Abstract
In this paper we deal with the product of two or three Cauchy differences equaled to zero. We show that in the case of two Cauchy differences, the condition of absolute continuity and differentiability of the two functions involved implies that one of them must be linear, i.e., we have a trivial solution. In the case of the product of three Cauchy differences the situation changes drastically: there exists non trivial \({\mathcal {C}}^{\infty }\) solutions, while in the case of real analytic functions we obtain that at least one of the functions involved must be linear. Some open problems are then presented.
Similar content being viewed by others
1 Introduction
We start by considering the following relation
where \(f:{\mathbb {R}} \rightarrow {\mathbb {R}}\) and we assume that f is absolutely continuous and has derivative at each point.
Following [2], the first step consists in giving a suitable representation of both f and \({\mathfrak {F}}\). Differentiating (1.1) with respect to x we obtain
and, setting \(x=0\),
Setting \(f'(0)=c\), \(c+\frac{\partial {\mathfrak {F}}(x,y)}{\partial x}|_{x=0}=:\phi (y)\) and substituting in (1.2) we obtain
and the following representation of \({\mathfrak {F}}\):
Thus, we conclude that if f and \({\mathfrak {F}}\) satisfy (1.1), then there exists a continuous function \(\phi \) such that (1.5) holds and
This simple result will be useful for solving the problem presented in the next section.
2 Alternative equation in two unknown functions
In 1978 Marek Kuczma [1] proposed to investigate the following Cauchy alternative functional equation in two unknown functions:
obviously there are the so–called trivial solutions, that is when one of the two functions is additive in the domain of our problem. Here we intend to deal with equation (2.1) where \(f,g:{\mathbb {R}} \rightarrow {\mathbb {R}}\) and \(f,g \in {\mathcal {C}}^1\). Equation (2.1) has been extensively studied by L. Paganoni and the present author [3,4,5] in a more general setting and under weaker regularity conditions and in [4] (Theorem 4)it has been proved that if the pair (f, g) of continuous functions is a solution of (2.1) then one of them is additive, in other words there are no non–trivial continuous solutions of (2.1). This result was obtained through a long procedure and here we will show that the requirement that the two functions are absolutely continuous and differentiable at each point makes the solution of (2.1) rather simple.
From (1.6) in the previous section we have the following:
where \(\phi , \gamma :{\mathbb {R}} \rightarrow {\mathbb {R}}\) are two continuous functions, \(\phi (0)=\gamma (0)=0\) and \(k_1, k_2\) arbitrary constants; moreover, we have \(f(0)=-k_1\) and \(g(0)=-k_2\) and, setting in (2.1) \(x=y=0\) we have \(f(0)g(0)=k_1k_2=0\).
Equation (2.1) can be written as
and after differentiating with respect to x we have
and, with \(x=0\), \(k_2\phi (y)+k_1\gamma (y)=0\). Since we have \(k_1k_2=0\), if \(k_2 \ne 0\), then \(k_1=0\) and \(\phi (y)=0\) for all \(y \in {\mathbb {R}}\); and by (2) this implies \(f(x+y)-f(x)-f(y)=0\), i.e., \(f(x)=ax\), \(x \in {\mathbb {R}}\), for some constant a. Obviously, if \(k_1 \ne 0\) and \(k_2=0\), we have \(g(x)=bx\).
Now, we assume \(k_1=k_2=0\). Denoting \(\Phi _y(t):=\phi (t+y)-\phi (t)\) and \(\Gamma _y(t):=\gamma (t+y)-\gamma (t)\), equation (2.2) becomes
Theorem 2.1
If (2.4) holds for every \(x,y \in {\mathbb {R}}\), then \(\Phi _y(x)\cdot \Gamma _y(x)=0\) for all \(x,y \in {\mathbb {R}}\).
Proof
Let us suppose that the statement is false, then there exist \(x_0, y_0 \in {\mathbb {R}}\) such that \(\Phi _{y_0}(x_0)\ne 0\) and \(\Gamma _{y_0}(x_0)\ne 0\). The continuity of \(\Phi _{y_0}(x)\) and \(\Gamma _{y_0}(x)\) in \(x_0\) implies the existence of a positive \(\delta \) such that \(\Phi _{y_0}(x)\ne 0\) and \(\Gamma _{y_0}(x)\ne 0\) for \(|x-x_0|<\delta \). But if
then
and, by (2.4), this implies
and from this we have that \(\Gamma _{y_0}(x)=0\) for \(|x-x_0|<\delta \): a contradiction. \(\square \)
So, from Theorem 2.1 we have
and, setting \(x=0\), \(\phi (y)\cdot \gamma (y)=0\), \(y \in {\mathbb {R}}\). Thus, developing (2.8) and considering the last equality, we obtain
If \(\phi (x_0)\ne 0\) for some \(x_0\), then \(\gamma (x_0)=0\) and from (2.9) we have \(\gamma (x_0+y)=0\) for all \(y \in {\mathbb {R}}\), hence \(\gamma \equiv 0\) and \(g(x)=bx\).
Summarizing we have the following
Theorem 2.2
A pair (f, g) of continuously differentiable functions is a solution of (2.1) if and only if at least one of the functions is linear.
Thus, we can say that equation (2.1) has only trivial \({\mathcal {C}}^1\) solutions, i.e., at least one of the factors must be identically zero.
3 Alternative equation in three unknown functions
Consider now the following alternative Cauchy equation in three unknown functions
In this case we have a completely different situation, i.e., there exist \({\mathcal {C}}^{\infty }\) functions solving (3.1) and neither of them is additive on the whole \({\mathbb {R}}\). An example is constructed as follows. Define the following three functions:
then we construct these other functions:
The three functions f, g and h are \({\mathcal {C}}^{\infty }\) on \({\mathbb {R}}\) and a simple (and tedious) check shows that the triple (f, g, h) is a solution of (3.1).
Things change completely if we increase the regularity of the functions. We assume that the triple (f, g, h) is a solution of equation (3.1), with all functions in \({\mathcal {C}}({\mathbb {R}})\) and at least one of them, say f, real analytic (i.e., it is a real function which is the restriction of an analytic function to the real line ). We intend to prove that if neither g nor h is additive, then f is additive, in other words, equation (3.1) has only trivial solutions.
Theorem 3.1
Assume that the triple (f, g, h) is a solution of equation (3.1) with g and h continuous and not linear, and f real analytic, than f is linear, i.e., equation (3.1) has only trivial solutions.
Proof
Since neither g nor h is additive, Theorem 2.2 in Sect. 2 implies that \({\mathfrak {G}}(x_0,y_0)\cdot {\mathfrak {H}}(x_0,y_0)\ne 0\) for some pair \((x_0,y_0)\in {\mathbb {R}}^2\) and, thanks to the continuity of g and h, there exists a square \(Q=(x_0-\delta ,x_0+\delta )\times (y_0-\delta ,y_0+\delta )\) such that \({\mathfrak {G}}(x,y)\cdot {\mathfrak {H}}(x,y)\ne 0\) for all \((x,y)\in Q\). Hence in Q we must have \({\mathfrak {F}}(x,y)=0\). We write the expansions of f with center in \(x_0\), \(y_0\) and \(x_0+y_0\), valid for \(|x-x_0|<\mu \) and \(|y-y_0|<\mu \), for a certain positive \(\mu \):
Since \({\mathfrak {F}}(x_0,y_0)=0\), from (3.2), setting for brevity \(x-x_0=t\) and \(y-y_0=v\), we have
We proceed by nullifying the coefficients of the previous power series; those of first degree are \(a_1-a_1^{\prime }\) and \(a_1-a_1^{\prime \prime }\) and by equalling them to 0 we get \(a_1=a_1^{\prime }=a_1^{\prime \prime }\). Going to the second degree we obtain
so \(a_2=a_2^{\prime }=a_2^{\prime \prime }=0\). Suppose \(a_n=a_n^{\prime }=a_n^{\prime \prime }=0\), then (3.3) becomes
and by nullifying the coefficients of terms of degree \(n+1\) we get
so \(a_{n+1}=a_{n+1}^{\prime }=a_{n+1}^{\prime \prime }=0\) and by induction we conclude that \(a_k=a_k^{\prime }=a_k^{\prime \prime }=0\) for all integers k and \(f(x)=f(x_0)+a_1(x-x_0)\). Because of the analyticity of f we have that it is affine on the whole \({\mathbb {R}}\), i.e., \(f(x)=\alpha +\beta x\). Since
we obtain that f is linear. \(\square \)
4 Further results and open problems
In this section we present some other results, namely some conditions giving only trivial solutions of equation (3.1) and we prove by giving examples that without them there exist non trivial solutions.
We start by using the simple result given in Sect. 1 and used in Sect. 2 for solving (2.1). Assume that the three functions f, g, h are absolutely continuous and differentiable. We have
where
By differentiating equation (4.1) with respect to x, we obtain
and setting \(x=0\) we have
where \(k_1=-f(0)\), \(k_2=-g(0)\) and \(k_3=-h(0)\). By setting \(x=y=0\) in equation (4.1), we have \(f(0)g(0)h(0)=-k_1k_2k_3=0\).
If one of the numbers \(k_1, k_2, k_3\), say \(k_1\), is zero and \(k_2k_3\ne 0\), then from equation (4.4) we have \(\phi (y)=0\) for all \(y \in {\mathbb {R}}\) and this implies \({\mathfrak {F}}(x,y)=0\) for all \(x, y \in {\mathbb {R}}\), hence f(x) is linear. If also \(k_2k_3=0\), equation (4.4) is identically satisfied.
We differentiate equation (4.3) again with respect to x to obtain
Setting \(x=0\), we obtain
Assume now that \(k_1=k_2=0\), then we have
so, if \(k_3 \ne 0\), \(\phi (x)\gamma (x)=0\) for all \(x \in {\mathbb {R}}\). Under these conditions, our original equation can be written as
Since the last factor for \(x=0\) is equal to \(k_3\), it cannot be always zero; if it is different from zero everywhere, then we come back to the alternative equation in two unknown functions studied in Sect. 2 and we conclude that either f or g is linear. Otherwise, let
clearly \({\mathcal {O}}\) is an open set, it contains all points of the forms (x, 0) and (0, y) and is symmetric with respect to the line \(y=x\). Thus, in \({\mathcal {O}}\) we must have
By repeating the proof of Theorem 2.1, from equation (4.10) we obtain that
By multiplying and using equation (4.7) we arrive at the following system:
Theorem 4.1
Suppose that \(x=0\) is an accumulation point of the set where \(\gamma \) is different from 0, then \({\mathfrak {F}}(x,y)=0\) for all \(x, y \in {\mathbb {R}}\), i.e., the function f is linear.
Proof
Take \(x_0 \in {\mathbb {R}}\) and consider the point \((0,x_0)\in {\mathcal {O}}\); since \({\mathcal {O}}\) is open, there exists \(\delta >0\) such that the open square \(Q_{\delta }(0,x_0)=(-\delta ,\delta )\times (x_0-\delta ,x_0+\delta )\) is contained in \({\mathcal {O}}\). By the hypotheses we can choose \(x_1\) with \((0,x_1)\in Q_{\delta }(0,x_0)\), \(\gamma (x_1)\ne 0\) and \(x_0\in (x_1+x_0-\delta ,x_1+x_0+\delta )\). From the system (4.12) we obtain
thus \(\phi (x_0)=0\) and this implies \(\phi \equiv 0\), so we are done. \(\square \)
The following example shows that without the condition stated in the previous proposition there exist non trivial solutions of equation (4.1) with \({\mathcal {C}}^{\infty }\) functions and \(k_3\ne 0\) (Note that the example presented in the previous section has \(k_1=k_2=k_3=0)\). Let h be the following function
where \(h_0(0)=-k_3 \ne 0\) and such that \(h \in {\mathcal {C}}^{\infty }\). The open set \({\mathcal {O}}\) is contained in the union of the three strips \({\mathbb {R}}\times (-1,1)\), \((-1,1)\times {\mathbb {R}}\) and that bounded by the lines \(y=1-x\) and \(y=-1-x\). Now, we must exhibit two \({\mathcal {C}}^{\infty }\) functions \(\phi \) and \(\gamma \) satisfying the system (4.12) and not identically zero. Take \(\phi (x)=0\) for \(x\le 5\), \(\phi (x)>0\) for \(x>5\), \(\gamma (x)=0\) for \(x\le 2\) and \(x\ge 3\) and \(\gamma (x) \ne 0\) for \(2<x<3\) and such that both functions are in \({\mathcal {C}}^{\infty }\). Obviously the first equation of system (4.12) is satisfied. The second equation is satisfied for \(x\le 2\) and for \(3\le x\le 5\) since both \(\phi \) and \(\gamma \) have value zero. If \(2<x<3\) then \(\phi (x)=0\) and \(\gamma (x)>0\), hence by the second equation of the system we must have \(\phi (x+y)=0\) for \((x,y) \in {\mathcal {O}}\): this is true since \(x+y<4\). If \(x>5\) then \(\gamma (x)=0\) and \(\phi (x)\ne 0\), hence we must have \(\gamma (x+y)=0\) for \((x,y) \in {\mathcal {O}}\): this is true since either \(-1<x+y<1\) or \(x+y>4\). The next step consists in proving that \({\mathfrak {F}}(x,y)\) and \({\mathfrak {G}}(x,y)\) are not identically zero and the triple f, g, h is a solution of the equation (4.1). We begin with \({\mathfrak {F}}(x,y)\): take \(x\le 5\) and \(y=1\), then
For \({\mathfrak {G}}(x,y)\) we take \(1<x\le 2\) and \(y=1\), then
To prove that the triple f, g, h is a solution of (4.1) we must prove that for each point \((x,y)\in {\mathcal {O}}\) either \({\mathfrak {F}}(x,y)\) or \({\mathfrak {G}}(x,y)\) is zero. Given the symmetry with respect to the line \(y=x\) it is enough to consider the two strips \((-1,1)\times {\mathbb {R}}\) and that bounded by the lines \(y=1-x\) and \(y=-1-x\) with \(x\ge 0\).
Let \(-1<x<1\) and \(x+y\le 5\), then
let now \(-1<x<1\) and \(x+y\ge 3\), then
and the first strip is covered. Now we go to the second strip; take (x, y) with \(1\le x\le 5\) and \(-1<x+y<1\), then
Take now \(x>5\) and \(-1<x+y<1\), then \(y<-4\), so
Now, we suppose that \(k_1=k_2=k_3=0\); in this case our equation becomes
We prove the following analogue of Theorem 2.1.
Theorem 4.2
Equation (4.20) implies that \(\Phi _y(x)\cdot \Gamma _y(x)\cdot \Psi _y(x)=0\) for all \(x,y \in {\mathbb {R}}\)
Proof
Suppose there exist \(x_0\) and \(y_0\) such that \(\Phi _{y_0}(x_0)\cdot \Gamma _{y_0}(x_0)\cdot \Psi _{y_0}(x_0)\ne 0\), then by continuity there exists \(\delta >0\) such that \(\Phi _{y_0}(x)\cdot \Gamma _{y_0}(x)\cdot \Psi _{y_0}(x)\ne 0\) for \(|x-x_0|<\delta \). If \(\int _{0}^{x_0} \Phi _{y_0}(t)dt\ne 0\), then for \(0<|x-x_0|<\delta \) it is \(\int _{0}^{x} \Phi _{y_0}(t)dt\ne 0\), so by equation (4.20) we have
and, by continuity,
If \(\int _{0}^{x_0} \Gamma _{y_0}(t)dt=0\), then we have \(\int _{0}^{x} \Gamma _{y_0}(t)dt\ne 0\) for \(0<|x-x_0|<\delta \), consequently \(\int _{0}^{x} \Psi _{y_0}(t)dt=0\) for \(0<|x-x_0|<\delta \) and, by continuity, \(\int _{0}^{x_0} \Psi _{y_0}(t)dt=0\): a contradiction since \(\Psi _{y_0}(x)\ne 0\) for \(|x-x_0|<\delta \). \(\square \)
Setting \(x=0\) in \(\Phi _y(x)\cdot \Gamma _y(x)\cdot \Psi _y(x)=0\), we get
Thus, we obtain the following system
The previous calculations show that a first step on a possible road towards the construction of all \({\mathcal {C}}^{\infty }\) solutions of equation (3.1) consists in solving the systems (4.12) and (4.24) which are consequences of the original one.
References
Fenyö, I., Forti, G.L.: On the inhomogeneous Cauchy functional equation. Istituto Matematico F. Enriques, Università degli Studi di Milano, Quaderno 46/S (1980)
Forti, G.L., Paganoni, L.: \(\Omega \)–additive functions on topological groups. In: Constantin Carathéodory: an International Tribute, 312–330, World Sci. Publishing, Singapore (1990)
Forti, G.L., Paganoni, L.: On an alternative Cauchy equation in two unknown functions. Some classes of solutions. Aequationes Math. 42, 271–295 (1991)
Forti, G.L., Paganoni, L.: Local solutions of an alternative Cauchy equation. Publ. Math. Debrecen 44, 51–65 (1994)
Kuczma, M.: Functional equations on restricted domains. Aequationes Math. 18, 1–34 (1978)
Funding
Open access funding provided by Universitá degli Studi di Milano within the CRUI-CARE Agreement.
Author information
Authors and Affiliations
Corresponding author
Additional information
Dedicated to Professor Ludwig Reich on the occasion of his 80th birthday, with esteem and friendship.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Forti, G.L. Alternative Cauchy equation in three unknown functions. Aequat. Math. 95, 1233–1242 (2021). https://doi.org/10.1007/s00010-021-00795-w
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00010-021-00795-w