If a $C^2$ function $u$ has a local maximum at $u(x,t)$, then
- $\nabla u = 0$, i.e. all of the partial derivatives of $u$ are zero (otherwise, you are not at a critical point: by moving "uphill" you increase $u$).
- $\nabla^2 u$, the Hessian of $u$, has no positive eigenvalues. Suppose otherwise, that $(v_x, v_t)$ is an eigenvector of $\nabla^2 u$ with $\lambda>0$ as an eigenvalue. Then consider $u(x+h v_x, t+h v_t)$ for small $h$. By Taylor's theorem
$$u(x+h v_x, t+h v_t) = u(x,t) + h\nabla u(x,t)\cdot v + \frac{h^2}{2} v^T \nabla^2 u v + h^2 R(h)$$
$$u(x+h v_x, t+h v_t) = u(x,t) + 0 + \frac{h^2}{2}\lambda + h^2 R(h),$$
with $\lim_{h\to 0} R = 0$. Therefore for sufficiently small $h$,
$$u(x+h v_x, t+h v_t) > u(x,t)$$
and $u$ does not have a maximum at $x,t$, a contradiction.
Since all of the eigenvalues of $\nabla^2 u$ are nonpositive, $\Delta u = \operatorname{tr}\nabla^2 u \leq 0.$