Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

I'm trying to prove that if $f\colon\mathbb{R}^n \to \mathbb{R}^n$ is a $\mathcal{C}^1$ mapping such that $f'(x)$ is a (linear) isometry for every $x \in \mathbb{R}^n$, then $f$ is an isometry. By an application of inverse mapping theorem and mean value theorem, we have that $|f(x) - f(y)| = |x-y|$ as long as $x$ and $y$ are sufficiently close. How to extend this to the whole space?

share|improve this question
    
I'm not really sure what you intend with $f'$ here, but do you really mean $f'(x)$ to be a map? Presumably evaluating a derivative at a point would result in an element of the field, but not a map. Please take some time to clarify these two points :) –  rschwieb May 28 '12 at 1:55
    
@rschwieb: For maps $f:\mathbb{R}^n\rightarrow \mathbb{R}^m$, the derivative at a point is defined to be a linear map satisfying certain properties. When viewed this way, for a function $f:\mathbb{R}\rightarrow\mathbb{R}$, the "usual" derivative at a point $x_0$ should be thought of as "multiplication by $f'(x_0)$." –  Jason DeVito May 28 '12 at 1:59
    
I suspect he means that $f'(x)$ is the differential of $f$ at the point $x$. For example, in one variable the functions equal to their own derivatives in this sense are really the linear functions $f(x) = mx$, not $f(x) = e^x$ (whose derivative at $x_0$ is the map $x \mapsto e^{x_0}x$ on the tangent space at $x_0$). Maybe one can use some kind of connectedness argument to show a local formula for $f$ as an orthogonal map composed with a translation has to work globally. –  KCd May 28 '12 at 2:00
    
His formulation is correct. The derivative is a linear transformation. This is the approach taken in Spivak's Calculus on Manifolds. In this case $f': \mathbb{R}^n \rightarrow L(\mathbb{R}^n)$. –  ncmathsadist May 28 '12 at 2:02
    
OK just checking :) I wasn't able to guess all of this. Thanks! –  rschwieb May 28 '12 at 2:07
add comment

1 Answer 1

up vote 2 down vote accepted

Edit

Here's a much simpler argument. Let $X\subseteq\mathbb{R}^n\times\mathbb{R}^n$ with $$X = \{(x,y)\in\mathbb{R}^n\times \mathbb{R}^n|d(x,y) = d(f(x),f(y)) \}.$$ Note that $(x,x)\in X$ for any $x$, so $X\neq \emptyset$. Your observation is equivalent to the statement that $X$ is open. To see $X$ is closed, note that if $g:\mathbb{R}^n\times\mathbb{R}^n\rightarrow\mathbb{R}$ with $g(x,y) = d(x,y) - d(f(x),g(y))$, then $g$ is continuous since $d$ and $f$ are, and $X = g^{-1}(0)$. Hence, $X$ is a nonempty clopen subset of $\mathbb{R}^n\times\mathbb{R}^n$, so $X = \mathbb{R}^n\times\mathbb{R}^n$, so $f$ is an isometry.

(End edit)

Here's one approach, borrowed from Riemannian geometry.

Let $\gamma:\mathbb{R}\rightarrow\mathbb{R}^n$ be any straight line paramaterized by arclength, meaning that $\|\gamma(t)-\gamma(s)\| = |t-s|$ for any $t$ and $s$. We will show that $f\circ \gamma$ is also a straight line parametrized by arclength.

Believing this for a second, for $x$ and $y$ in $\mathbb{R}^n$, if $\gamma$ is chosen to be the line going through $x$ and $y$ with $\gamma(t) = x$ and $\gamma(s) = y$, then we have \begin{align*} d(x,y) &= \|\gamma(t)-\gamma(s)\|\\\ &= |t-s|\\\ &= \|f(\gamma(t))-f(\gamma(s))\| \\\ &= d(f(x),f(y)), \end{align*} establishing what we want.

So, why is $f\circ \gamma$ a straight line parameterized by arclength? This follows from your observation that $\|f(x)-f(y)\| = \|x-y\|$ for $x$ and $y$ close together. More specifically, looking at the point $\gamma(t)$, we know that for $s$ close to $t$ (and therefore $\gamma(s)$ close to $\gamma(t)$), that $\|f(\gamma(t)) - f(\gamma(s))\| = \|\gamma(t)-\gamma(s)\|$. This implies that for the line segment of points near $\gamma(t)$, that $f($segment$)$ is another line segment, parametrized by arclength.

Since $\mathbb{R}$ is connected, so is $f(\gamma(\mathbb{R}^n))$. This implies that $f(\gamma(\mathbb{R}^n))$ is a union of line segments end point to end point where each segment is parameterized by arclength. This is a straight line iff there are no corners. But if $f(\gamma(t_0))$ is a corner, then $f$ brings points near $\gamma(t_0)$, but on either side of it, closer together, contradicting your earlier observation that locally $f$ preserves distance.

share|improve this answer
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.