Helmholtz Diffusion equation with reaction term:
$$
k\Delta u + u = f ~ \text{in} ~\Omega
$$
$$
\nabla u \cdot \mathbf{n} = 0 ~ \text{in} ~\partial \Omega
$$
For sufficiently small $k$ (relative to the mesh), there can be oscillations (with Lagrange elements) that return a negative $u$ in some regions, specially when $f$ is a step function. I tried to apply a DG method with the Symmetric Interior Penalty formulation, and I am still seeing these negative values. Is this possible to happen? I was told that DG methods preserve positivity and cannot return negative values. Is there any way to preserve positivity in DG methods for elliptic equations?
Preserve positivity: If $f(x)>0 ~\forall x\in\Omega$, then $u(x)>0 ~\forall x\in\Omega$
Edit again: What if the boundary conditions are of zero flux? Edited the problem.