On applications of diophantine approximations (Q3680919)

From MaRDI portal

!

This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:

scientific article; zbMATH DE number 3902757
Language Label Description Also known as
default for all languages
No label defined
    English
    On applications of diophantine approximations
    scientific article; zbMATH DE number 3902757

      Statements

      On applications of diophantine approximations (English)
      0 references
      1984
      0 references
      arithmetic properties
      0 references
      values of Siegel G-functions
      0 references
      measure of linear independence
      0 references
      algebraic independence
      0 references
      Padé approximants of the first kind
      0 references
      system of simultaneous forms
      0 references
      Padé approximants of the second kind
      0 references
      Galochkin condition
      0 references
      0 references
      In this ingenious note the author gives a complete proof of a remarkable theorem on values of Siegel \(G\)-functions. For simplicity we restrict our enunciation to the rational field \(\mathbb{Q}\). A \(G\)-function \(f(x)\) is then a power series of the form \(\sum^{\infty}_{n=0}a_n x^n\) with the property that for some \(C\) and each \(n\geq 0\) there is a positive integer \(d_n\) such that \(d_na_0, \ldots, d_na_n\) are integers of absolute values at most \(C^n\). Suppose \(f_1(x),\ldots, f_n(x)\) are such \(G\)-functions satisfying a system of differential equations \[ f'_i(x) = \sum^n_{j=1} A_{ij}(x) f_j(x)\quad\text{for }A_{ij}(x)\text{ in }\mathbb{Q}(x). \tag{*} \] Assume also that the functions \(1, f_1(x),\ldots, f_n(x)\) are linearly independent over \(\mathbb{Q}(x)\). Then if \(r=a/b\neq 0\) is a rational number with denominator \(b>0\) sufficiently large with respect to \(| a|\), the values \(1, f_1(r), \ldots, f_n(r)\) are linearly independent over \(\mathbb{Q}\). (In fact the author proves a measure of linear independence with exponent arbitrarily close to \(n\); and also a similar result on algebraic independence. Such results are comparatively straightforward consequences of his method, and we refer the reader to the original paper for details.) This kind of result had been obtained for special functions by \textit{C. L. Siegel} in his famous paper [Abh. Preuß. Akad. Wiss., Phys.-Math. Kl. 1929, No. 1, 70 p. (1929; JFM 56.0180.05)], and it was to a certain extent generalized by \textit{M. S. Nurmagomedov} [Math. USSR, Sb. 14 (1971), 329--357 (1972); translation from Mat. Sb., Nov. Ser. 85 (127), 339--365 (1971; Zbl 0225.10037)], \textit{A. I. Galochkin} [Math. USSR, Sb. 24 (1974), 385-407 (1976); translation from Mat. Sb., Nov. Ser. 95 (137), 396--417 (1974; Zbl 0311.10035)], and \textit{E. Bombieri} [Recent progress in analytic number theory, Vol. II, 1--67 (1981; Zbl 0461.10031)]. The last two authors imposed an extra technical condition on the functions; this can be stated either in terms of \(p\)-adic radius of convergence (Bombieri) or in terms of a certain iteration property of the system (*) (Galochkin -- see below). The author's main innovation is therefore to remove entirely this technical condition. Let us try to indicate how this is achieved, for simplicity assuming that the \(A_{ij}(x)\) in (*) are polynomials in \(R=\mathbb{Z}[x]\). In earlier work an approximating form \(\Lambda = P_1f_1+\cdots+P_nf_n\) is constructed for \(P_1,\ldots,P_n\) in \(\mathbb R\) (so-called Padé approximants of the first kind). This gives rise to a linear map \(\mathcal D: \mathbb R^n\to \mathbb R^n\) defined by \(\mathcal D(P_1,\ldots, P_n)=(P^*_1,\ldots, P^*_n)\), where \(d\Lambda/dx = P^*_1f_1+\cdots+P^*_nf_n\). The Galochkin condition referred to above amounts to requiring that the iterates \(\mathcal D^k/k!\) be essentially integral, i.e., there is a positive integer \(e_k\leq c^k\) such that \(e_k \mathcal D^ k/k!\) takes \(\mathbb R^n\) into \(\mathbb R^n.\) The present author constructs instead of a single form a system of simultaneous forms of the shape \(\Lambda_i = Qf_i - P_i\) \((1\leq i\leq n)\) (so-called Padé approximants of the second kind). In this case it is easy to verify that, at least modulo the ideal \(I\) generated over \(\mathbb R\) by \(\Lambda_1,\ldots,\Lambda_n\), we have \(d\Lambda_i/dx\equiv Q^*f_i - P^*_i\) \((1\leq i\leq n)\) with \(Q^* = dQ/dx\). Thus modulo \(I\) the effect of the corresponding iterated \(\mathcal D^k/k!\) on \(Q\) at least is to replace it by \((1/k!) d^kQ/dx^k\), which is still in \(\mathbb R\) if \(Q\) is. It turns out that these weaker properties suffice. Finally the author shows how to recover Padé approximants of the first kind from those of the second kind. From this he deduces that the Galochkin condition does in fact always hold; in his earlier terminology [Journées arithmétiques, Exeter 1980, Lond. Math. Soc. Lect. Note Ser. 56, 11--82 (1982; Zbl 0489.10027)] all \(G\)-functions are \((G,C)\)-functions.
      0 references

      Identifiers