Dynamics on character varieties and Malgrange irreducibility of Painlevé VI equation (Q962068)

The \(SL(2,{\mathbb C})\) representation of the fundamental group of the four punctured sphere (which is routinely identified with the monodromy representation associated with the classical sixth Painlevé equation \(P_{VI}\)) modulo a conjugation by an element of \(SL(2,{\mathbb C})\) is isomorphic to a six-dimensional hypersurface of \({\mathbb C}^7\) called the character variety. The latter is defined by the quartic equation \(x^2+y^2+z^2+xyz=Ax+By+Cz+D\), \(A=ab+cd\), \(B=ad+bc\), \(C=ac+bd\), \(D=4-a^2-b^2-c^2-d^2-abcd\), where \(a,b,c,d\) are traces of the monodromy matrices along four basic loops \(\alpha, \beta, \gamma, \delta\), while \(x,y,z\) are the traces of the monodromy matrices corresponding to the loop products \(\alpha\beta, \beta\gamma, \gamma\alpha\), respectively. The famous Riemann-Hilbert correspondence provides a diffeomorphism of the moduli space of the logarithmic connections with simple poles at \(\{0,t,1,\infty\}\) and the respective local exponents \((\pm\theta_{\alpha}/2, \pm\theta_{\beta}/2, \pm\theta_{\gamma}/2, \pm\theta_{\delta}/2)\) onto the above cubic surface with parameters \((a,b,c,d)=(2\cos(\pi\theta_{\alpha}), 2\cos(\pi\theta_{\beta}), 2\cos(\pi\theta_{\gamma}), 2\cos(\pi\theta_{\delta}))\). The nonlinear monodromy of \(P_{VI}\) obtained by an analytic continuation of the sixth Painlevé transcendent around \(0\) and \(1\) generates an automorphism of \(S_{(A,B,C,D)}\) which is an element of the modular group \(\Gamma_2 \in PSL (2,{\mathbb{Z}})\) generated by three automorphisms \(g_x=s_zs_y\), \(g_y=s_xs_z\), \(g_z=s_ys_x\). The main concern of the paper is the dynamics related to the polynomial automorphism subgroup \(\Gamma_2^{\pm}\in PGL(2,{\mathbb{Z}})\). The latter is the free product of 3 involutions \(s_x,s_y,s_z\), preserving \(a,b,c,d\) and sending \((x,y,z)\) to \((-x-yz+A,y,z)\), \((x,-y-zx+B,z)\) and \((x,y,-z-xy+C)\), respectively. The authors prove that, if the smooth part of \(S_{(A,B,C,D)}\) with the real parameters \((A,B,C,D)\) has a bounded component, then all preimages \((a,b,c,d)\) of \((A,B,C,D)\) are real as well, and, depending on the choice of \((a,b,c,d)\), the bounded component of \(S_{(A,B,C,D)}\) stands for either \(SU(2)\) or \(SL(2,{\mathbb R})\) representation (both cases occur for a given surface \(S_{(A,B,C,D)}\)). This is important because the boundedness of an infinite \(\Gamma_2^{\pm}\)-orbit of a point \(m\in S_{(A,B,C,D)}\) implies that the parameters \((A,B,C,D)\) are real, while the orbit itself is contained and dense in the unique bounded connected component of the smooth part of \(S_{(A,B,C,D)}\) (fixed points of \(\Gamma_2^{\pm}\)-action are the singular points of \(S_{(A,B,C,D)}\) and correspond to one-parameter Riccati solutions of \(P_{VI}\); finite orbits correspond to algebraic solutions of \(P_{VI}\)). The main result of the paper is the assertion that the group \(\Gamma_2^{\pm}\) does not preserve holomorphic geometric structures like a finite type holomorphic curve, singular holomorphic web or an affine structure on a smooth part of \(S_{(A,B,C,D)}\) except for the case of the Cayley cubic, \((A,B,C,D)=(0,0,0,4)\), \((a,b,c,d)=(0,0,0,0)\) or \((2,2,2,-2)\), modulo Okamoto symmetries. This theorem allows the authors to prove that equation \(P_{VI}\) is irreducible in the sense of Malgrange except for the case of the Cayley cubic (the Malgrange reducibility of the Cayley cubic case follows from the computation of the invariant polynomial affine structure by Casale; for those parameters, \(P_{VI}\) can be integrated in terms of elliptic functions giving rise to the Picard solution but in a way that is non classical with respect to Nishioka-Umemura definition of irreducibility).