Conclusion

The dissertation set out to show that the polynomial method, as it has developed from Chevalley and Warning through Alon to Dvir, Guth–Katz, and Ellenberg–Gijswijt, rests on three lemmas, and that the principal theorems of the subject can be derived from those lemmas with complete proofs and explicit constants. Chapter 2 stated and proved the lemmas: counting, with and without multiplicity; zeros, in the Schwartz–Zippel, Nullstellensatz, and multiplicity forms; and rank, with the Croot–Lev–Pach expansion. Chapters 3 to 6 derived Theorems A to F from them, each chapter beginning by naming the lemmas it used and each proof identifying the problem-specific bridge. Chapter 7 proved one new theorem, on punctured Kakeya sets in the plane, by a combinatorial argument that the polynomial method does not reach, and Chapter 8 marked the boundary of what the lemmas can do.

Three conclusions may be drawn.

The first is that the method is elementary in a precise sense: nothing beyond the algebra of polynomials over a field and the dimension of a vector space enters any proof. The only numerical facts used are that factorials of integers less than $p$ are prime to $p$, that $\sum_{t\in\mathbb F_q}t^a$ vanishes for $0\le a<q-1$, and that $4t^2+t-2$ has root $(\sqrt{33}-1)/8$. This elementary character is what allowed Dvir's two-page paper to settle a problem on which analysts had worked for a decade, and it is what makes the method teachable: a reader with a first course in algebra can follow every step of Chapters 2 to 6.

The second is that the method's power and its limits have the same source. Its power comes from the rigidity of polynomials: a polynomial that vanishes at $q$ points of a line vanishes on the line, and a polynomial of degree less than $q$ that vanishes at every direction vanishes everywhere. Its limits come from the same rigidity: the method sees only zero sets, and when the object of interest is not close to a zero set, as a Besicovitch set in $\mathbb R^n$ is not, or when the question is about structure rather than size, the method has nothing to say. The computations in Appendix A make this concrete. In the plane Kakeya problem the bound is within $(q-1)/2$ of the truth, because a minimal Kakeya set there very nearly is a zero set of degree $q$; in the joints problem the constant is off by a factor of $36$, because the pruning and the dimension count each lose a constant that the true extremal configuration does not.

The third is methodological. The theorems of this dissertation are presented in the literature as separate achievements with separate proofs, and they were discovered separately. Seen from the three lemmas they are one argument with four bridges, and the bridges are where the mathematics specific to each problem lives. This suggests that new applications of the method should be sought by looking for new bridges: configurations on which a low-degree polynomial is forced, by the configuration's structure, to vanish on a larger set, or to have a diagonal matrix, or to have derivatives that vanish on the same set. The multiplicity layer of Chapter 2 is available for any bridge that produces vanishing along lines, and the slice-rank layer for any bridge that produces a diagonal tensor. Which further configurations admit such bridges is, in the author's view, the most promising direction for the method, and it is one that the three-lemma organisation is designed to make visible.