Averages over curves with torsion (Q1297507)

This paper is concerned with the problem of establishing the correct amount of smoothing in \(L^p\) for the operation of convolution with a smooth measure on the helix \(\{ (\cos t, \sin t, t) \}\) in \(\mathbb R^3\). In \(L^2\) one gains \(1/3\) degree of regularity, while in \(L^\infty\) no regularity is obtained. This trivially gives \(1/6\) smoothing in \(L^4\), but it is conjectured [\textit{D. Oberlin} and \textit{H. F. Smith}, Proc. Am. Math. Soc. 127, No. 10, 2911-2915 (1999; preceding review)] that this can be improved to \(1/4\). This conjecture, together with what can be obtained by duality and interpolation, was shown to be sharp in the paper just mentioned. In this paper the authors show that the sharp conjecture would follow from the optimal square function estimate for cone multipliers. The ``standard'' square function estimate on the cone due to Mockenhaupt loses a factor of \(N^{1/4}\) at frequency \(N\), which corresponds to the \(1/6\) smoothing result, however any improvement on Mockenhaupt's estimate leads to a new result on the helix smoothing problem. This estimate has been improved by several authors, including Bourgain, Tao, and Vargas; the best result to date (Feb. 2000) is due to Wolff, and loses \(N^{1/4 - 1/44 + \varepsilon}\), which corresponds to a smoothing of \(1/6 + 1/132 - \varepsilon\) for the helix.