Flag Algebra Methods

• Flag Algebra Methods is a rigorous framework that encodes densities of substructures in graphs and hypergraphs using types, flags, and algebraic relations.
• The approach transforms complex extremal problems into semidefinite programming tasks, producing sum-of-squares certificates and utilizing symmetry reduction to simplify computations.
• Tools like Flagmatic and FlagAlgebraToolbox automate these methods, achieving best-known bounds for Turán-type densities and contributing to stability results in extremal combinatorics.

Flag algebra methods are a suite of algebraic and optimization-based techniques developed to formalize and automate the derivation of asymptotically sharp inequalities for densities of small substructures in large, constrained combinatorial objects, most notably graphs and hypergraphs. Introduced by Razborov in 2007, the flag algebra framework enables the encoding of extremal (often Turán-type) problems as structured semidefinite programs (SDPs), systematically leveraging symmetries and polynomial identities to yield computer-assisted proofs and certificates for density bounds in extremal combinatorics (Silva et al., 2016, Jeong et al., 19 Jan 2026, Raymond et al., 2015).

1. Algebraic Framework: Types, Flags, and Densities

The core of flag algebra methods is the formal algebraic encoding of densities of small substructures ("flags") inside large forbidden-configuration-free objects. Fix a combinatorial theory $T$ (e.g., undirected graphs, $r$-uniform hypergraphs). For a fixed vertex set size $N$, denote by $\mathcal{F}_n(T)$ the set of all $n$-vertex $T$-structures (up to isomorphism).

• Type: A type $\sigma$ of order $k$ is a $T$-structure with $k$ labeled vertices.
• Flag: A $\sigma$-flag of order $n\geq k$ is a $T$-structure $F$ on $[n]$ in which the first $k$ vertices induce $\sigma$.

Densities are defined as follows: $$p(F;G) = \frac{|\{\varphi:[k]\hookrightarrow [N] \text{ injective} : \varphi(F) \subseteq G\}|}{\binom{N}{k}}$$ for induced density of $F$ in $G$, and for flags: $$p_\sigma(F;G) = \Pr_{\psi:[k]\hookrightarrow V(G)\;\text{preserving}\;\sigma}\bigl(\text{extending}\;\psi\;\text{to}\;[n]\;\text{yields a copy of}\;F\bigr)$$

A flag algebra $\mathcal{A}^\sigma(T)$ is the $\mathbb{R}$-vector space with basis symbols $[F]$ for all $\sigma$-flags $F$ (over all relevant $n$), modulo the chain-rule and normalization relations: $$[F] = \sum_{F_1\in\mathcal{F}_a^\sigma,\, F_2\in\mathcal{F}_b^\sigma} p_\sigma(F_1, F_2; F) [F_1]\cdot [F_2]$$ with $[\,\sigma\,]$ serving as the multiplicative unit. The algebraic multiplication is defined with combinatorial structure constants $c_{F_1,F_2}^H$ related to flag extensions (Silva et al., 2016, Jeong et al., 19 Jan 2026, Bodnár, 10 Jan 2026).

Limit homomorphisms $\phi_G:\mathcal{A}^\sigma(T)\to\mathbb{R}$ are given by $\phi_G([F]) = \lim p_\sigma(F;G)$ for sequences of growing $T$-structures $G$.

2. Extremal Problem Formulation and Semidefinite Programming

Extremal problems (e.g., Turán-type density questions) are encoded as optimization problems over these limit homomorphisms. For a linear combination $$f = \sum_i \alpha_i [F_i] \in \mathcal{A}^\sigma(T)$$, one seeks

$\sup_\phi \phi(f)$

subject to zero density constraints for forbidden substructures.

Razborov's sum-of-squares (SOS) framework provides a certificate: If one can write

$$\lambda - f = \sum_j G_j^\top G_j \quad\text{in }\mathcal{A}^\sigma(T)$$

for $G_j$ in the algebra, then for every positive homomorphism $\phi$, $\phi(f) \le \lambda$ (Silva et al., 2016, Jeong et al., 19 Jan 2026, Bodnár, 10 Jan 2026).

The SOS condition reduces the search for bounds to a semidefinite program:

• Choose a finite basis of $\sigma$-flags of size $n$, enumerate all, and let $M$ be the moment (Gram) matrix whose entries are

$$M_{ij} = \langle [F_i], [F_j] \rangle = \text{coefficient of}\;[\,\sigma\,]\;\text{in}\;[F_i]\cdot [F_j]$$

The constraints $M \succeq 0$ (positive semidefiniteness) define the feasible region.

The SDP is: $$\max \sum_i \alpha_i \phi([F_i]) \quad \text{subject to}\;\phi([\,\sigma\,])=1,\;M(\phi)\succeq 0,\;\text{and zero or positive density constraints}$$ (Silva et al., 2016, Bodnár, 10 Jan 2026).

3. Symmetry Reduction and Computational Techniques

Many flags are equivalent under the action of automorphisms of the type $\sigma$ or the symmetric group $S_n$ acting on labels. Implementations exploit this by block-diagonalizing the moment matrix $M$ using group representation theory, significantly reducing computational complexity and storage requirements (Raymond et al., 2015).

Symmetric SDPs resulting from the flag algebra setting can be reformulated as symmetry-adapted SDPs in the sense of Gatermann–Parrilo, with blocks indexed by partitions $\lambda$ of $n$ satisfying $\lambda \unrhd (n-|\sigma|,1^{|\sigma|})$. This result, formalized by Raymond–Singh–Thomas, provides that only blocks above a certain "hook-shape threshold" contribute, yielding SDPs whose size is bounded independently of $n$ for fixed flag and type sizes (Raymond et al., 2015).

4. Software: Flagmatic and FlagAlgebraToolbox

Automated frameworks such as Flagmatic and FlagAlgebraToolbox enable the systematic construction, solution, and verification of flag algebra SDPs (Falgas-Ravry et al., 2011, Bodnár, 10 Jan 2026). Flagmatic, implemented in Python, specializes in $r$-graph density problems, supports convenient notation for forbidden families, rational certificate output, and verification routines. FlagAlgebraToolbox is an extension of SageMath implementing analogous workflows for a broader class of combinatorial theories, automating the assembly, numeric solution, rational rounding, and verification of exact flag algebra certificates.

Typical computational workflow:

1. Enumerate all admissible small subgraphs free of the forbidden family.
2. Enumerate types and flags, compute densities.
3. Assemble the SDP and apply block-diagonalization.
4. Solve numerically (e.g., using CSDP), producing floating-point solutions.
5. Round to rationals, reconstruct kernel projections as needed, and verify all constraints in exact arithmetic.
6. Output a certificate for independent checking (Falgas-Ravry et al., 2011, Bodnár, 10 Jan 2026).

5. Concrete Applications and Extremal Results

Flag algebra methods have resolved or provided best-known bounds for a variety of longstanding Turán-type and inducibility problems. Highlights include:

• Mantel’s Theorem: maximum edge density in triangle-free graphs is $1/2$.
• Multiple exact Turán densities for 3-graphs with various forbidden configurations: e.g., $\pi(K_4^-, C_5, F_{3,2}) = 12/49$, $\pi(K_4^-, F_{3,2}) = 5/18$, $\pi(J_4, F_{3,2}) = 3/8$, $\pi(F_{3,2}, \text{induced} K_4^-) = 3/8$, $$\pi(K_5, \text{induced 5-set spanning 8 edges}) = 3/4$$ (Falgas-Ravry et al., 2011).
• Classical results such as $\pi(K_4, \text{induced } G_1) = 5/9$ (Razborov), and non-principal pair examples showing strict inequality for pairs of forbidden configurations.

The methodology is broadly applicable to graphs, hypergraphs, tournaments, permutations, and other finite relational structures (Silva et al., 2016, Jeong et al., 19 Jan 2026).

6. Stability, Limitations, and the Complexity Barrier

Extensions of the flag algebra method enable the automated derivation of stability results: for certain extremal problems, all nearly extremal objects are forced to approximate a specific "blow-up" configuration up to small edit distance (Pikhurko et al., 2017). Sufficient conditions for such "perfect stability" (including $\lambda$-minimality, strictness of the optimum, and resistance to local flips) can be checked directly from flag algebra certificates using computational scripts (Pikhurko et al., 2017).

Despite its strength, the method encounters a complexity barrier:

• The SDP size scales with the number of admissible subgraphs, which is polynomial for "simple" blow-up configurations but super-polynomial (and thus infeasible) for "complex" problems involving iterated blow-ups or multiple non-isomorphic extremal constructions.
• Unstable problems (multiple extremal models) and those with infinite families of extremal examples (e.g., Sidorenko-type families) currently elude exact flag algebra solutions (Falgas-Ravry et al., 2011).

Key open directions include adapting flag algebra techniques to "complex" Turán problems and refining the understanding of densities in more intricate forbidden-subgraph settings.

7. Logical and Proof-Theoretic Perspective

Flag algebra reasoning can be viewed as a formal proof system for asymptotic substructure inequalities, with a syntactic layer (density expressions and assertions), an algebraic semantics via positive homomorphisms (evaluation on limits), and a proof strategy comprising labelled sum-of-squares (SOS) certificates and downward transfer to the unlabelled setting via adjoint pairs—analogous to constructs in categorical logic and program semantics (Jeong et al., 19 Jan 2026).

This framework clarifies the skeletal steps involved: encoding the extremal problem, constructing a labelled SOS, transferring to the unlabelled algebra using the downward operator (built from an adjoint pair), and extracting an explicit, verifiable inequality for use in density bounds and combinatorial proofs.

---

References

• (Silva et al., 2016) "Flag Algebras: A First Glance"
• (Raymond et al., 2015) "Symmetry in Turán Sums of Squares Polynomials from Flag Algebras"
• (Falgas-Ravry et al., 2011) "On applications of Razborov's flag algebra calculus to extremal 3-graph theory"
• (Jeong et al., 19 Jan 2026) "An Introduction to Razborov's Flag Algebra as a Proof System for Extremal Graph Theory"
• (Pikhurko et al., 2017) "Strong Forms of Stability from Flag Algebra Calculations"
• (Bodnár, 10 Jan 2026) "FlagAlgebraToolbox: Flag Algebra Computations in SageMath"