Sums of squares on real algebraic surfaces (Q2491169)

The existence of standard representations of positive semi-definite polynomial functions has been studied intensely for a long time. Given polynomials \(F_1,\dots, F_r\in\mathbb{R}[X]= \mathbb{R}[X_1,\dots, X_n]\) one defines the basic closed semi-algebraic set \(K= \{x\in\mathbb{R}^n\mid\forall i: F_i(x)= 0\}\) and the preorderings \[ PO(F_1,\dots, F_r)= \{\sum s_e\cdot F^{e_1}_1\cdot\cdots\cdot F^{e_r}_r\mid\forall e: s_e\in \sum \mathbb{R}[X]^2\}, \] \[ P(K)= \{F\in \mathbb{R}[X]\mid\forall x\in K: F(x)\geq 0\}, \] (where \(\sum\mathbb{R}[X]^2\) is the semi-ring of sums of squares in the polynomial ring). It is trivially true that \(PO(F_1,\dots,F_r)\subseteq P(K)\). The main concern of the current paper is the question: When is it true that the preorderings are equal? Many partial answers have been known previously, in particular for the case that \(\dim(K)\neq 2\). The author develops localization techniques that allow him to present various 2-dimensional examples where \(PO(F_1,\dots,F_r)= P(K)\).