6
$\begingroup$

Here is a problem from my homework, which asks me to show a Gronwall-type inequality.

Let $1\leq\beta<\gamma\leq\infty, 0<T\leq\infty$ and let $f\in L^\rho(0,T)$, where $1\leq\rho<\infty$ is defined by $\frac1\rho=\frac1\beta-\frac1\gamma$. If $\eta\geq0$ and $\varphi\in L_{\text{loc}}^\gamma([0,T))$ satisfy $$\|\varphi\|_{L^\gamma(0,t)}\leq\eta+\|f\varphi\|_{L^\beta(0,t)},$$ for all $0<t<T$. Prove that $$\|\varphi\|_{L^\gamma(0,t)}\leq\eta\Phi(\|f\|_{L^\rho(0,t)}),$$ for all $0<t<T$, where $\Phi(s)=2\Gamma(3+2s)$ and $\Gamma$ is the Gamma function.

I cannot see why the Gamma function appears here. Applying Hölder's inequality to the assumption gives that $$\|\varphi\|_{L^\gamma(0,t)}\leq\eta+\|f\|_{L^\rho(0,t)}\|\varphi\|_{L^\gamma(0,t)}.$$ Now we can do the iteration, but this process fails to give the desired result.

Any help would be appreciated.

| cite | improve this question | | | | |
$\endgroup$
  • $\begingroup$ there's no further context to the question? $\endgroup$ – Calvin Khor Apr 5 at 5:37
  • $\begingroup$ @CalvinKhor No. I’m afraid the above is all that I can provide. $\endgroup$ – Feng Shao Apr 6 at 8:43
1
$\begingroup$

I’m grateful for all efforts people made to solve my problem. With the help of my teacher, I finally find a solution. And it turns out that there is some thing wrong in the original problem.

Let $0=x_0<x_1<\cdots<x_n=T$ be such that $\|f\|_{L^\rho(x_{k-1},x_k)}=1/2$ for each $1\leq k\leq n-1$ and $\|f\|_{L^\rho(x_{n-1},x_n)}\leq1/2$. A simple calculation gives that $\|f\|_{L^\rho(0,x_k)}=\frac{k^{1/\rho}}2$ for $1\leq k\leq n-1$.

Next we do the iteration. By Minkowski's inequality and Hölder's inequality \begin{align*} \|\varphi\|_{L^\gamma(0,x_{k+1})}&\leq\eta+\|f\varphi\|_{L^\beta(0,x_{k+1})}\\ &\leq\eta+\|f\varphi\|_{L^\beta(0,x_{k})}+\|f\varphi\|_{L^\beta(x_k,x_{k+1})}\\ &\leq\eta+\|f\|_{L^\rho(0,x_k)}\|\varphi\|_{L^\gamma(0,x_k)}+\|f\|_{L^\rho(x_k,x_{k+1})}\|\varphi\|_{L^\gamma(x_k,x_{k+1})}\\ &\leq\eta+\frac{k^{1/\rho}}2\|\varphi\|_{L^\gamma(0,x_k)}+\frac12\|\varphi\|_{L^\gamma(x_k,x_{k+1})}\\ &\leq\eta+\frac{k^{1/\rho}}2\|\varphi\|_{L^\gamma(0,x_k)}+\frac12\|\varphi\|_{L^\gamma(0,x_{k+1})}. \end{align*} Note that $1\leq \rho<\infty$, we thus have $$\|\varphi\|_{L^\gamma(0,x_{k+1})}\leq 2\eta+k^{1/\rho}\|\varphi\|_{L^\gamma(0,x_k)}\leq 2\eta+k\|\varphi\|_{L^\gamma(0,x_k)}.$$ By induction we can easily deduce that $$\|\varphi\|_{L^\gamma(0,x_k)}\leq 2\eta (k+1)!$$

Fix $t\in(0,T)$, then there is some $1\leq k\leq n$ such that $t\in [x_{k-1},x_k)$, so \begin{align*} \|\varphi\|_{L^\gamma(0,t)}&\leq 2\eta (k+1)!\\ &=2\eta \Gamma(k+2)\\ &=2\eta \Gamma\left(3+\frac{k-1}2\cdot2\right). \end{align*} Since $\|f\|_{L^\rho(0,t)}\geq \|f\|_{L^\rho(0,x_{k-1})}=\frac{(k-1)^{1/\rho}}2$, we conclude that \begin{align*} \|\varphi\|_{L^\gamma(0,t)}&\leq 2\eta\Gamma\left(3+2(2\|f\|_{L^\rho(0,t)})^\rho\right)\\ &\leq\eta\Phi\left(2^\rho\|f\|_{L^\rho(0,t)}^\rho\right). \end{align*}

| cite | improve this answer | | | | |
$\endgroup$

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.