Counting Lattice Point and a failure of Strichartz estimate on periodic domain

60 Views Asked by At

My question comes from the expression (4.48) of the textbook of Linares and Ponce which states that the direct imitation of Strichartz estimate for $(t,x) \in [0,2\pi]^2$ is not true.

I don't know how to show that counterexample:

There is a constant $c>0$ such that for all large enough integers $N$, $\| \sum_{k=1}^N e^{i (k^2 t + kx)}\|^6_{L^6_{t,x}} \geq c N^3 \log N$.

A direct computation shows that this problem is equivalent to show the following problem of counting lattice point has a lower bound $c N^3 \log N$:

$ \#\{(k_1,j_1,k_2,j_2,k_3,j_3) \in \mathbb{N}^6_{\leq N}: k_1-j_1+k_2-j_2+k_3-j_3 = k_1^2 - j_1^2 + k_2^2 - j_2^2 + k_3^2 - j_3^2 = 0 \} \geq cN^3 \log N $.

My attempt

For each fixed $(j_1,j_2,j_3)$ (maybe each $j_i \geq N/2$ is needed),

I conjecture that the number of lattice points on intersection of the sphere $k_1^2+k_2^2+k_3^2 = j_1^2 + j_2^2+ j_3^2$ and the plane $k_1+k_2+k_3 = j_1 + j_2+ j_3$ is greater than some thing of order $\log N$. But I don't know how to show this.

Many thanks for any discussion or idea!

1

There are 1 best solutions below

0
On BEST ANSWER

I find out the above $L^6$ inequality is answered in the notes of Tadairo Oh (maybe earlier in a classical paper of Bouragin ) by a method of computing the Weyl sum $\sum_{n =1}^N e^{i (nx + n^2 t)}$ (sometimes called the exponential sum).

So we can prove that inequality without solving the counting lattice problem I post.

Conversely, I learn that some counting lattice problem can be solved by using suitable Weyl sum.