Many Sums and Products Conjecture

• The Many Sums and Products Conjecture is a central problem in additive and multiplicative combinatorics, positing that large finite sets cannot simultaneously exhibit small k-fold sumsets and product sets.
• Recent counterexamples in the reals use high-degree number fields and structured lattice constructions to demonstrate that the conjectured polynomial lower bounds fail under sophisticated combinatorial and incidence-geometric techniques.
• Extensions of the conjecture to p-adic fields, finite fields, and function fields, along with positive partition-regular results in the rationals, highlight its deep connections and diverse implications in modern arithmetic research.

The Many Sums and Products Conjecture, a far-reaching and multifaceted problem in additive and multiplicative combinatorics, explores the simultaneous largeness of multi-fold sumsets and product sets—both in their classic Erdős–Szemerédi form for sums and products of real or integer sets, and in various strengthened, partition-regular, and structural variants. Its recent dramatic falsification in the real numbers, parallel partition-regular consequences in the rationals, and related conjectural analogues in other structures represent distinct, deeply-intertwined strands of modern arithmetic combinatorics.

1. Formal Statement and Historical Context

Let $A$ be a finite subset of $\mathbb{R}$ and $k \geq 2$. Define the $k$-fold sumset

$kA = \{a_1 + \cdots + a_k : a_i \in A\}$

and the $k$-fold product set

$A^{(k)} = \{a_1 \cdots a_k : a_i \in A\}.$

The Many Sums and Products Conjecture (Erdős–Szemerédi 1977) asserts that for every $\varepsilon > 0$ and fixed $k$,

$$\max(|kA|, |A^{(k)}|) \gg_{k,\varepsilon} |A|^{k - \varepsilon}$$

should hold for all sufficiently large $\mathbb{R}$0.

This conjecture generalizes the classical sum-product phenomenon, originally conjectured to hold for $\mathbb{R}$1, to arbitrary $\mathbb{R}$2, predicting, in essence, that no large finite set can have both small $\mathbb{R}$3-fold sumsets and $\mathbb{R}$4-fold product sets—thus exhibiting simultaneous additive and multiplicative structure is conjecturally impossible.

There are also “weak” and “strong” forms: the weak form prescribes largeness for $\mathbb{R}$5; the strong “edge-labeled” form asserts largeness of the set of $\mathbb{R}$6 pairs for all (sufficiently dense) graphs over $\mathbb{R}$7’s indices, which has itself been explicitly refuted (Alon et al., 2018).

2. Counterexamples and Disproof over $\mathbb{R}$8

The Many Sums and Products Conjecture for reals was decisively disproved in (Bloom et al., 27 May 2026). Here, for every $\mathbb{R}$9, the authors construct arbitrarily large sets $k \geq 2$0, in fact subsets of number fields of degree $k \geq 2$1, such that

$k \geq 2$2

for some absolute constant $k \geq 2$3. This construction builds $k \geq 2$4 as a direct product of two structured lattices in a high-degree totally real field—balancing an “additive box” (of elements with controlled archimedean size) and a “multiplicative box” (units with vector-valued logarithms in a hypercube).

The argument crucially exploits the independence of the additive and multiplicative structures inside these number fields, combined with Schinzel's unit separation lemma and precise count estimates from Minkowski’s geometry of numbers and Dirichlet’s unit theorem. The construction ensures that the “directness” gives $k \geq 2$5 (where $k \geq 2$6 and $k \geq 2$7 are the multiplicative and additive boxes, respectively), while both $k \geq 2$8 and $k \geq 2$9 are dominated by the ambient lattice’s exponential size in $k$0.

The resulting upper bound is exponentially sub-optimal compared to the conjectured polynomial regime, and no stronger-than-logarithmic exponent in $k$1 is possible over $k$2. Prior lower bounds (e.g., by Mudgal–GGMT) only gave $k$3, much weaker than the conjectural $k$4 (Bloom et al., 27 May 2026).

3. Extensions to Other Algebraic Structures

The mechanism behind the real counterexamples generalizes to $k$5-adic fields $k$6, finite fields $k$7, and global function fields:

Setting	Construction	Resulting Bound
$k$8	Embed $k$9 into $kA = \{a_1 + \cdots + a_k : a_i \in A\}$0 via splitting fields	$kA = \{a_1 + \cdots + a_k : a_i \in A\}$1, $kA = \{a_1 + \cdots + a_k : a_i \in A\}$2
$kA = \{a_1 + \cdots + a_k : a_i \in A\}$3	Reduce $kA = \{a_1 + \cdots + a_k : a_i \in A\}$4 mod prime; use completely split primes	$kA = \{a_1 + \cdots + a_k : a_i \in A\}$5 for $kA = \{a_1 + \cdots + a_k : a_i \in A\}$6
$kA = \{a_1 + \cdots + a_k : a_i \in A\}$7	Use sections on curves, implement lattice approach	$kA = \{a_1 + \cdots + a_k : a_i \in A\}$8, $kA = \{a_1 + \cdots + a_k : a_i \in A\}$9

For $k$0, one selects number fields in which $k$1 splits completely; in function fields, one utilizes high-genus curves with many rational points, translating the direct-product lattice method into this context.

4. Partition-Regularity and Ramsey-Type Results

In contrast to the negative structural results over the reals, fully positive results are available in the field of partition-regularity for the rationals. Alweiss (Alweiss, 2023) proves:

For each $k$2 and any finite coloring $k$3, there exist nonzero $k$4 such that all subset sums and all subset products (over all nonempty $k$5), i.e., the set

$k$6

are monochromatic in $k$7.

The proof uniquely exploits the divisibility structure of $k$8—scaling and shifting allow all patterns to be simultaneously monochromatized via repeated applications of the multidimensional polynomial van der Waerden theorem and algebraic reduction. Notably, the corresponding question over the integers $k$9 remains open even for $A^{(k)} = \{a_1 \cdots a_k : a_i \in A\}.$0; the failure of universal divisibility is the central obstacle.

5. Connections to Infinite Multiple Zeta-Star Values

A related but distinct “Many Sums and Products Conjecture” arises in the study of infinite multiple zeta-star values (MZSVs) (Li et al., 27 Mar 2026). Here, the set

$A^{(k)} = \{a_1 \cdots a_k : a_i \in A\}.$1

serves as an analogue of continued fractions with bounded partial quotients. Conjectures in this setting assert:

• The algebraic points in $A^{(k)} = \{a_1 \cdots a_k : a_i \in A\}.$2 consist only of integers $A^{(k)} = \{a_1 \cdots a_k : a_i \in A\}.$3 for $A^{(k)} = \{a_1 \cdots a_k : a_i \in A\}.$4, and $A^{(k)} = \{a_1 \cdots a_k : a_i \in A\}.$5 contains no other algebraic values for larger $A^{(k)} = \{a_1 \cdots a_k : a_i \in A\}.$6;
• The closure of $A^{(k)} = \{a_1 \cdots a_k : a_i \in A\}.$7, where the indices are bounded below, forms a thin Cantor-type set whose sumset $A^{(k)} = \{a_1 \cdots a_k : a_i \in A\}.$8 has empty interior (and measure zero for $A^{(k)} = \{a_1 \cdots a_k : a_i \in A\}.$9).

Explicitly, for $\varepsilon > 0$0, $\varepsilon > 0$1, but for bounded-below indices $\varepsilon > 0$2, the sumset never contains an interval, sharply contrasting the behavior of continued fraction analogues (Li et al., 27 Mar 2026).

6. Constructive, Incidence-Geometric, and Combinatorial Techniques

The recent refutations of many-sums-and-products conjectures, and the best known positive results, depend on a combination of lattice-counting and direct-product constructions (over number fields or function fields), together with sophisticated combinatorics and incidence geometry.

Key technical ingredients include:

• Lattice-point enumerations: estimating the size of boxes in additive and multiplicative lattices, using Minkowski's theorem and Dirichlet's unit theorem;
• Unit separation: identifying when the product representation is “direct” to ensure $\varepsilon > 0$3;
• Incidence geometry: employing Elekes-style point-line incidence bounds and the Szemerédi–Trotter theorem to provide lower bounds, and matching these with explicit upper bound constructions (Alon et al., 2018);
• Ramsey theory and van der Waerden-type results: key in partition-regular settings, leveraging polynomial progressions and divisibility arguments (Alweiss, 2023);
• Extensions to finite fields: replacing geometric with Stepanov-type bounds and adapting the constructions to subgroups and coset structures.

7. Broader Structural and Computational Aspects

Computational and structural analysis (cf. “Visualizing the Sum-Product Conjecture” (O'Bryant, 2024)) provides both small-$\varepsilon > 0$4 ground truths and empirical evidence about extremal sets. Explicit datasets up to $\varepsilon > 0$5 show no evidence for the original Erdős–Szemerédi drift towards quadratic sum-product largeness in $\varepsilon > 0$6; the minimum observed exponent for $\varepsilon > 0$7 is $\varepsilon > 0$8 at $\varepsilon > 0$9, with the lower envelope largely realized by divisor-closed or smooth number sets.

Open conjectures in this field include explicit polynomial lower bounds (e.g., $k$0, possible only for geometric progressions), and the “golden ratio sum-product bound” $k$1.

8. Ongoing Challenges and Open Questions

While the strong many-sums-and-products regime is now falsified over $k$2 and its analogues, significant challenges remain:

• Precise exponent determination: There is an order-of-magnitude gap between constructive upper bounds (arising from sophisticated number-theoretic constructions) and current lower bounds (from incidence geometry) across regimes (Alon et al., 2018, Bloom et al., 27 May 2026).
• Behavior in other algebraic settings: Partition-regular consequences (as over $k$3) lack analogues over $k$4; the extension of explicit constructions to “genuinely nonlinear” configurations or to finite fields is incomplete.
• Structural characterization: The classification of sets simultaneously almost sum-free and product-free at scale (beyond the structured direct-product model) is open.

In summary, the Many Sums and Products Conjecture illustrated how deeply the interplay of additive and multiplicative structures influences the shape of sets and progressions in diverse algebraic contexts, and its recent dramatic refutations and positive consequences define an active front in combinatorial and arithmetic research (Bloom et al., 27 May 2026, Alweiss, 2023, O'Bryant, 2024, Li et al., 27 Mar 2026, Alon et al., 2018).