Core idea

For smooth divergence-free velocity and constant ν\nu, the is equivalent to

∂tu+∇⋅(u⊗u+pI−ν(∇u+(∇u)T))=f.\partial_tu+\nabla\cdot \bigl(u\otimes u+pI-\nu(\nabla u+(\nabla u)^{\mathsf T})\bigr)=f.

The divergence of a matrix is taken row by row: (∇⋅J)i=∑j∂jJij(\nabla\cdot J)_i=\sum_j\partial_jJ_{ij}.

Verification

The product rule yields

∇⋅(u⊗u)=(u⋅∇)u+u(∇⋅u),\nabla\cdot(u\otimes u)=(u\cdot\nabla)u+u(\nabla\cdot u),

while equality of mixed derivatives gives

∇⋅(∇u+(∇u)T)=Δu+∇(∇⋅u).\nabla\cdot(\nabla u+(\nabla u)^{\mathsf T}) =\Delta u+\nabla(\nabla\cdot u).

Both extra terms vanish under incompressibility. Also ∇⋅(pI)=∇p\nabla\cdot(pI)=\nabla p.

Simpler equivalent flux

For divergence-free velocity, u⊗u+pI−ν∇uu\otimes u+pI-\nu\nabla u has the same divergence as the symmetric-stress flux. The flux tensors differ, but their local momentum equations coincide under this constraint.