Integration on the Surreals: a Conjecture of Conway, Kruskal and Norton

Abstract

In his monograph On Numbers and Games, J. H. Conway introduced a real-closed field No of surreal numbers containing the reals and the ordinals, as well as a vast array of less familiar numbers. A longstanding aim has been to develop analysis on No as a powerful extension of ordinary analysis on the reals. This entails finding a natural way of extending important functions f from the reals to the reals to functions f* from the surreals to the surreals, and naturally defining integration on the f*. The usual square root, log, and exp were naturally extended to No by Bach, Conway, Kruskal, and Norton, retaining their usual properties. Later Norton also proposed a treatment of integration, but Kruskal discovered it has flaws. In his recent survey [2, p. 438], Siegel characterizes the question of the existence of a reasonable definition of surreal integration as "perhaps the most important open problem in the theory of surreal numbers." This paper addresses this and related unresolved issues with positive and negative results. In the positive direction, we show that semi-algebraic, semi-analytic, analytic, meromorphic, or more generally Écalle-Borel transseriable functions extend naturally to No, and an integral with good properties exists on them. In the negative direction, we show there is a fundamental set-theoretic obstruction to naturally extending many larger families of functions.