Extension and averaging operators for finite fields

Using the definition of the inverse Fourier transform and the orthogonality relations of the nontrivial additive character \chi of {\mathbb F_q}, we see

Use the complete square method (?), compute the sums over x_j\in {\mathbb F_q} and then obtain that

Case I. Suppose that d\geq 3 is odd. Then \eta^d\equiv \eta, because \eta is the multiplicative character of order two. Therefore, if \frac{m_1^2}{a_1}+\cdots+\frac{m_d^2}{a_d}=0, then the proof is complete, because \sum_{s\in {\mathbb F_q}\setminus \{0\}}\eta(s)=0. On the other hand, if \frac{m_1^2}{a_1}+\cdots+\frac{m_d^2}{a_d}\neq 0, then the statement follows from using a change of variable,-\frac{1}{4s}\left(\frac{m_1^2}{a_1}+\cdots+\frac{m_d^2}{a_d}\right) \to s, and the facts that \eta(4)=1, \eta(s)=\eta(s^{-1}) for s\neq 0, and G_1=\sum_{s\neq 0} \eta(s)\chi(s).
Case II. Suppose that d\geq 2 is even. Then \eta^{d}\equiv 1. The proof is complete, because \sum_{s\neq 0} \chi(as)=-1 for all a\neq 0, and \sum_{s\neq 0} \chi(as)=(q-1) if a=0.

Recall that the Fourier inverse transform of the surface measure d\sigma is given by the relation

where m\in ({\mathbb F_q^d},dm). Therefore, it is clear that (d\sigma)^\vee(0,\ldots,0)=1 for all d\geq 2. If we compare this with the values (d\sigma)^\vee(0,\ldots,0) given by Lemma 4.1, then we see that |S|\sim q^{d-1} for d\geq 3. Since the absolute of the Gauss sum G_1 is exactly q^{1/2}, the statements in Corollary 4.2 follows immediately from Lemma 4.1.

By duality, it suffices to prove that the following restriction estimate holds: for every function g defined on ({\mathbb F_q^d},dm), we have

By the orthogonality principle and Hölder’s inequality, we see

It therefore suffices to show that for every function g on ({\mathbb F_q^d},dm),

Define K=(d\sigma)^\vee -\delta_0. Since (d\sigma)^\vee(0,\dots,0)=1, we see that K(m)=0 if m=(0,\dots,0), and K(m)=(d\sigma)^\vee(m) if m\in {\mathbb F_q^d}\setminus \{(0,\dots,0)\}. It follows that

where the inequality follows from the fact that dm is the counting measure and \frac{2(\alpha +2)}{\alpha}\geq \frac{2(\alpha +2)}{\alpha+4}. Thus, it is enough to show that for every g on ({\mathbb F_q^d},dm),

We now claim that the following two estimates hold: for every function g on ({\mathbb F_q^d},dm),

and

Note that the estimate (Proof) follows by interpolating (Proof) and (Proof). It therefore remains to show that both (Proof) and (Proof) hold. Using Plancherel, the inequality (Proof) follows from the following observation:

where the last line is due to the observation that for each x\in ({\mathbb F_q^d},dx)
\widehat{K}(x)= d\sigma(x)-\widehat{\delta_0}(x)=q^d|S|^{-1} S(x)-1 \lesssim q. On the other hand, the estimate (Proof) follows from Young’s inequality and the assumption on the Fourier decay estimates away from the origin. Thus, the proof is complete.

Consider a function K on ({\mathbb F_q^d}, dm) defined as K=(d\sigma)^\vee -\delta_0. We want to prove that for every function f on ({\mathbb F_q^d},dx),

Since d\sigma= \widehat{K}+\widehat{\delta_0}= \widehat{K}+1 and \|f\ast 1\|_{L^{\alpha+2}({\mathbb F_q^d},dx)}\lesssim \|f\|_{L^{\frac{\alpha+2}{\alpha+1}}({\mathbb F_q^d},dx)}, it suffices to show that for every f on ({\mathbb F_q^d},dx),

Notice that this can be done by interpolating the following two estimates:

and

The inequality (Proof) follows from the Plancherel theorem, the size assumption of |(d\sigma)^\vee|, and the definition of K. On the other hand, the inequality (Proof) follows from Young’s inequality and the observation that \|\widehat{K}\|_{L^\infty({\mathbb F_q^d},dx)}\lesssim q. Thus, the proof of Lemma 6.1 is complete.

It is clear from the Plancherel theorem that

It therefore remains to show that

Since the space ({\mathbb F_q^d},dx) is isomorphic to its dual space ({\mathbb F_{q}^d},dm) as an abstract group, we may identify the space ({\mathbb F_q^d},dx) with the dual space ({\mathbb F_{q}^d},dm). Thus, they possess same algebraic structures. Recall that we have endowed them with different measures: the counting measure dm for ({\mathbb F_q^d},dm) and the normalized counting measure dx for ({\mathbb F_{q}^d},dx). For these reasons, the inequality (Proof) is essentially same as the following: for every subset E of ({\mathbb F_{q}^d},dm)

where S is considered as

By duality (?), the statement (2.1) in Theorem 2.1 implies that the following restriction estimate holds: for every function g on ({\mathbb F_q^d},dm),

If we take g(m)=E(m), then we have

Since |S|\sim q^{d-1}, (Proof) holds and the proof is complete.