In the FTA project in Nijmegen we have formalized a constructive proof of the Fundamental Theorem of Algebra. In the formalization, we have first defined the (constructive) algebraic hierarchy of groups, rings, fields, etcetera. For the reals we have then defined the notion of real number structure, which is basically a Cauchy complete Archimedean ordered field. This boils down to axiomatizing the constructive reals. The proof of FTA is then given from these axioms (so independent of a specific construction of the reals), where the complex numbers are defined as pairs of real numbers. The proof of FTA that we have chosen to formalize is the one in the seminal book by Troelstra and van Dalen [17], originally due to Manfred Kneser [12]. The proof by Troelstra and van Dalen makes heavy use of the rational numbers (as suitable approximations of reals), which is quite common in constructive analysis, because equality on the rationals is decidable and equality on the reals isn't. In our case, this is not so convenient, because the axiomatization of the reals doesn't `contain' the rationals. Moreover, we found it rather unnatural to let a proof about the reals be mainly dealing with rationals. Therefore, our version of the FTA proof doesn't refer to the rational numbers. The proof described here is a faithful presentation of a fully formalized proof in the Coq system.

Is it reasonable to do constructive mathematics without the axiom of countable choice? Serious schools of constructive mathematics all assume it one way or another, but the arguments for it are not compelling. The fundamental theorem of algebra will serve as an example of where countable choice comes into play andhow to proceedin its absence. Along the way, a notion of a complete metric space, suitable for a choiceless environment, is developed. By constructive mathematics I mean, essentially, mathematics that is developed along the lines proposed by Errett Bishop [1]. More precisely, I mean mathematics that is done in the context of intuitionistic logic — without the lawof excluded middle. My reasons for identifying these notions are discussed in [9] and [10], the basic contention being that constructive mathematics has the same subject matter as classical mathematics. Ruitenburg [11] treated the fundamental theorem of algebra in a choiceless