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.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I want to show the following statement:

if $u\in W^{1,p}(\Omega)$, then $u_-$,$u_+$ and $|u|\in W^{1,p}(\Omega)$ with $$D(u_+) = Du\cdot I_{u>0}\qquad\text{and}\qquad D(|u|)=Du\cdot \text{sign}(u)\qquad \text{a.e.}$$ Furthermore, if $u,\, v \in W^{1,p}(\Omega)$, then $\max\{u,v\} \in W^{1,p}(\Omega)$.

My hunch: the main difficulty is to show $u_+ \in W^{1,p}(\Omega).$ The rest can be shown by vector space property: $$u_- = -u+u_+,\\ |u| = u_-+u_+,\\ \max\{u,v\} = \frac{u+v}{2}+\frac{|u-v|}{2}.$$

To show $u_+ \in W^{1,p}(\Omega)$, I construct an estimation function: $$u_\epsilon = (\sqrt{u^2+\epsilon^2}-\epsilon)\cdot I_{u>0}.$$

My first question:

How can I show such an estimation function is in $W^{1,p}(\Omega)$ for any $\epsilon>0$?

My second question:

How to use the estimation function to show $u_+ \in W^{1,p}(\Omega)$? By sending $\epsilon \downarrow 0$?

My third question:

Could it be true for general $W^{1,p}(\Omega)$?

share|cite|improve this question
up vote 2 down vote accepted

Sorry fot the mess in notation $\partial_i=(\cdot)_{x_i}$

1) Let $F:\mathbb{R}\to\mathbb{R}$, with $F\in C^1$, $|F'|\leq C$ and $F(0)=0$, then for every $u\in W^{1,p}(\Omega)$ we have $F(u)\in W^{1,p}(\Omega)$ and $(F(u))_{x_i}=F'(u)u_{x_i}$.

To prove this let $u_k \in C^{\infty}(\Omega)\cap W^{1,p}(\Omega)$ with $\| u_k-u\|_{1,p}\to 0$, then since $F(0)=0$ we have $|F(u)|\leq C|u|$ and so $F(u)\in L^p$, moreover $|F(u_k)-F(u)|\leq C|u_k-u|$ so $F(u_k) \to F(u)$ in $L^p$. Now since the $u_k$ are smooth the chain rule applies and gives $|(F(u_k))_{x_i}|=|F'(u_k)\partial_iu_k| \leq C|\partial_iu_k|$, (notice that this last sequence has an $L^p$ limit, namely $|\partial_iu|$) and since $F'$ is continous and (a subsequence) $u_k(x) \to u(x)$, that $F(u_k(x))\to F(u(x))$, $\partial_iu_k(x)\to \partial_iu$ a.e. and by a generalization of DCT that $F'(u_k)\partial_iu_k \to F'(u)\partial_iu$ in $L^p$. This gives the desired conclusion.

2) In your problem let $F_{\varepsilon}(t)=\sqrt{t^2+\varepsilon^2}-\varepsilon$ and apply the same arguments, noting that $|F_{\varepsilon}'|\leq 1$ and $F_{\varepsilon}(t) \to tI_{(0,\infty)}$ and $F'_{\varepsilon} \to I_{(0,\infty)}$. More expicitly $|F_\varepsilon(u)|\leq |u|$ and $|F_\varepsilon' (u)\partial_iu|\leq |\partial_iu|$.

share|cite|improve this answer
thanks for the answer. How did you obtain $|F(u)|\leq|u|$? I guess you mean $|F(u)|\leq C\cdot |u|$ – newbie May 8 '12 at 2:16
@newbie Yes, I've edited. Hopefully that's the only mistake... – Jose27 May 8 '12 at 2:21
I think the subsequent inequality $|F(u_k)-F(u)|\leq \cdots$, but it isn't substantial. I have updated the question a little bit, please help me out if you don't mind. Cheers. – newbie May 8 '12 at 2:30
@newbie The answer to the third question is yes, and it'll become clear once you do the second question. For this use the hints I gave in 2), the goal is to prove that if $v_\varepsilon=F_\varepsilon(u)$ then $v_\varepsilon \to u_+$ in $L^p$ and $v_\varepsilon' \to DuI_{u>0}$ in $L^p$, and the ingredients are all there, just mimic the arguments given in 1) to prove said convergence. – Jose27 May 8 '12 at 2:56

Your Answer


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.