Logarithmic-exponential series (Q5942884)

This paper develops the theory of logarithmic-exponential series (LE-series) over the real field \({\mathbb R}\), or more generally over any logarithmic-exponential ordered field \(k\). This structure was introduced by the same authors in [`Logarithmic-exponential power series', J. Lond. Math. Soc. (2) 56, 417--434 (1997; Zbl 0924.12007)], with a slightly different notation. With \(k\) as above, the authors define a field \(k((G))\) of series over \(k\) with monomials in \(G\) (where \(G\) is any multiplicatively written ordered abelian group). They then define, by successive \(\omega\)-chain constructions, the ordered exponential field \(k((x^{-1}))^E\), and the field of LE-series \(k((x^{-1}))^{LE}\), also written \(k((t))^{LE}\). The field \(k((t))^{LE}\) embeds into a field \(k((G^{LE}))\) of generalised power series. Hence, there is a notion of the support of an element \(f\in k((t))^{LE}\). This will be reverse well-ordered, and the corresponding leading monomial function yields a non-archimedean absolute value, with a corresponding valuation. The support of an element of \({\mathbb R}((t))^{LE}\) is shown to be countable (and the field has cardinality \(2^{\aleph_0}\)). The field \(k((t))^{LE}\) has a natural derivation, defined first via a derivation on \(k((t))^E\). The elements of \(k\) are exactly those with derivative 0. If \({\mathbb R}_{\text{an, exp}}\) denotes the o-minimal expansion of \({\mathbb R}\) with restricted analytic functions and exponentiation, then there is a model-theoretically well-behaved embedding into \({\mathbb R}((t))^{LE}\) of the Hardy field \(H({\mathbb R}_{\text{an, exp}}) \) of germs at infinity. It turns out that the derivative on \({\mathbb R}((t))^{LE}\) (or more generally \(k((t))^{LE}\)) inherits the expected valuation-theoretic properties of Hardy fields, so the valuation on \(k((t))^{LE}\) is a differential valuation in the sense of Rosenlicht. It is shown in Section 5 that \(k((t))^{LE}\) is closed under integration (defined as reverse differentiation). A necessary and sufficient condition is given for a second order ODE in one variable to have a non-trivial solution. Then in Section 6, a composition operation \(k((t))^{LE} \times k((t))^{LE}_{\infty} \rightarrow k((t))^{LE}\) is defined, where \(k((t))^{LE}_{\infty}\) denotes \(\{g\in k((t))^{LE}:g>k\}\). Natural properties of this are demonstrated; in particular the chain rule holds, and computational inverses exist, so \(k((t))^{LE}_{\infty}\) is a group under composition. The final section develops the theory of certain subfields \(k((t))^{LE,{\mathcal F}}\); here \({\mathcal F}=({\mathcal F}_m: m\in \omega)\), and each \({\mathcal F}_m \subseteq k[[X_1,\ldots,X_m]]\) has a certain closure property with respect to partial derivatives. Such fields are closed under differentiation, composition, and `truncation'. In the case when \({\mathcal F}_m=k[[X_1,\ldots,X_m]]\) for each \(m\), the resulting field is independently identified as \(k((t))^{LE,\text{ft}}\), the field of \(LE\) series with support of hereditarily finite type. It is shown that this is closed under integration, and has the appropriate closure property under compositional inverse. The field \({\mathbb R}((t))^{LE,\exp}\) is Ecalle's trigèbre \({\mathbb R}[[[X]]]\) of transseries.