Improved rank bounds for design matrices and a new proof of Kellys theorem

We study the rank of complex sparse matrices in which the supports of different columns have small intersections. The rank of these matrices, called design matrices, was the focus of a recent work by Barak et. al. (BDWY11) in which they were used to answer questions regarding point configurations. In this work we derive near-optimal rank bounds for these matrices and use them to obtain asymptotically tight bounds in many of the geometric applications. As a consequence of our improved analysis, we also obtain a new, linear algebraic, proof of Kelly’s theorem, which is the complex analog of the Sylvester-Gallai theorem.

💡 Research Summary

The paper investigates the rank of a special class of sparse complex matrices known as design matrices, where each column has a bounded number of non‑zero entries and any two distinct columns intersect in at most a few rows. This model was introduced by Barak, Dvir, Wigderson, and Yehudayoff (BDWY11) and was instrumental in addressing combinatorial geometry questions such as variants of the Sylvester‑Gallai theorem. The authors improve upon the BDWY rank bound, which contained a logarithmic factor, by developing a refined linear‑algebraic analysis that eliminates the log term and yields a near‑optimal linear dependence on the matrix dimensions.

The technical core consists of two innovations. First, the authors apply a bidirectional scaling (row‑ and column‑wise diagonal matrices) to bring the design matrix into a doubly‑balanced form where every row and column has unit ℓ₂‑norm. This normalization equalizes the contribution of each row and column to the spectrum, allowing a precise lower bound on the largest singular values. Second, they reinterpret the column‑intersection condition as a graph on the columns: vertices correspond to columns and an edge is present whenever two columns share a row. By bounding the average degree of this “intersection graph” in terms of the design parameters (q = number of non‑zeros per column, t = maximum intersection size), they derive a new inequality that directly controls the off‑diagonal entries of AᵀA. Combining the spectral bound with the graph‑theoretic inequality yields the main rank estimate

\