Freiman Bihomomorphisms in Additive Combinatorics

• Freiman bihomomorphisms are functions defined on subsets of product groups that preserve additive relations in each coordinate, structuring dense additive sets.
• Researchers use iterative averaging, Bohr set techniques, and cocycle extensions to extend these maps from finite vector spaces to arbitrary abelian groups.
• The study links structured additive patterns with inverse theorems for uniformity norms and approximate polynomial classification, impacting higher-order combinatorics.

A Freiman bihomomorphism is a function defined on a subset of a direct product of two abelian groups (or vector spaces) that simultaneously behaves as a Freiman homomorphism of order 2 in each coordinate separately. This class of maps arises naturally as a structural tool in higher-order additive combinatorics, playing a pivotal role in the inverse theory of uniformity norms, approximate polynomial classification, and the analysis of dense additive subsets. Recent developments have expanded the scope from finite vector spaces over a prime field (Gowers et al., 2020) to arbitrary finite abelian groups (Milićević, 4 Jan 2026), revealing new algebraic phenomena and quantitative techniques crucial for understanding multilinear and approximate-algebraic structures.

1. Definition and Fundamental Properties

Let $G_1$, $G_2$, and $H$ be finite abelian groups or vector spaces, and let $A \subset G_1 \times G_2$. A map $\phi: A \to H$ is a Freiman bihomomorphism if for every fixed $y \in G_2$, the map $x \mapsto \phi(x, y)$ is a Freiman 2-homomorphism on those $x$ with $(x, y) \in A$, and symmetrically, for every fixed $x \in G_1$, the map $y \mapsto \phi(x, y)$ is a Freiman 2-homomorphism on those $y$ with $(x, y) \in A$. Explicitly, the defining conditions are:

• For any quadruple $(x_1, y), (x_2, y), (x_3, y), (x_4, y)\in A$ with $x_1+x_2=x_3+x_4$,

$$\phi(x_1,y) + \phi(x_2,y) = \phi(x_3,y) + \phi(x_4,y)$$

• For any quadruple $(x, y_1), (x, y_2), (x, y_3), (x, y_4)\in A$ with $y_1+y_2=y_3+y_4$,

$$\phi(x,y_1) + \phi(x,y_2) = \phi(x,y_3) + \phi(x,y_4)$$

In finite vector spaces, one often restricts attention to dense domains $A$ and studies how “approximate bihomomorphisms” must be close to genuine bilinear maps. The order-preserving variant, a Freiman 2-isomorphism, additionally respects order and additive relations (Amirkhanyan et al., 2014).

2. Structural Theorems and Quantitative Inverse Results

The central structural result, extending the vector-space case (Gowers et al., 2020) to general abelian groups (Milićević, 4 Jan 2026), asserts that for a dense Freiman bihomomorphism $\phi$ on $A \subset G_1 \times G_2$, there exist:

• Bohr sets $B_1 \subset G_1$, $B_2 \subset G_2$ of controlled codimensions;
• Shifts $s \in G_1$, $t \in G_2$;
• A rank-$r$ subgroup $E \subset H$ of bounded rank;
• An $E$–bihomomorphism $\Phi: B_1 \times B_2 \to H$,

such that $\Phi(x, y) = \phi(x+s, y+t)$ for a dense fraction of $(x, y) \in B_1 \times B_2$. When $E = \{0\}$, $\Phi$ is a genuine bilinear map. This construction uses iterative averaging, Bohr set analysis, cocycle extension, and dependent random choice—a blend of combinatorial and Fourier-analytic methods. In finite vector spaces, the explicit bound on the agreement density is tower-type: for $A$ of density $\delta$, the bihomomorphism agrees with an affine map $\Phi$ on at least $\varepsilon |G_1||G_2|$ points, with

$$\varepsilon = (\exp^{(O(1))}(O_p(\delta^{-1})))^{-1}$$

where $\exp^{(t)}$ is the iterated exponential (Gowers et al., 2020).

3. Proof Techniques and Extension Phenomena

The main proof steps diverge markedly between the vector-space and general group settings (Milićević, 4 Jan 2026). While vector spaces permit exact extension of Freiman-linear maps, arbitrary abelian groups require E-homomorphisms: maps satisfying the homomorphism law up to a small error set $E$. The eight critical proof stages involve:

• Reduction to low-rank systems via Balog–Szemerédi–Gowers and Bogolyubov-type lemmas.
• Promotion of small-rank conditions to Bohr-respectedness through dependent random choice.
• Construction and densification of bilinear Bohr varieties supporting the bihomomorphism.
• Abstract regularity and algebraic integration steps, introducing cocycle identities and controlling error sets.

A novel aspect is the necessity to extend bihomomorphisms from Bohr slices to full products at the cost of a finite error set; this phenomenon does not appear in vector spaces but is crucial for achieving quantitative global structure in general groups (Milićević, 4 Jan 2026).

4. Connections to Additive Combinatorics and Uniformity Norms

Freiman bihomomorphisms underlie quantitative inverse theorems for Gowers uniformity norms ($U^3$, $U^4$) and the structure of approximate polynomials. Key implications include:

• Functions with large uniformity norms correlate with structured forms (quadratic, cubic, almost-cubic polynomials) where the controlling map arises as a bihomomorphism on a large dense subset (Gowers et al., 2020, Milićević, 4 Jan 2026).
• Dense subspaces in products $G_1 \times G_2$ that are subspaces along principal directions can be covered by bilinear varieties of bounded codimension (Gowers et al., 2020).
• Order-preserving Freiman 2-isomorphisms facilitate interval condensation for sets of small doubling, with applications to additive energy, extremal combinatorics, and diagonal-set constructions (Amirkhanyan et al., 2014).

The following table organizes principal results relating Freiman bihomomorphisms to combinatorial applications:

Application Area	Key Result/Map Type	Reference
Gowers $U^k$ inverse problems	Approximate bilinear/cubic	(Gowers et al., 2020, Milićević, 4 Jan 2026)
Dense additive subspaces	Bilinear variety structure	(Gowers et al., 2020)
Small doubling sets in integers	Order-preserving 2-isomorphism	(Amirkhanyan et al., 2014)
Additive energy refinement	Interval condensation + $EI$	(Amirkhanyan et al., 2014)

5. Generalizations, Open Problems, and Limitations

A significant direction is the uniform treatment of bihomomorphism structure theorems across all finite abelian groups, crucial for resolving the Jamneshan–Shalom–Tao conjecture on $U^k$ inverse theory (Milićević, 4 Jan 2026). Current quantitative results in cyclic groups yield doubly-exponential bounds not yet matched in full generality. The complexity of cocycle identities and error-control intensifies outside bounded-exponent settings; for groups of exponent $2^d$, extension and identification of genuine bihomomorphisms become tractable, as error sets can often be absorbed into the group structure (Milićević, 4 Jan 2026).

The machinery of Freiman bihomomorphisms is expected to interface fruitfully with nilspace and nilmanifold theories, though the present treatment remains entirely quantitative and avoids ergodic theory. Among further challenges is sharpening regularity and Bogolyubov-type lemmas for improved bounds, and understanding higher-dimensional equidistribution phenomena for multi-homomorphisms (Amirkhanyan et al., 2014).

6. Historical Evolution and Bibliographical Perspective

The theory of Freiman bihomomorphisms is rooted in the classical Freiman isomorphism and the structural study of sets with small doubling. The order-preserving variant and interval condensation trace to Amirkhanyan–Bush–Croot (Amirkhanyan et al., 2014). The application to uniformity norms and multiaffine maps is developed in Milićević’s work on finite fields (Gowers et al., 2020) and extended to arbitrary abelian groups in the recent quantitative inverse theory (Milićević, 4 Jan 2026). The methods are informed by Fourier-analytic and combinatorial regularity paradigms and build on extensive foundational results by Gowers, Manners, and others in the study of approximate polynomials and uniformity norms.

The development of E-homomorphism extension techniques, cocycle identities, and the interplay with Bohr sets and coset progressions constitutes a substantial advance, offering new perspectives for future research in additive structure and higher-order algebraic analysis.