1

Can please somebody tell me, how solve this problem ?

We say that the uniformly elliptic operator $$Lu\ =\ -\sum_{i,j=1}^na^{ij}u_{x_ix_j}\ +\ \sum_{i=1}^nb^iu_{x_i}\ +\ cu$$ satisfies the weak maximum principle if for all $u\in C^2(U)\cap C(\bar{U})$ $$\left\{\begin{array}{rl} Lu \leq 0 & \mbox{in } U\\ u \leq 0 & \mbox{on} \partial U \end{array}\right.$$ implies that $u\leq 0$ in $U$.

Suppose that there exists a function $v\in C^2(U)\cap C(\bar{U})$ such that $Lv \geq 0$ in $U$ and $v > 0$ on $\bar{U}$. Show the $L$ satisfies the weak maximum principle.

(Hint: Find an elliptic operator $M$ with no zeroth-order term such that $w := u/v$ satisfies $Mw \leq 0$ in the region $\{u > 0\}$. To do this, first compute $(v^2w_{x_i})_{x_j}$.)

Thanks so much in advance.

Cookie
  • 13,952
FASCH
  • 1,762
  • 1
  • 21
  • 31
  • 1
    You're presenting a (graduate level?) qualifying exam, but you can't explain anything about what you've tried and what your thoughts are? – Erick Wong Jan 04 '14 at 00:36

2 Answers2

5

First, assume that you found the operator $M$ in the hint. If $w$ has a local maximum, all first derivatives are zero and all second derivatives are negative, so $Mw$ is positive, a contradiction. Thus, all maxima are on the boundary, so $w\leq 0$. (Edit: I implicitly assumed that the mixed partial terms were 0, but the result still holds because the second derivative terms can be diagonalized)

But $v$ is positive, so $u=vw$ must also be nonnegative.

This assumes that you completed the hint. Would you like help with that part?

Brian Rushton
  • 13,375
  • Thank you for your answer. Actually, the hint is the difficult part for me, could you please help me with that? – FASCH Jan 04 '14 at 01:58
  • @FASCH In order for someone to help you with that, you might want to explain what you're having trouble with. – Erick Wong Jan 04 '14 at 05:26
  • 1
    Dear Brian Rushton, thanks again for you answer. I already got the operator $M$ for myself. – FASCH Jan 05 '14 at 01:10
  • @BrianRushton If you can elaborate the hint, please let me know. My question (http://math.stackexchange.com/questions/1148857/weak-maximum-principle) concerns precisely this part only. I do believe I can complete the exercise on my own after figuring out the operator $M$. – Cookie Feb 16 '15 at 22:44
  • @Cookie First of all I apologize for searching for you here in a place where it is not supposed for me to comment here. But, please, help me with this? I only would like at least a hint, a small tip. Please, help me, I know from your profile that you are good at pde classic theory. https://math.stackexchange.com/questions/4797950/what-hypothesis-can-i-assum-in-order-to-obtain-infinity-solutions-for-this-probl – Silvinha Nov 05 '23 at 10:15
1

$Mw=-\sum_{ij=1}^{n}a^{ij}w_{ij}+\sum_{i=1}^{n}(b_i -\frac{2}{w}\sum_{k=1}^{n}a^{ik}v_k)w_i$

$Mw=\frac{Lu}{v}-\frac{uLv}{v^2}$

Use this we can solve the problem, I find in Han Qing's book.

Kira Yamato
  • 1,364
  • 1
    It may be $Mw=-\sum_{ij=1}^{n}a^{ij}w_{ij}+\sum_{i=1}^{n}(b_i -\frac{2}{\color{red}{v}}\sum_{k=1}^{n}a^{ik}v_k)w_i$ – Hans May 16 '15 at 14:00