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

net/publication/240154255

On the stability of the generalized cosine functional equations

Article · January 2001

CITATIONS READS

32 160

2 authors, including:

Gwang Hui Kim

Kangnam University
114 PUBLICATIONS 861 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Young Whan Lee View project

All content following this page was uploaded by Gwang Hui Kim on 19 May 2014.

The user has requested enhancement of the downloaded file.

Pl. Kannappan and G.H. Kim

On the stability of the generalized cosine

functional equations

Abstract. The aim of this paper is to study the stability

problem of the generalized cosine functional equations for
complex and vector valued functions.

1. Introduction.
One of the trigonometric functional equations studied extensively is
the equation

f (x + y) + f (x − y) = 2f (x)f (y), for x, y ∈ G (C)

(Wilson [8], Kannappan [6], Dacić [3]) known as the cosine or d’Alembert’s
functional equation where f : G → C, G, a group (not necessarily
Abelian) and C, the set of complex numbers. It is known (see [6]) that
if f satisfies (C) and the condition

f (x + y + z) = f (x + z + y), for x, y, z ∈ G, (K)

then there is a homomorphism m : G → C

m(x + y) = m(x)m(y), for x, y ∈ G, (1)

such that f has the form

1
f (x) = (m(x) + m(−x)), for x ∈ G. (2)
2
Ever since Ulam [7] in 1940 raised the stability problem of the Cauchy
equation f (x + y) = f (x) + f (y), many authors (see Hayer [5], Ger [4],
etc.) treated the stability problem for many other functional equations.
Baker [2] proved the result.

1
 Let ε ≥ 0 be a given number and let G(+) be an Abelian group.
Then any unbounded solution f : G → C of the inequality

|f (x + y) + f (x − y) − 2f (x)f (y)| ≤ ε, for x, y ∈ G

satisfies the cosine equation (C).

Badora [1] presented a new, short proof of Baker’s result.
The aim of this paper is to investigate the stability problem of the
functional equations

f (x + y) + f (x − y) = 2g(x)f (y), x, y ∈ G (3)

and

f (x + y) + f (x − y) = 2f (x)g(y), x, y ∈ G (4)

in the next two sections modelled after [1], where G is a group.

2. Stability of (3) and (4) for complex functions

In this section we will consider the stability of (3) and (4) and their
variants. First we will take up (3) and prove the following theorem.
Theorem 1.
Let ε ≥ 0 and f, g : G → C satisfy the inequality

|f (x + y) + f (x − y) − 2g(x)f (y)| ≤ ε (3)0

with f and g satisfying the (K) condition, where G(+) is a group. Then
either f and g are bounded or g satisfies (C) and f and g satisfy (3) and
(4). Further, in the latter case there exists a homomorphism m : G → C
satisfying (1) such that
b 1
f (x) = (m(x) + m(−x)) and g(x) = (m(x) + m(−x)), (5)
2 2
for x ∈ G, where b is a constant.
Proof. We will consider only the nontrivial f (that is, f 6= 0). Put y = 0
in (3)0 to get
ε
|f (x) − g(x)f (0)| < , for x ∈ G. (6)
2
2
If g is bounded, then using (6), we have

|f (x)| = |f (x) − g(x)f (0) + g(x)f (0)|

ε
≤ + |g(x)f (0)|,
2
which shows that f is also bounded. On the other hand if f is bounded,
choose y0 such that f (y0 ) 6= 0 and then use (3)0 ,
f (x + y0 ) + f (x − y0 ) f (x + y0 ) + f (x − y0 ) ε
|g(x)|− ≤ − g(x) ≤
2f (y0 ) 2f (y0 ) 2|f (y0 )|
to get that g is also bounded on G.
It follows easily now that if f (or g) is unbounded, then so is g (or
f ). Let f and g be unbounded. Then there are sequences {xn } and
{yn } in G such that g(xn ) 6= 0, |g(xn )| → ∞ as n → ∞ and f (yn ) 6=
0, lim |f (yn )| = ∞.
n
First we will show that g indeed satisfies (C).
¿From (3)0 with y = yn we obtain
f (x + yn ) + f (x − yn ) ε
− g(x) ≤ ,
2f (yn ) 2|f (yn )|
that is,
f (x + yn ) + f (x − yn )
lim = g(x). (7)
n 2f (yn )
Using (3)0 again and (K) we have

|f (x + (y + yn )) + f (x − (y + yn )) − 2g(x)f (y + yn )
+ f (x + (y − yn )) + f (x − (y − yn )) − 2g(x)f (y − yn )| ≤ 2ε

so that
f ((x + y) + yn ) + f ((x + y) − yn ) f ((x − y) + yn ) + f ((x − y) − yn )
+
2f (yn ) 2f (yn )
f (y + yn ) + f (y − yn ) ε
−2g(x) ≤ for x, y ∈ G,
2f (yn ) |f (yn )|
which with the use of (7) implies that

|g(x + y) + g(x − y) − 2g(x)g(y)) ≤ 0,

3
that is g is a solution of (C). Since g satisfies also (K), from [6] we see
that there exists a homomorphism m : G → C satisfying the second part
of (5).
As before applying (3)0 twice and the (K) condition, first we have

f (xn + y) + f (xn − y)
lim = f (y), (8)
n 2g(xn )
and then

|f ((xn + x) + y) + f ((xn + x) − y) − 2g(xn + x)f (y)

+ f ((xn − x) + y) + f ((xn − x) − y) − 2g(xn − x)f (y)| ≤ 2ε

so that
f (xn + (x + y)) + f (xn − (x + y)) f (xn + (x − y)) + f (xn − (x − y))
+
2g(xn ) 2g(xn )
g(xn + x) + g(xn − x) ε
−2· f (y) ≤ .
2g(xn ) |g(xn )|

¿From (8) and g satisfying (C), it follows that

|f (x + y) + f (x − y) − 2g(x)f (y)| ≤ 0,

that is, f and g are solutions of (3).

Finally, applying (3)0 , (7) and (K), we get

|f ((xn + y) + x) + f ((xn + y) − x) − 2g(xn + y)f (x) + f ((xn − y) + x)

+ f ((xn − y) − x) − 2g(xn − y)f (x)| ≤ 2ε

and that
f (xn + (x + y)) + f (xn − (x + y)) f (xn + (x − y)) + f (xn − (x − y))
+
2g(xn ) 2g(xn )
g(xn + y) + g(xn − y) ε
− 2f (x) · ≤
2g(xn ) |g(xn )|

resulting to (4). From (3) and (4), it is easy to see that f (x) = bg(x),
for some constant b.
This proves the theorem.

4
 We now consider a slight variation of (3)0 .
Theorem 2.
Let ε ≥ 0. Let fn : G → C (where G is a group) be a sequence of
functions converging uniformly to f on G. Suppose f, g, fn : G → C
satisfy the inequality

|f (x + y) + f (x − y) − 2g(x)fn (y)| ≤ ε, for x, y ∈ G (3)00

with f satisfying (K). Then either f is bounded or g satisfies (C) and f

and g satisfy (3).
Proof. Since {fn } is uniformly convergent to f, there is a positive integer
N such that

|fk (x) − f (x)| < ε for all x ∈ G, k ≥ N. (9)

Suppose that f is unbounded. Then (9) shows that so is fk , for k ≥ N.

Further, g is also unbounded. Indeed, y = 0 in (3)00 yields |f (x) −
g(x)fk (0)| ≤ 2ε .
If fk (0) = 0, then f is bounded which is not the case. Thus, g is also
unbounded.
Choose k ≥ N. Then there exist two sequences {xn } and {yn } in G
such that g(xn ) 6= 0, |g(xn )| → ∞, fk (yn ) 6= 0 and |fk (yn )| → ∞.
Inequality (3)00 with y = yn implies that

f (x + yn ) + f (x − yn ) ε
− g(x) ≤ , x ∈ G, k ≥ N.
2fk (yn ) 2|fk (yn )|

Hence we have
f (x + yn ) + f (x − yn )
lim = g(x). (10)
n 2fk (yn )

Now applying (3)00 again, we have

|f (x + (y − yn )) + f (x − (y + yn )) − 2g(x)fk (y + yn ) + f (x + (y − yn ))
+ f (x − (y − yn )) − 2g(x)fk (y − yn )| ≤ 2ε.

5
Using the (K) condition for f and dividing by fk (yn ), we get

f (x + y + yn ) + f (x + y − yn ) f (x − y + yn ) + f (x − y − yn )
+
2fk (yn ) 2fk (yn )
fk (y + yn ) + fk (y − yn ) ε
−2g(x) ≤
2fk (yn ) |fk (yn )|

for x, y ∈ G, k ≥ N.
Now
fk (y + yn ) + fk (y − yn ) f (y + yn ) + f (y − yn )
− g(y) = − g(y)
2fk (yn ) 2fk (yn )
fk (y − yn ) − f (y − yn ) + fk (y + yn ) − f (y + yn )
+ .
2fk (yn )

→ 0 as n → ∞, since {fn } converges uniformly to f, that is, use (9) and

(10).
Hence |g(x + y) + g(x − y) − 2g(x)g(y)| ≤ 0, that is, g is a solution
of (C).
Utilizing (3)00 with x = xn , for k ≥ N, we obtain

f (xn + y) + f (xn − y) ε
− fk (y) ≤ . (11)
2g(xn ) 2|g(xn )|

Again applying (3)00 and the (K) condition, we get

|f ((xn + x) + y) + f ((xn + x) − y) − 2g(xn + x)fk (y)

+ f ((xn − x) + y) + f ((xn − x) − y) − 2g(xn − x)fk (y)| ≤ 2ε

or
f (xn + x + y) + f (xn − (x + y)) f (xn + x − y) + f (xn − (x − y))
+
2g(xn ) 2g(xn )
g(xn + x) + g(xn − x) ε
− 2fk (y) · ≤ .
2g(xn ) |g(xn )|

By (11) and (C) (for g) we see that

|fk (x + y) + fk (x − y) − 2fk (y)g(x)| ≤ 0

6
for x, y ∈ G, k ≥ N. Since {fk } converges uniformly to f, we conclude
that f and g are solutions of (3). This completes the proof.
Now we take up the stability of (4). We prove the following theorems.
Theorem 3.
Let ε ≥ 0 and G be a group. Suppose f, g : G → C satisfy the
inequality
|f (x + y) + f (x − y) − 2f (x)g(y)| ≤ ε, for x, y ∈ G (4)0
with f even (that is, f (−x) = f (x)) and f satisfies (K). Then either f
and g are bounded or f and g are unbounded and g satisfies (C) and f
and g are solutions of (4).
Proof. We consider only nontrivial f, that is, f 6= 0. When f is bounded,
choose x0 such that f (x0 ) 6= 0 and use (4)0 to get
|f (x0 + y) + f (x0 − y)| f (x0 + y) + f (x0 − y)
|g(y)| − ≤ − g(y)
2|f (x0 )| 2f (x0 )
ε
≤ ,
2|f (x0 )|
which shows that g is also bounded.
Suppose f is unbounded. Choose x = 0 in (4)0 to have |f (y)+f (−y)−
2f (0)g(y)| ≤ ε, that is, |f (y) − f (0)g(y)| ≤ 2ε (this is the only place we
use that f is even). Since f is unbounded, f (0) 6= 0. Hence g is also
unbounded.
Let f and so g be unbounded. Then there exist sequences {xn } and
{yn } in G such that f (xn ) 6= 0, |f (xn )| → ∞, g(yn ) 6= 0, |g(yn )| → ∞.
Applying twice the inequality (4) and using (K) for f twice, first we
get
f (xn + y) + f (xn − y) ε
− g(y) ≤
2f (xn ) 2|f (xn )|
that is,
f (xn + y) + f (xn − y)
lim = g(y), for y ∈ G, (12)
n 2f (xn )
and then we obtain
|f ((xn + x) + y) + f ((xn + x) − y) − 2f (xn + x)g(y) + f ((xn − x) + y)
+ f ((xn − x) − y) − 2f (xn − x)g(y)| ≤ 2ε,

7
that is,

f (xn + (x + y)) + f (xn − (x + y) f (xn + (x − y)) + f (xn − (x − y))

+
2f (xn ) 2f (xn )
f (xn + x) + f (xn − x) ε
− 2g(y) ≤
2f (xn ) |f (xn )|

which by (12) leads to |g(x + y) + g(x − y) − 2g(y)g(x)| ≤ 0, so that g

satisfies (C).
Again applying the inequality (4)0 twice and using (K) condition for
f twice, first we have

f (x + yn ) + f (x − yn ) ε
− f (x) ≤ (13)
2g(yn ) 2|g(yn )|

and then we get

|f (x + (yn + y)) + f (x − (yn + y)) − 2f (x)g(yn + y) + f (x + (yn − y))

+ f (x − (yn − y) − 2f (x)g(yn − y)| ≤ 2ε

that is,

f ((x + y) + yn ) + f ((x + y) − yn ) f ((x − y) + yn ) + f ((x − y) − yn )

+
2g(yn ) 2g(yn )
g(yn + y) + g(yn − y) ε
−2f (x) ≤
2g(yn ) |g(yn )|

which by (13) and (C) yields

|f (x + y) + f (x − y) − 2f (x)g(y)| ≤ 0,

so that f and g are solutions of (4). This completes the proof of this
theorem.
Note that the evenness of f is used to prove that g is unbounded
when f is and nowhere else.
Theorem 4.
Let ε ≥ 0. Let fn : G → C (where G is a group) be a sequence of

8
functions converging uniformly to f on G. Suppose f, fn , g : G → C
satisfy the inequality
|f (x + y) + f (x − y) − 2fn (x)g(y)| ≤ ε, for x, y ∈ G (4)00
with f even and satisfies (K). Then either f is bounded or g satisfies (C)
and f and g are solutions of (4).
Proof. Since {fn } converges uniformly to f, there exists a positive integer
N such that
|fk (x) − f (x)| < ε for all x ∈ G, k ≥ N. (9)0
Suppose that f is unbounded. From (9)0 it follows that fk is unbounded
for all k ≥ N. With x = 0, (4)00 gives since f is even (this is the only
place we use the evenness to conclude that g is also unbounded) we have
|f (y) − 2fk (0)g(y)| ≤ 2ε . fk (0) = 0 implies that f is bounded which is
not the case. Hence fk (0) 6= 0 and g is also unbounded. Take k ≥ N.
Choose sequences {xn } and {yn } in G such that fk (xn ) 6= 0, |fk (xn )| →
∞, g(yn ) 6= 0 and |g(yn )| → ∞.
Apply the inequality (4)00 twice and f satisfies (K) to get
f (xn + y) − f (xn − y) ε
− g(y) ≤ , (14)
2fk (xn ) 2|fk (xn )|
and
|f ((xn + x) + y) + f ((xn + x) − y) − 2fk (xn + x)g(y) + f ((xn − x) + y)
+ f ((xn − x) − y) − 2fk (xn − x)g(y)| ≤ 2ε,
that is,
f (xn + (x + y)) + f (xn − (x + y)) f (xn + (x − y)) + f (xn − (x − y))
+
2fk (xn ) 2fk (xn )
fk (xn + x) + fk (xn − x) ε
− 2g(y) ≤ ,
2fk (xn ) |fk (xn )|
for x, y ∈ G, k ≥ N. Further,
fk (xn + x) + fk (xn − x) f (xn + x) + f (xn − x)
− g(x) = − g(x)
2fk (xn ) 2fk (xn )
fk (xn + x) − f (xn + x) + fk (xn − x) − f (xn − x)
+
2fk (xn )

9
→ 0 as n → ∞ because of (9)0 and (14). Thus,

|g(x + y) + g(x − y) − 2g(y)g(x)| ≤ 0,

showing thereby that g is cosine.

Finally, applying (4)00 twice and the fact that f satisfies (K), we obtain
f (x + yn ) + f (x − yn ) ε
− fk (x) ≤ , for x ∈ G, (15)
2g(yn ) 2|g(yn )|
and

|f (x + (yn + y)) + f (x − (yn + y)) − 2fk (x)g(yn + y) + f (x + (yn − x)

+ f (x − (yn − y)) − 2fk (x)g(yn − y)| ≤ 2ε,

that is
f ((x + y) + yn ) + f ((x + y) − yn ) f ((x − y) + xn ) + f ((x − y) − xn )
+
2g(yn ) 2g(yn )
g(yn + y) + g(yn − y) ε
−2fk (x) ≤
2g(yn ) |g(yn )|
which yields by (15) and (C) that

|fk (x + y) + fk (x − y) − 2fk (x)g(y)| ≤ 0.

Since {fn } converges uniformly to f, we see that f and g are solutions

of (4). This completes the proof.

3. Stability of (3) and (4) for vector valued functions

In [1] Badora gave a counter-example to illustrate the failure of the
superstability of the cosine functional equation (C) in the case of the
vector valued mappings. Therefore there is a need to consider the vector
valued functions separately. We prove the following two theorems in this
section. Let G be a group and A be a complex normed algebra with
identity.
Theorem 5.
Suppose f, g : G → A satisfy the inequality

kf (x + y) + f (x − y) − 2g(x)f (y)k ≤ ε, (3)000

10
for x, y ∈ G with f satisfying (K) and

kf (x) − f (−x)k ≤ η, for x ∈ G, (16)

for some ε, η ≥ 0. Suppose there is a z0 ∈ G such that g(z0 )−1 exists and
kf (x)g(z0 )k is bounded for x ∈ G. Then there is a m : G → A such that

km(x + y) − m(x)m(y)k ≤ a1 , for x, y ∈ G (17)

and
1
f (x) − (m(x) + m(−x)) ≤ a2 , for x ∈ G (18)
2
for some constants a1 and a2 .
Proof. Let sup kf (x)g(z0 )k = M. Then using (3)000 and (16), we get by
x∈G
using (K)

kf (x)g(−z0 )k ≤ kf (−x)g(z0 )k + kf (x)g(−z0 ) − f (−x)g(z0 )k

1
≤ M + kf (z0 − x) + f (z0 + x) − 2g(z0 )f (−x)
2
− (f (−z0 + x) + f (−z0 − x) − 2g(−z0 )f (x))
− (f (z0 + x) − f (−z0 − x) + f (z0 − x) − f (−z0 + x)k
≤ M + ε + η. (19)

Define a function h : G → A given by

1
h(x) = (f (x) + f (−x)), for x ∈ G.
2
Then h is even, that is, h(−x) = h(x),
η
kh(x) − f (x)k ≤ for x ∈ G, kh(x)g(z0 )k ≤ M. (20)
2
Define a function m : G → A by

m(x) = h(x) + ig(z0 ), for x ∈ G.

11
Utilizing (20), we get
km(x + y) − m(x)m(y)k = kh(x + y) + ig(z0 ) − h(x)h(y)
+ i(h(x) + h(y))g(z0 ) + g(z0 )2 k
≤ kh(x + y)k + kh(x)h(y)k
+ k(h(x) + h(y))g(z0 )k + kg(z0 )k + kg(z0 )k2
≤ kh(x + y) − f (x + y)k + kf (x + y)k
+ kh(x)h(y)g(z0 )2 · g(z0 )−2 k + kh(x)g(z0 )k
+ kh(y)g(z0 )k + kg(z0 )k + kg(z0 )k2
η
≤ + M kg(z0 )k−1 + M 2 kg(z0 )k−2
2
+ 2M + kg(z0 )k + kg(z0 )k2
= a1 (say) which is (17).
Finally by (20), we have
1
f (x) − (m(x) + m(−x)) = f (x) − h(x) + h(x)
2
1
− (h(x) + h(−x)) − ig(z0 )
2
η
≤ + kg(z0 )k = a2 (say), which is (18).
2
This proves the theorem.
Lastly we prove the following theorem.
Theorem 6.
Let f, g : G → A satisfy the inequality
kf (x + y) + f (x − y) − 2f (x)g(y)k ≤ ε, x, y ∈ G, (4)000
with f satisfying (K) and
kf (x) − f (−x)k ≤ η, for x ∈ G,
for some nonnegative ε and η. Suppose there exists a z0 ∈ G such that
g(z0 )−1 exists and kf (x)g(z0 )k is bounded over G. Then there exists a
mapping m : G → A such that
km(x + y) − m(x)m(y)k ≤ a1 , for x, y ∈ G

12
and
1
f (x) − (m(x) + m(−x)) ≤ a2 , for x ∈ G,
2
for some constants a1 and a2 .
The proof runs parallel to that of Theorem 5.

References
[1] R. Badora, On the stability of cosine functional equation, Rocznik
Naukowo-Dydak., Prace Mat. XV (1998), 1–14.
[2] J.A. Baker, The stability of the cosine equation, Proc. Amer. Math.
Soc. 80 (1980), 411–416.
[3] J.R. Dacić, The cosine functional equation for groups, Mat. Vesnik
6.21 (1969), 339–342.
[4] R. Ger, A survey of recent results on stability of functional equa-
tions, Proceedings of the 4th International Conference on Func-
tional Equations and Inequalities, Pedagogical University in Krakow
(1994), 5–36.
[5] D.H. Hyers, On the stability of the linear functional equation, Proc.
Nat. Acad. Sci. U.S.A. 27 (1941), 222–224.
[6] Pl. Kannappan, The functional equation f (xy)+f (xy −1 ) = 2f (x)f (y)
for groups, Proc. Amer. Math. Soc. 19 (1968), 69–74.
[7] S.M. Ulam, A collection of mathematical problems, Interscience
Tracts in Pure and Applied Mathematics 8, Interscience Publishers,
New York-London, 1960.
[8] W.H. Wilson, On certain related functional equations, Bull. Amer.
Math. Soc. 26 (1920), 300–372.
Pl. Kannappan
Department of Pure Mathematics
University of Waterloo
Waterloo, ON N2L 3G1
Canada
E-mail: plkannappan@watdragon.uwaterloo.ca

13
 G.H. Kim
Department of Mathematics
Kangnam University
Suwan, 449-702 Korea
E-mail: ghkim@kns.kangnam.ac.kr

14

View publication stats