A domain decomposition method for conformal mapping onto a rectangle (Q805786)

The authors consider a quadrilateral Q of the following form. The domain is given by \[ G=\{(x,y):\;0<x<1,\quad -\tau_ 1(x)<y<\tau_ 2(x)\}, \] and the corners of Q are \(z_ 1=-\tau_ 1(0)i\), \(z_ 2=1-\tau_ 1(1)i\), \(z_ 3=1+\tau_ 2(1)i\), \(z_ 4=\tau_ 2(0)i\). Let \(h=m(Q)\) be the modulus of Q. In the decomposition method proposed by the authors, Q is decomposed by the interval [0,1] into two smaller quadrilaterals \(Q_ 1\), \(Q_ 2\) with modules \(h_ 1=m(Q_ 1)\) and \(h_ 2=m(Q_ 2)\), and the central question is now: How well is h approximated by \(h_ 1+h_ 2\) provided that \(h^*=\min (h_ 1,h_ 2)\) is large? The authors give quantitative estimates for \(h-(h_ 1+h_ 2)\) under certain assumptions on \(\tau_ 1\), \(\tau_ 2\), using the Garrick integral equation for the conformal mapping of ring domains. Further, the conformal map of the rectangle \(R=(0,1)\times (0,h)\) onto G can be approximated by the conformal maps \(g_ 1\), \(g_ 2\) of the subrectangles \((0,1)\times (0,h_ 1)\) and \((0,1)\times (0,h_ 2)\) onto the two subdomains \(G_ 1\) and \(G_ 2\), again under certain assumptions on \(\tau_ 1\) and \(\tau_ 2\). Numerical examples show that these could probably be omitted, which - in the case of the estimate for \(h-(h_ 1+h_ 2)\)- was confirmed in a recent paper by the reviewer and \textit{W. Hayman} [ibid. 7, No.4, 453-467 (1991; Zbl 0729.30008)].