Convergence analysis of the lowest order weakly penalized adaptive discontinuous Galerkin methods