arXiv++ Combinatorics

We obtain restriction estimates of $ε$-removal type for the set of $k$-th powers of integers, and for discrete $d$-dimensional surfaces of the form \[ \{ (n_1,\dots,n_d,n_1^k + \dotsb + n_d^k) \,:\, |n_1|,\dots,|n_d| \leq N \}, \] which we term '$k$-paraboloids'. For these surfaces, we obtain a satisfying range of exponents for large values of $d,k$. We also obtain estimates of $ε$-removal type in the full supercritical range for $k$-th powers and for $k$-paraboloids of dimension $d < k(k-2)$. We rely on a variety of techniques in discrete harmonic analysis originating in Bourgain's works on the restriction theory of the squares and the discrete parabola.

We study translation-invariant additive equations of the form $\sum_{i=1}^s λ_i \mathbf{P}(\mathbf{n}_i) = 0$ in variables $\mathbf{n}_i \in \mathbb{Z}^d$, where the $λ_i$ are nonzero integers summing to zero, and $\mathbf{P}$ is a system of homogeneous polynomials such that the above equation is invariant by translation. We investigate the solvability of this equation in subsets of density $(\log N)^{-c(\mathbf{P},\mathbfλ)}$ of a large box $[N]^d$, via the energy increment method. We obtain positive results in roughly the number of variables currently needed to derive a count of the solutions in the complete box $[N]^d$, for the curve $\mathbf{P} = (x,\dots,x^k)$ and the multidimensional systems of large degree studied by Parsell, Prendiville and Wooley, using only a weak form of restriction estimates. We also obtain results for the $(d+1)$-dimensional parabola $\mathbf{P}=(x_1,\dots,x_d,x_1^2+\dotsb+x_d^2)$ that rely on the recent Strichartz estimates of Bourgain and Demeter.

We show that the equation $λ_1 n_1^2 + ... + λ_s n_s^2 = 0$ admits non-trivial solutions in any subset of $[N]$ of density $(\log N)^{-c_s}$, provided that $s \geq 7$ and the coefficients $λ_i$ sum to zero and satisfy certain sign conditions. This improves upon previous known density bounds of the form $(\log\log N)^{-c}$.

Consider a translation-invariant system of linear equations $V x = 0$ of complexity one, where $V$ is an integer $r \times t$ matrix. We show that if $A$ is a subset of the primes up to $N$ of density at least $C(\log\log N)^{-1/25t}$, there exists a solution $x \in A^t$ to $V x = 0$ with distinct coordinates. This extends a quantitative result of Helfgott and de Roton for three-term arithmetic progressions, while the qualitative result is known to hold for all systems of equations of finite complexity by the work of Green and Tao.

We show that if a finite, large enough subset A of an arbitrary abelian group satisfies the small doubling condition |A + A| < (log |A|)^{1 - epsilon} |A|, then A must contain a three-term arithmetic progression whose terms are not all equal, and A + A must contain an arithmetic progression or a coset of a subgroup, either of which of size at least exp^[ c (log |A|)^{delta} ]. This extends analogous results obtained by Sanders and, respectively, by Croot, Laba and Sisask in the case where the group is that of the integers or a finite field.

Our main result states that when A, B, C are subsets of Z/NZ of respective densities α,β,γ, the sumset A + B + C contains an arithmetic progression of length at least e^{c(\log N)^c} for densities α> (\log N)^{-2 + ε} and β,γ> e^{-c(\log N)^c}, where c depends on ε. Previous results of this type required one set to have density at least (\log N)^{-1 + o(1)}. Our argument relies on the method of Croot, Laba and Sisask to establish a similar estimate for the sumset A + B and on the recent advances on Roth's theorem by Sanders. We also obtain new estimates for the analogous problem in the primes studied by Cui, Li and Xue.