I'm reading the paper "Global existence and scattering for rough solutions of a nonlinear Schrödinger equation on $\mathbb{R}^3$ by Colliander, Keel, Staffilani, Takaoka and Tao. They study the following NLS $$i\partial_t\phi(x,t)+\Delta\phi(x,t)=\vert\phi(x,t)\vert^2\phi(x,t)$$ with initial datum $\phi_0\in H^s(\mathbb{R}^3)$. At the beginning of the paper they state that conservation laws and the localintime theory immediately yield globalintime wellposedness for $s\geq 1$. I understand the case $s=1$ just using the conservation of the energy. But now, let's consider for instance the case $s=2$. In particular an $H^2$solution is an $H^1$solution, but how to control that the solution doesn't blowup in the $H^2$norm?

5$\begingroup$ Could you change your title to something more descriptive, As it is now, your title is useless, and would describe a nontrivial amount of the questions here. Imagine a world where every fifth question here is titled "Help in understanding a statement in a paper" $\endgroup$ – Paul Plummer May 27 '15 at 15:56

3$\begingroup$ I took a stab at a better title. Feel free to change it if there is something else you prefer. I also corrected the spelling of "Schrödinger". $\endgroup$ – Nate Eldredge May 27 '15 at 16:35
What you need is "persistence of regularity".
As you said, data in $H^1$ lead to global solutions that remain bounded in $H^1$ by conservation of mass and energy.
If you take data in $H^2$, say, you just need to check that the solution does not blow up its $H^2$norm in finite time. (Of course, it may be the case that $\lim_{t\to\infty} \u(t)\_{H^2} = \infty$.)
To do this, you can run a bootstraptype argument using Strichartz estimates. For example, on an interval $[0,T]$ you could use the $L_t^2 L_x^{6/5}$ endpoint and Sobolev embedding to estimate
\begin{align*} \u\_{L_t^\infty H_x^2} \lesssim \u_0\_{H^2} + T^{1/2}\u\_{L_t^\infty H_x^1}^2 \u\_{L_t^\infty H_x^2}. \end{align*}
Choosing $T$ sufficiently small depending on $\u\_{L_t^\infty H_x^1}$ (which is bounded purely in terms of the mass and energy), you can deduce that $$ \u\_{L_t^\infty H_x^2([0,T]\times\mathbb{R}^3)} \lesssim 2\u_0\_{H^2}. $$ In particular, $$ \u(T)\_{H^2} \lesssim 2 \u_0\_{H^2}. $$
Now you can run the same argument on $[T,2T]$. In particular, you see that the $H^2$norm at most doubles on intervals of length $T$. Thus $u$ remains in $H^2$ throughout its lifespan (although the $H^2$norm may blow up in "infinite time").
Modifications of this type of argument also give continuity in $H^2$.