Assume that
\(n\geq2\),
\(\phi_{i}\)
\((i=1,2,3,\ldots ,n)\) satisfies (H1)–(H3). For
\(i=1,2,3,\ldots ,n\), let
\(t_{i}(x)\) be a measurable function on
\(\mathbb{R}^{n}\) with
\(t_{i}(x)\in\mathbb{R}\). Denote
\(t(x)=(t_{1}(x),t_{2}(x),\ldots,t_{n}(x))\), we set
$$\begin{aligned} S_{t(x),\varPhi}f(x)={}&(2\pi)^{-n} \int_{\mathbb{R}^{n}} e^{ix\cdot\xi}e^{i(t_{1}(x) \phi_{1}(|\xi_{1}|)+t_{2}(x)\phi_{2}(|\xi_{2}|)+ \cdots+t_{n}(x)\phi_{n}(|\xi_{n}|))}\hat{f}(\xi)\,d\xi,\\ & x\in\mathbb{R}^{n}, f\in\mathcal{S}\bigl(\mathbb{R}^{n}\bigr). \end{aligned}$$
For
\(4\leq q<\infty\) and
\(s=n(\frac{1}{2}-\frac{1}{q})\), that is,
\(\frac{n}{4}\leq s<\frac{n}{2}\) and
\(q=\frac{2n}{n-2s}\). By linearizing the maximal operator (see [
30]) to prove the global estimate (
1.8) holds, it suffices to show that
$$ \Vert S_{t(x),\varPhi}f \Vert _{L^{q}(\mathbb{R}^{n})}\leq C \Vert f \Vert _{\dot {H}^{s}}=C \biggl( \int_{\mathbb{R}^{n}} \vert \xi \vert ^{2s} \bigl\vert \hat{f}(\xi) \bigr\vert ^{2}\,d\xi \biggr)^{1/2}. $$
(2.1)
To prove (
2.1) it suffices to prove that
$$ \Vert S_{t(x),\varPhi}f \Vert _{L^{q}(\mathbb{R}^{n})}\leq C \biggl( \int_{\mathbb{R}^{n}} \vert \xi_{1} \vert ^{\frac{2s}{n}} \vert \xi_{2} \vert ^{\frac{2s}{n}} |\cdots \vert \xi_{n} \vert ^{\frac{2s}{n}} \bigl\vert \hat{f}(\xi) \bigr\vert ^{2} \,d\xi \biggr)^{1/2}. $$
(2.2)
Let
\(g(\xi)=|\xi_{1}|^{\frac{s}{n}}|\xi_{2}|^{\frac{s}{n}} \cdots|\xi_{n}|^{\frac{s}{n}}\hat{f}(\xi)\), then we have
$$\begin{aligned} S_{t(x),\varPhi}f(x)={}& \int_{\mathbb{R}^{n}} e^{ix\cdot\xi}e^{i(t_{1}(x) \phi_{1}( \vert \xi_{1} \vert )+t_{2}(x)\phi_{2}( \vert \xi_{2} \vert )+ \cdots+t_{n}(x)\phi_{n}( \vert \xi_{n} \vert ))} \vert \xi_{1} \vert ^{-\frac{s}{n}} \vert \xi_{2} \vert ^{-\frac{s}{n}} \cdots \vert \xi_{n} \vert ^{-\frac{s}{n}}g(\xi)\,d \xi \\ ={}&R_{\varPhi}g(x), \end{aligned}$$
(2.3)
where
$$R_{\varPhi}g(x)= \int_{\mathbb{R}^{n}} e^{ix\cdot\xi}e^{i(t_{1}(x) \phi_{1}( \vert \xi_{1} \vert )+t_{2}(x)\phi_{2}( \vert \xi_{2} \vert )+ \cdots+t_{n}(x)\phi_{n}( \vert \xi_{n} \vert ))} \vert \xi_{1} \vert ^{-\frac{s}{n}} \vert \xi_{2} \vert ^{-\frac{s}{n}} \cdots \vert \xi_{n} \vert ^{-\frac{s}{n}}g(\xi)\,d \xi. $$
To prove (
2.2) it suffices to prove that
$$ \Vert R_{\varPhi}g \Vert _{L^{q}(\mathbb{R}^{n})}\leq C \Vert g \Vert _{L^{2}(\mathbb{R}^{n})} $$
(2.4)
for
g continuous and rapidly decreasing at infinity. We take a real-valued function
\(\rho\in C_{0}^{\infty}(\mathbb{R}^{n})\) such that
\(\rho(x)=1\) if
\(|x|\leq1\) and
\(\rho(x)=0\) if
\(|x|\geq2\). And we choose a real-valued function
\(\psi\in C_{0}^{\infty}(\mathbb{R})\) such that
\(\psi(x)=1\) if
\(|x|\leq1\) and
\(\psi(x)=0\) if
\(|x|\geq2\), and set
\(\sigma(\xi )=\psi(\xi_{1})\psi(\xi_{2}) \cdots\psi(\xi_{n})\). For
\(\xi\in\mathbb{R}^{n}\) and
\(N=1,2,3,\ldots\) , we set
\(\rho_{N}(x)=\rho(\frac{x}{N})\) and
\(\sigma_{N}(\xi)=\sigma(\frac{\xi}{N})\). For
\(x\in\mathbb{R}^{n}\),
\(g\in L^{2}(\mathbb{R}^{n})\), and for
\(N=1,2,3,\ldots\) , we define
$$\begin{aligned} R_{N,\varPhi}g(x)={}&\rho_{N}(x) \int_{\mathbb{R}^{n}} e^{ix\cdot\xi}e^{i(t_{1}(x) \phi_{1}( \vert \xi_{1} \vert )+t_{2}(x)\phi_{2}( \vert \xi_{2} \vert )+ \cdots+t_{n}(x)\phi_{n}( \vert \xi_{n} \vert ))} \vert \xi_{1} \vert ^{-\frac{s}{n}} \vert \xi_{2} \vert ^{-\frac{s}{n}} \cdots\\ &{}\times \vert \xi_{n} \vert ^{-\frac{s}{n}} \sigma_{N}(\xi)g(\xi)\,d\xi. \end{aligned}$$
The adjoint of
\(R_{N,\varPhi}\) is given by
$$\begin{aligned} R^{\prime}_{N,\varPhi}h(\xi)={}&\sigma_{N}(\xi) \vert \xi_{1} \vert ^{-\frac{s}{n}} \vert \xi_{2} \vert ^{-\frac{s}{n}} \cdots\\ &{}\times \vert \xi_{n} \vert ^{-\frac{s}{n}} \int_{\mathbb{R}^{2}} e^{-ix\cdot\xi}e^{-i(t_{1}(x) \phi_{1}( \vert \xi_{1} \vert )+t_{2}(x)\phi_{2}( \vert \xi_{2} \vert )+ \cdots+t_{n}(x)\phi_{n}( \vert \xi_{n} \vert ))} \rho_{N}(x)h(x)\,dx, \end{aligned}$$
where
\(\xi\in\mathbb{R}^{n}\) and
\(h\in L^{2}(\mathbb{R}^{n})\). To prove (
2.4) it is sufficient to prove that
$$ \Vert R_{N,\varPhi}g \Vert _{L^{q}(\mathbb{R}^{n})}\leq C \Vert g \Vert _{L^{2}(\mathbb{R}^{n})}. $$
(2.5)
By duality, to prove (
2.5) it suffices to show that
$$ \bigl\Vert R_{N,\varPhi}^{\prime}h \bigr\Vert _{L^{2}(\mathbb{R}^{n})}\leq C \Vert h \Vert _{L^{q^{\prime}}(\mathbb{R}^{n})}, $$
(2.6)
where
\(\frac{1}{q}+\frac{1}{q^{\prime}}=1\). Thus, we have
$$ \bigl\Vert R_{N,\varPhi}^{\prime}h \bigr\Vert _{L^{2}(\mathbb{R}^{n})}^{2} = \int \bigl\vert R^{\prime}_{N,\varPhi}h(\xi) \bigr\vert ^{2}\,d\xi = \int_{\mathbb{R}^{n}} \int_{\mathbb{R}^{n}} K_{N}(x,y)\rho_{N}(x) \rho_{N}(y)h(x)\overline{h(y)}\,dx\,dy, $$
(2.7)
where
$$ K_{N}(x,y)=K^{1}_{N}(x,y) K^{2}_{N}(x,y)\cdots K^{n}_{N}(x,y) $$
(2.8)
and
$$ K^{i}_{N}(x,y)= \int_{\mathbb{R}} \vert \xi_{i} \vert ^{-\frac{2s}{n}} e^{i(y_{i}-x_{i})\xi_{i}} e^{i(t_{i}(y)-t_{i}(x))\phi( \vert \xi_{i} \vert )} \psi_{N}(\xi_{i})^{2} \,d\xi_{i}, $$
(2.9)
where
\(i=1,2,\ldots,n\) and
\(N=1,2,\ldots\) . Since
\(\frac{n}{4}\leq s <\frac{n}{2}\), we have
\(\frac{1}{2}\leq \frac{2s}{n}<1\). Therefore, by Lemma
1.3, (2.9), and (2.8), we obtain
$$ \bigl\vert K_{N}(x,y) \bigr\vert \leq C \frac{1}{ \vert x_{1}-y_{1} \vert ^{1-\frac{2s}{n}}} \frac{1}{ \vert x_{2}-y_{2} \vert ^{1-\frac{2s}{n}}} \cdots\frac{1}{ \vert x_{n}-y_{n} \vert ^{1-\frac{2s}{n}}}. $$
(2.10)
We define
$$P_{i}f(x_{1},x_{2},\ldots, x_{n})= \int_{\mathbb{R}} \frac{1}{ \vert x_{i}-y_{i} \vert ^{1-\frac{2s}{n}}} f(x_{1}, \ldots,x_{i-1},y_{i}, x_{i+1},\ldots,x_{n}) \,dy_{i}, $$
\(i=1,2,\ldots,n\). Thus, by (2.7) and (
2.10), we obtain
$$\begin{aligned} &\int \bigl\vert R_{N,\varPhi}^{\prime}h(\xi) \bigr\vert ^{2}\,d\xi \\ &\quad \leq C \int_{\mathbb{R}^{n}} \int_{\mathbb{R}^{n}} \frac{1}{ \vert x_{1}-y_{1} \vert ^{1-\frac{2s}{n}}} \frac{1}{ \vert x_{2}-y_{2} \vert ^{1-\frac{2s}{n}}} \cdots \frac{1}{ \vert x_{n}-y_{n} \vert ^{1-\frac{2s}{n}}} \bigl\vert h(x) \bigr\vert \bigl\vert h(y) \bigr\vert \,dx\,dy \\ &\quad = C \int_{\mathbb{R}^{n}} \biggl( \int_{\mathbb{R}} \int_{\mathbb{R}} \int_{\mathbb{R}} \int_{\mathbb{R}}\frac{1}{ \vert x_{n}-y_{n} \vert ^{1-\frac{2s}{n}}} \frac{1}{ \vert x_{n-1}-y_{n-1} \vert ^{1-\frac{2s}{n}}}\cdots \frac {1}{ \vert x_{3}-y_{3} \vert ^{1-\frac{2s}{n}}} \\ &\qquad{}\times\frac {1}{ \vert x_{2}-y_{2} \vert ^{1-\frac{2s}{n}}} \biggl( \int\frac {1}{ \vert x_{1}-y_{1} \vert ^{1-\frac{2s}{n}}} \bigl\vert h(y_{1},y_{2}, \ldots, y_{n}) \bigr\vert \,dy_{1} \biggr) \,dy_{2}\,dy_{3}\cdots \,dy_{n} \biggr) \bigl\vert h(x) \bigr\vert \,dx \\ &\quad =C \int_{\mathbb{R}^{n}}P_{n}P_{n-1}\cdots P_{2}P_{1} \vert h \vert (x) \bigl\vert h(x) \bigr\vert \,dx. \end{aligned}$$
(2.11)
Invoking Hölder’s inequality, we get
$$ \int \bigl\vert R_{N,\varPhi}^{\prime}h(\xi) \bigr\vert ^{2}\,d\xi\leq C \bigl\Vert P_{n}P_{n-1}\cdots P_{2}P_{1} \vert h \vert \bigr\Vert _{L^{q}(\mathbb{R}^{n})} \Vert h \Vert _{L^{q^{\prime}}(\mathbb{R}^{n})}. $$
(2.12)
Since
\(q=\frac{2n}{n-2s}\), it follows that
\(q^{\prime}=\frac {2n}{n+2s}\) and the fact
\(\frac{1}{q}=\frac{1}{q^{\prime}}-\frac{2s}{n}\). Denote by
\(I_{\sigma}\) the Riesz potential of order
σ, which is defined by
$$I_{\sigma}(f) (u)= \int_{\mathbb{R}}\frac{f(v)}{ \vert u-v \vert ^{1-\sigma}}\,dv. $$
Applying the fact
\(I_{s}\) is bounded from
\(L^{q^{\prime}}(\mathbb{R})\) to
\(L^{q}(\mathbb{R})\), we have
$$ \biggl( \int_{\mathbb{R}} \bigl\vert P_{j}h(x) \bigr\vert ^{q}\,dx_{j} \biggr) ^{1/q}\leq C \biggl( \int_{\mathbb{R}} \bigl\vert h(x) \bigr\vert ^{q^{\prime}} \,dx_{j} \biggr) ^{1/q^{\prime}}, $$
(2.13)
where
\(j=1,2,\ldots,n\). By (
2.13) and Minkowski’s inequality, we have
$$ \bigl\Vert P_{n}P_{n-1}\cdots P_{2}P_{1} \vert h \vert \bigr\Vert _{L^{q}(\mathbb{R}^{n})} \leq C \Vert h \Vert _{L^{q^{\prime}}(\mathbb{R}^{n})}. $$
(2.14)
Therefore, (
2.6) follows from (
2.12) and (
2.14). Now we complete the proof of Theorem
1.1.