A Note on Adjoint Error Estimation for One-Dimensional Stationary Balance Laws with Shocks

Abstract

We consider one-dimensional steady-state balance laws with discontinuous solutions. Giles and Pierce [J. Fluid Mech., 426 (2001), pp. 327--345] realized that a shock leads to a new term in the adjoint error representation for target functionals. This term disappears if and only if the adjoint solution satisfies an internal boundary condition. Curiously, most computer codes implementing adjoint error estimation ignore the new term in the functional, as well as the internal adjoint boundary condition. The purpose of this note is to justify this omission as follows: if one represents the exact forward and adjoint solutions as vanishing viscosity limits of the corresponding viscous problems, then the internal boundary condition is naturally satisfied in the limit.