1: \begin{proof}

2: Suppose that there exists $f$ such that $t^a=\nabla^af$ is future-directed timelike. If we consider an arbitrary future-directed causal curve $\gamma$ with tangent vector $v^a$, we have $g_{ab}v^a\nabla^bf>0$ and thus $v(f)>0$. As consequence there can be no closed causal curves in $(\mathscr{M},g)$ since $f$ cannot return to its initial value because it is monotonically decreasing along the curve $\gamma$.\\

3: Now let $t^a=\nabla^a f$ and set $\tilde{g}_{ab}$ as in \eqref{eqn:stca}. It is easy to see that the inverse of $\tilde{g}_{ab}$ is given by 

4: \begin{equation*}

5: \tilde{g}^{ab}=g^{ab}+t^at^b/(1-t^ct_c).

6: \end{equation*}

7: Then we obtain 

8: \begin{equation*}

9: \tilde{g}^{ab}\nabla_af\nabla_bf=t^at_a+\left(t_at^a\right)^2/(1-t^ct_c)=t^at_a/(1-t^ct_c)>0.

10: \end{equation*}

11: Hence $\tilde{g}^{ab}\nabla_bf$ is a timelike vector in the metric $\tilde{g}_{ab}$. By repeating the previous argument it follows that the space-time $(\mathscr{M},\tilde{g})$ contains no closed causal curves. Thus $(\mathscr{M},g)$ is stably causal.

12: The converse is more complicated to show and its rigorous proof can be found in \cite{HawEll}.

13: \end{proof}

14: