Singularities of vector fields arising from one-dimensional Riemann problems (Q1181699)

From the authors' introduction: The authors are motivated by the search for shock-wave solutions of two-dimensional Riemann problems arising as limits of solutions of an associated parabolic system. The conservation equations associated to the problem are of the form \(U_t+f(U)_x=0\), where \(U\) is a vector in the plane. The Riemann problem asks for a solution \(U(x,t)\) of the above equation for \(x\) in the line and positive \(t\) satisfying the initial condition \[ U(x,0)=U_0\text{ for } x<0,\text{ or } = U_1\text{ for } x>0, \] where \(U_0\) and \(U_1\) are constant vectors. An elementary shock-wave solution is given by \[ U(x,t)=U_0\text{ for } x-st<0,\text{ or } = U_1\text{ for } x- st>0, \tag{*} \] whenever the Rankine-Hugoniot condition, \(f(U_1)- f(U_0)=s(U_1-U_0)\), holds. One familiar way of searching for shock-wave solutions which are plausible physically is the viscosity method: one considers for \(\varepsilon\to 0\) limits of solutions of \(U_t+f(U)_x=\varepsilon(PU_x)_x\), where \(P\) is a smooth function taking values on \(2\times 2\) invertible matrices and \(\varepsilon>0\) [cf. \textit{D. G. Schaeffer} and \textit{M. Shearer}, Trans. Am. Math. Soc. 304, 267--306 (1987; Zbl 0656.35081)]. \textit{I. M. Gel'fand} [Transl., II. Ser. Am. Math. Soc. 29, 295--381 (1963); translation from Usp. Mat. Nauk 14, No. 2(86), 87--158 (1959; Zbl 0127.04901)] gave a criterion to decide whether a solution given in \((*)\) can be obtained by the viscosity method: this happens if and only if \(U_ 1\) is a singularity of the vector field \(X_{s,U_0}(U)=P^{-1}((f(U)-sU)- (f(U_0)-sU_0))\) in the plane, and there is an orbit \(\gamma(\tau)\) of \(X\) such that \(\lim_{\tau\to-\infty}\gamma(\tau)=U_ 0\) and \(\lim_{\tau\to\infty}\gamma(\tau)=U_1\). Note that the singularities of \(X_{s,U_0}\) are exactly the solutions \(U_1\) of the Rankine-Hugoniot condition for \(U_0\) and \(s\) fixed. Thus in order to understand solutions \((*)\) it is natural to study the number and type of singularities of the parametrized family of vector fields \(X_{s,U_0}\), \(U_0\in\mathbb{R}^2\) and \(s\in\mathbb{R}\). If we set \(F_s(U)=f(U)-sU\) then \(X_{s,U_0}=P^{-1}(F_s-F_s(U_0))\) and hence, since \(P\) is invertible, the singularities of \(X_{s,U_0}\) are the solutions of the equation \(F_s(U)=F_s(U_0)\). In the present paper the authors consider the following questions: what is the number of solutions of the equation \(F_s(U)=F_s(U_0)\) and how does it change with \(U_0\)? What are the bifurcations of singularities of the family of vector fields \(X_{s,U_0}\)? What are the singularities joined by an orbit \(\gamma(\tau)\) for the family \(X_{s,U_0}\) and how do they bifurcate?