If you are on a compact domain $\Omega$in $\mathbb{R}^n$ you need to impose some elliptic boundary conditions or it is difficult to make predictions. Assume for simlicity that $f$ satisfies the Dirichlet boundary conditions, i.e., it vanishes on the boundary of $\Omega$. Then $$\lambda_3\Vert f\Vert_{L^\infty}=\sup_{x\in\Omega} |D_3f(x)| \leq C\lambda^{3+\frac{n-1}{2}} \Vert f\Vert_{L^2}\leq C'\lambda^{3+\frac{n-1}{2}} \Vert f\Vert_{L^\infty}$$ This gives you $$\lambda_3\leq C' \lambda^{3+\frac{n-1}{2}}. $$ I personally have serious reasons to believe that the exponent $3+\frac{n-1}{2}$is the best that you can hope for for an *arbitrary* $D_3$.