Let $f: \mathbb{R}^d \to \mathbb{R}^m $ be a function sucht that $Df(a) = 0 $ for all $a \in U \subseteq \mathbb{R}^d$. Also, suppose $U$ is open and path-connected.
Question: IS $f$ constant on $U$ ??
Let $f: \mathbb{R}^d \to \mathbb{R}^m $ be a function sucht that $Df(a) = 0 $ for all $a \in U \subseteq \mathbb{R}^d$. Also, suppose $U$ is open and path-connected.
Question: IS $f$ constant on $U$ ??
Here is little more precise answer: Let $a, b \in U$ be arbitrary. We are done if we can show $f(a) = f(b)$. Since $U$ is path-connected, we can find a $C^1$-path (obtained from a continuous path by a suitable regularization procedure) $\gamma :[0, 1] \to U$ from $a$ to $b$. Then
\begin{align*} f(b) -f(a) &= f(\gamma(1))- f(\gamma (0)) = \int_0^1 \frac{\operatorname{d}}{\operatorname{d}t} f(\gamma(t)) \operatorname{d}t = \int_0^1 Df_{\gamma(t)} (\dot{\gamma}(t)) \operatorname{d}t = 0, \end{align*} where we have used the chain rule and $Df =0$. This argument also works if the path is only piecewise $C^1$ by applying it to the segments.
Although for the function of several variables there is no Mean Value Theorem, there is another similar result.
Suppose $f$ maps a convex open set $E \subset \mathbb{R}^d$ into $\mathbb{R}^m$, $f$ is differentiable in $E$, and there is a real number $M$ s.t. $\left\| {f'\left( x \right)} \right\| \leqslant M$ for every $x \in E$. Then $\left| {f\left( a \right) - f\left( b\right)} \right|\leqslant M\left| {a - b} \right|$ for all $a,b \in E$.
In this case, we can put $M=0$ since $Df(a) = 0$.
Yes: pick any two points and join them with a piecewise-linear path. On each segment, $f$ is constant, and hence the values at the initial and final points coincide.
This answer is similar to that of Dragoslav, except it uses the mean value theorem.
We want to show that every coordinate function $f_i : U \to \mathbb{R}$ is constant. Since $U$ is path-connected, it is enough to show that $$ f_i(\gamma(0)) = f_i(\gamma(1)) $$ holds for every path $\gamma : [0,1] \to U$. The map $f_i \circ \gamma$ is a function of a single real variable, so we can apply the ordinary mean value theorem to obtain $$ (f_i \circ \gamma)(1) - (f_i \circ \gamma)(0) = (f_i \circ \gamma)'(\xi)(1 - 0) $$ for some $\xi \in [0,1]$. Furthermore, the chain rule for total derivatives yields $$ (f_i \circ \gamma)'(\xi) = \underbrace{Df_i(\gamma(\xi))}_{= \, 0} \gamma'(\xi) = 0 $$ (note that the $f_i$ are totally differentiable iff $f$ is totally differentiable).