Abstract:We identify an assumption on linear forms $\phi_1, \dots, \phi_k: \mathbb{F}_p^n \to \mathbb{F}_p$ that is much weaker than approximate joint equidistribution on the Boolean cube $\{0,1\}^n$ and is in a sense almost as weak as linear independence, but which guarantees that every subset of $\{0,1\}^n$ on which none of $\phi_1, \dots, \phi_k$ has full image has a density which tends to 0 with $k$. This density is at most quasipolynomially small in $k$, a bound that is necessarily close to sharp.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is that under certain conditions, the density of subsets defined on the Boolean cube $\{0, 1\}^n$ will tend to 0, and the rate of this density decrease is quasipolynomial. Specifically, the paper explores how the density of any subset of $\{0, 1\}^n$ that does not contain the complete images of these linear forms tends to 0 as $k$ increases when the linear forms $\varphi_1,\ldots,\varphi_k:\mathbb{F}_p^n\rightarrow\mathbb{F}_p$ satisfy certain assumptions.
### Background of the Paper
1. **Restrictions of Mod - $p$ Linear Forms**:
- The paper studies the restrictions of mod - $p$ linear forms on $\{0, 1\}^n$.
- These restricted forms have been discussed in several recent studies, for example, in the general settings of polynomials and Abelian group homomorphisms.
2. **Obstacles to Combinatorial Statements**:
- These restricted forms also appear in the obstacles to Hales - Jewett - type combinatorial statements.
3. **Distribution Problems**:
- The distribution of mod - $p$ linear forms and their $k$-tuples on $\{0, 1\}^n$ is more complex than on the entire $\mathbb{F}_p^n$.
- On $\mathbb{F}_p^n$, linear independence and joint uniform distribution are equivalent, but this is not the case on $\{0, 1\}^n$.
### Main Results
1. **Density Decrease under Weak Conditions**:
- The paper proposes a weaker assumption than those in previous works, that is, the support size of the linear combination of linear forms is at least a certain constant $r$.
- This assumption ensures that the density of subsets of $\{0, 1\}^n$ that satisfy a large number of non - trivial conditions will tend to 0 as $k$ increases.
2. **Theorem 1.4**:
- Let $p$ be a prime number and $S$ be a subset of $\mathbb{F}_p$ with size at least 2.
- There exists a constant $a = a(p, S)>0$ such that if $k\geq1$, $E_1,\ldots,E_k$ are strict subsets of $\mathbb{F}_p$, and $\varphi_1,\ldots,\varphi_k$ are mod - $p$ forms, satisfying
\[
|Z(\varphi_j-\varphi_i)|\geq2\max_{1\leq t\leq k}L(S, E_t)- 1
\]
for any $1\leq i < j\leq k$, then the density of points $x\in S^n$ satisfying
\[
\varphi_1(x)\in E_1,\ldots,\varphi_k(x)\in E_k
\]
is at most $O(k^{-a/\log\log k})$ as $k$ tends to infinity.
### Proof Methods
1. **Sunflower Structure**:
- Prove the exponentially small density by constructing a sunflower structure.
- A sunflower structure means that the central form of a set of linear forms and the supports of each form are pairwise disjoint.
2. **From Sunflower to Small Balls**:
- Relax the assumptions to these forms being contained in small balls of bounded radius.
- Use induction to prove the existence of a sunflower containing many forms.
3. **Completion of the Proof**:
- Given any set of mod - $p$ forms, either this set of forms is jointly approximately uniformly distributed, or it can be reduced to forms contained in small balls of bounded radius.
### Open Problems
1. **Quantitative Bounds**:
- Example 2.4 shows that the bounds in Theorem 1.4 and Corollary 1.5 cannot be improved.