This final chapter marks a departure from the main framework of the book by putting emphasis on hermitian forms over the complex field rather than symmetric forms over the real field. The passage is both natural and necessary. To give a simple motivation: polynomial or rational functions with real coefficients, so much praised in the preceding chapters, may very well have complex roots or complex poles. Taking them into account greatly simplifies computations and conceptual thinking, as we all remember from elementary algebra. A second important observation goes back to the dictionary between elementary functions and matrices: by writing in complex coordinates a real valued polynomial (in any number of variables) p(z, z) = c αβ z α z β uniquely determines the hermitian matrix (c αβ ), while a similar decomposition q(x) = γ αβ x α+β , with real coefficients γ αβ , so much needed for semidefinite programming, has a clear ambiguity. The appearance at this late stage of the book of imaginary "ghosts" related to the basic entities encountered so far should not discourage the truly real and very applied reader.
IntroductionA question arises from the very beginning: how much of the vast theory of hermitian forms (in a finite or infinite number of variables) should the student or practitioner in applied areas of real algebra, functional analysis, algebraic geometry, or optimization theory know? Due to the depth and wide ramifications of hermitian forms (over the complex field) versus forms over real fields, the answer is: quite a lot! The good news is that the material, old and new, either is well known, circulating in part as folklore, or is accessible, due to a century and a half of continuous development † Mihai Putinar was supported by NSF grant DMS-1001071.