4
$\begingroup$

Is there a necessary and sufficient condition so that the arc length of a convergent sequence of functions converges to the arc length of the limit of the function?

I know that if $f'_n$ is continuously differentiable on the interval $[a,b]$ and $f$ converges for one $x \in [a,b]$. If $f'_n$ converges uniformly, then $f'_n$ converges to $f'$.

So for a normal Cartesian coordinate system the arc length is $\int\sqrt{1+(f')^{2}}$ $dx$ and we have that $f'_n$ converges uniformly to $f'$ and if the arc length of $f_n$ is exists for all $n$ and the arc length of $f$ exists, then the arc length of $f_n$ converges to $f$.

But, I know that its possible for a convergent sequence of piecewise differentiable functions $f_n$ still have it's arc length converges to $f$ when $f_n$ converges to $f$.

So does there exist some kind of most weak condition that gives the arc length of a convergent sequence of functions converges to the arc length of the limit and does there exists a converse condition that will give an if and only if statement?

$\endgroup$

1 Answer 1

5
$\begingroup$

A sufficient condition is that $f_n'\to f'$ in $L^1$, that is $\int_a^b |f_n'-f'|\to 0$. The justification for this (unexciting) observation is that the function $t\mapsto \sqrt{1+t^2}$ has derivative bounded by $1$, hence $|\sqrt{1+t^2}-\sqrt{1+s^2}|\le |t-s|$ for all $t,s\in\mathbb R$, which implies $$ \left| \int_a^b \sqrt{1+{f_n'}^2} - \int_a^b \sqrt{1+{f'}^2}\right| \le \int_a^b |f_n'-f'| $$


Within some classes of functions one can give weaker sufficient conditions. For example, if every $f_n$ is convex, and $f_n\to f$ uniformly, then the lengths of graphs converge.


Another relevant fact is that mere pointwise convergence $f_n\to f$ suffices to conclude that $$\int_a^b \sqrt{1+{f'}^2}\le\liminf_{n\to\infty} \int_a^b \sqrt{1+{f_n'}^2}$$ (the upper semicontinuity of length). To see this, pick a partition $t_j$ of $[a,b]$ that almost realizes the length of the graph of $f$. For sufficiently large $n$, the differences $|f_n(t_j)-f(t_j)|$ are as small as we want. Since the length of the graph of $f_n$ between $(t_{j-1}, f_{n}(t_{j-1}))$ and $(t_j, f_n(t_j))$ is no less than the length of line segment connecting these points, the claim follows.


All that said, I don't think there is any non-tautological necessary and sufficient condition for the general case.

$\endgroup$

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.