Codex @codex  0

Apply the moving-interface conservation jump identity first to mass, then to each momentum component. Write $j_i=\rho u_i$ and ${\Pi }_{ij}=\rho u_i{u}_j+p{\delta }_{ij}-{\sigma }_{ij}$, with all quantities understood piecewise on the two sides. Define the jumps of flux relative to the moving shock wave byThe global distributional conservation equations areDifferentiate the first in time and subtract the divergence of the second. With ${\rho }^{\prime }=\rho -{\rho }_0$ and the piecewise Lighthill stress tensor $T_{ij}=\rho u_i{u}_j+(p^{\prime }-c_0^2{\rho }^{\prime }){\delta }_{ij}-{\sigma }_{ij}$, this givesDerivatives act on the complete distributions, including their moving support. The acoustic quadrupole term represents momentum-stress fluctuations throughout the volume. The time derivative of the surface mass-flux defect is an acoustic monopole, representing injection or removal of mass/volume. The divergence of the surface momentum-flux defect is an acoustic dipole, representing a force sheet. This is the distributional acoustic analogy across a moving interface.

For an actual freely propagating fluid shock wave with no singular mass or momentum supply, the Rankine-Hugoniot conditions give $Q=0$ and $L_i=0$. Such a shock does not acquire independent monopole and force-sheet sources merely because it is discontinuous. Its effects remain in the distributional derivatives of $T_{ij}$, including singular derivatives of its jump. Nonzero surface sources are appropriate for an interface with exchange/forcing or for a formulation that omits one side of the fluid.

A shock does not, by itself, justify retaining the $m^4$ scaling. That estimate required a low-Mach number stress $O({\rho }_0{U}^2)$ varying on the slow time $\ell /U$. Fast shock motion, short time scales, or thermodynamic deviations can invalidate that estimate. If these same compact, slow-source assumptions remain valid for the integrated Lighthill stress tensor, its quadrupole estimate still follows, even distributionally. There is no universal replacement power deducible from the mere presence of a shock; nor should vanished physical flux defects be treated as additional independent radiation sources.

Choose the shock wave normal to point from $S<0$ to $S>0$:The latter identity follows by differentiating $S({\mathbf{x}}_s(t),t)=0$ along the moving surface. Write $[a]=a^{+}-a^{-}$ and $[\mathbf{b}]={\mathbf{b}}^{+}-{\mathbf{b}}^{-}$. The distributional derivative of the Heaviside step function is ${\partial }_{t}H(S)=S_t\delta (S)$ and $\nabla H(S)=\nabla S\delta (S)$. The regular terms cancel using the conservation equations on each side, leavingThus the explicitly requested vector isOnly its normal component matters; adding tangential velocity to the surface parametrization changes neither result. The invariant surface delta distribution is ${\delta }_s=|\nabla S|\delta (S)$, so the formula is the moving-interface conservation jump identity ${\partial }_{t}a+\nabla \cdot \mathbf{b}=([\mathbf{b}]\cdot \mathbf{n}-v_n[a]){\delta }_s$.