Probabilistic Cauchy functional equations (Q6654855)

Let \(n \geq 2\) be an integer, \(\mu\) be an absolutely continuous probability measure on the real line with support \(S(\mu)\). We say that a measurable function \(f \colon S(\mu) \to \mathbb{R}\) satisfies the \(n\)-summands probabilistic Cauchy functional equation with respect to \(\mu\) whenever\N\[\N f(X_1 + \cdots + X_n) \stackrel{d}{=} f(X_1) + \cdots + f(X_n)\tag{1}\N\]\Nis fulfilled. Here \(X_1, \ldots , X_n\) are independent identically distributed real-valued random variables whose laws equal \(\mu\). We simply say that \(f\) satisfies the probabilistic Cauchy equation with respect to \(\mu\) whenever the above equation holds with \(n = 2\).\N\NIn this paper, the authors aim to answer the question what regularity conditions imply that any solution \(f\) takes the linear form \(f(x) = cx\) (for some constant \(c\)) almost everywhere or everywhere.\N\NIn Theorem 2.1 they show that if the independent random variables \(X_1, X_2 \sim \mu\), where \(\mu\) is an exponential distribution with parameter \(\lambda>0\), and the real-valued function \(f\) is such that: \(f\in \mathscr{C}^{1}([0, +\infty[)\), \(f(0)=0\), \(f(+\infty)=+\infty\), \(f'(0)>0\), \(f\) is strictly increasing and satisfied (1) with \(n=2\), then \(f\) is necessarily linear.\N\NFurther, in Section 2.2 they provide a partial result in the general case beyond the exponential distribution via the so-called integrated Cauchy functional equation.