A continuous, constructive solution to Hilbert's \(17^{th}\) problem (Q798716)

Let K be an ordered field, and \(\bar K\) its unique real closure. A polynomial \(F\in K[x_ 1,...x_ n]\) is called positive semi-definite (psd) if \(f\geq 0\) in \(\bar K\). Roughly speaking, Artin's proof shows that f is a sum of squares (SOS) of rational functions \(r_ i\in K(x_ 1,...x_ n)\) provided f is psd and each positive element of K is an SOS. More elegantly, now for all K, f is a weighted SOS: \(p_ ir^ 2_ i\), where \(p_ i\in K^+\) and f is psd. During the fifties to the dependence of \(p_ ir^ 2_ i\) on f, that is, on its variables x and coefficients c, it was given attention. The best result, by Daykin, has remained unpublished. It provides finitely many representations of f by SOS with terms \(p_ ir^ 2_ i\) such that (i) each \(p_ i\) depends polynomially on c, (ii) each \(p_ ir^ 2_ k\) is rational in both x and c, and (iii) if f is psd then, for one of the representations, all \(p_ i\geq 0.\) The author gives a new proof of this result, but, in contrast to Daykin, without attention to bounds on the number of terms and their degrees (depending on n and the degree of f). He also treats a natural topological variant where, in place of (ii), \(p_ ir^ 2_ i\) depends rationally on x, and continuously (for the order topology) on both x and c. His principal result achieves this for real closed K, where \(p_ i\) can be absorbed (since \(| x|\) is semi-algebraic and continuous), and \(r_ i\) is semi-algebraic in c. Having previously excluded a rational representation without Daykin's case distinctions, he conjectures that a continuous, piecewise rational solution is possible for all K. Here the 'pieces' are (basic) semi-algebraic sets. - As the author observed in a later preprint, the last line of p. 366 is false, with a counter-example obtained from the identity on 1.12 of p. 369. However, the false statement is not used in the paper at all, and should be omitted.