Regular splittings and monotone iteration functions (Q1061451)

Let X be a set with a partial ordering, introduced by a reflexive, antisymmetric and transitive relation. Let Y be a partially ordered set, which is directed downwards and upwards. For a function \(f: X\to Y\), a function \(g: X\times Y\to X\) is called a monotone iteration (MI) function belonging to f, if it has the following properties: (M1) g is monotone increasing with respect to both arguments, \((M2)\quad g(x,f(x))=x,\) \(\forall x\in X\), \((M3)\quad x\leq g(x,y)\Rightarrow f(g(x,y))\leq y,\quad x\geq g(x,y)\Rightarrow f(g(x,y))\geq y,\) \(\forall x\in X\), \(y\in Y\), (M4) if \(x_ k\uparrow x\), or \(x_ k\downarrow x\), then \(\forall y\in Y\Rightarrow g(x_ k,y)\to g(x,y).\) The first author [Z. Angew. Math. Mech. 52, 233-238 (1972; Zbl 0258.65057)] has introduced the notion of regular splitting of a function f. This regular splitting generalizes in a natural way the regular splitting for a linear function, introduced by \textit{R. S. Varga} [Boundary Probl. Diff. Equations, Proc. Sympos., Madison, April 20-22, 1959, 121-142 (1960; Zbl 0100.125)]. In this work, the authors prove necessary and sufficient conditions in order that a given MI-function is (in a precisely defined sense) at least as fast as a second one. Proving that regular splittings of linear functions are generating MI- functions (for nonlinear functions this is generally not true), the authors show that the concept of an MI-function is more general than the concept of regular splitting. As example an iteration method for computing the square root is given.