Frobenius and the symbolical algebra of matrices (Q2468668)

In this paper --- which is the latest in a series of detailed studies by the same author on the subject of matrix algebra in the nineteenth century --- Thomas Hawkins focuses on two problems solved by the German mathematician G.~F. Frobenius (1849--1917). The first of the problems discussed was solved by Frobenius in 1883 based on previous work by Weierstrass and Weber and others. Frobenius concerned himself with a situation encounted by Kronecker in connection with complex multiplication of abelian functions. The problem deals with what Hawkins has termed \textit{abelian matrices of order \(n\)}, which are \(2g\times 2g\) matrices \(\tilde{A}\) with integer coefficients such that \[ \tilde{A}^t J \tilde{A} = nJ\text{, \quad where \(J = \begin{pmatrix} 0 & I_g \cr -I_g & 0\end{pmatrix}\).} \] By a ``fertile combination of matrix algebra and Weierstrass' elementary divisor theory'' (p.~41), Frobenius was led to a solution that completely characterized so-called \textit{principal} abelian matrices of order \(n\), i.e. abelian matrices \(\tilde{A}\) of the form \(\tilde{A}= \begin{pmatrix} A & B \cr \Gamma & \Delta \end{pmatrix}\) --- where \(A,B,\Gamma, \Delta\) are \(g\times g\) matrices with integer coefficients --- such that the equation \[ T = (A+T\Gamma)^{-1}(B+T\Delta) \] has a solution \(T=\Phi+i\Psi\) in the space of complex symmetric \(g\times g\) matrices for which \(\Psi\) is positive definite. The problem of characterizing principal abelian matrices arose from the study of abelian functions that was highly fashionable in the nineteenth century. Thus, Frobenius' solution formed part of the integration of different approaches to the study of abelian functions and at the same time added to the emerging independent study of matrix algebra. The second problem solved by Frobenius and discussed by Hawkins arose from research on transformations of bilinear forms undertaken by Weierstrass and Kronecker. Frobenius defined two \(n\times n\) matrices \(A,B\) to be \textit{congruent} if \[ R^t A R = B\text{\quad for some nonsingular \(R\).} \] The core of the problem was then to prove that \(\lambda A -B\) is congruent to \(\lambda C-D\) when they have the same elementary divisors. In 1896, Frobenius found an approach to obtain this if he could prove that if \(U\) is a square matrix with \(\det U \neq 0\) and if \(m\) is the degree of the minimal polynomial of \(U\), then a polynomial \(\chi(z)\) of degree \(m-1\) exists such that \(\chi(U)^2 = U\). His proof of this ``square root theorem'' then followed by a simple and general route. In his paper, Hawkins presents these two problems and their solutions as obtained by Frobenius as well as their mathematical contexts within the problem fields from which they arose. The paper contains proofs of the key results, either as they were carried out in the papers or as they can be reconstructed from other sources. As a consequence of its approach and depth, Hawkins' paper is not an easy paper to read. Its historical contextualisation within the problem fields is enlightening yet internal. His attempts to indicate the influences on Frobenius sometimes reduces to listing similar, previous developments acknowledging the lack of direct evidence that Frobenius was aware of them. More is done by way of tracing influences when Hawkins in the final section analyses the impact on Frobenius' ideas on subsequent developments. The main strength of Hawkins' paper is its presentation of the technical aspects and the elucidation of internal motivations. Furthermore, it indirectly touches upon a number of important historical issues such as changing standards of rigor, the emergence and development of a Berlin-centred approach to mathematics, and the close interactions and relations between some of the main mathematical theories of the nineteenth century such as matrix algebra, abelian functions and Riemann surfaces. These aspects are deep and rich and must be taken up elsewhere based on in-depth studies similar to this one. However, Hawkins does elaborate on one such issue, namely that of ``generic'' and ``non-generic'' argumentation in the field of matrix algebra. The notion of ``generic'' argumentation covers generalizations from generic cases with either \(2\times 2\) or \(3\times 3\) matrices that was favoured by Cayley and Laguerre. Hawkins contrasts this with the ``non-generic'' argumentation style praised and practiced by Berlin-based algebraists that derived from Weierstrass' theory of elementary divisors. Thus, Hawkins argues, Frobenius' contributions were also important for showing the power of the non-generic approach. Thereby, it also helped solidify new standards of rigorous proof in the field of matrix algebra.