On a characterization of the logarithm by a mean value property

Any real polynomial f(x) = ax2 + bx + c, x ∈ IR, has the property that f (x)-f (y) x-y for every (x, y) ∈ IR, x ꞊ y. It turns out that that particular form of the Lagrange mean value theorem characterizes polynomials of at most second degree. Much more can be proved: J. Aczél [1] has shown that, with no regularity assumptions, a triple (/, g, h) of functions mapping IR into itself satisfies the equation f(x)-g(y) x-y= h(x + y) for all (x, y) ∈ IR, x ≠ y, if and only if there exist real constants a, 6, c such that f (x) = g(x) = ax2 + b, x + c, x ∈ IR, and h(x) = ax + b, x ∈ IR. Generalizations involving weighted arithmetic means were also considered (see e.g. M. Falkowitz [3] and the references therein) and characterizations of polynomials of higher degrees (in the same spirit) were obtained (see [4] and [5], for instance). In what follows we are going to characterize the logarithm in a similar way. To this end, denote by D the open first quadrant of the real plane IR2 with the diagonal removed, i.e. D := (O, ∞)2 \ {(x, x) e IR2 : x ∈ (0, ∞) }.Applying the classical Lagrange mean value theorem to the logaritmic function we derive the existence of a function D 3 (x, y) -> £(x,y) € intcony {x, y} such that the equality log a:-log y x-y £(z,y)