arXiv subjects
Publications and source records attributed to Michael Christ.
We establish a stable form of the general Euclidean Brascamp-Lieb inequality in all cases in which the Lebesgue exponents are strictly between 1 and 2, asserting that all near-extremizers are nearly Gaussian.
Examples are constructed of infinite-dimensional subspaces $V\subset L^2(\mu)$ with the property that for any $f,g\in V$, if $|f|$ is approximately equal to $|g|$ with respect to the $L^2$ norm, then there exists a unimodular scalar $z$ such that $f$ is approximately equal to $zg$.
Let $B$ be a ball in ${\mathbb R}^2$. For $j=1,2,3$ let $\varphi_j:B\to{\mathbb R}^1$ be real analytic submersions, and let $a_j$ be real analytic coefficient functions. To any $\varepsilon>0$ and any Lebesgue measurable functions $f_j:{\mathbb R}^1\to {\mathbb C}$ associate the sublevel set $S = S(f_1,f_2,f_3,\varepsilon) = \{x\in B: |\sum_{j=1}^3 a_j(x)(f_j\circ\varphi_j)(x)|<\varepsilon\}$. Let $S' = \{x\in S: \max_j|f_j\circ\varphi_j(x)|\ge 1\}$. Our main result is an upper bound, under certain hypotheses on the data $\varphi_j,a_j$ for the Lebesgue measure of $S'$ of the form $|S'| \le c\varepsilon^\gamma$ for some constants $c,\gamma>0$ that depend on the data $a_j,\varphi_j$ but not on the functions $f_j$ or parameter $\varepsilon$. The main hypothesis is that in any connected open subset of $B$, the only real analytic solution $(f_1,f_2,f_3)$ of $\sum_j a_j(x)(f_j\circ\varphi_j)(x)\equiv 0$ is the trivial solution $f_k=0\ \forall\,k$. Certain auxiliary hypotheses, which hold for generic $\varphi_j,a_j$, are also imposed. The case in which all coefficients $a_j$ are constant was previously known. This result is a principal ingredient in an analysis, in a companion paper, of related implicitly oscillatory integrals with four factors $f_j$. Certain related results are also discussed. In particular, a generalization to arbitrarily many summands f_j is obtained for the special case in which all mappings $\varphi_j$ are linear.
For quadrilinear functionals $\int_B \prod_{j=1}^4 (f_j\circ\varphi_j)$, where $B\subset{\mathbb R}^2$ is a ball, $\varphi_j:B\to{\mathbb R}^1$ are real analytic submersions, and $f_j\in L^\infty({\mathbb R}^1)$ are bounded and measurable, we seek a majorization of the integral by a product of negative order Sobolev norms of the factors $f_j$. An obvious necessary condition is that any smooth solution of $\sum_j (g_j\circ\varphi_j)\equiv 0$, in any connected open set, must be constant. Assuming this condition and certain auxiliary hypotheses, we establish an upper bound of the desired type. The proof relies in part on a three term sublevel set inequality established in a companion paper.
Lebesgue space bounds $L^{p_1}({\mathbb R}^1) \times L^{p_2}(^1) \to L^q({\mathbb R}^1)$ are established for certain maximal bilinear operators. The proof combines a trilinear smoothing inequality with Calder\'on-Zygmund theory. A reference to overlapping work of other authors on one observation has been added.
We prove a.e. convergence of continuous-time quadratic averages with respect to two commuting $\mathbb{R}$-actions, coming from a single jointly measurable measure-preserving $\mathbb{R}^2$-action on a probability space. The key ingredient of the proof comes from recent work on multilinear singular integrals; more specifically, from the study of a curved model for the triangular Hilbert transform.
Lebesgue space inequalities are proved for a variant of the triangular Hilbert transform involving curvature. The analysis relies on a crucial trilinear smoothing inequality developed herein, and on bounds for an anisotropic variant of the twisted paraproduct. The trilinear smoothing inequality also leads to Lebesgue space bounds for a corresponding maximal function and a quantitative nonlinear Roth-type theorem concerning patterns in the Euclidean plane.
Inequalities are established for certain trilinear scalar-valued functionals. These functionals act on measurable functions of one real variable, are defined by integration over two- or three-dimensional spaces, and are controlled in terms of Lebesgue space norms of the functions, and of negative powers of large parameters describing a degree of oscillation. Related sublevel set inequalities are a central element of the analysis. The main results and the main lines of their proofs are largely unchanged in this draft, but some details have been corrected. The analysis has already been extended in work of the author, Durcik, and Roos.
We prove a sharp, global-in-time Strichartz estimate for the Schr\"odinger equation on the cylinder $\mathbb{R}\times\mathbb{T}$.
An inequality of Brascamp-Lieb-Luttinger and of Rogers states that among subsets of Euclidean space $\mathbb{R}^d$ of specified Lebesgue measures, balls centered at the origin are maximizers of certain functionals defined by multidimensional integrals. For $d>1$, this inequality only applies to functionals invariant under a diagonal action of $\text{Sl}(d)$. We investigate functionals of this type, and their maximizers, in perhaps the simplest situation in which $\text{Sl}(d)$ invariance does not hold. Assuming a more limited symmetry involving dilations but not rotations, we show under natural hypotheses that maximizers exist, and moreover, that there exist distinguished maximizers whose structure reflects this limited symmetry. For small perturbations of the $\text{Sl}(d)$--invariant framework we show that these distinguished maximizers are strongly convex sets with infinitely differentiable boundaries. It is shown that maximizers fail to exist for certain arbitrarily small perturbations of $\text{Sl}(d)$--invariant structures.
A version of the Riesz-Sobolev convolution inequality is formulated and proved for arbitrary compact connected Abelian groups. Maximizers are characterized and a quantitative stability theorem is proved, under natural hypotheses. A corresponding stability theorem for sets whose sumset has nearly minimal measure is also proved, sharpening recent results of other authors. For the special case of the group $\mathbb{R}/\mathbb{Z}$, a continuous deformation of sets is developed, under which an appropriately scaled Riesz-Sobolev functional is shown to be nondecreasing.
A symmetrization inequality of Rogers and of Brascamp-Lieb-Luttinger states that for a certain class of multilinear integral expressions, among tuples of sets of prescribed Lebesgue measures, tuples of balls centered at the origin are among the maximizers. Under natural hypotheses, we characterize all maximizing tuples for these inequalities for dimensions strictly greater than 1. We establish a sharpened form of the inequality.
We establish optimal Lebesgue estimates for a class of generalized Radon transforms defined by averaging functions along polynomial-like curves. The presence of an essentially optimal weight allows us to prove uniform estimates, wherein the Lebesgue exponents are completely independent of the curves and the operator norms depend only on the polynomial degree. Moreover, our weighted estimates possess rather strong diffeomorphism invariance properties, allowing us to obtain uniform bounds for averages on curves satisfying a natural nilpotency hypothesis.
An inequality of Brascamp-Lieb-Luttinger generalizes the Riesz-Sobolev inequality, stating that certain multilinear functionals, acting on nonnegative functions of one real variable with prescribed distribution functions, are maximized when these functions are symmetrized. It is shown that under certain hypotheses, when the functions are indicator functions of sets of prescribed measures, then up to the natural translation symmetries of the inequality, the maximum is attained only by intervals centered at the origin. Moreover, a quantitative form of this uniqueness is established, sharpening the inequality. The hypotheses include an auxiliary genericity assumption which may not be necessary.
Young's convolution inequality provides an upper bound for the convolution of functions in terms of $L^p$ norms. It is known that for certain groups, including Heisenberg groups, the optimal constant in this inequality is equal to that for Euclidean space of the same topological dimension, yet no extremizing functions exist. For Heisenberg groups we characterize ordered triples of functions that nearly extremize the inequality. The analysis relies on a characterization of approximate solutions of a certain class of functional equations. A result of this type is developed for a class of such equations.
The Riesz-Sobolev inequality provides an upper bound, in integral form, for the convolution of indicator functions of subsets of Euclidean space. We formulate and prove a sharper form of the inequality. This can be equivalently phrased as a stability result, quantifying an inverse theorem of Burchard that characterizes cases of equality.
A bilinear inequality of Geba, Greenleaf, Iosevich, Palsson, and Sawyer for the Fourier transform is shown to be equivalent to a simpler linear inequality, and the range of exponents is extended. Related mixed-norm inequalities are discussed.
A set subset of Euclidean space whose indicator function has maximal Gowers norm, among all sets of equal measure, is an ellipsoid up to Lebesgue null sets. If the indicator function has nearly maximal Gowers norm then the set nearly coincides with an ellipsoid.