Lagrange and the solution of numerical equations (Q5953250)

J.-L. Lagrange's work on the solution of numerical polynomial equations occupies the bulk of this article, following on a brief reference to his writing on algebraic solutions in \textit{Réflexions sur la Résolution Algébrique des Equations}, 1770. The work on numerical equations extended over a long period (1769/1770-1798, revised 1808) and developed a general, systematic algorithm for detecting, isolating and approximating, with arbitrary precision, all real and complex roots of a polynomial equation with real coefficients. Convergence is guaranteed (unlike in Newton's method) and to accelerate this, Lagrange used nonsimple continued fractions and Möbius transformations (well ahead of his time). In order to obtain a better algorithm for isolating roots, Lagrange worked in a quotient ring of a polynomial ring. All this is well outlined in the abstract. The authors, while pointing to secondary literature in which much of Lagrange's work has been discussed, deal in detail with aspects which have not and which they consider (rightly) to be important mathematically and historically. Over and above the content outlined above, Lagrange's basic approximation algorithm used continued fractions and this method is now almost forgotten. Details are presented of the method used by Lagrange for finding a lower bound on the differences of the real roots of a given real polynomial by computing the coefficients of an auxiliary equation, the ``equation of differences'' whose roots are the differences of all the distinct roots of the original equation. Following on this, two other methods are given, first published in 1795 and 1798, in which a lower bound is found more efficiently without the actual computation of the coefficients of the equation of differences. The authors show how Lagrange used algebraic congruences 50 years before this concept appeared in the work of Kummer and Cauchy. Finally, the article discusses improvements made by Lagrange to his approximation algorithm by the use of continued fractions, thereby accelerating convergence. It is the other that uses Möbius transformations long before Möbius to speed calculations. The paper is very well referenced and Lagrange's mathematics is very clearly explicated. It is high time that the subject-matter was brought to light and, in so doing, the authors have provided a very useful service to historical mathematical scholarship.