Discontinuous solutions of the Navier-Stokes equations for compressible flow (Q808325)

The one-dimensional Navier-Stokes equations for a compressible fluid capable of conducting heat satisfy the system \[ v_ t-u_ x=0,\quad u_ t+p_ x=(\epsilon u_ x/v)_ x, \] \[ (e+{1\over 2} u^ 2)_ t+(up)_ x=((\epsilon uu_ x+\lambda T)/v)_ x, \] with initial date \((v,u,e)=(v_ 0,u_ 0,e_ 0)\), where v is the specific volume, u the velocity, p the pressure, e the internal energy, T the temperature, \(\epsilon\) and \(\lambda\) positive constants, and x the Lagrangean coordinate. The initial data \(v_ 0,u_ 0,e_ 0\) are \(L^ 2\)- functions and \(u_ 0,v_ 0\) are BV functions since, in several practical instances, the motion of the fluid is originated by discontinuous initial perturbations. A classical question is that of predicting the evolution in time of these initial discontinuities, whether they become smooth or persist all time and whether they are convected along particle paths. The fundamental result is that u and T become regular after positive times and \(v,p,u_ x,T_ x\) preserve their discontinuities for all time. In order to prove these properties it is necessary to establish two integral estimates on \(u_ x\) and \(T_ x\) having the forms \(\int^{\infty}_{-\infty}u_ xdx\sim t^{-{1\over 2} }\), \(\int^{\infty}_{-\infty}T_ xdx\sim t^{-1}\), which are also optimal in the sense that the exponents of t cannot be decreased. As for the jumps of \(v,p,e,u_ x,T_ x\), they satisfy a relation of the type \[ [L(t)]=[L(0)]\exp [\epsilon^{-1}\int^{1}_{0}\alpha (s)ds], \] where [L] is a well defined function of the jumps and \(\alpha\) a function depending on the initial jumps. When \(\alpha\) is negative the discontinuities decay exponentially in time. These discontinuities must not be confused with those of inviscid flows, that is, when \(\epsilon =\lambda =0.\) The method used to obtain the integral estimates is that of semidiscrete approximation, showing that analogous estimates hold for the approximate solutions up to a time \(t_ 0\) independent of the discretization as the size of the mesh tends to zero.