f or
some p and is summable(C, 1) to
sumS,
then
S.
and alsoSmk
converges to S.2.
Suppose
that the abovetrigonometrical
series is the Fourier series of anintegrable
functionf(x).
This series possesses in-finitely
many gaps(nv, nV+l)
such thatn,,,/n,,
> A > 1 for some03BB.
Write,
for agiven point
xo ,Since
holds for almost all x, the Fourier series of an
integrable
func-tion is therefore almost
everywhere
summable(C, 1) by Fejér’s
theorem. From this
fact,
we drawimmediately
thefollowing
well-known
KOLMOGOROFF’S THEOREM
[1, 2]. Il
the Fourier serieso f
anintegrable function f (x )
possessesin f initely
many gaps(nv, nV+l)
such that
n"+l/n"
> ,u > 1, then thepartial sum Sny
converges almosteverywhere
to1(x).
We can also conclude
that,
at agiven point
xo at whichq*(t) =0(1) (t
-+0),
ifnV+l/nv
> Â > 1, thenSnv(aeo)
-f(xo)
as v 00.
29 3. We write
In this note, we
replace
the conditioncp*(t)
=0(l) by
theweaker
conditioncpl (t)
=0(1) (t -> + 0).
Wedevelop Kolmogoroff’s
theorem into the
following
manner.THEOREM.
Il
thelacunary
Fourier seriesof
anintegrable f unction f(x)
possessesin f initely
many gaps( n,, , nV+l)
such thatnV+l/nV
> Â> 1, andi f,
at agiven point
xo,( i ) CPl(t)
= 0(1) (t
-+0)
andwhen 0
t q f or
some ~, thenSnv(xo)
~f(x0)
as v --> oo.4.
Now,
we are in aposition
to prove the theorem.Take,
forinstance,
n" = n, n"+l = m and dénoterespectively by Dn(t)
and
Kn(t)
Dirichlet’s andFejér’s
kernels. ThenThen,
from theidentity
and in virtue of the
special property
of thelacunary
Fourierséries,
we obtainWrite
Intégration by parts gives
say. We are
going
to estimate the orders of theintegrals I1
andI. respectively.
We writesay. Since
( 2
sint/2)-I-t-1
isbounded,
weobtain, by
Riemann-Lebesgue’s theorem,
as m - oo
by
De la Vallée Poussin’s test[8,
p.88, f 2.8]
for theconvergence of Fourier series at a
given point by
the condition(ii) and ~2 (t)
= 0(1)
as 1 - + 0.Similarly,
31
as n - oo
by
the same test. It follows thatas n - oo. In
estimating
the order of1",
let us writesay. We have
say.
Now,
Considering
thatfor 0 t
n/2
and oc > 0, weget
Therefore,
Moreover,
we haveLast, by
the second mean valuetheorem,
we obtain’where A is an absolute constant. From the above
analysis,
itfollows that
For a
given a
> 0, we can choose a ô so small thatby
the condition(i).
Afterfixing
ô, we take asufficiently large
88
integer n
which makes11.1, 1141
and(m-n)-16-S
all less than e.Thus,
we obtainfinally
Since e is an
arbitrary
smallquantity, letting
it tend to zero,we
get
This proves the theorem.
REFERENCES.
A. KOLMOGOROFF
[1] Une contribution à l’étude de la convergence des séries de Fourier, Fund.
Math., 5 (1924), 96-97.
J. MARCINKIEWICZ
[2] A new proof of a theorem of Fourier series, Journ. London Math. Soc., 8
(1933), 279.
A. ZYGMUND
[8] Trigonometrical Series, Warsaw (1985).
National Taiwan University, Taipei, China.