Proof of Fixed Point for nonlinear systems in Pazy's Semigroups of linear operators and applications to PDEs book

36 Views Asked by At

In Theorem 1.2 of Pazy's book called Semigroups of linear operators and applications to PDEs, a fixed point method is used to show that the nonlinear problem has a solution. The map $F:C([t_0,T];X) \to C([t_0, T];X)$ is defined by

$$(Fu)(t)=T(t-t_0)u_0 + \int_{t_0}^t T(t-s) f(s,u(s)) ds, t_0 \leq t\leq T.$$

Using the assumption that $$\|T(r)v\|_X \leq M\|v\|_X$$ for all $r \in [0,T]$ and the fact that $f$ is continuous in $t$ and uniformly Lipshitz continuous with constant $L$ on $X$, Pazy shows $$\|(Fu)(t)-(Fv)(t)\| \leq ML(t-t_0)\|u-v\|.$$ This part is understandable to me. He then says using this inequality and the definition of $F$ you can show using induction that $$||(F^nu)(t)-(F^n)v(t)||\leq \frac{(ML(t-t_0))^n}{n!}||u-v||.$$

To me this part is clear without the factorial, but I am not sure how we get the bound divided by $n!$. I'm assuming it has something to do with the properties of $T(t)$ since $T(t)$ is a $C_0$ semigroup, but I still don't see it. Does anyone have any ideas where this came from?

1

There are 1 best solutions below

0
On BEST ANSWER

It is a standard kind of estimate used, for example, in the proof of Picard's existence theorem. In order to get the factorial, you have to use the integral equality:

$$\begin{aligned} \|(F^2u)(t)-(F^2v)(t)\|&\leq\int_{t_0}^t \|T(t-s) [f(s,(Fu)(s))-f(s,(Fv)(s))] \|\,ds\\ &\leq M \int_{t_0}^t \|f(s,(Fu)(s))-f(s,(Fv)(s)) \|\,ds\\ &\leq M L \int_{t_0}^t \|(Fu)(s)-(Fv)(s) \|\,ds\\ &\leq M L \int_{t_0}^t ML(s-t_0)\|u-v\|\,ds\\ &= M^2 L^2\|u-v\|\int_{t_0}^t (s-t_0)\,ds\\ &= M^2 L^2\|u-v\|\frac{(s-t_0)^2}{2}\Bigg|_{s=t_0}^{s=t}\\ &= \frac{(M L(t-t_0))^2}{2}\|u-v\|\\ \end{aligned}$$

Analogously, the general case is proved by induction.