1: \begin{abstract}

2: We study the fractal pointwise convergence for the equation 

3: $i\hbar\partial_tu + P(D)u = 0$, where the symbol $P$ is real, homogeneous and non-singular.

4: We prove that for initial data $f\in H^s(\R^n)$ with $s>(n-\alpha+1)/2$

5: the solution $u$ converges to $f$ $\mathcal{H}^\alpha$-a.e, where

6: $\mathcal{H}^\alpha$ is the $\alpha$-dimensional Hausdorff measure.

7: We improve upon this result depending on the dispersive strength of $P$.

8: On the other hand, for a family of polynomials $P$ and given $\alpha$, 

9: we exploit a Talbot-like effect 

10: to construct initial data whose solutions $u$ diverge 

11: in sets of Hausdorff dimension $\alpha$. 

12: To compute the dimension of the sets of divergence, 

13: we adopt the Mass Transference Principle from Diophantine approximation.

14: We also construct counterexamples for quadratic symbols like the saddle

15: to show that 

16: our positive results are sometimes best possible.

17: \end{abstract}

18: