Volume 2013, Issue 1 380743
Research Article
Open Access

About the Local Stability of the Four Cauchy Equations Restricted on a Bounded Domain and of Their Pexiderized Forms

Fulvia Skof

Corresponding Author

Fulvia Skof

Dipartimento di Matematica, Università di Torino, Via Carlo Alberto 10, 10123 Torino, Italy unito.it

Search for more papers by this author
First published: 16 December 2013
Academic Editor: Janusz Brzdęk

Abstract

The Ulam-Hyers stability of functional equations is widely studied from various points of view by many authors. The present paper is concerned with local stability of the four Cauchy equations restricted on a bounded domain. These results can be easily adapted to the corresponding Pexiderized equations.

1. Introduction

After this introduction, in Section 2 the local stability of the additive equation
()
and, as a consequence, of the logarithmic equation
()
both restricted on a bounded domain in R2 is studied.

It is well known that the problem of stability was posed, for the additive equation, by S. Ulam and was solved by Hyers [1] in 1941, with reference to the equation valid on the whole space. Afterwards, stability was widely studied by many authors, from various points of view, considering further equations on the whole space or putting them in very general settings (see, for instance, [2, 3]).

As for the “local” stability of equations on a restricted domain, first results can be found in [4, 5] (see also [6]) and they concern substantially the set of functions f : DfR → (S,  ∥·∥), satisfying the condition of δ-additivity
()
restricted either on the triangular domain
()
for some given r > 0, or the unbounded domain
()
for some given a > 0.
In the present paper (Section 2) the bounded restricted domain of inequality (3) will be assumed to be the triangle
()
for some given (a, b) ∈ R2 and r > 0.

It has to be remarked that in the classical paper [1] by Hyers as well as in case of restricted domains studied in [4, 5], the solutions of the “equation” correlated to the inequality (3) on the given domain are either additive functions on the whole space (in cases of the general result by Hyers and of the domain Ka defined in (5)) or the restrictions to Df = [0, r) of functions F additive on R2 (in case of a domain like E0 defined in (4)).

On the contrary, when a restricted domain like that in (6) is assumed, the local solution of the corresponding additive equation, restricted to the same set, is different from the restriction on the domain of f of some function, which is additive in the whole space R2 (see [7]).

Therefore, in order to adhere to the sense of Ulam’s question in case of a restricted domain like E(a, b; r) too, the locally δ-additive function f has to be compared with the local solution of the corresponding exact equation restricted to the same set E(a, b; r).

In this frame, in Section 2, first the local stability of the additive Cauchy equation restricted to E(a, b; r) will be proved (Theorem 1); then, as a consequence of this result, the local stability of the logarithmic Cauchy equation (2) will be proved (Theorem 6).

These results can be easily extended to the Pexiderized forms of the same equations.

Notice that the problem of the “local” stability for the remaining two Cauchy equations,
()
()
restricted to bounded domains, requires a suitable slightly different approach because of the peculiar properties of the local solutions of such equations when they are restricted on bounded domains (see [8, 9]).

This problem will be the object of Section 3 of the present paper, where results of local stability of (7) and (8) will be proved (Theorems 15 and 16, resp.).

2. About the Additive and the Logarithmic Cauchy’s Equations

2.1. A Result on Local Stability of the Additive Equation

In the set of functions f of a real variable with values in a normed space S = (S, ∥·∥), let us consider the inequality
()
for some δ > 0 and (x, y) ∈ E(a, b; r), defined in (6), for given (a, b) ∈ R2 and r > 0.
As usual, the projections of E will be denoted by
()

For E(a, b; r) as defined in (6), the projections are Ex = [a, a + r), Ey = [b, b + r),  Ex+y = [a + b, a + b + r), and the domain Df of functions f satisfying (9) is Df = ExEyEx+y.

We purpose to check whether each f :   Df → (S, ∥·∥) satisfying (9) for (x, y) ∈ E(a, b; r) is uniformly approached on Df by some φ :   DfS satisfying the additive equation restricted to E(a, b; r), namely, by a function φ of the following form (see [7]):
()
where h0 : RS is additive on the whole space R2 and α, βS are constant.

A positive answer is given by the following.

Theorem 1. Let f :   Df → (S, ∥·∥), (S, ∥·∥) being a Banach space, satisfy (9) for some δ > 0 and every (x, y) ∈ E(a, b; r), defined in (6), for given (a, b) ∈ R2 and r > 0; Df = ExEyEx+y.

Then there exists (at least) a function H :   RS, additive on R2, such that the function F :   DfS defined by

()
has both the following properties:
  • (i)

    F  is a (local) solution of the additive equation restricted on E(a, b; r);

  • (ii)

    F approaches uniformly f on Df, and

    ()
    holds for every tDf.

In order to prove Theorem 1, let us premise two lemmas.

Lemma 2. Let f :   Df = ExEyEx+y → (S, ∥·∥) satisfy the condition (9) restricted to the set E = E(a, b; r) defined in (6). Then the functions γi : [0, r) → S, i = 1,2, 3, defined by

()
for t ∈ [0, r) satisfy both the following inequalities:
()
()

Proof of Lemma 2. Let us prove (15) firstly for i = 1, j = 2. For t ∈ [0, r) the points (a + t, b) and (a, b + t) belong to E(a, b; r); hence, from (9) both the following inequalities hold

()
whence,
()
namely, (15) for i = 1, j = 2.

Similarly we prove (15), for i = 1, j = 3, assuming the pairs (a, b) and (a + t, b) with t ∈ [0, r), from the formulas

()
and for i = 2,   j = 3, assuming the pairs (a, b), (a, b + t), and t ∈ [0, r).

In order to prove (16) let us assume (ξ, η) ∈ E0 = E(0,0; r).

For i = 1, from (15) and (9) we get

()
since (a + ξ + η, b) ∈ E(a, b; r) and (a + ξ, b + η) ∈ E(a, b; r), it follows from (9) that
()
with ∥σ∥ < 2δ, ∥ρk∥ < δ for k = 1,2, whence (16) for i = 1.

Similarly, for i = 2, a and b interchanged.

As for γ3 with (ξ, η) ∈ E0 = E(0,0; r), from (9)

()
where ∥τk∥ < δ, k = 1,2, 3,4, whence ∥Φ3∥ < 4δ.

Lemma 2 is proved.

Lemma 3. If f :   Df = ExEyEx+y → (S, ∥·∥), S being a Banach space, satisfies (9) on E = E(a, b; r), then each of the functions γi(t), i = 1,2, 3, defined in (14) for t ∈ [0, r), is uniformly approached on [0, r) by the restriction of a function Hi : RS additive on R2.

The following inequalities hold:

()
()

Proof of Lemma 3. Since each function γ1, γ2, γ3 is 4δ-additive on E(0,0; r) (from Lemma 2), formula (23) follows immediately from a known result [4, Lemma 2]:

Letf : [0, d) → (X, ∥·∥), X   a Banach space, be   δ-additive on   Ed = {(x, y) ∈ R2 : 0 ≤ x < d,   0 ≤ y < d,   x + y < d}.

Then there exists at least one additive function   L : RX  such that

()
Moreover, for every i, j = 1,2, 3 and t ∈ [0, r), formula (24) follows from (15) and (23); in fact
()
This means that the restriction to [0, r) of each additive Hj, j = 1,2, 3, approaches uniformly on [0, r) each function γ1, γ2, γ3.

Lemma 3 is proved.

Proof of Theorem 1. According to (24) in Lemma 3, each function γi(t), i = 1,2, 3, defined in (14) for t ∈ [0, r), namely,

()
is uniformly approached on [0, r) by each of the additive functions Hj : RS, j = 1,2, 3.

Let us define Fi : DfS, i = 1,2, 3, as follows:

()
Such functions Fi, i = 1,2, 3, satisfy obviously the additive equation restricted to E(a, b; r).

Moreover, thanks to Lemmas 2 and 3, each function Fi approaches uniformly f on Df as in formula (13); in fact, for arbitrary (x, y) ∈ E(a, b; r),

()
similarly for yEy, γ2(yb) = f(y) − f(b).

On the projection Ex+y, where x + y = a + b + t and γ3(x + yab) = f(x + y) − f(a + b), we get from Lemma 3 and formula (9)

()
Therefore, each function Fi : DfS, i = 1,2, 3, satisfies (13), and Theorem 1 is proved.

Remark 4. The foregoing study was developed as though the projections Ex, Ey, and Ex+y were pairwise disjoint.

If two of them overlap, for instance ExEy is nonempty, in every common point the values given by the different parts of formulas of approximating function F have to be the same.

More in particular, if the set Df = ExEyEx+y is connected, in (28) the equations

()
hold, whence f(a) − Hi(a) = f(b) − Hi(b) = 0. In this case, the locally δ-additive f is uniformly approached on the whole Df by the restriction to Df of a function H : RS, additive on R2 (H = Hi, i either =1, or =2, or =3).

2.2. A Result on Local Stability of the Logarithmic Equation

On the ground of the results in Section 2.1 it is easy to prove the local stability of the logarithmic Cauchy equation (2) restricted to the bounded domain
()
for given (a, b) ∈ R2 and r > 0.
The projections Jx, Jy, Jxy of J are given by
()
Since the local stability of (2) depends on the comparison of every f satisfying
()
for (x, y) ∈ J(a, b; r) with some solution φ of the corresponding equation (2) restricted to J(a, b; r), let us premise (Lemma 5) the local solution of (2).

Lemma 5. Let S be a real linear space; if φ : Df = JxJyJxyS satisfies (2) on the bounded domain J = J(a, b; r) defined in (32), then there exists a function h0 : RS, additive on R2, such that

()

Proof of Lemma 5. By the usual substitutions

()
the domain J(a, b; r) is transformed into a set E(a, b; r) like the one defined in (6). Let us consider
()
Equation (2) is transformed into
()
Therefore, using Theorem 1, we obtain (35) with additive h0.

The local stability of the logarithmic equation (2) is stated by the following.

Theorem 6. Let (S, ∥·∥) be a Banach space; if f : Df = JxJyJxy → (S, ∥·∥) satisfies

()
for some δ > 0 and every (x, y) ∈ J(a, b; r), defined in (32), for given (a, b) ∈ R2 and r > 0, then there exists (at least) a function H :   RS, additive on R2, such that the function L :   DfS defined by
()
satisfies both the following properties:
  • (i)

    L is a local solution of the logarithmic equation on the restricted domain J(a, b; r);

  • (ii)

    L approaches uniformly fon Df, and

    ()
    holds for every tDf.

Proof of Theorem 6. The usual substitutions x = eu, y = ev (like in proof of Lemma 5) transform the inequality (39) restricted to the set J(a, b; r) into

()
restricted to the set E = E(a, b; r), defined in (6).

Now, we can follow the same line of proof as in Section 2.1 by defining the functions γi : [0, r) → S,  i = 1,2, 3, related to g; namely,

()
then there exist functions Hi : RS, i = 1,2, 3, additive on R2, such that each of the functions Gi, i = 1,2, 3, 
()
is a local solution of the equation g(u + v) = g(u) + g(v) restricted to E(a, b; r), and
()
holds for i = 1,2, 3.

Now let us come back to f, by the substitutions which transformed (39) into (42), beginning by the transformation of functions Gi defined in (44); on Jx (from Eu)

()
similarly, for Gi(ln y) on Jy (from Ev) and for Gi(ln xy) on Jxy (from Eu+v).

By the definition

()
formula (44) changes into
()
Obviously each Li(τ), i = 1,2, 3, satisfies the logarithmic equation restricted to J(a, b; r).

In order to prove the approximation stated in (41), let us begin by the projection Jx: for xJx; then u = ln x, g(u) = f(x), and u ∈ [a, a + r); hence, from (44)

()
and from (45)
()
Similarly for f on Jy and on Jxy.

Therefore, (41) is true with L = L1 or L = L2 or L = L3, and Theorem 6 is proved.

Remark 7. Remarks about the consequence of a possible overlapping of the projections of the given restricted domain, like those in Remark 4, could be repeated here.

2.3. About the Pexiderized Forms of the Foregoing Equations

Stability results for the Pexiderized forms of the additive and the logarithmic equations, namely,
()
can be easily stated on the ground of the foregoing Theorems 1 and 6.
In fact, when the inequality
()
is satisfied for every (x, y) ∈ E(a, b; r), the statement of Theorem 1 can be easily adapted to the condition (52), because the functions λ(x), μ(y) and ψ(x + y) play a role like that of the restrictions of f to the projections Ex, Ey, Ex+y in case of a unique function f.
Similarly for
()
restricted to J(a, b; r), by use of Theorem 6.

Remark 8. In case of a Pexiderized equation on restricted domain, overlapping of the projections of the given bounded domain obviously produces no changes in the result.

3. About the Remaining Two Cauchy Equations (7) φ(x + y) = φ(x)φ(y) and (8) φ(xy) = φ(x)φ(y) on a Bounded Restricted Domain

3.1. Preliminaries

As for (7) φ(x + y) = φ(x)φ(y) the restricted domain is assumed to be E(a, b; r) defined in (6); the domain of (8) φ(xy) = φ(x)φ(y) is J(a, b; r) defined in (32) for fixed real a, b and r > 0.

Let us premise the local solutions of the above equations (see papers [8, 9] and [10], resp.).

Lemma 9. Let φ : Dφ = ExEyEx+yR satisfy (7) restricted to E(a, b; r) defined in (6).

If and only if there exists some (x, y) ∈ E(a, b; r), (xEx, yEy) such that φ(x) ≠ 0 and φ(y) ≠ 0, the following properties (P1), (P2), (P3) hold:

  •  (P1)

    φ(t) ≠ 0 for every tEx+y (hence for every tDφ);

  •  (P2)

    sgn  φ(t) is constant on each projection Ex,   Ey,   Ex+y (not necessarily the same in different projections);

  •  (P3)

    φ(t) is given on Dφ by the following formulas:

    •  (i)

      if tEx, φ(t) = AeG(t)−G(a),

    •  (ii)

      if tEy, φ(t) = BeG(t)−G(b),

    •  (iii)

      if tEx+y, φ(t) = ABeG(t)−G(a+b),

    • where G : RR is additive on R2; A ≠ 0, B ≠ 0 are constant.

Remark 10. Notice that φ restricted to each of the projections Ex, Ey, Ex+y is the restriction of a solution Φ : RR of the equation

()
valid on the whole R2, for suitable K ≠ 0.

Since this equation can be written as

()
we get , for some additive G(t), whence formulas in (P3) of Lemma 9 for
()
respectively, in Ex, Ey, Ex+y.

Lemma 11 (see [10].)The general nowhere vanishing solution φ : Dφ = JxJyJxyR of (8) restricted to the set J(a, b; r) defined in (32) is given by the following formulas:

()
where h : RR is additive on R2.

Remark 12. As in Remark 10, we can see that the local solution φ of (8), restricted to each of the projections Jx, Jy, Jxy, is the restriction of a solution Ψ : R+R of a more general equation

()
for suitable K ≠ 0.

From KΨ(xy) = KΨ(x)KΨ(y), it follows that Ψ(t) = (1/K)eh(ln t), with

()

3.2. How the Question of Local Stability of (7) or (8) Has to Be Properly Formulated?

The foregoing Remarks 10 and 12, which point out a connection of the restricted equation under consideration with more general equations, namely,
()
suggest the following forms of perturbation of such equations:
for −1 < θ(x, y) < 1 and some fixed δ > 0.

Moreover, it is known (see [8, 9]) that the local solutions of the restricted equations (7) or (8), which vanish somewhere, are expressed by formulas containing arbitrary functions; therefore, the problem of the local stability seems to be significant in the set of nowhere vanishing functions f.

In this frame, the perturbed forms (7)δ and (8)δ can be written equivalently as

The stability results which follow are framed in this context.

3.3. A Sign Property concerning the Perturbed Forms of the Exponential Equation and the Power Equation

Here, we will be concerned with the condition (7)δ′, (x, y) ∈ E(a, b; r) defined in (6) for some fixed δ > 0, in the set of functions f : Df = ExEyEx+yR, such that f(t) ≠ 0 for every tDf.

Let us premise a remark about signs of nowhere vanishing functions f satisfying (7)δ′ on E(a, b; r). From Lemma 9, Property (P2), it is known that every nowhere vanishing solution of the exponential Cauchy equation restricted to E(a, b; r) keeps a constant sign in each of the projections Ex, Ey, Ex+y of E(a, b; r).

We will see that a similar property is true also for every solution of the restricted condition (7)δ′, which is rewritten here as follows:
From (7)δθ′, assuming (x, y) = (a, b + ξ) and (x, y) = (a + ξ, b) with 0 ≤ ξ < r, we get
()
()
whence
()
()

Moreover, from (7)δθ′ for (x, y) = (a + ξ, b + ξ), 0 ≤ ξ < r/2, θ = θ(a + ξ, b + ξ), f(a + b + 2ξ) = ; hence, f(t) has constant sign in Ex+y.

As a consequence, from (62), (64) it follows that f has constant signs also in Ex and in Ey (the signs of f(a) and of f(b), resp.).

This proves the following.

Lemma 13. Every nowhere vanishing function f : Df = ExEyEx+yR satisfying (7)δ in E(a, b; r) keeps constant sign in each of the projections Ex, Ey, Ex+y of E(a, b; r).

Similarly, we can prove a sign property concerning the perturbed form of the power equation.

Let us consider now the condition (8)δ′; namely,
()
for some fixed δ > 0, assuming f(t) ≠ 0 for every tDf = JxJyJxy.
The usual substitutions of variables x, y allow us to use the foregoing results about the exponential equation. Put
()
whence f(xy) = :g(u + v), a + bu + v < a + b + r.
Then (65) is transformed into
()
namely

Therefore, from Lemma 13, it follows that g(u) has constant sign (=sgn g(a)) in [a, a + r), whence f(x) has constant sign (=sgn f(ea)) in [ea, ea+r); similarly for g(v) in [b, b + r), namely for f(y) in [eb, eb+r) and for g(u + v) in [a + b, a + b + r), namely f(xy) in [ea+b, ea+b+r).

Hence, the following result is proved.

Lemma 14. Every nowhere vanishing function f : Df = JxJyJxyR satisfying (65) restricted to J(a, b; r) has constant signs in each of the projections Jx, Jy, Jxy of J(a, b; r).

3.4. A Result of Local Stability for the Exponential Cauchy Equation

In the set of functions f : Df = ExEyEx+yR such that f(t) ≠ 0 for every tDf, let us consider the inequality  (7)δ′, with (x, y) ∈ E(a, b; r) for some fixed δ > 0.

From (7)δ′
()
since
()
f satisfies
()
namely for λ(t): = ln |f(t)|,
()
On the ground of Theorem 1, there exists (at least) one additive function H :  RR, such that the function L :  DfS defined by
()
is a local solution of the additive Cauchy equation restricted to E(a, b; r), such that
()
Since |f(t)| = eλ(t), whence,
()
we get
()
substitution of L(t) by its explicit formulas gives the following:
()
and similarly
()
By defining
()
it is easily proved that
()
Moreover, from (78),
()
This means that the values of f(t) in Df are “near” (in dependence on δ) the values of a local solution F(t) of the corresponding equation restricted to the same domain E(a, b; r) and give the following theorem of local stability.

Theorem 15. If the function f : Df = ExEyEx+yR, is nowhere vanishing in its domain Df and satisfies  (7)δ′, for some given δ > 0 and every (x, y) ∈ E(a, b; r) defined in (6) for given (a, b) ∈ R2 and r > 0, then there exists (at least) an additive function H : RR such that the function F :   DfR

()
has both the properties:
  • (i)

    F is a nowhere vanishing local solution of the exponential Cauchy equation F(x + y) = F(x)F(y) restricted to E(a, b; r);

  • (ii)

    the values F(t) are near the values f(t) on Df; more exactly

    ()

3.5. A Result on Local Stability of the Power Cauchy Equation

In the set of nowhere vanishing functions f : Df = JxJyJxyR, let us consider the inequality (65), (x, y) ∈ J(a, b; r) ⊂ R2 defined in (32), for some given δ > 0.

The usual substitutions
()
transform the condition (65) into
Hence, thanks to Theorem 15 (referred to g(t) instead of f(t)), there exists (at least) one additive function H : RR such that the function G : DgR defined by
()
satisfies the exponential equation G(u + v) = G(u)G(v) restricted to E(a, b; r) and approaches g on Dg = EuEvEu+v as follows:
Formula (82)g can be rewritten as
for tEuEvEu+v.
From the definition of G,
  • for uEu = [a, a + r) then g(u) = f(x), G(u) = G(ln x) = f(ea) · eH(a) · eH(ln x),

  • similarly for vEv = [b, b + r) then g(v) = f(y), G(ln y) = f(eb) · eH(b) · eH(ln y),

  • and for u + vEu+v = [a + b, a + b + r): G(ln (xy)) = f(ea)f(eb) · eH(a)−H(b) · eH(ln xy).

Hence, by defining Φ : DfR as follows:
()
we get
()
namely,
()

This proves the following property of local stability of the “power” Cauchy equation.

Theorem 16. Let the nowhere vanishing function f : DfR+R satisfy the condition (65) for some given δ > 0 and every (x, y) ∈ J(a, b; r) defined in (32), for given (a, b) ∈ R2 and r > 0; Df = JxJyJxy; then there exists (at least) an additive function H : RR such that the function Φ :   DfR defined by

()
has both the following properties:
  • (i)

    Φ is a local solution of the Cauchy equation Φ(xy) = Φ(x)Φ(y) restricted to J(a, b; r);

  • (ii)

    the values of Φ(t) are near the values f(t) in Df; more exactly

    ()

3.6. Remark about the Pexiderized Forms of the Foregoing Equations

According to the remarks at the end of Section 2, the stability results given by Theorems 15 and 16 can be easily adapted to the Pexiderized forms of the corresponding equations, namely, to
()
for nowhere vanishing functions
()

    The full text of this article hosted at iucr.org is unavailable due to technical difficulties.