Maximizers for the Stein–Tomas Inequality

Abstract

Abstract. We give a necessary and sufficient condition for the precompactness of all optimizing sequences for the Stein-Tomas inequality. In particular, if a wellknown conjecture about the optimal constant in the Strichartz inequality is true, we obtain the existence of an optimizer in the Stein-Tomas inequality. Our result is valid in any dimension. Main resultA fundamental result in harmonic analysis is the Stein-Tomas theorem [30,36], which states that if f ∈ L 2 (S N −1 ), N ≥ 2, then the inverse Fourier transformf of f dω, with dω the surface measure on S N −1 , that is, and its L q (R N ) norm is bounded by a constant times the L 2 (S N −1 ) norm of f . Moreover, it is well known that the exponent q is optimal (smallest possible) for this to hold for any f ∈ L 2 (S N −1 ). In this paper we are interested in the optimal Stein-Tomas constant,where · denotes the norm in L 2 (S N −1 ). The value of R N and optimizing functions are only known in dimension N = 3 due to a remarkable work of Foschi [15]; see [11] for partial progress in N = 2. Our main concern here is whether the supremum defining R N is attained and, more generally, the description of maximizing sequences for R N . These questions were recently considered in fundamental papers by Christ and Shao, where the existence of a maximizer for N = 3 [12] and N = 2 [29] was shown, as well as a precompactness result for maximizing sequences for N = 3 [13]. What makes dimensions N = 2 and 3 special is that the exponent q in (1.1) is an even integer, so that one can multiply out |f | q . Our results will be valid in any dimension.c 2016 by the authors. This paper may be reproduced, in its entirety, for non-commercial purposes. Christ and Shao discovered that for the problem of existence of an maximizer for R N a key role is played by the Strichartz inequality [32]. The optimal constant in this inequality isThe overall factor (2π) −(d+2)/d and the factor 1/2 in front of the Laplacian are not important, but simplify some formulas below. We say that a sequence (The following is our main result.up to modulations and, in particular, there is a maximizer for R N .Clearly, the optimization problem for R N is invariant under modulations, so precompactness up to modulations is the best one can expect. Our theorem says that assumption (1.2) is sufficient for this. In fact, it is easy to see that (1.2) is also necessary for the precompactness modulo modulations of all maximizing sequences. We will comment on this in Remark 2.5, where we will also see that (1.2) holds with ≥ instead of >.As we will argue below, in dimensions N = 2 and N = 3, the strict inequality (1.2) holds and so we recover the Christ-Shao results on the existence of optimizers [12,29] and precompactness in N = 3 [13] and we obtain, for the first time, precompactness of maximizing sequences for N = 2.We believe, but cannot prove, that the strict inequality (1.2) holds in any dimension. To verify it, it seems natural to first compute S N −1 and then to use a perturbation argument to establish (1.2). In ...

Sharp Strichartz inequalities for fractional and higher-order Schrödinger equations

“…This variant was already observed in [33,Proposition 2.31] for the case of the cone, and the proof follows similar lines to that of [13, Proposition 1.1]. Note that the function Θ may depend on the sequence tf n u, but not on n. The following proof is inspired by [19,Proposition 2.2].…”

“…Various results of a similar flavour to that of Theorem 1.3 have appeared in the recent literature. They are typically derived from a sophisticated application of concentrationcompactness techniques [9,39], a full profile decomposition [24,25,38], or the missing mass method as in [19,20]. We introduce a new variant which follows the spirit of the celebrated works of Lieb [4,27] and Lions [28,29].…”

“…Odd curves are of independent interest, in particular because a new phenomenon emerges: caps centered around points with parallel tangents interact strongly, regardless of separation between the points. This mechanism was discovered in [9], and further explored in [5,17,19,20,39]. Some of these works include a symmetrization step which relies on the convolution structure of the underlying inequality.…”

Extremizability of Fourier restriction to the paraboloid

“…Applying a symmetry, we then have an extremizing sequence in L p ξ that is nearly bounded with compact support, which means that it is almost in L 2 ξ . (This part of the argument is closely related to Lieb's method of missing mass [18]; see also [11].) Truncating introduces some error, but lets us apply the L 2 ξ -based profile decomposition.…”

Extremizers for Fourier restriction on hyperboloids

“…Several authors have investigated the interface between bilinear restriction theory and these extremal questions, both from the restriction side and the partial differential equations point of view. Here we mention the works [1,2,5,6,7,10,14], all of which deal with these connections. Many other authors have contributed to the development of the area, and we refer the reader to [3] for an exposition of related literature on sharp Fourier restriction theory.…”