• No results found

Hyperstability and Stability of a Logarithm type Functional Equation

N/A
N/A
Protected

Academic year: 2020

Share "Hyperstability and Stability of a Logarithm type Functional Equation"

Copied!
8
0
0

Loading.... (view fulltext now)

Full text

(1)

Hyperstability and Stability of a Logarithm-type

Functional Equation

Young Whan Lee

1

,

Gwang Hui Kim

2,∗

1DepartmentofComputerHackingandInformationSecurity,CollegeofEngineeringDaejeonUniversity,Daejeon,34520,Korea 2DepartmentofMathematics,KangnamUniversity,Yongin,Gyeonggi,16979,Korea

Copyright c2019 by authors, all rights reserved. Authors agree that this article remains permanently open access under the terms of the Creative Commons Attribution License 4.0 International License

Abstract

In 2001, Maksa and P´ales [12] introduced a new type’s stability: hyperstability for a class of linear functional equationf(x) +f(y) = n1Pn

i=1f(xϕi(y)). Riedel and Sahoo [14] have generalized a functional equation associated with

the distance between the probability distributions, which isf(pr, qs) +f(ps, qr) = 2M(rs)f(p, q) + 2M(pq)f(r, s). Elfen etc. [7] obtained the solution of the functional equationf(pr, qs) +f(ps, qr) = 2f(p, q) + 2f(r, s)on semigroupG. The aim of this paper is to investigate the hyperstability and the Hyers-Ulam stability for the above Logarithm-type functional equation considered by Elfen, etc. Namely, iff is an approximative equation related to the above equation, then it is a solution of this equation which exists withinε−bound of a given approximative functionf.

Keywords

Information Measure, Distance Measure, Superstability, Multiplicative Function, Stability of Functional Equation

AMS Subject Classification:39B82, 39B52

1

Introduction

The following stability problem is well-known as Ulam’s stability problem [16]:

LetG1be a group and letG2be a metric group with a metricd(·,·). Given >0, does there exist a δ >0

such that if a mappingh:G1→G2satisfies the inequalityd(h(xy), h(x)h(y))< δfor allx,y∈G1, then there

exists a homomorphismH :G1→G2withd(h(x), H(x))< for allx∈G1?

In next year, Hyers [11] proved a first partial answer to Ulam’s problem for an additive mapping on a Banach space. D. G. Bourgin obtained many excellent results for the stability ([3], [4]).Hyers’ theorem was generalized by Aoki [1] for the case bounded by variables, and their results are improved by Rassias [13] to the case of the linear mapping and by Ger [9] . G˘avruta [8] proved a further generalization of the Rassias’ theorem by using a general control function.

The superstability phenomenon of the exponential equationf(x+y) =f(x)f(y)was discovered by Baker, Lawrence, and Zorzitto [2] in 1979. The superstability for asymptotic phenomenon of the exponential equation was discovered by Ger [9].

In 2001, Maksa and P´ales [12] proved a new type’s stability for a class of linear functional equation f(x) +f(y) = 1

n

n

X

i=1

f(xϕi(y)), (1)

wheref is a real-valued mapping defined on a semigroupS, and the mappingsϕ1, ϕ2,· · ·, ϕn : S → Sare pairwise distinct

automorphisms. That is as following:

Letε:S×S→Rbe a function such that there exists a sequenceukthat satisfies

lim

k→∞ε(uks, t) = 0 (s, t∈S).

Assume thatf :S →Xsatisfies the stability inequality

f(s) +f(t)−

1 n

n

X

i=1

f(sϕi(t))

(2)

whereX is a real normed space. Then,fis a solution of (1).

Such a phenomenon is called the hyperstability of the functional equation. Gselmann [10], Brazde¸k and Ciepli´nski [5] investigated the hyperstability of functional equations. A similar concept was introduced by Sirouni and Kabbaj [15].

Riedel and Sahoo [14] solved a functional equation associated with the distance between the probability distributions. Let M : (0.1)→Cbe a given multiplicative function. Then, iff : (0.1)2

Csatisfies the functional equation f(pr, qs) +f(ps, qr) = 2M(rs)f(p, q) + 2M(pq)f(r, s)

if and only if

f(p, q) =M(p)M(q)hL(p) +L(q) +l(p q,

p q)

i

,

whereM : (0.1)→Cis an arbitrary logarithmic function andl: (0.1)2

Cis a bilogarithmic function. Thus, we will call it a logarithm-type functional equation

In addition, Elfen, Riedel and Sahoo [7] solved a functional equation

f(pr, qs) +f(sp, rq) = 2f(p, q) + 2f(r, s)

on semigroupG. Its solution type offon ¯G is given by

f(p, q) =A(p) +A(q) +ψ(pq−1, pq−1), whereA:G→Cis a homomorphism andψ:G→Cis a symmetric bi-homomorphism.

Now we consider the logarithm-type functional equation given by 1

2[f(pr, qs) +f(ps, qr)] =f(p, q) +f(r, s). (2) For example, iff(x, y) = lnxy, thenf is a solution of the equation (2). In this paper, we investigate the hyperstability and stability of the functional equation (2). Namely, we prove that iff satisfies a stability inequality for the equation (2), then it is also a solution of this equation and also we can find an another solution of it which has anε−error bound forf.

2

Hyperstability of the logarithm-type functional equation

In this section, we investigate the hyperstability of the equation (2). Throughout this section, let(G,·)denote a noncommu-tative semigroup,Xa real normed space, andRthe set of real numbers. And letR+denote the set of positive real numbers.

Theorem 1. Letε:G2×G2−→Rbe a function such that there exists a sequenceuk ∈Gthat satisfies

lim

k→∞ε(uk(p, q),(r, s)) = 0

for allp, q, r, s∈G. Assume thatf :G×G−→Xsatisfies the stability inequality

1 2

h

f(pr, qs) +f(ps, qr)i−f(p, q)−f(r, s)

≤ε((p, q),(r, s)) (3) for allp, q, r, s∈G.Then,

1 2

h

f(pr, qs) +f(ps, qr)i=f(p, q) +f(r, s).

Proof. Define a functionF :G2×G2−→Xby

F((p, q),(r, s)) =f(p, q) +f(r, s)−1

2

h

f(pr, qs) +f(ps, qr)i.

Then, for allp, q, r, s, v, w∈G, we have F((p, q),(r, s)) +1

2

h

F((pr, qs),(v, w)) +F((ps, qr),(v, w))i

=f(p, q) +f(r, s) +f(v, w)

−1

2

h

(3)

And also, for allp, q, r, s, v, w∈G, we have F((r, s),(v, w)) +1

2

h

F((p, q),(rv, sw)) +F((p, q),(rw, sv))i

=f(p, q) +f(r, s) +f(v, w)

−1

2

h

f(prv, qsw) +f(psw, qrv) +f(prw, qsv) +f(psv, qrw)i.

Thus,F satisfies the following functional equation F((p, q),(r, s)) +1

2

h

F((pr, qs),(v, w)) +F((ps, qr),(v, w))i (4) =F((r, s),(v, w)) +1

2

h

F((p, q),(rv, sw)) +F((p, q),(rw, sv))i.

By (3), we get

||F((p, q),(r, s))|| ≤ε((p, q),(r, s)),

and with the assumed sequence{uk}, we obtain

lim

k→∞F(uk(p, q),(r, s))≤klim→∞ε(uk(p, q),(r, s)) (5)

for allp, q, r, s∈G. The equation (4) implies

F((r, s),(v, w)) (6) =F((p, q),(r, s)) +1

2

h

F((pr, qs),(v, w)) +F((ps, qr),(v, w))i

−1

2

h

F((p, q),(rv, sw)) +F((p, q),(rw, sv))i.

Letr, s, v, w, p0, q0be fixed. Applying the norm and substitutingp=ukp0, q =ukq0in (6) , and ask→ ∞, respectively,

we obtain

|| lim

k→∞F(r, s),(v, w))||

=|| lim

k→∞

h

F(uk(p0, q0),(r, s)) +

1 2

h

F(uk(p0r, q0s),(v, w)) +F(uk(p0s, q0r),(v, w))

i

−1

2

h

F(uk(p0, q0),(rv, sw)) +F(uk(p0, q0),(rw, sv))

ii

||.

By applying of (5) and the triangle inequalities, we obtain

||F(r, s),(v, w))||

lim

k→∞

h

ε(uk(p0, q0),(r, s)) +

1 2

h

ε(uk(p0r, q0s),(v, w)) +ε(uk(p0s, q0r),(v, w))

ii

+ lim

k→∞

1 2

h

ε(uk(p0, q0),(rv, sw)) +ε(uk(p0, q0),(rw, sv))

i

.

Hence, we obtain from the assumed sequence{uk}the required result

F((r, s),(v, w)) = 0

for anyr, s, v, w∈G.

Corollary 2. Assume thatf :R+×R+−→Xsatisfies the stability inequality

1 2

h

f(pr, qs) +f(ps, qr)i−f(p, q)−f(r, s)

rs

pq or pqrs (7)

for allp, q, r, s∈R+. Then,

1 2

h

f(pr, qs) +f(ps, qr)i=f(p, q) +f(r, s).

Proof. Letε((p, q),(r, s)) = rspqanduk =ak fora >1, orε((p, q),(r, s)) =pqrsanduk =ak for0< a <1. Then, we

obtain

lim

k→∞ε(uk(p, q),(r, s)) = 0,

(4)

3

Stability of the logarithm-type functional equation

In this section, we investigate the stability of the equation (2). Throughout this section, let(G,·) denote a commutative semigroup,N the set of natural numbers, andXa Banach space.

Theorem 3. Letε >0.Assume thatf :G×G−→Xsatisfies the stability inequality

1 2

h

f(pr, qs) +f(ps, qr)i−f(p, q)−f(r, s)

≤ε (8)

for allp, q, r, s∈G.Then there exists a functionF :G×G−→X such that

1 2

h

F(pr, qs) +F(ps, qr)i=F(p, q) +F(r, s)

and||F(p, q)−f(p, q)|| ≤ 39ε

40 for anyp, q∈G, whereFis defined by

F(p, q) := lim

n→∞

h 1

22nf(p 2n, q2n)

+

n−2

X

i=0

1

2i·22n−i +

1 2n+1

f((pq)2n−1,(pq)2n−1)i

for anyp, q, r, s∈G.

Proof. Lettingr=p, s=qin (8) and dividing it by2, we have

1 22

h

f(p2, q2) +f(pq, pq)i−f(p, q)

ε

2. (9)

And also, lettingp=q=r=sin (8) and dividing it by2, we have

1 2f(p

2, p2)f(p, p)

ε

2. (10)

Let us show that the following inequality holds for everyn∈N:

1 2f(p

2n, p2n)f(p2n−1, p2n−1)

ε

2. (11)

Replacingpbyp2andqbyq2in (9) respectively, and dividing22, we have

1

222

h

f(p22, q22) +f((pq)2,(pq)2)i−f(p

2, q2)

22

ε

2·22. (12)

Thus by (9),(10), and (12), we have

1

222f(p

22, q22) + ( 1

222+

1 22

1 2)f((pq)

2,(pq)2)f(p, q)

1

222

h

f(p22, q22) +f((pq)2,(pq)2)i−f(p

2, q2)

22

+

1 22

h

f(p2, q2) +f(pq, pq)i−f(p, q)

(13)

+ 1 22

f((pq)2,(pq)2)

2 −f(pq, pq)

≤ ε

2·22 +

ε 2 +

ε 2·22 =

ε 2+

ε 22.

In addition, by lettingpbyp2n−1

andqbyq2n−1

in (9), and dividing22n−1

, the following inequality holds for everyn∈N:

1 22n

h

f(p2n, q2n) +f((pq)2n−1,(pq)2n−1)i−f(p

2n−1, q2n−1)

22n−1

≤ ε

(5)

By (10),(12), and (14), we have 1

223f(p

23, q23) + 1

223 + (

1

222+

1 23)

1 2

f((pq)22,(pq)22)−f(p, q) ≤ 1 223 h

f(p23, q23) +f((pq)22,(pq)22)i−f(p

22, q22)

222 + 1

222f(p

22, q22) + ( 1

222 +

1

23)f((pq)

2,(pq)2)f(p, q)

+ ( 1 222+

1 23)

f((pq)22,(pq)22)

2 −f((pq)

2,(pq)2)

(15)

≤ ε

2·222+

ε 2 +

ε 22 +

ε

222+

ε 23 1 2 =ε 2 + ε 22 +

ε

222 +

ε 24

=ε 2 +

ε 22 +

ε 23.

Note that

n

X

j=4

j−2

X

i=2

ε

2i·22j−i +

n

X

i=1

ε

22i−1 +

n

X

i=3

ε 2i+1

≤ ε

22

1

222+

1

223 +

1

224 +· · ·

+ ε 23

1

222+

1

223 +

1

224 +· · ·

+· · ·+ ε 2n−1

1

222 +

1

223 +

1

224+· · ·

(16) +ε

2+ ε 22+ε

1

222 +

1

223 +

1

224+· · ·

+ ε 24

1 + 1 2+

1 22 +

1 23+· · ·

=ε 1 222+

1

223+

1

224 +· · ·

1 + 1 22+

1 23+

1 24+· · ·

+ε 2+

ε 22+

ε 24

1 +1 2 +

1 22 +

1 23 +· · ·

=ε· 1

15· 3 2+ ε 2+ ε 22 +

ε 23

=39ε 40.

Suppose that the following inequality holds forn≥4and for anyp, q∈G:

1 22nf(p

2n, q2n) +

n−2

X

i=0

1

2i·22n−i +

1 2n+1

f((pq)2n−1,(pq)2n−1)−f(p, q)

≤ n X j=4

j−2

X

i=2

ε

2i·22j−i +

n

X

i=1

ε

22i−1 +

n

X

i=3

ε

2i+1. (17)

Note that

n−1

X

i=0

1

2i·22n+1−i =

1

22n+1 +

1 2

n−2

X

i=0

1

(6)

for alln∈N. Then, for anyp, q∈G, based on (14) and (18), we obtain 1

22n+1f(p 2n+1

, q2n+1)

+

n−1

X

i=0

1

2i·22n+1−i +

1 2n+2

f((pq)2n,(pq)2n)−f(p, q)

≤ 1

22n+1

h

f(p2n+1, q2n+1) +f((pq)2n,(pq)2n)i−f(p

2n, q2n)

22n

+ 1 22nf(p

2n, q2n) (19)

+

n−2

X

i=0

1

2i·22n−i +

1 2n+1

f((pq)2n−1,(pq)2n−1)−f(p, q)

+

n−2

X

i=0

1

2i·22n−i +

1 2n+1

f((pq)2n,(pq)2n)

2 −f((pq)

2n−1,(pq)2n−1)

= ε

2·22n +

n

X

j=4

j−2

X

i=2

ε

2i·22j−i +

n

X

i=1

ε

22i−1

+

n

X

i=3

ε 2i+1 +

nX−2

i=0

1

2i·22n−i +

1 2n+1

ε 2 = n X j=4

j−2

X

i=2

ε

2i·22j−i +

n−1

X

i=2

ε

2i·22n+1−i

+ ε 22n +

n

X

i=1

ε

22i−1

+ n X i=3 ε 2i+1 +

ε 2n+2

=

n+1

X

j=4

j−2

X

i=2

ε

2i·22j−i +

n+1

X

i=1

ε

22i−1 +

n+1

X

i=3

ε 2i+1.

Thus, by induction, inequality (17) holds for alln≥4and for anyp, q∈G. Now forn≥4, we have

1 22nf(p

2n, q2n) +

n−2

X

i=0

1

2i·22n−i +

1 2n+1

f((pq)2n−1,(pq)2n−1)

− 1

22n−1f(p 2n−1

, q2n−1)

n−3

X

i=0

1

2i·22n−1−i +

1 2n

f((pq)2n−2,(pq)2n−2)

(20) ≤ 1 22n

h

f(p2n, q2n) +f((pq)2n−1,(pq)2n−1)i− 1

22n−1f(p

2n−1, q2n−1)

+

n−3

X

i=0

1

2i·22n−1−i

f((pq)2n−1,(pq)2n−1)

2 −f((pq)

2n−2

,(pq)2n−2) + 1

2n+1f((pq)

2n−1,(pq)2n−1) 1

2nf((pq)

2n−2,(pq)2n−2)

≤ ε

2·22n−1 +

n−3

X

i=0

1

2i·22n−1−i ·

ε 2 +

ε 2·2n

≤ ε

2·22n−1 +

1 22n−1

n−3

X

i=0

1 2i+1 ·

ε 2 +

ε

2·2n −→ 0

asn→ ∞. Thus, if we let

Yn =

1 22nf(p

2n, q2n) +

n−2

X

i=0

1

2i·22n−i +

1 2n+1

(7)

then{Yn}is a Cauchy sequence due to (20), and so we can define a functionF :G×G−→Xby

F(p, q) := lim

n→∞

h 1

22nf(p 2n, q2n)

+

n−2

X

i=0

1

2i·22n−i +

1 2n+1

f((pq)2n−1,(pq)2n−1)i. (21) Then, due to (16), (17), and (21), we have

||F(p, q)−f(p, q)|| ≤ 39ε

40 ∀p, q∈G. Finally, the functionFdefined in (21) holds the required equation (2) as follows:

1 2

h

F(pr, qs) +F(ps, qr)i−F(p, q)−F(r, s)

≤ lim

n→∞

1 2

1

22nf((pr)

2n,(qs)2n)

+1 2

nX−2

i=0

1

2i·22n−i +

1 2n+1

f((pqrs)2n−1,(pqrs)2n−1)

+1 2

1 22nf((ps)

2n,(qr)2n)

+1 2

nX−2

i=0

1

2i·22n−i +

1 2n+1

f((pqrs)2n−1,(pqrs)2n−1)

− 1

22nf((p)

2n,(q)2n)

n−2

X

i=0

1

2i·22n−i +

1 2n+1

f((pq)2n−1,(pq)2n−1)

− 1

22nf((r)

2n,(s)2n)

n−2

X

i=0

1

2i·22n−i +

1 2n+1

f((rs)2n−1,(rs)2n−1)

≤ lim

n→∞

1 22n

1 2

h

f((pr)2n,(qs)2n) +f((ps)2n,(qr)2n)i

−f(p2n, q2n)−f(r2n, s2n)

+ lim

n→∞

nX−2

i=0

1

2i·22n−i +

1 2n+1

×

1 2

h

f((pqrs)2n−1,(pqrs)2n−1) +f((pqrs)2n−1,(pqrs)2n−1)i

−f((pq)2n−1,(pq)2n−1)−f((rs)2n−1,(rs)2n−1)

≤ lim

n→∞

ε

22n + limn→∞

nX−2

i=0

1

2i·22n−i +

1 2n+1

ε

= 0.

Corollary 4. Letε >0.Assume thatf :G×G−→Xsatisfies the stability inequality

1 22

h

f(p2, q2) +f(pq, pq)i−f(p, q)

≤ε (22)

for allp, q∈G.Then there exists a functionF :G×G−→Xsuch that

1 22

h

F(p2, q2) +F(pq, pq)i=F(p, q)

and||F(p, q)−f(p, q)|| ≤ 39ε

(8)

REFERENCES

[1] T. Aoki,On the stability of the linear transformation in Banach spaces, J. Math. Soc. Japan, 2 (1950), 64-66.

[2] J. A. Baker, J. Lawrence and F. Zorzitto,The stability of the equationf(x+y) =f(x)f(y), Proc. Amer. Math. Soc., 74 (1979), 242–246.

[3] D. G. Bourgin,Multiplicative transformations, Proc. Natl. Acad. Sci. USA, 36 (1950), 564-570.

[4] D. G. Bourgin,Approximately isometric and multiplicative transformations on continuous function rings, Duke Math. J., 16, (1949), 385-397.

[5] J. Brazde¸k and K. Ciepli´nski,Hyperstability and superstability, Abstr.Appl. Anal., Special Issue. (2013) 1-13.

[6] J. K. Chung, Pl. Kannappan, C. T. Ng and P. K. Sahoo,Measures of distance between probability distributions, J. Math. Anal. Appl., 138 (1989), 280-292.

[7] H. H. Elfen, T. Riedel and P. K. Sahoo,A variant of the quadratic functional equation on groups and an applicationBull. Korean Math. Soc. 54(6), (2017), 2165-2182

[8] P. Gavruta,A Generalization of the Hyers-Ulam-Rassias Stability of Approximately Additive Mapping, J. Math. Anal. Appl., 184 (1994), 431-436.

[9] R. Ger,Superstability is not natural, Roczik Naukowo-Dydaktyczny WSP w Krakowie, Prace Mat. 159, (1994) 109–123. [10] E. Gselmann,Hyperstability of functional equation, Acta Math. Hungar., 124(1-2), 2009, 179-188.

[11] D. H. Hyers,On the stability of the linear functional equation, Proc. Natl. Acad. Sci. 27(1941), 222-224.

[12] G. Maksa and Z. P´ales,Hyperstability of a class of linear functional equations, Acta Math. Acad. Paedagog. Nyhazi. (N. S.) 17 (2001) 107–112.

[13] Th. M. Rassias,On the stability of the linear mapping in Banach spaces, Proc. Amer. Math. Soc. 72(1978), 297-300. [14] T. Riedel and P. K. Sahoo,On a generalization of a functional equation associated with the distance between the probability

diastributions, Publ. Math. Debrecen, 46 (1995), 125-135.

[15] Sirouni and S. Kabbaj,The E-hyperstability of an Euler-Lagrange type quadratic functional equation in Banach spaces, British J. Math. Computer Science, 6(6) (2015), 481-493.

References

Related documents

This paper analyzes the application of network digital audio transmission technology, points out the significance of the application of network digital audio transmission technology

Segmental challenge with the individual top-dose of 100 μg carbon nanoparticles showed a significant relative increase of neutrophils (p = 0.05) in peripheral blood as compared

ABSTRACT: In present work, a new vanishing cream using a natural base from palm oil and palm kernel oil and standard vanishing cream using stearic acid were

of both the drugs. The absorbance of resulting solution was measured at 278 nm and 319 nm for both Ciprofloxacin HCl and Ornidazole, and calibration curves

Intraday precision was determined by for three times in the same day analyzing Capecitabine (morning, afternoon, evening) at respective absorption maxima using respective

One way to bridge gender gap in mathematics is to give them equal opportunity of learning via an interesting medium like the computer and it is in this view that this

In Chapter 4, we develop an algorithm for an improved autoregressive filter; in particular, we will implement the linear regression with consistency constraints to choose