10
$\begingroup$

Let $n \geq 3$. Does there exist a Lipschitz function $f: \mathbb R^n \to \mathbb R$ such that the following conditions hold?

  1. The origin is a weak Lebesgue point of $\nabla f$, in the sense that the following limit exists:

$$L := \lim_{\delta \to 0_+} \frac{1}{|B_\delta (0)|} \int_{B_{\delta}(0)}\nabla f.$$

  1. $f$ is differentiable at $0$.

  2. $\nabla f(0)$ is not equal to its Lebesgue average $L$.

$\endgroup$

1 Answer 1

12
$\begingroup$

No. Subtracting a constant and a linear function, we may assume that $f(0)=0$ and $\nabla f(0)=0$. In other words, $f(x)=o(|x|)$ for $x\to 0$. Then, fixing a coordinate direction, say, of $n$-th coordinate, we look at the $n$—th coordinate of the integral $\int_{B_\delta(0)} \nabla f$, it equals $\int_{B_\delta(0)} \partial f/\partial x_n$. We organize integration in the following way: first by $x_n$ for fixed other coordinates, then by other coordinates. On the first step each integral over $x_n$ is the difference of values of $f$ at two points in the ball, thus it is uniform $o(\delta)$. After integration by other coordinates you get $o(\delta^n)$. Thus, $n$-th coordinate of $L$ equals 0. Every other coordinate too.

$\endgroup$
5
  • $\begingroup$ Ah, the uniform $o(|x|)$ really works well here. Thank you for the answer! $\endgroup$
    – Nate River
    Commented Sep 1 at 16:24
  • $\begingroup$ Just adding a note for myself and others who may be curious - we use the Lipschitz property crucially on the first step when we integrate over $x_n$, in particular we need the absolute continuity on lines to apply the FTC. I wonder if this holds for Sobolev functions, which are only absolutely continuous on almost every line. $\endgroup$
    – Nate River
    Commented Sep 1 at 22:26
  • $\begingroup$ Hmm, is not it just integration by parts which is a definition of weak derivative? $\endgroup$ Commented Sep 2 at 5:24
  • $\begingroup$ Maybe… I am not sharp enough to do this mentally haha. I am just concerned about this step: “On the first step, each integral over $x_n$ is the difference of values of $f$ at two points in the ball”. This will only hold for $\mathcal L^{n-1}$-almost every other coordinate for functions in $W^{1, p}$ (finite $p$), but maybe there is no problem. $\endgroup$
    – Nate River
    Commented Sep 2 at 5:51
  • $\begingroup$ Yes it should still be fine since the integral over the other coordinates is unaffected. $\endgroup$
    – Nate River
    Commented Sep 2 at 6:08

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .