Toroidal circle planes that are not Minkowski planes (Q1281743)

A toroidal circle plane \((T,C)\) is a hyperbola structure or \((B^*)\)-geometry on the torus \(T={\mathbb S^1}\times {\mathbb S^1}\) with circle set \(C\) consisting of graphs of homeomorphisms of the 1-sphere \({\mathbb S^1}\). The basic axiom in a toroidal circle plane is the axiom of joining, that is, any three mutually non-parallel points are contained in a unique circle, where two points of \(T\) are called parallel if and only if they have the same first or second coordinates. One obtains a 2-dimensional Minkowski plane if also the axiom of touching is satisfied, that is, given any circle \(c\) and two non-parallel points \(p\) and \(q\) such that \(p\) lies on \(c\) and \(q\) does not, there is a unique circle that contains both points and intersects \(c\) only in \(p\). Although there are many models for 2-dimensional Minkowski planes, none were known for toroidal circle planes. This lack of examples is remedied in the paper under review. The author first observes that the circle space of a toroidal circle plane has two connected components, one (positive half) consisting of the graphs of orientation preserving homeomorphisms of \({\mathbb S^1}\), the other (negative half) of graphs of orientation reversing homeomorphisms. Furthermore, both halves of the circle space of a toroidal circle plane are rather independent of each other, that is, one can combine the positive half from one toroidal circle plane and the negative half from a different toroidal circle plane and obtain again a toroidal circle plane. In his construction of toroidal circle planes that are not Minkowski planes the author modifies the negative halves of the circle spaces of certain 2-dimensional Minkowski planes. The derived affine plane at a distinguished point of such a Minkowski plane is the Euclidean plane with parallel classes represented by Euclidean lines of slopes \(\pm 1\). This representation, which differs from the one that is usually used where parallel classes are the horizontal and vertical lines, allows to describe circles in the negative half as graphs of two functions defined on \({\mathbb R}\). Based on an earlier construction of the author [Result. Math. 30, 122-135 (1996; Zbl 0871.51008)] all circles not passing through the distinguished point are obtained from a family of circles each of which have the diagonals of the Euclidean plane as asymptotes by using the shift group \((x,y)\mapsto (x+b,y+c)\) for \(b,c\in {\mathbb R}\). The trick, of course, is to define the circles in this family in such a way that the axiom of touching is no longer satisfied. The geometric idea behind the construction that achieves this goal is to collapse two symmetric strips of width 2 about the coordinate axes onto the respective axes. Starting from a circle that is symmetric about the \(y\)-axis this process yields a circle that is bent at the \(y\)-axis and therefore allows more than one tangent line at the point of intersection with the \(y\)-axis. In terms of the associated functions one can start with a symmetric, differentiable function \(m:{\mathbb R}\to{\mathbb R}\) such that \(\lim_{x\to\pm\infty} m(x)\mp x=0\) and such that the derivative of \(m\) is an increasing homeomorphism from \({\mathbb R}\) onto the open interval \(]-1,1[\). Then circles in the above family are the graphs \(f_a:{\mathbb R}\to{\mathbb R}\), \(a>0\), defined by \(f_a(x)={1\over a}m(a(x-1))-1\) and \(f_a(x)={1\over a}m(a(x+1))-1\) for \(x\leq 0\) and \(x\geq 0\), respectively , and \(x\mapsto -f_a(-x)\). Besides providing specific examples the author also discusses a rather general setting using the shift group and characterises those functions by analytical conditions that lead to toroidal circle planes or 2-dimensional Minkowski planes.