The discretized sum-product and projection theorems (Q628365): Difference between revisions

The author proves: Given \(0<\sigma<1\) and \(\kappa>0\), there exist \(\varepsilon_0>0\) and \(\varepsilon_1>0\) such that the following holds for \(\delta>0\) sufficiently small. Let \(\mu\) be a probability measure on \([0,1]\) satisfying \[ \mu(I)<\rho^\kappa \] whenever \(I\) is a size-\(\rho\) interval such that \(\delta<\rho<\delta^{\varepsilon_0}\). Let \(A\subset [1,2]\) be a discrete set consisting of \(\delta\)-separated points \(| A|>\delta^{-\sigma}\) such that also \[ | A\cap I|<\rho^\kappa| A| \] whenever \(I\) is a size-\(\rho\) interval such that \(\delta<\rho<\delta^{\varepsilon_0}\). Then there exists \(x\in\text{supp}\mu\) such that \[ N(A+xA,\delta)>\delta^{-\varepsilon_1}| A|, \] where \(N(X,\delta)\) denotes the covering number of the set \(X\) by \(\delta\)-balls. Given \(0<\alpha<2\), \(\beta>0\), \(\kappa>0\), there exist \(\tau_0>0\) and \(\eta>\frac \alpha 2\) such that the following holds. Let \(\mu_1\) be a probability measure on \(S^1\) such that \[ \max_\theta \mu_1([\theta-\rho,\theta+\rho])<C\rho^\kappa. \] Let \(\delta>0\) be chosen sufficiently small and let \( {\mathcal A}\subset [1,2]\times [1,2]\) be a union of size-\(\delta\) squares satisfying \[ | {\mathcal A}|=\delta^{2-\alpha} \] and \[ \max_x | {\mathcal A}\cap {\mathcal B}(x,\rho)|<\delta^\beta| {\mathcal A}|\text{ for }\delta <\rho<\delta^{\tau_0}. \] Then there exists \(\theta\in \text{supp}\mu_1\) such that \[ | \pi_\theta({\mathcal A})|>\delta^{1-\eta}, \] where \(\pi_\theta\) denotes the orthogonal projection on the line \(y=(\tan\theta)x\). Given \(0<\alpha<2\), \(\kappa>0\), there exists \(\eta>\frac \alpha 2\) such that, if \({\mathcal A} \subset {\mathbb R}\times {\mathbb R}\) is a set of Hausdorff dimension \[ H-\dim {\mathcal A}>\alpha, \] then \[ H-\dim \pi_\theta({\mathcal A})\geq \eta \] for all \(\theta\in S^1\) except in an exceptional set \(E\) satisfying \[ H-\dim E\leq \kappa. \] Let \({\mathbb P}^k\) be the \(k\)-dimensional projective space and for \(\bar{y}\in {\mathbb P}^k,\rho>0\), let \[ V(y^{\perp},\rho)=\left\{\bar{x}\in {\mathbb P}^k:\frac{|\langle x,y\rangle|}{| x|\cdot| y|}<\rho\right\} . \] For \(d\geq 2\), given \(0<\alpha<d\) and \(\kappa>0\), there exist \(\tau_0\), \(\varepsilon>0\), and \(\eta>\frac{\alpha}{d}\) such that the following holds for small \(\delta>0\): Let \(\mu_1\) be a probability measure on \({\mathbb P}^{d-1}\) such that \[ \max_{\bar{y}}\mu_1(V(y^{\perp},\rho))<\rho^k\,\text{if}\,\delta<\rho<\delta^{\tau_0}. \] Let \(E\subset [0,1]^d\) be a union of size-\(\delta\) balls and satisfying \[ | E|>\delta^{d-\alpha} \] and a non-concentration property \[ \max_x | E\cap B(x,\rho)|<\rho^k| E|\text{ if }\delta <\rho<\delta^{\kappa}. \] Then there exist \(D\subset {\mathbb P}^{d-1}\) and \(E'\subset E\) with the following properties: \[ \eta(D)>1-\delta^\varepsilon, \] \[ | E'| >\delta^\varepsilon| E|, \] and \[ |\pi_\theta(E'') |>\delta^{1-\eta} \] whenever \(\theta\in D\) and \(E''\subset E'\) satisfies \[ | E''|>\delta^{2\varepsilon}| E|. \] For all \(\kappa>0\) and \(C>1\), there exist \(k\in{\mathbb Z}_+\) and \(C'>1\) such that the following holds. Let \(\mu\) be a probability measure on \([\frac 12,1]\) satisfying, for \(\delta<\rho<1\), \[ \max_a\mu(B(a,\rho))<C\rho^\kappa. \] Denote by \(\nu\) the image measure of \(\mu\times\ldots \times\mu\) (\(k^2\)-fold) under the map \[ (x_1,\ldots,x_{k^2})\mapsto (x_1\ldots x_k)+(x_{k+1}\ldots x_{2k})+\ldots+(x_{k^2-k+1}\ldots x_{k^2}) . \] Then for all \(\delta<\rho<1\), \[ \max_a\nu(B(a,\rho))<C'\rho. \] Let \(\mu\) and \(\nu\) be (possibly discrete) probability measures on \([0,1]\) satisfying, for all \( \delta\leq \rho\leq 1\), the conditions \[ \max_a \mu(B(a,\rho)) \lesssim\rho^\alpha \] and \[ \max_a \nu(B(a,\rho)) \lesssim\rho^\beta. \] Assume \(\alpha+\beta>1\). Let \(\psi:[0,1]\times [0,1]\to [0,1], (x,y)\mapsto xy\) and \(\psi[\mu\otimes\nu]\) be the image measure. Then \[ | (\psi[\mu\otimes\nu])^\wedge (\xi)| \lesssim |\xi|^{\frac{1-\alpha-\beta}{2}} \text{ for } 1\leq|\xi|\leq \frac{1}{\delta}. \]