Restriction estimates for space curves with respect to general measures (Q2445944)

The authors consider the adjoint of the Fourier restriction operator to the curve \(\gamma(t), t \in I = [0,1]\), which is an oscillatory integral, \[ T_{\lambda}^{\gamma} (f)(x) = \int e^{-i \lambda x\cdot \gamma(t)} f(t) \, dt \] and conditions under which it satisfies the estimate \[ \| T_{\lambda}^{\gamma} (f) \|_{L^q(d \nu)} \leq C \lambda^{- \beta} \| f \|_{L^p(I)}, \] where \(\nu\) is a measure on \(\mathbb{R}^d\). They first discuss the situation in which \(\nu\) is the Lebesgue measure, where the estimate is known to be related to curvature conditions on \(\gamma\). Under the condition \[ \det(\gamma^{\prime}(t), \gamma^{\prime \prime}(t), \dots, \gamma^{(d)}(t) )\neq 0, \] where \(d\) is the dimension of the range of \(\gamma\), the estimate holds with \(\beta = d/q\) if \[ \frac{ d(d + 1) }{2q} + \frac1{p} \leq 1 \text{\;and \;} q > \frac{d^2 + d + 2}{2}. \] They discuss the history of this estimate and remark that it can be generalized by replacing \(x \cdot \gamma(t)\) by \(\phi(x,t)\) and the curvature condition by \[ \det(\partial_t \nabla_x(\phi(x,t), \partial_t^2 \nabla_x(\phi(x,t), \dots, \partial_t^d \nabla_x(\phi(x,t) ) \neq 0. \] The authors consider the case in which the Lebesgue measure is replaced by a measure \(\mu\) where for \(0< \alpha \leq d\), \[ \mu(B(x, r)) \leq C_{\mu} r^{\alpha}, \] where the constant \(C_{\mu}\) depends on \(\mu\), but not on \(x\) and \(r\). They remark that the corresponding inequality must have \(\beta\) replaced by \(\frac{\alpha}{q}\), and consider the range of \((p, q)\) for which the estimate \[ \| T_{\lambda}^{\gamma} (f) \|_{L^q(d \nu)} \leq C \lambda^{- \frac{\alpha}{q} } \| f \|_{L^p(I)} \] holds. The result is quite complicated. Let \(\beta = \beta(\alpha)\) be defined by \[ \beta(\alpha) = (j + 1) \alpha + \frac{ (d - j - 1)(d - j)}{2} \] if \(d - j - 1 < \alpha \leq d - j\), for \(j = 0, \dots, d - 1\). \(\beta\) increases as \(\alpha\) increases. Their main result is that if \(\mu\) is as above, and \(\gamma\) satisfies the curvature condition, then if \(\frac{d}{q} \leq (1 - \frac{1}{p}), q \geq 2d\), and \[ \frac{\beta(\alpha) }{q} + \frac{1}{p} < 1, \;\;q > \beta(\alpha) + 1, \] the estimate \[ \| T_{\lambda}^{\gamma} (f) \|_{L^q(d \nu)} \leq C \lambda^{- \frac{\alpha}{q} } \| f \|_{L^p(I)} \] holds if \(\lambda \geq 1\). This is the main result of the paper, but they also make remarks on the critical case \[ \frac{\beta(\alpha) }{q} + \frac{1}{p} = 1 \] and prove a result on a degenerate case in which the determinant of some collection of derivatives other than the first \(d\) is nonzero. The result in that case is similar to the above with more restrictions on \((p,q)\) if appropriate weights are introduced.