Let $f:\mathbb{R^2}\mapsto\mathbb{R}$ be continuous and have partial derivatives in $D=\{(x,y):x^2+y^2\leq1\}$, and let $\mathscr{H}$ the set of such functions for which $\sup_D |f|\leq1$.
Could someone calculate $$ \sup\limits_{f \in \mathscr{H}} \inf\limits_{z\in D} [f_x^2(z)+f_y^2(z)]\;? $$
Current work:
I have an rigorous proof that it is less than 16.
Proof as follows:
Define $F(x,y)=f(x,y)+2(x^2+y^2)$. We have $F(0,0)\leq1$ and $F|_{\partial D}\geq1$, which mean that minimun can be obtained in the interior. Therefore, we have an interior point $(x_0,y_0)$ s.t. $$F_x(x_0,y_0)=F_y(x_0,y_0)=0. $$ Then $f_x(x_0,y_0)=-4x_0,f_y(x_0,y_0)=-4y_0$ $$f_x^2(x_0,y_0)+f_y^2(x_0,y_0)=16(x_0^2+y_0^2)<16.$$
Q.E.D.
Intuition:
However, the upper bound may be too rough. We could always find some point where $f_x^2+f_y^2$ is small.
If there is some place slopes fast, then the restriction of $f$ means there is always somewhere plain.
Examples like $f(x,y)=x, y $ or $\displaystyle\frac{x+y}{\sqrt{2}}$ all have $f_x^2+f_y^2$ with 1.