A Symmetric Proof of the Riemann-Roch Theorem, and a New Form of the Unit Theorem (Q5811277)

A proof of the Riemann-Roch theorem for the classical case and several remarks concerning the theorem are given. Let \(u\) be the algebraic function of degree \(n\) with \(z\) as the independent variable. As the Riemann-Roch theorem the authors adopt the form \(N((\tau)) + \tfrac12 \sum\sum \tau\,\nu = N((\bar \tau)) + \tfrac12 \sum\sum \bar\tau\,\nu\). By an order basis \(((\tau))\) or a divisor \(A\), we understand a system of individual orders \(\tau\) referring respectively to given \(u\)-branches at given \(z\)-values. \(\tau\) is a multiple of \(1/\nu\) where \(\nu\) is the order of the \(u\)-branch, and only a finite number of \(\tau\) are different from zero. \(N((x))\) is \(\dim (A^{-1})\) in the usual notation. Order bases \(((\tau))\) and \(((\bar\tau))\) are said to be complementary, to the level of a rational function \(S(z,u)\) with the order basis \(((\tau))\), provided \(\tau + \bar\tau = \sigma - 1 + 1/\nu\) for all \(u\)-branches at all finite \(z\)-values, while \(\tau + \bar\tau = \sigma + 1 + 1/\nu\) for all \(u\)-branches at the infinite \(z\)-value. In the notation of divisor this means \(\bar A=A^{-1}(S(z,u) W^{-1}\) \((W\) differential divisor), instead of usual \(\bar A=A^{-1}(S(z,u) W\). Double Summation \(\sum\sum\) extends first over \(u\)-branches with fixed \(z\), and then over \(z\)-values. The author denotes left hand side of the required equality by \(RR((\tau))\) and calls it the Riemann-Roch expression for the order basis \(((\tau))\). He shows first the existence of order basis \(((t))\) with \(N((t)) = N((\bar t)) = 0\) and \(RR((t)) - RR((\bar t)) = RR((\tau)) - RR((\bar\tau))\). Let \((\tau)\) be the part of \(((\tau))\) that refers to the individual \(z\)-value \(z = a\), and let the polynomials in \(u\) of degree \(i\) \(w_i(z,u) = u^i/(z-a)^{\alpha_i} + \ldots\) \((i = 0, 1, \ldots, n-1)\) be a local function basis for all the rational functions of \((z,u)\) with orders equal to or greater than \(\tau\), for \(z = a\). He then proves \(N((r)) > n + \sum\sum \alpha\). From this and an equality in a previous note of him we have \(RR((t)) = RR((\bar t))\), and hence \(RR((\tau)) = RR((\bar\tau))\).