Using results from the generalised moment problem and duality from convex optimization the author shows that given a nonnegative polynomial $f$ on $\Bbb R^n$ simple sum of squares (sos) perturbations $f_\varepsilon$ arbitrarily near to $f$ can be constructed. This is important since sos representations can be found relatively fast by semidefinite programming and are a surrogate for deciding the NP hard problem of nonnegativity of a polynomial. Furthermore, certain minimization problems for polynomials can be related to problems for positive semidefinite polynomials by observing that $\min\{f(x): x\in K\}=\max \{\lambda: f(x)-\lambda \geq 0, \forall x\in K\},$ where $K\subseteq \Bbb R^n$ is suitable. For $f=\sum_\alpha f_\alpha x^\alpha\in \Bbb R[x_1,\dots,x_n],$ define $\vert \vert f\vert \vert _1=\sum_\alpha \vert f_\alpha\vert $ and denote by {\bf P} the problem $f^*:=\inf\{f(x): x\in \Bbb R^n\},$ and by ${\cal P}_M$ the problem $\inf \{\int f \,d\mu: \int \sum_{i=1}^n e^{x_i^2} d\mu \leq ne^{M^2}\}.$ Here $\mu\in {\cal P}(\Bbb R^n), $ the space of probability measures on $\Bbb R^n.$ Let $\inf {\cal P}_M$ be the optimal value of ${\cal P}_M.$ The multiindices $\alpha \in \Bbb N^n$ are considered suitably endowed with a natural linear order. P3.2: Assume $-\infty < f^*.$ Then $\inf {\cal P}_M \downarrow f^*$ as $M\rightarrow \infty.$ Next the author introduces for $r\geq \deg f/2,$ $r\in \Bbb N$ a semidefinite programming problem $Q_r$ that is a relaxation of ${\cal P}_M.$ Let its dual be $Q_r^*.$ In $Q_r$ one has to minimize $\sum_\alpha f_\alpha y_\alpha,$ given certain semidefiniteness constraints on the reals $y_\alpha.$ T3.3: Assume $f$ has a global minimum $f^*>-\infty.$ Then $\max Q_r^*=\min Q_r \uparrow \inf {\cal P}_M$ for all admissible $r$ as $r \rightarrow \infty.$ Let ${\bold y}^{(r)}= \{y_\alpha^{(r)}\}$ be an optimal solution of $Q_r$ and embed it naturally into Banach space $l_\infty.$ Then every pointwise accumulation point {\bf y}$^*$ of the sequence $\{ {\bold y^{(r)}} \}$ of optimal solutions admits a representation $y_\alpha^*=\int x^\alpha d\mu^* $ with a unique $\mu^*\in {\cal P}(\Bbb R^n)$ that itself is an optimal solution to ${\cal P}_M.$ As a consequence one can approximate the optimal value $f^*$ of ${\bold P}$ as closely as desired by solving SDP relaxations $Q_r$ for sufficiently large values of $r$ and $M.$ $M$ can be fixed whenever whenever a global minimizer $x^*$ of $f$ can be shown to have $\infty$-norm$\leq M.$ This method is simpler than the procedure proposed in {\it J. B. Lasserre} [SIAM J. Optim. 11, No. 3, 796--817 (2001;

Zbl 1010.90061)]. Another important result is T4.1: Assume $0\leq f^*=\min f(x).$ For every $\varepsilon >0$ there is some $r(f,\varepsilon) \in \Bbb N$ such that $f_\varepsilon =f+ \varepsilon \sum_{k=0}^{r(f,\varepsilon)} \frac{1}{k!}(x_1^{2k}+\cdots+x_n^{2k})$ is sum of squares. In particular $\vert \vert f-f_\varepsilon\vert \vert _1 \rightarrow 0$ as $\varepsilon \downarrow 0.$ Define the semialgebraic set ${\bold K}= \{x\in \Bbb R^n: g_j(x)\geq 0, j=1,\dots ,m\}.$ C4.3: If the $g_j$ are concave (hence ${\bold K}$ convex) and satisfy Slater’s condition, $f$ is convex, and ${\bold K}$ compact, then $f_\varepsilon=f_0+\sum_{j=1}^m \lambda_j g_j$ with some sos $f_0$ and nonnegative $\lambda_j.$ \quad In particular this shows a simplified Putinar representation for $f_\varepsilon,$ see {\it M. Putinar} [Indiana Univ. Math. J. 42, No. 3, 969--984 (1993;

Zbl 0796.12002)] or {\it A. Prestel} and {\it C. N. Delzell} [Positive polynomials. From Hilbert’s 17th problem to real algebra. Springer Monographs in Mathematics. Berlin: Springer (2001;

Zbl 0987.13016)]. In a footnote the reader is informed that this well written and interestingly motivated paper also appeared in SIAM J. Optim. 16, 751--765 (2006), see above.