For a differentiable function $f : \mathbb{R} \to \mathbb{R}$ with $|f'(x)|\leq 1$ and any set $E \subset \mathbb{R}$

$$m^*(f(E)) \leq m^*(E)$$

where $m^*$ is the outer Lebesgue measure

First, the definition of the outer measure

$$m^*(E) = \inf \bigg\{ \sum_{n=1}^\infty(b_n - a_n): \bigcup_{n=1}^\infty(a_n, b_n) \supset E \bigg\}$$

From the definition of derivative, we have that

$$|f'(x)| = \lim_{x\to y}\bigg|\frac{f(x)-f(y)}{x-y} \bigg| \leq 1$$

Which gives that $\forall x \in E \Rightarrow |f(x)-f(y)| \leq |x-y|$

Therefore, when calculating $m^*(f(E))$, we know that $f(E) \subset f(\bigcup_{n=1}^\infty (a_n, b_n)) = \bigcup_{n=1}^\infty (f((a_n, b_n))) = \bigcup_{n=1}^\infty (f(a_n), f(b_n))$


$$\sum_{n=1}^\infty (f(b_n) - f(a_n)) \leq \sum_{n=1}^\infty (b_n - a_n)$$

Which, by definition means that

$$m^*(f(E)) \leq m^*(E)$$

Is this good enough? I think I might be missing something or not making this rigurous enough.

I don't knoe if this applies for any arbitrary subset $E \subset \mathbb{R}$ or just an interval. If it is just an interval I can define $E$ as a countable union of disjoint intervals (previous problem) $E = \bigcup_{n=1}^\infty I_n$ which gives

$$m^*(f(E)) = m^* \big( f \big( \bigcup_{j=1}^\infty I_j\big)\big) = m^*\big[ \bigcup_{j=1}^\infty f(I_j) \big] \leq \sum_{n=1}^\infty m^*(f(I_j)) \leq \sum_{n=1}^\infty m^*(I_j)$$

But I am lost in this step. Thanks for the help! :)


You have made a mistake in writing $f(a_n,b_n)$ as $(f(a_n),f(b_n))$. To correct this error you gave to argue that $f(a_n,b_n) \subset [f(c_n),f(d_n)]$ where $c_n$ is the point in $[a_n,b_n]$ where $f$ attains its minimum and where $d_n$ is the point in $[a_n,b_n]$ where $f$ attains its maximum. Hence $m^{*}(f(a_n,b_n)) \leq m^{*} [f(c_n),f(d_n)] =f(d_n)-f(c_n)\leq b_n-a_n$ by MVT.

  • $\begingroup$ Oh ok! Thanks! Also, is this the only problem with my proof? $\endgroup$ – The Bosco Feb 14 at 11:53
  • 1
    $\begingroup$ @TheBosco Yes. Your proof is fine except for this. $\endgroup$ – Kavi Rama Murthy Feb 14 at 11:54

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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