Stability of the Cauchy type equations in the class of differentiable functions (Q1288884)

This article considers an adaptation of a stability problem posed by Ulam in the 1940's. Given a real normed space \(X\) and a real Banach space \(Y\), the authors ask whether, for every \(\varepsilon>0\), there exists \(\delta>0\) such that, if \(f:X\to Y\) satisfies \(\|f(x+y)-f(x)-f(y)\|<\delta\) for all \(x\) and \(y\) in \(X\), then there exists an additive function \(A:X\to Y\) for which \(\|f(x)-A(x)\|<\varepsilon\) for all \(x\) in \(X\). Accordingly, for a function \(f:X\to Y\), the Cauchy difference of \(f\) is defined by \({\mathcal C}f(x, y):=f(x+y)-f(x)-f(y)\) for all \(x\) and \(y\) in \(X\). The authors then restrict their attention to functions \(f\) belonging to certain subspaces of \(C^n(X,Y)\) on which a special sort of semi-norm can be defined. The semi-norms of interest include, for instance, \(\|f\|:= \sum_{i=0}^{n-1}{\|D^if(0)\|+\|D^nf\|_{\sup}}\) and \(\|f\|:= \max_{i=0, \ldots, n}{\|D^if\|_{\sup}}\). The main stability result is Theorem 4.1 which asserts that, if \({\mathcal C}f\) belongs to a subspace of \(C^n(X\times X,Y)\) equipped with one of the selected semi-norms, then there exists an additive function \(A:X\to Y\) such that \(f-A\) belongs to the corresponding subspace of \(C^n(X,Y)\) and \(\|f-A\|\leq \|{\mathcal C}f\|\). A key step toward establishing this theorem is Theorem 2.1 which shows that the space \(C^n(X,Y)\) has the so-called `` double difference property''. Namely, if \({\mathcal C}f\) is in \(C^n(X\times X,Y)\), then there exists a unique additive function \(A:X\to Y\) such that \(f-A\) is in \(C^n(X,Y)\) and \(D(f-A)(0)=0\). Moreover, the theorem shows that \(\|D^k(f-A)(0)\|\leq \|D^k{\mathcal C}f(0)\|\) and \(\|D^k(f-A)\|_{\sup}\leq \|D^k{\mathcal C}f\|_{\sup}\) for all \(k\leq n\). The authors also establish analogous results, including a stability property (Theorem 4.2), for the Jensen difference of \(f\) defined by \({\mathcal J}f(x,y):= f({{x+y}\over{2}}) -{{f(x)+f(y)}\over{2}}\). The relation between \({\mathcal C}f\) and \({\mathcal J}f\) given in Theorem 3.1 should read \[ {\mathcal C}f(x,y) = {\mathcal J}f(2x, 2y)-{\mathcal J}f(2x, 0)-{\mathcal J}f(0, 2y) - f(0), \] but the error doesn't affect the results.