See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/48188700

Twin problems from the Monthly and the Stolz-Cesaro Lemma

Article · January 2011

Source: arXiv

CITATIONS READS
0 750

1 author:

Eugen Ionascu
Columbus State University
75 PUBLICATIONS 376 CITATIONS

SEE PROFILE

All content following this page was uploaded by Eugen Ionascu on 20 May 2014.

The user has requested enhancement of the downloaded file.

 TWIN PROBLEMS FROM THE MONTHLY AND THE STOLZ-CESÀRO
LEMMA

EUGEN J. IONASCU
arXiv:1101.3278v1 [math.CA] 17 Jan 2011

1. INTRODUCTION. In this note1 we present solutions to two problems which appeared

in the American Mathematical Monthly. Although it may appear that one problem is more
general than the other, the two problems seem to cover different situations but both give
sufficient conditions for a real valued function not to be periodic over the real line. However
the two problems are related most intimately because they can be proved using essentially the
same technique. In the end, we introduce a new problem which actually implies both. These
problems appeared in ‘04-‘05 and their solutions in ‘06 and ‘07 (see [4], [5], [7] and [10]). The
proofs included here are based on a particular case of the well-known Stolz-Cesàro Lemma
and on the fact that every continuous periodic function on R must be uniformly continuous.
The use of the latter idea is not new as it was used in the published solutions of these
problems. On the other hand, the use of Stolz-Cesàro Lemma, is just another good example
where an old tool of analysis appears unexpectedly (see [2], [6], [8], [9], and [13]). L’Hospital’s
rule, which is very well known to calculus students is its “differentiable” counterpart.
The version of Stolz-Cesàro Lemma we are going to employ here is stated next.

Lemma 1. Let {an } and {bn } be two sequences such that {bn } is increasing and convergent
an an+1 − an
to infinity. If lim = ∞ then lim sup = ∞.
n→∞ bn n→∞ bn+1 − bn

We are going to include its classical idea of proof for completeness. Let us assume to the
an+1 − an
contrary that γ := lim sup < ∞. Then for an arbitrary ǫ > 0 there exists n0 ∈ N
n→∞ bn+1 − bn
such that
an+1 − an
≤ γ + ǫ,
bn+1 − bn
or

Date: January 18, 2011.

1991 Mathematics Subject Classification. (MSC2010) 54C30,42A75, 43A60 .
Key words and phrases. uniform continuity, periodic functions, almost periodic functions.
1A close version to this note appeared in Crux Mathematicorum with Mathematical Mayhem, vol. 34,
Issue 7, (2008), pp. 424-429. We are posting this here with the permission of the Canadian Mathematical
Society.
1
2 EUGEN J. IONASCU

(1) an+1 − an ≤ (bn+1 − bn )(γ + ǫ),

for all n ≥ n0 . Adding up inequalities as in (1) for n = k...l, l > k ≥ n0 , we obtain
al+1 − ak ≤ (bl+1 − bk )(γ + ǫ).
Eventually bl+1 is going to be a positive number so we can divide the last inequality by bl+1
and then let l → ∞. Using the hypothesis we obtain ∞ ≤ γ + ǫ which is a contradiction.
Next, we are including the two original problems.

Problem 11111 . Let f and g be nonconstant, continuous periodic functions mapping R

into R. Is it possible that the function h on R given by h(x) = f (xg(x)) is periodic?

The second problem seems to be more general but it is not clear to us at this point if this is
indeed the case. We are going to discuss the relationship between the two problems briefly
but it is not our purpose to get into the details of a thorough analysis.

Problem 11174. Let f and g be nonconstant, continuous functions mapping R into R

satisfying the following conditions:
1. f is periodic.
g(xn )
2. There is a sequence {xn }n≥1 such that lim xn = ∞ and lim = ∞.
n→∞ n→∞ xn
3. f ◦ g is not constant on R.
Determine whether h = f ◦ g can be periodic.

Both problems have a negative answer. The function h1 (x) = sin(x cos x) gives obvious
choices for f and g that satisfy the conditions in the first problem but it does not seem to be
an example (at least in an obvious way) good for the second problem. On the other hand the
function h2 (x) = sin(x2 ) gives rise to an f and a g that satisfy the conditions of the second
problem but it is hard to imagine that h2 (x) = fˆ(xĝ(x)) for some fˆ and ĝ nonconstant,
continuous periodic functions. It is an interesting question whether or not, for example, h1
can be covered by Problem 11111.
The conditions in Problem 11174 can be weakened to obtain:
Theorem 1. Let f and g be nonconstant, continuous functions mapping R into R and
satisfying the following conditions:
(i) f is periodic.
(ii) there exist sequences {xn }n≥1 and {yn }n≥1 such that
g(xn ) − g(yn)
inf |xn − yn | > 0 and lim = ∞.
n n→∞ xn − yn
 TWIN PROBLEMS FROM THE MONTHLY AND THE STOLZ-CESÀRO LEMMA 3

Under these assumptions the function h = f ◦ g cannot be periodic.

2. SOME FACTS FROM REAL ANALYSIS. Let us begin with this next fact about
continuous functions on compact sets (Theorem 4.19 in [12]).
Theorem 2. Every continuous function on a compact set in a metric space is uniformly
continuous.
We recall that a function on some domain D(f ) ⊂ R is uniformly continuous if for each
ǫ > 0 there exists a δ > 0 such that for every x, y ∈ D(f ) for which |x − y| < δ we have
|f (x) − f (y)| < ǫ.
An easy consequence of this theorem is:
Corollary 1. Every continuous periodic function on R is uniformly continuous.
This can be seen by applying Theorem 2 to the restriction of a periodic continuous function
on R to the compact set [0, 2T ] where T > 0 is a period of f and then taking the δ ′ given by
the theorem corresponding to an arbitrary ǫ. This δ ′ is good for the interval [0, 2T ] as the
theorem insures but then δ := min{δ ′ , T } is actually good for R as one can easily verify.
The idea of our proofs is to show that the function h is not uniformly continuous. As a
result of Corollary 1 we see that h cannot be periodic.
Let us see how Problem 11111 follows from Theorem 1. Since g is assumed to be continuous
and periodic but not constant we can find a and b such that g(a) − g(b) 6= 0. Assume T > 0
is a period of g. Then we consider xn = a + nT and yn = b + nT . Then |xn − yn | = |a − b| > 0
and

|xn g(xn ) − yn g(yn )|

lim = |a − b|−1 lim |ag(a) − bg(b) + nT (g(a) − g(b)| = ∞,
n→∞ |xn − yn | n→∞

ĝ
which says that x → xg(x) and f satisfy the conditions (i) and (ii) in Theorem 1 and so
f ◦ ĝ = h is not periodic. So we have a solution for Problem 11111.
To show that Problem 11174 follows from Theorem 1 we need the weaker version of the
Stolz-Cesàro Lemma as stated in the introduction as Lemma 1.
Now, let us assume the f , g and {xn } satisfy the conditions 1-3 in Problem 11174. We
can find a subsequence {xnk } of {xn }, so that xnk+1 − xnk ≥ 1 for all k, and for which either
g(xnk ) g(xnk )
lim = ∞ or lim = −∞. Without loss of generality we may assume the first
k→∞ xnk k→∞ xnk
situation because the other case is going to follow from this one by changing g with −g and
4 EUGEN J. IONASCU

fˆ
f with x → f (−x) (x ∈ R). By Lemma 1 we see that
g(xnk+1 ) − g(xnk )
lim sup =∞
k→∞ xnk+1 − xnk
which proves the existence of the two sequences in (ii) as in Theorem 1. Hence, Theorem 1
can be applied to f and g and get that h = f ◦ g is not periodic. This settles Problem 11174.

3. PROOF OF THEOREM 1. Let us start with f and g satisfying (i) and (ii) of
Theorem 1. Because g is continuous and by property (ii) we see that the interval In :=
g([xn , yn ]) (or In := g([yn , xn ]), for n large enough, must be an interval that has length
greater than the period T of f . Hence the range of f is the same as the range of h = f ◦ g.
Since f is assumed nonconstant then h is nonconstant. Therefore we can choose α and β
such that f (g(α)) 6= f (g(β)) and then we let ǫ0 = |f (g(α)) − f (g(β))| > 0. As we said in
the introduction the key idea is to prove that h is not uniformly continuous. More precisely,
we want to show that the definition of uniform continuity is not satisfied for this ǫ0 .
We fix n ∈ N large enough to insure that |In | > 2T and denote by ♯(g(α)) the number of
integer values of k for which g(α) + kT is in In . Then, it is easy to see that
|g(xn ) − g(yn )|
♯(g(α)) > − 1 > 1.
T
Similarly we denote by ♯(g(β)), the number of integers k for which g(β) + kT is in In .
Again, we have ♯(g(β)) > |g(xn )−g(y
T
n )|
− 1 > 1.
It is clear that the values g(α) + kT (k ∈ Z) interlace with those of g(β) + kT (k ∈ Z).
Using again the fact that g is continuous, by repeated application of the Intermediate Value
Theorem we can find two finite sequences uk and vk in the interval [xn , yn ] (or [yn , xn ])
both increasing and interlacing such that g(uk ) = g(α) + lk T and g(bk ) = g(β) + sk T with
lk , sk ∈ Z. The number of the intervals of the form [uk , vk ) (or [vk , uk ), [vk , uk+1), etc.) is at
least
M := min[2(♯(g(α) − 1), 2(♯(g(β) − 1)] ≥ 2.
These intervals form a partition of a subinterval of Jn := [xn , yn ] (or Jn := [yn , xn ]) of
length |xn − yn |. It follows that at least one of these intervals has to have length less than
or equal to |xnM
−yn |
.
We denote such an interval by [ζn , ηn ] and notice that

|xn − yn | |xn − yn | 1
(2) |ζn − ηn | ≤ < |g(xn )−g(yn )| = 2 |g(xn )−g(yn )|
→ 0 as n → ∞,
M 2 −4 − 4
T T |xn −yn | |xn −yn |
 TWIN PROBLEMS FROM THE MONTHLY AND THE STOLZ-CESÀRO LEMMA 5

and |f (g(ζn)) − f (g(ηn )| = ǫ0 . For an arbitrary but fixed δ > 0, we choose n even bigger
so that |ζn − ηn | < δ. This can be done because of (2). For such an n we still have
|h(ζn ) − h(ηn )| ≥ ǫ0 which proves that h is not uniformly continuous.
In the end we would like to leave the reader with a natural question: can Theorem 1 be
generalized to almost periodic functions? There are various concepts of almost periodicity
but we are going to include here as an example only Bohr’s definition:
A continuous real valued function ̥ defined on R is said to be almost periodic if for
each ǫ > 0 there exists an L > 0 such that every interval of length L contains an ǫ-period,
i.e. a number T such that |̥(x + T ) − ̥(x)| < ǫ for all x ∈ R.
What we find encouraging when it comes to new developments, related to the above ques-
tions, is the fact that every almost periodic function is also uniformly continuous ([3]).
Acknowledgements: We thank professor Albert VanCleave who gave us helpful sugges-
tions after reading an earlier version of this note.

References
[1] Art of problem Solving, http://www.artofproblemsolving.com/
[2] D. M. Batinetu-Giurgiu, Siruri (Sequences), Editura Albatros, Bucuresti 1979.
[3] C. Corduneanu, Almost Periodic Functions. New York: Wiley Interscience, 1961.
[4] P.P. Dalyay, Problem 11111, Amer. Math. Monthly, Vol. 111, no. 9, 2004, p. 822.
[5] P.P. Dalyay, Problem 11174, Amer. Math. Monthly, Vol. 112, no. 8, 2005, p. 749.
[6] E. J. Ionascu and P. Stanica, Effective Asymptotics for Some Nonlinear Recurrences and Almost Doubly-
Exponential Sequences, Acta Math. Univ. Comenian. (N.S.) no. 73 (2004), no. 1, 75–87.
[7] GCHQ Problem Solving Group, Amer. Math. Monthly, Vol. 113, no. 5, 2006, p. 467.
[8] A. Mannino, Some classic Stolz-Cesro theorems. (Italian), Atti Accad. Sci. Lett. Arti Palermo Parte I
(4) 37 (1977/78).
[9] G. Nagy, The Stolz-Cesaro Theorem http://www.math.ksu.edu/∼nagy/snippets/stolz-cesaro.pdf
[10] National Security Agency Problems Group, Amer. Math. Monthly, Vol. 114, no. 9, 2007, p. 836.
[11] K. A. Ross, Elementary Analysis: The theory of calculus, Springer-Verlag, 1980
[12] W. Rudin, Principles of Mathematical Analysis, Third Edition, McGraw-Hill, Inc. 1964.
[13] Gh. Sireţchi, The Toeplitz theorem and some of its consequences (Romanian), Gaz. Mat. (Bucharest)
90 (1985), no. 3, 65–70.
Department of Mathematics, Columbus State University
Columbus, GA 31907; ionascu eugen@colstate.edu
Honorific Member of the Romanian Institute of Mathematics “Simion Stoilow”

View publication stats