1: \begin{proof}

2: At a formal level, since $u$ is a very singular subsolution, its

3: initial condition is somehow concentrated at $x=0$. It thus

4: ``vanishes'' outside the origin and the conditions on the initial

5: data in Proposition~\ref{pr.wp2} are fulfilled. As more regularity

6: on the initial condition is needed to apply this result, we provide

7: a rigorous proof now. Consider $\tau>0$. According to

8: \eqref{regvss} and \eqref{comp.VSS}, $u(\tau)$ satisfies \eqref{wp1} and \eqref{wp2} with $\kappa=\gamma$ and we infer from Proposition~\ref{pr.wp2} that the solution $u^\tau$ to \eqref{eq1} in $(\tau,\infty)\times\real^N$ with initial condition $u^\tau(\tau)=u(\tau)$ satisfies

9: \begin{align*}

10: (t-\tau)^{\alpha-N\beta} \|u^\tau(t)\|_{1} + (t-\tau)^{\alpha} \|u^\tau(t)\|_{\infty} \leq & K_\gamma, \\

11: (t-\tau)^{\alpha+\beta} \|\nabla u^\tau(t)\|_{\infty} \leq & K_\gamma,

12: \end{align*}

13: for $t>\tau$. Now, if $u$ is a very singular subsolution to \eqref{eq1}, the comparison principle gives $u\le u^\tau$ in $(\tau,\infty)\times\real^N$ and \eqref{decayVSS} follows at once from the previous estimate after letting $\tau\to 0$. Next, if $u$ is a very singular subsolution to \eqref{eq1}, we obviously have $u^\tau=u$ and thus \eqref{gradestVSS}.

14: \end{proof}

15: