On reduction of linear two variable functional equations to differential equations without substitutions (Q1198198)

General methods for solving functional equations are few and far in between. The oldest one is reduction to differential equation. Going back at least to d'Alembert, it was systematically described by \textit{N. H. Abel} [Oeuvres complètes (Grondahl \& Son, Christiania, 1881; JFM 13.0020.01), pp. 389-398, 1-10], cf. also \textit{H. Kiesewetter} [Publ. Math. Debrecen 5, 265-268 (1958; Zbl 0095.269)]. In Abel's words (freely translated) ``For any functional equation in two free variables one can find, by repeated differentiation with respect to the two variables, as many equations as needed to eliminate any unknown function. In this manner one obtains an equation which contains only a single one of these functions and which, in general, will be a differential equation of a certain order. Thus it is in general possible to find all the unknown functions from a single equation \dots''. Easier said than done (notice the cautionary ``in general''-s). Abel gives, however, several examples where his method works. In most such cases new variables have to be introduced and thus substitutions with the accompanying problem of what domains are covered. While the author does not quote Abel, that is what ``without substitution'' in the title may allude to. Ideed, the author's method avoids such substitutions but it reduces functional equations to differential-functional equations rather than to differential equations. It applies to functional equations of the form \[ h_ 0(x,y)f_ 0(g_ 0(x,y))+\sum_{j=1}^ n h_ j(x,y)f_ j(g_ j(x,y))=F(x,y) \] and constructs a linear partial differential operator which eliminates all but the first term on the left hand side, thus obtaining a linear differential-functional equation for \(f_ 0\). The theoretical underpinnings are carefully worked out. The method requires differentiability in \((2n-1)\)-st order of the unknown functions but results of \textit{A. Járai} [Aequationes Math. 25, 52-66 (1982; Zbl 0517.39009); ibid. 30, 21-54 (1986; Zbl 0589.39012)] show that measurability suffices if the known functions are sufficiently regular. Of course, applying a general method to a specific equation is lengthier than ad-hoc methods. The author shows how his method works on the example \[ f(x+y)+g(x-y)+h(xy)=0, \] also coming from Abel (loc. cit. p. 9). For a direct way of solution, without any regularity condition, see e.g. the reviewer [Bull. Am. Math. Soc. 20. No. 2, 153-163 (1989; Zbl 0676.39004), in particular p. 158]. \{On p. 240, line 15 (6) should stand rather than (5).\}.