If I have a given continuous nonlinear map $T:\mathbb{R}^2\rightarrow \mathbb{R}^2$, and a region $D \subset \mathbb{R}^2$, is it necessarily true that $T(\partial D)=\partial T(D)$? That is, do boundary points of D get mapped to the boundary of the image of D after applying T?
I can see how this does not hold if $T$ is discontinous, but I can't think of a continuous $T$ where this does not hold. It also "feels right," but that's gotten people in trouble before!
I was attempting to prove this by looking at the effect on the open neighborhood around a boundary point after applying $T$... but couldn't make it very far.
Thanks for your help!
Your instincts are right: this statement is not true.:) A couple of counterexamples:
Let $D = {\mathbb R}^2$, and let $T$ be this projection: $$ T(x, y) = x^3. $$ Then $D$ has no boundary, while the image of $T$ is a line, hence is its own boundary.
Let $D = \{ (x, y) \; : \; -\pi \leq x \leq \pi, \; 1 < y < 2 \}$, and let $$ T(x, y) = \left(\; y \cos(x), \; y \sin(x) \; \right). $$ This maps $D$ onto the open annulus centered at the origin and with radii 1, 2. However, the boundary points of $D$ with $|x|=\pi$ are mapped into interior points of the annulus.
Let $D$ be the open strip $$ \{(x, y) \; : \; -\pi/2 < x < \pi/2, \; y \in {\mathbb R} \}. $$ Its boundary is the union of the two lines $|x| = \pi/2$. Now let $\tan$ be the principal branch of the tangent function and let--you guessed it!)-- $$ T(x, y) = (\tan(x), y). $$ Then $T(D)$ is all of ${\mathbb R}^2$, hence has no boundary at all.