For a , write e=u2/2e=|u|^2/2. Assume uLloc2\nabla u\in L^2_{\rm loc} and e,eu,pu,fue,eu,pu,f\cdot u are . The local energy inequality is

te+div((e+p)u)νΔe+νu2fu\partial_t e+\operatorname{div}((e+p)u)-\nu\Delta e +\nu|\nabla u|^2\leq f\cdot u

in distributions. Explicitly, every nonnegative ϕCc(Q)\phi\in C_c^\infty(Q) satisfies

νQu2ϕQe(tϕ+νΔϕ)+(e+p)uϕ+fuϕ.\nu\int_Q|\nabla u|^2\phi\leq\int_Q e(\partial_t\phi+\nu\Delta\phi)+(e+p)u\cdot\nabla\phi+f\cdot u\,\phi.

It permits extra energy loss relative to the .

Local and global statements

Tests have compact support inside space-time. Obtaining a global energy inequality requires cutoff limits and endpoint information; the two definitions are not interchangeable without those arguments.