Killing the Sum-Product Conjecture: An Unexpected Twist in Additive Combinatorics
28 May 2026, Yanjiang
Algebraic integers in high-degree number fields simultaneously keep sums and products small, toppling the 40-year-old sum-product conjecture in a stunning counterexample.
For forty years, a deceptively simple rule stood like bedrock in combinatorial number theory: for any finite set of real numbers, either the set of all pairwise sums must balloon to nearly the square of the original size, or the set of all pairwise products must. It was called the sum‑product conjecture, and as far as anyone could tell, it was a law of the mathematical universe—a statement about the fundamental incompatibility of addition and multiplication. Until now. In a preprint (arXiv:2605.28781), four mathematicians have constructed explicit sets of real numbers where both the sumset and the product set remain stubbornly small, violating the conjecture by a substantial margin and collapsing a decades‑old pillar.
To appreciate what has fallen, we need to see what the sum‑product conjecture actually claimed. Given a finite set (A) of numbers, let (A+A = {a+b : a,b\in A}) be the set of all pairwise sums, and (AA = {ab : a,b\in A}) the set of all pairwise products. The conjecture, first voiced by Erdős and Szemerédi in 1983 and sharpened through decades of incremental work, asserted that (\max(|A+A|,|AA|)) is at least (|A|^{2-o(1)}) — meaning that for large sets, either the sumset or the product set must have size roughly the square of the original. Put simply: you cannot be simultaneously additively and multiplicatively structured; one of the two operations must blow up the set almost quadratically.
Think of a rubber sheet. If you stretch it in one direction—say, addition—it expands, its surface area growing. If you try to stretch it just as forcefully in the orthogonal direction—multiplication—at the same time, the sheet should tear, or at least some part must bulge disproportionately. The conjecture said the sheet cannot resist a double stretch without creating an enormous total area; the two pulls are adversarial by nature. The counterexample of Thomas F. Bloom at the University of Manchester, together with Will Sawin of Princeton University, Carl Schildkraut of Stanford University, and Dmitrii Zhelezov, reveals a fabric so cleverly woven that both pulls produce only controlled, small deformations, leaving the sheet nearly flat.
How could such a fabric exist? The answer lies deep in algebraic number theory, a territory that additive combinatorics had largely kept at arm’s length. The team’s construction uses numbers that are algebraic integers—roots of polynomial equations with whole‑number coefficients—living inside high‑degree number fields. A number field is a generalisation of the rational numbers, and you can picture its elements as vectors in a space of many dimensions: a field of degree (d) is like a (d)-dimensional coordinate system, each number having (d) pieces of data. By carefully selecting a field whose degree grows with the size of the set (A), the authors engineer a collection of numbers that lies simultaneously in an arithmetic progression—keeping sums compact—and inside a multiplicative group of units—numbers whose inverses are also algebraic integers, which keeps the product set tight. The elements are built as products (u \cdot p), where (u) is drawn from a carefully chosen set of units—numbers whose multiplicative behaviour is tightly controlled—and (p) belongs to a box-like set of algebraic integers that provides additive structure; the key insight is that in a number field of sufficiently high degree, both lattices can be engineered to stay simultaneously compact.
The effect is striking. The construction produces arbitrarily large sets (A) for which (\max(|A+A|,|AA|) \le |A|^{2-c}) for a fixed positive constant (c). That is, both the sumset and the product set are sub‑quadratic by a constant exponent gap—not just a logarithmic sliver, but a genuine, uniform shortfall. For a conjecture that aimed for an exponent arbitrarily close to (2), any positive (c) is lethal. The authors do not stop there; they also dismantle the many sums and products conjecture, which extended the same intuition to (k)-fold sums and products. They show that for any (k\ge 3) one can build sets where both (|kA|) and the (k)-fold product set (A^{(k)}) are bounded above by (|A|^{C \log k / \log \log k}), a growth far slower than the conjectured (|A|^{k-\varepsilon}). The firewall was never about the number of operations; it was about the hidden harmony between addition and multiplication that number fields can unlock.
But there is a crucial nuance, one that both saves the spirit of the original conjecture and opens a deeper inquiry. The sets that break the conjecture are not arbitrary collections of real numbers plucked off the line; every element is an algebraic integer in a number field of degree roughly (\log|A|). The counterexamples are soaked in algebraic structure—they are not random‑looking, not typical. This raises a natural question that earlier work on universal lower bounds sharpened: does the sum‑product conjecture still hold for sets that lack this number‑theoretic scaffolding? Recently, Cushman (arXiv:2512.13849) proved that for any finite set (A) of real numbers, (\max(|A+A|,|AA|)) is at least roughly (|A|^{4/3+\delta}) for some tiny positive (\delta). That lower bound is far from quadratic, but it is unconditional—it demands no special structure, and it remains the best known universal result even after the conjecture’s demise. Bloom and his colleagues show that one cannot push such a universal bound to an exponent of (2), because algebraic integers provide a loophole. Yet the door remains open that for “generic” sets of real numbers—those without conspiratorial number‑theoretic origin—the sum‑product conjecture might still be essentially true. The tension between structure and randomness, a perennial theme in combinatorics, has been thrown into sharp relief.
The paper also extinguishes the conjecture’s analogues in other universes: over (p)-adic numbers, over finite fields, and in function fields of positive characteristic, similar algebraic constructions bring the edifice down. The fault was never a peculiarity of the real line; it was the insufficient imagination of the original conjecture’s scope. Where researchers once believed that no set could be simultaneously additively and multiplicatively slim, the new work shows that number fields bristle with such sets—if you know where to look. In earlier investigations of integers with few prime factors (Agrawal et al., arXiv:2512.04931), mathematicians had already seen hints that algebraic multiplicativity could soften sum‑product type bounds; the present breakthrough transforms that hint into a principled design.
The sum‑product conjecture was a boundary marker, a statement about the impossibility of certain kinds of order. Its demise is humbling, but it is also exhilarating—it reminds us that even the most intuitive mathematical principles can house hidden exceptions, and that the real prize is understanding the map of where and how those exceptions occur. The quest now shifts from proving a universal law to charting the precise territory where addition and multiplication can conspire to stay small. How close can one push the exponent toward (2) while still using algebraic integers? Can one construct a counterexample over the rational numbers, or are number fields essential? What is the largest possible (c)? These questions will steer the next generation of additive combinatorics, a field that has just been handed a new continent to explore.
What Bloom and his collaborators have done is not merely produce a counterexample; they have opened a window onto a landscape where addition and multiplication are not always at war. The rubber sheet, it turns out, can be woven to withstand a double stretch without tearing—provided the threads are spun from algebraic gold. The sum‑product conjecture may be dead, but from its passing a more nuanced, more fascinating picture of structure and disorder is taking shape.
— Yanjiang
Yanjiang is an online editor of LoomSci.com.
References
- Thomas F. Bloom et al., The sum-product conjecture is false for real numbers, arXiv:2605.28781
- Agrawal et al., More on the sum-product problem for integers with few prime factors, arXiv:2512.04931
- Cushman, A Note on the Sum-Product Problem and the Convex Sumset Problem, arXiv:2512.13849