We investigate the stability of the functional equation 2f(x+y)+f(x-y)+f(y-x)-3f(x)-f(-x)-3f(y)-f(-y)=0 by using the fixed point theory in the sense of Cădariu and Radu.
1. Introduction
In 1940, Ulam [1] raised a question concerning the stability of homomorphisms as follow. Given a group G1, a metric group G2 with the metric d(·,·), and a positive number ε, does there exist a δ>0 such that if a mapping f:G1→G2 satisfies the inequality d(f(xy),f(x)f(y))<δ,
for all x,y∈G1 then there exists a homomorphism F:G1→G2 with d(f(x),F(x))<ε,
for all x∈G1? When this problem has a solution, we say that the homomorphisms from G1 to G2 are stable. In the next year, Hyers [2] gave a partial solution of Ulam's problem for the case of approximate additive mappings under the assumption that G1 and G2 are Banach spaces. Hyers' result was generalized by Aoki [3] for additive mappings and by Rassias [4] for linear mappings by considering the stability problem with unbounded Cauchy’s differences. The paper of Rassias had much influence in the development of stability problems. The terminology Hyers-Ulam-Rassias stability originated from this historical background. During the last decades, the stability problems of functional equations have been extensively investigated by a number of mathematicians, see [5–12].
Almost all subsequent proofs, in this very active area, have used Hyers' method of [2]. Namely, the mapping F, which is the solution of a functional equation, is explicitly constructed, starting from the given mapping f, by the formulae F(x)=limn→∞(1/2n)f(2nx) or F(x)=limn→∞2nf(x/2n). We call it a direct method. In 2003, Cădariu and Radu [13] observed that the existence of the solution F for a functional equation and the estimation of the difference with the given mapping f can be obtained from the fixed point theory alternative. This method is called a fixed point method. In 2004, they applied this method [14] to prove stability theorems of the Cauchy functional equation: f(x+y)-f(x)-f(y)=0.
In 2003, they [15] obtained the stability of the quadratic functional equation: f(x+y)+f(x-y)-2f(x)-2f(y)=0,
by using the fixed point method. Notice that if we consider the functions f1,f2:ℝ→ℝ defined by f1(x)=ax and f2(x)=ax2, where a is a real constant, then f1 satisfies (1.3), and f2 holds (1.4), respectively. We call a solution of (1.3) an additive map, and a mapping satisfying (1.4) is called a quadratic map. Now we consider the functional equation: 2f(x+y)+f(x-y)+f(y-x)-3f(x)-f(-x)-3f(y)-f(-y)=0,
which is called the Cauchy additive and quadratic-type functional equation. The function f:ℝ→ℝ defined by f(x)=ax2+bx satisfies this functional equation, where a,b are real constants. We call a solution of (1.5) a quadratic-additive mapping.
In this paper, we will prove the stability of the functional equation (1.5) by using the fixed point theory. Precisely, we introduce a strictly contractive mapping with the Lipschitz constant 0<L<1. Using the fixed point theory in the sense of Cădariu and Radu, together with suitable conditions, we can show that the contractive mapping has the fixed point. Actually the fixed point F becomes the precise solution of (1.5). In Section 2, we prove several stability results of the functional equation (1.5) using the fixed point theory, see Theorems 2.3, 2.4, and 2.5. In Section 3, we use the results in the previous sections to get a stability of the Cauchy functional equation (1.3) and that of the quadratic functional equation (1.4), respectively.
2. Main Results
We recall the following result of the fixed point theory by Margolis and Diaz.
Theorem 2.1 (see [16] or [17]).
Suppose that a complete generalized metric space (X,d), which means that the metric d may assume infinite values, and a strictly contractive mapping J:X→X with the Lipschitz constant 0<L<1 are given. Then, for each given element x∈X, either
d(Jnx,Jn+1x)=+∞,∀n∈N∪{0}
or there exists a nonnegative integer k such that
d(Jnx,Jn+1x)<+∞ for all n≥k,
the sequence {Jnx} is convergent to a fixed point y* of J,
y* is the unique fixed point of J in Y∶={y∈X,d(Jkx,y)<+∞},
d(y,y*)≤(1/(1-L))d(y,Jy) for all y∈Y.
Throughout this paper, let V be a (real or complex) linear space, and let Y be a Banach space. For a given mapping f:V→Y, we use the following abbreviation: Df(x,y)∶=2f(x+y)+f(x-y)+f(y-x)-3f(x)-f(-x)-3f(y)-f(-y),
for all x,y∈V. If f is a solution of the functional equation Df≡0, see (1.5), we call it a quadratic-additive mapping. We first prove the following lemma.
Lemma 2.2.
If f:V→Y is a mapping such that Df(x,y)=0 for all x,y∈V∖{0}, then f is a quadratic-additive mapping.
Proof.
By choosing x∈V∖{0}, we get
8f(0)=Df(3x,2x)-Df(4x,x)-2Df(3x,x)-Df(2x,x)+Df(-3x,-2x)-Df(-4x,-x)-2Df(-3x,-x)-Df(-2x,-x)+4Df(x,-x)=0.
Since f(0)=0, we easily obtain Df(x,y)=0 for all x,y∈V.
Now we can prove some stability results of the functional equation (1.5).
Theorem 2.3.
Let φ:(V∖{0})2→[0,∞) be a given function. Suppose that the mapping f:V→Y satisfies
‖Df(x,y)‖≤φ(x,y),
for all x,y∈V∖{0}. If there exist constants 0<L,L′<1 such that φ has the property
1L′φ(x,y)≤φ(2x,2y)≤2Lφ(x,y),
for all x,y∈V∖{0}, then there exists a unique quadratic-additive mapping F:V→Y such that
‖f(x)-F(x)‖≤3(φ(x,x)+φ(-x,-x))16(1-L),
for all x∈V∖{0}. In particular, F is represented by
F(x)=limn→∞(f(2nx)+f(-2nx)2⋅4n+f(2nx)-f(-2nx)2n+1),
for all x∈V.
Proof.
It follows from (2.5) that
limn→∞φ(2nx,2ny)2n=0,
for all x,y∈V∖{0}, and
‖8f(0)‖=limn→∞‖Df(3x2n,2x2n)-Df(4x2n,x2n)-2Df(3x2n,x2n)-Df(2x2n,x2n)+Df(-3x2n,-2x2n)-Df(-4x2n,-x2n)-2Df(-3x2n,-x2n)-Df(-2x2n,-x2n)+4Df(x2n,-x2n)‖≤limn→∞(φ(3x2n,2x2n)+φ(4x2n,x2n)+2φ(3x2n,x2n)+φ(2x2n,x2n)+φ(-3x2n,-2x2n)+φ(-4x2n,-x2n)+2φ(-3x2n,-x2n)+φ(-2x2n,-x2n)+4φ(x2n,-x2n))≤limn→∞L′n(φ(3x,2x)+φ(4x,x)+2φ(3x,x)+φ(2x,x)+φ(-3x,-2x)+φ(-4x,-x)+2φ(-3x,-x)+φ(-2x,-x)+4φ(x,-x))=0,
for all x∈V∖{0}. From this, we know that f(0)=0. Let S be the set of all mappings g:V→Y with g(0)=0. We introduce a generalized metric on S by
d(g,h)=inf{K∈R+∣‖g(x)-h(x)‖≤K(φ(x,x)+φ(-x,-x))∀x∈V∖{0}}.
It is easy to show that (S,d) is a generalized complete metric space. Now we consider the mapping J:S→S, which is defined by
Jg(x)∶=g(2x)-g(-2x)4+g(2x)+g(-2x)8,
for all x∈V. Notice
Jng(x)=g(2nx)-g(-2nx)2n+1+g(2nx)+g(-2nx)2⋅4n,
for all n∈ℕ and x∈V. Let g,h∈S, and let K∈[0,∞] be an arbitrary constant with d(g,h)≤K. From the definition of d, we have
‖Jg(x)-Jh(x)‖=38‖g(2x)-h(2x)‖+18‖g(-2x)-h(-2x)‖≤12K(φ(2x,2x)+φ(-2x,-2x))≤KL(φ(x,x)+φ(-x,-x)),
for all x∈V∖{0}, which implies that
d(Jg,Jh)≤Ld(g,h),
for any g,h∈S. That is, J is a strictly contractive self-mapping of S with the Lipschitz constant L. Moreover, by (2.4), we see that
‖f(x)-Jf(x)‖=116‖-3Df(x,x)+Df(-x,-x)‖≤316(φ(x,x)+φ(-x,-x)),
for all x∈V∖{0}. It means that d(f,Jf)≤3/16<∞ by the definition of d. Therefore, according to Theorem 2.1, the sequence {Jnf} converges to the unique fixed point F:V→Y of J in the set T={g∈S∣d(f,g)<∞}, which is represented by (2.7) for all x∈V. By the definition of F, together with (2.4) and (2.7), it follows that
‖DF(x,y)‖=limn→∞‖Df(2nx,2ny)-Df(-2nx,-2ny)2n+1+Df(2nx,2ny)+Df(-2nx,-2ny)2⋅4n‖≤limn→∞2n+12⋅4n(φ(2nx,2ny)+φ(-2nx,-2ny))=0,
for all x,y∈V∖{0}. By Lemma 2.2, we have proved that
DF(x,y)=0,
for all x,y∈V.
Theorem 2.4.
Let φ:(V∖{0})2→[0,∞) be a given function. Suppose that the mapping f:V→Y satisfies (2.5) for all x,y∈V∖{0}. If there exists a constant 0<L<1/2 such that φ has the property
φ(2x,2y)≤2Lφ(x,y)
for all x,y∈V∖{0}, then there exists a unique quadratic-additive mapping F:V→Y satisfying (2.6) for all x∈V∖{0}. Moreover, if φ(x,y) is continuous, then f itself is a quadratic-additive mapping.
Proof.
It follows from (2.18) that
limn→∞φ(2nx,2ny)=limn→∞(2L)nφ(x,y)=0,
for all x,y∈V∖{0}, and
‖8f(0)‖=limn→∞‖Df(3⋅2nx,2n+1x)-Df(2n+2x,2nx)-2Df(3⋅2nx,2nx)-Df(2n+1x,2nx)+Df(-3⋅2nx,-2n+1x)-Df(-2n+2x,-2nx)-2Df(-3⋅2nx,-2nx)-Df(-2n+1x,-2nx)+4Df(2nx,-2nx)‖≤limn→∞(φ(3⋅2nx,2n+1x)+φ(2n+2x,2nx)+2φ(3⋅2nx,2nx)+φ(2n+1x,2nx)+φ(-3⋅2nx,-2n+1x)+φ(-2n+2x,-2nx)+2φ(-3⋅2nx,-2nx)+φ(-2n+1x,-2nx)+4φ(2nx,-2nx))≤limn→∞(2L)n(φ(3x,2x)+φ(4x,x)+2φ(3x,x)+φ(2x,x)+φ(-3x,-2x)+φ(-4x,-x)+2φ(-3x,-x)+φ(-2x,-x)+4φ(x,-x))=0,
for all x∈V∖{0}. By the same method used in Theorem 2.3, we know that there exists a unique quadratic-additive mapping F:V→Y satisfying (2.6) for all x∈V∖{0}. Since φ is continuous, we get
limn→∞φ((a1⋅2n+a2)x,(b1⋅2n+b2)y)≤limn→∞(2L)nφ((a1+a22n)x,(b1+b22n)y)=0⋅φ(a1x,b1y)=0,
for all x,y∈V∖{0} and for any fixed integers a1,a2,b1,b2 with a1,b1≠0. Therefore, we obtain
2‖f(x)-F(x)‖≤limn→∞(‖(F-f)(-(2n+1+1)x)‖‖Df((2n+1)x,-2nx)-DF((2n+1)x,-2nx)‖+‖(F-f)((2n+1+1)x)‖+‖(F-f)(-(2n+1+1)x)‖+3‖(f-F)((2n+1)x)‖+‖(f-F)(-(2n+1)x)‖+3‖(f-F)(-2nx)‖+‖(f-F)(2nx)‖‖(F-f)(-(2n+1+1)x)‖)≤limn→∞φ(‖(F-f)(-(2n+1+1)x)‖(2n+1)x,-2nx)+38(1-L)limn→∞(ψ((2n+1+1)x)+2ψ((2n+1)x)+2ψ(2nx)ψ((2n+1+1)x))=0,
for all x∈V∖{0}, where ψ(x) is defined by ψ(x)=φ(x,x)+φ(-x,-x). Since f(0)=0=F(0), we have shown that f≡F. This completes the proof of this theorem.
We continue our investigation with the next result.
Theorem 2.5.
Let φ:(V∖{0})2→[0,∞). Suppose that f:V→Y satisfies the inequality ∥Df(x,y)∥≤φ(x,y) for all x,y∈V∖{0}. If there exists 0<L<1 such that the mapping φ has the property
Lφ(2x,2y)≥4φ(x,y)
for all x,y∈V∖{0}, then there exists a unique quadratic-additive mapping F:V→Y such that
‖f(x)-F(x)‖≤L8(1-L)(φ(x,x)+φ(-x,-x)),
for all x∈V∖{0}. In particular, F is represented by
F(x)=limn→∞(2n-1(f(x2n)-f(-x2n))+4n2(f(x2n)+f(-x2n))),
for all x∈V.
Proof.
By the similar method used to prove f(0)=0 in the proof of Theorem 2.3, we can easily show that f(0)=0. Let the set (S,d) be as in the proof of Theorem 2.3. Now we consider the mapping J:S→S defined by
Jg(x)∶=g(x2)-g(-x2)+2(g(x2)+g(-x2)),
for all g∈S and x∈V. Notice that
Jng(x)=2n-1(g(x2n)-g(-x2n))+4n2(g(x2n)+g(-x2n))
and J0g(x)=g(x), for all x∈V. Let g,h∈S, and let K∈[0,∞] be an arbitrary constant with d(g,h)≤K. From the definition of d, we have
‖Jg(x)-Jh(x)‖=3‖g(x2)-h(x2)‖+‖g(-x2)-h(-x2)‖≤4K(φ(x2,x2)+φ(-x2,-x2))≤LK(φ(x,x)+φ(-x,-x)),
for all x∈V∖{0}. So
d(Jg,Jh)≤Ld(g,h),
for any g,h∈S. That is, J is a strictly contractive self-mapping of S with the Lipschitz constant L. Also we see that
‖f(x)-Jf(x)‖=12‖Df(x2,x2)‖≤12(φ(x2,x2)+φ(-x2,-x2))≤L8(φ(x,x)+φ(-x,-x)),
for all x∈V∖{0}, which implies that d(f,Jf)≤L/8<∞. Therefore, according to Theorem 2.1, the sequence {Jnf} converges to the unique fixed point F of J in the set T∶={g∈S∣d(f,g)<∞}, which is represented by (2.25). Since
d(f,F)≤11-Ld(f,Jf)≤L8(1-L),
the inequality (2.24) holds. From the definition of F(x), (2.4), and (2.23), we have
‖DF(x,y)‖=limn→∞‖(Df(x2n,y2n)+Df(-x2n,-y2n))+4n2(Df(x2n,y2n)+Df(-x2n,-y2n))2n-1(Df(x2n,y2n)-Df(-x2n,-y2n))+4n2(Df(x2n,y2n)+Df(-x2n,-y2n))‖≤limn→∞2n+4n2(φ(x2n,y2n)+φ(-x2n,-y2n))=0,
for all x,y∈V∖{0}. By Lemma 2.2, F is quadratic additive.
Remark 2.6.
If φ satisfies the equality φ(x,y)=φ(-x,-y) for all x,y∈V∖{0} in Theorems 2.3, 2.4, and 2.5, then the inequalities (2.6) and (2.24) can be replaced by
‖f(x)-F(x)‖≤φ(x,x)4(1-L),‖f(x)-F(x)‖≤Lφ(x,x)8(1-L),
for all x∈V∖{0}, respectively.
3. Applications
For a given mapping f:V→Y, we use the following abbreviations: Af(x,y)∶=f(x+y)-f(x)-f(y),Qf(x,y)∶=f(x+y)+f(x-y)-2f(x)-2f(y),
for all x,y∈V. Using Theorems 2.3, 2.4, and 2.5 we will show the stability results of the additive functional equation Af≡0 and the quadratic functional equation Qf≡0 in the following corollaries.
Corollary 3.1.
Let fi:V→Y,i=1,2,3, be mappings for which there exist functions ϕi:(V∖{0})2→[0,∞),i=1,2,3, such that
‖Afi(x,y)‖≤ϕi(x,y),
for all x,y∈V∖{0}. If there exists 0<L<1 such that
1Lϕ1(x,y)≤ϕ1(2x,2y)≤2Lϕ1(x,y),ϕ2(2x,2y)≤Lϕ2(x,y),4ϕ3(x,y)≤Lϕ3(2x,2y)
for all x,y∈V∖{0}, then there exist unique additive mappings Fi:V→Y,i=1,2,3, such that
‖f1(x)-F1(x)‖≤3(ϕ1(x,x)+ϕ1(x,-x)+ϕ1(-x,x)+ϕ1(-x,-x))8(1-L),‖f2(x)-F2(x)‖≤3(ϕ2(x,x)+ϕ2(x,-x)+ϕ2(-x,x)+ϕ2(-x,-x))8(1-L/2),‖f3(x)-F3(x)‖≤L(ϕ3(x,x)+ϕ3(x,-x)+ϕ3(-x,x)+ϕ3(-x,-x))4(1-L),
for all x∈V∖{0}. In particular, the mappings Fi,i=1,2,3, are represented by
F1(x)=limn→∞f1(2nx)2n,F2(x)=limn→∞f2(2nx)2n,F3(x)=limn→∞2nf3(x2n),
for all x∈V. Moreover, if ϕ2(x,y) is continuous, then f2 is itself an additive mapping.
Proof.
Notice that
Dfi(x,y)=2Afi(x,y)+Afi(x,-y)+Afi(y,-x),
for all x,y∈V and i=1,2,3. Put
φi(x,y)∶=2ϕi(x,y)+ϕi(x,-y)+ϕi(y,-x),
for all x,y∈V and i=1,2,3, then φ1 satisfies (2.5), φ2 satisfies (2.18), and φ3 satisfies (2.23). Therefore, ∥Dfi(x,y)∥≤φi(x,y), for all x,y∈V∖{0} and i=1,2,3. According to Theorem 2.3, there exists a unique mapping F1:V→Y satisfying (3.6), which is represented by (2.7). Observe that, by (3.2) and (3.3),
limn→∞‖f1(2nx)+f1(-2nx)2n+1‖=limn→∞‖f1(2nx)+f1(-2nx)-f1(0)2n+1‖=limn→∞12n+1‖Af1(2nx,-2nx)‖≤limn→∞12n+1ϕ1(2nx,-2nx)≤limn→∞Ln2ϕ1(x,-x)=0,
as well as
limn→∞‖f1(2nx)+f1(-2nx)2⋅4n‖≤limn→∞2nLn2⋅4nϕ1(x,-x)=0,
for all x∈V∖{0}. From this and (2.7), we get (3.9). Moreover, we have
‖Af1(2nx,2ny)2n‖≤ϕ1(2nx,2ny)2n≤Lnϕ1(x,y),
for all x,y∈V∖{0}. Taking the limit as n→∞ in the above inequality and using F1(0)=0, we get
AF1(x,y)=0,
for all x,y∈V. According to Theorem 2.4, there exists a unique mapping F2:V→Y satisfying (3.7), which is represented by (2.7). By using the similar method to prove (3.9), we can show that F2 is represented by (3.10). In particular, if ϕ2(x,y) is continuous, then φ2 is continuous on (V∖{0})2, and we can say that f2 is an additive map by Theorem 2.4. On the other hand, according to Theorem 2.5, there exists a unique mapping F3:V→Y satisfying (3.8) which is represented by (2.25). Observe that, by (3.2) and (3.5),
limn→∞22n-1‖f3(x2n)+f3(-x2n)‖=limn→∞22n-1‖Af3(x2n,-x2n)‖≤limn→∞22n-1ϕ3(x2n,-x2n)≤limn→∞Ln2ϕ3(x,-x)=0,
as well as
limn→∞2n-1‖f3(x2n)+f3(-x2n)‖≤limn→∞Ln2n+1ϕ3(x,-x)=0,
for all x∈V∖{0}. From these and (2.25), we get (3.11). Moreover, we have
‖2nAf3(x2n,y2n)‖≤2nϕ3(x2n,y2n)≤Ln2nϕ3(x,y),
for all x,y∈V∖{0}. Taking the limit as n→∞ in the above inequality and using F3(0)=0, we get
AF3(x,y)=0,
for all x,y∈V.
Corollary 3.2.
Let fi:V→Y,i=1,2,3, be mappings for which there exist functions ϕi:(V∖{0})2→[0,∞),i=1,2,3, such that
‖Qfi(x,y)‖≤ϕi(x,y),
for all x,y∈V∖{0}. If there exists 0<L<1 such that the mapping ϕ1 satisfies (3.3), ϕ2 satisfies (3.4), and ϕ3 satisfies (3.5) for all x,y∈V∖{0}, then there exist unique quadratic mappings Fi:V→Y,i=1,2,3, such that
‖f1(x)-F1(x)‖≤3(ϕ1(x,x)+ϕ1(x,-x)+ϕ1(-x,x)+ϕ1(-x,-x))16(1-L),‖f2(x)-F2(x)‖≤3(ϕ2(x,x)+ϕ2(x,-x)+ϕ2(-x,x)+ϕ2(-x,-x))16(1-L/2),‖f3(x)-F3(x)‖≤L(ϕ3(x,x)+ϕ3(x,-x)+ϕ3(-x,x)+ϕ3(-x,-x))8(1-L),
for all x∈V∖{0}. In particular, the mappings Fi,i=1,2,3, are represented by
F1(x)=limn→∞f1(2nx)4n,F2(x)=limn→∞f2(2nx)4n,F3(x)=limn→∞4nf3(x2n),
for all x∈V. Moreover, if ϕ2(x,y) is continuous, then f2 itself is a quadratic mapping.
Proof.
Notice that
Dfi(x,y)=Qfi(x,y)+12(Qfi(x,-y)+Qfi(y,-x)),
for all x,y∈V and i=1,2,3. Put φi(x,y)∶=ϕi(x,y)+(1/2)(ϕi(x,-y)+ϕi(y,-x)), for all x,y∈V and i=1,2,3, then φ1 satisfies (2.5), φ2 satisfies (2.18), and φ3 satisfies (2.23). Moreover,
‖Dfi(x,y)‖≤φi(x,y),
for all x,y∈V∖{0} and i=1,2,3. According to Theorem 2.3, there exists a unique mapping F1:V→Y satisfying (3.23) which is represented by (2.7). Observe that
limn→∞‖f1(2nx)-f1(-2nx)2n+1‖=limn→∞12n+1‖Qf1(2n-1x,-2n-1x)-Qf1(-2n-1x,2n-1x)‖≤limn→∞12n+1(ϕ1(2n-1x,-2n-1x)+ϕ1(-2n-1x,2n-1x))≤limn→∞Ln2(ϕ1(x2,-x2)+ϕ1(-x2,x2))=0,
as well as
limn→∞‖f1(2nx)-f1(-2nx)2⋅4n‖≤limn→∞Ln2n+1(ϕ1(x2,-x2)+ϕ1(-x2,x2))=0,
for all x∈V∖{0}. From this and (2.7), we get (3.26) for all x∈V. Moreover, we have
‖Qf1(2nx,2ny)4n‖≤ϕ1(2nx,2ny)4n≤Ln2nϕ1(x,y),
for all x,y∈V∖{0}. Taking the limit as n→∞ in the above inequality, we get
QF1(x,y)=0
for all x,y∈V∖{0}. Using F1(0)=0, we have
QF1(x,0)=0,QF1(0,y)=-F1(y2,-y2)+F1(-y2,y2)=0,
for all x,y∈V∖{0}. Therefore, QF1(x,y)=0 for all x,y∈V.
Next, by Theorem 2.4, there exists a unique mapping F2:V→Y satisfying (3.24), which is represented by (2.7). By using the similar method to prove (3.26), we can show that F2 is represented by (3.27). In particular, ϕ2(x,y) is continuous, then φ2 is continuous on (V∖{0})2, and we can say that f2 is a quadratic map by Theorem 2.4. On the other hand, according to Theorem 2.5, there exists a unique mapping F3:V→Y satisfying (3.25) which is represented by (2.25). Observe that
4n‖f3(x2n)-f3(-x2n)‖=4n‖Qf3(x2n+1,-x2n+1)-Qf3(-x2n+1,x2n+1)‖≤4n(ϕ3(x2n+1,-x2n+1)+ϕ3(-x2n+1,x2n+1))≤Ln(ϕ3(x2,-x2)+ϕ3(-x2,x2)),
for all x∈V∖{0}. It leads us to get
limn→∞4n(f3(x2n)-f3(-x2n))=0,limn→∞2n(f3(x2n)-f3(-x2n))=0,
for all x∈V∖{0}. From these and (2.25), we obtain (3.28). Moreover, we have
‖4nQf3(x2n,y2n)‖≤4nϕ3(x2n,y2n)≤Lnϕ3(x,y),
for all x,y∈V∖{0}. Taking the limit as n→∞ in the above inequality and using F3(0)=0, we get
QF3(x,y)=0,
for all x,y∈V.
Now, we obtain Hyers-Ulam-Rassias stability results in the framework of normed spaces using Theorems 2.3 and 2.4.
Corollary 3.3.
Let X be a normed space, and let Y be a Banach space. Suppose that the mapping f:X→Y satisfies the inequality
‖Df(x,y)‖≤θ(‖x‖p+‖y‖p)
for all x,y∈X∖{0}, where θ≥0 and p∈(-∞,0)∪(0,1)∪(2,∞). Then there exists a unique quadratic-additive mapping F:X→Y such that
‖f(x)-F(x)‖≤{θ2p-4‖x‖pifp>2,θ2-2p‖x‖pif0<p<1
for all x∈X∖{0}. Moreover if p<0, then f is itself a quadratic-additive mapping.
Proof.
This corollary follows from Theorems 2.3, 2.4, and 2.5, and Remark 2.6, by putting
φ(x,y)∶=θ(‖x‖p+‖y‖p),
for all x,y∈X∖{0} with L=2p-1<1 if p<1, L=22-p<1 if p>2, and L′=2-p<1 if p>0.
Corollary 3.4.
Let X be a normal space let and Y be a Banach space. Suppose that the mapping f:X→Y satisfies the inequality
‖Df(x,y)‖≤θ‖x‖p‖y‖q,
for all x,y∈X∖{0}, where θ≥0 and p+q∈(-∞,0)∪(0,1)∪(2,∞). Then there exists a unique quadratic-additive mapping F:X→Y such that
‖f(x)-F(x)‖≤{θ‖x‖p+q2(2p+q-4)ifp+q>2,θ‖x‖p+q2(2-2p+q)if0<p+q<1,
for all x∈X∖{0}. Moreover, if p+q<0, then f is itself a quadratic-additive mapping.
Proof.
This corollary follows from Theorems 2.3, 2.4, 2.5, and Remark 2.6, by putting
φ(x,y)∶=θ‖x‖p‖y‖q,
for all x,y∈X∖{0} with L=2p+q-1<1 if p+q<1, L=22-p-q<1 if p+q>2, and L′=2-p-q<1 if p+q>0.
UlamS. M.1968New York, NY, USAInterscience PublishersHyersD. H.On the stability of the linear functional equation1941272222240004076 (2,315a)10.1073/pnas.27.4.222ZBL0061.26403AokiT.On the stability of the linear transformation in Banach spaces1950264660040580 (12,717a)10.2969/jmsj/00210064ZBL0040.35501RassiasT. M.On the stability of the linear mapping in Banach spaces197872229730010.2307/2042795507327 (80d:47094)ZBL0398.47040GăvruţaP.A generalization of the Hyers-Ulam-Rassias stability of approximately additive mappings1994184343143610.1006/jmaa.1994.12111281518 (95e:47089)ZBL0818.46043JunK.-W.LeeY.-H.A generalization of the Hyers-Ulam-Rassias stability of the Pexiderized quadratic equations. II2007471911032309803 (2008a:39043)ZBL1143.39014KimG. H.On the stability of functional equations with square-symmetric operation2001422572661823746 (2003d:39044)ZBL0990.39028KimH.-M.On the stability problem for a mixed type of quartic and quadratic functional equation2006324135837210.1016/j.jmaa.2005.11.0532262476 (2007f:39052)ZBL1106.39027LeeY.-H.On the stability of the monomial functional equation200845239740310.4134/BKMS.2008.45.2.3972419086ZBL1152.39023LeeY.-H.JunK.-W.A generalization of the Hyers-Ulam-Rassias stability of Jensen's equation1999238130531510.1006/jmaa.1999.65461711432 (2000m:39062)ZBL0933.39053LeeY.-H.JunK.-W.A generalization of the Hyers-Ulam-Rassias stability of the Pexider equation2000246262763810.1006/jmaa.2000.68321761952 (2001k:39044)ZBL0957.39008LeeY.-H.JunK.-W.On the stability of approximately additive mappings200012851361136910.1090/S0002-9939-99-05156-41641128 (2000j:39023)ZBL0961.47039CădariuL.RaduV.Fixed points and the stability of Jensen's functional equation200341, article 4ZBL1043.39010CădariuL.RaduV.On the stability of the Cauchy functional equation: a fixed point approach2004346Graz, AustriaKarl-Franzens-Universitaet Graz4352Grazer Mathematische BerichteCădariuL.RaduV.Fixed points and the stability of quadratic functional equations200341125482245911 (2007d:39038)MargolisB.DiazJ. B.A fixed point theorem of the alternative, for contractions on a generalized complete metric space1968743053090220267 (36#3333)10.1090/S0002-9904-1968-11933-0ZBL0157.29904RusI. A.Principles and applications of fixed point theory1979