Small Set Expansion in High Dimensional Expanders

• Small set expansion in high dimensional expanders is defined as the robust boundary expansion of small k-face subsets, generalizing graph edge expansion.
• Techniques include fat machinery, global averaging, and bridging approaches that yield explicit expansion parameters and improved unique-neighbor bounds.
• These expansion properties underpin advances in quantum LDPC codes, locally testable codes, and hardness results for constraint satisfaction problems.

Small set expansion in high dimensional expanders formalizes the robust boundary expansion behavior of small subsets of $k$-faces in simplicial complexes, extending the foundational phenomenon of edge expansion in graphs into higher dimensions. The sharp understanding and explicit realization of small set expansion in these complexes is central to topological combinatorics, coding theory—especially quantum LDPC codes—and the construction of locally testable codes, as well as the development of new hardness-of-approximation frameworks in constraint satisfaction problems.

1. Core Definitions and Expansion Notions

A $d$-dimensional simplicial complex $X$ is a downward-closed set of subsets (“faces”) over a ground set $V$, with $X(k)$ denoting its $(k+1)$-element faces. The principal object of study is the expansion profile of small subsets $A \subseteq X(k)$:

• (α,ε)-Parity small-set expansion: $X$ is an $(\alpha,\epsilon)$-parity expander at dimension $k$ if, for every $A \subseteq X(k)$ with $|A| \leq \alpha|X(k)|$, at least $\epsilon|A|$ of the $(k+1)$-faces contain an odd number of members of $A$.
• (α,ε)-Unique-neighbor-like ($\delta_1$) expansion: For $A \subseteq X(k)$ as above, $X$ is an $(\alpha,\epsilon)$-$\delta_1$ expander if at least $\epsilon|A|$ of the $(k+1)$-faces contain exactly one $k$-face from $A$, formalizing the unique-neighbor (singleton) boundary property familiar in expander graphs.

Unique-neighbor-like expansion is strictly stronger: every $(\alpha,\epsilon)$-$\delta_1$ expander is an $(\alpha,\epsilon)$-parity expander, but not conversely. Constructions exist where parity expansion holds but $\delta_1$-expansion fails, e.g., when $A$ decomposes into disjoint triples such that every $(k+1)$-face meets $A$ in $0$ or $3$ $k$-faces (Kaufman et al., 2022).

2. Techniques and Expansion Regimes

Several approaches have been developed to establish small set expansion in high dimensional expanders, with different strengths and domains of applicability:

• Fat machinery / combinatorial local methods: This approach inductively classifies “fat” lower-dimensional faces and establishes expansion for extremely small sets, leveraging the propagation of local density conditions [KKL14, EK16, KM21 cited in (Kaufman et al., 11 Dec 2025)].
• Global averaging and double-balance: “Double balanced sets” are defined via a multi-level local combinatorial condition: for $f \subseteq X(k)$, in every link, the density of $f$ is controlled (up to a parameter $\alpha$) by the densities in links of lower dimension. This concept generalizes product-space pseudorandomness and ensures strong $\delta_1$-expansion for very small sets in one-sided local spectral expanders, without global Fourier structure (Kaufman et al., 2022). Every nontrivial cohomology class in a cosystolic expander is shown to be double balanced, yielding exponentially improved minimal distance bounds for these codes.
• Global averaging operators: These act between cochains on different dimensions to relate global expansion to local expansion in links; previous implementations covered an exponentially larger regime of set sizes but only showed weak (nonvanishing) expansion [KM22, DD24 as cited in (Kaufman et al., 11 Dec 2025)].
• Bridging approaches: By combining the local "fat machinery" and the global averaging tactics into a two-case analysis (distinguishing whether many faces are "heavy" or not), significant uniform improvements are obtained: strong expansion is established for sets exponentially larger than in previous approaches, and the expansion guarantee is explicit ($\phi(S)\geq\frac{\beta^k}{k!(k+1)^4}$ for set sizes up to normalized weight $\beta^k/(k+2)!$) (Kaufman et al., 11 Dec 2025).

3. Explicit Constructions and Combinatorial Bounds

Recent constructions have achieved explicit families of high dimensional expanders, often based on Ramanujan complexes:

Result/Construction	Key Expansion Achieved	Regime/Notes
Ramanujan clique complex (Hsieh et al., 2024)	Unique-neighbor expansion ≈ 0.6d (beyond ½-barrier)	Small sets in (bi)regular graphs via incidence graphs
Explicit (α,ε)-δ₁ expanders (Kaufman et al., 2022)	Group-independent cosystolic expansion	Ramanujan complexes, λ-local spectral property
Double-balanced sets in one-sided expanders (Kaufman et al., 2022)	δ₁-expansion ≈ (k+2) for tiny sets	One-sided local spectral complexes

The key combinatorial innovation in (Hsieh et al., 2024) is the use of small-set triangle-density bounds in the 4D Ramanujan clique complex, enabling a tripartite line-product construction yielding biregular graphs with two-sided unique-neighbor expansion exceeding the ½ spectral barrier. These tight triangle-density controls prevent the formation of clusters of high internal intersection that would otherwise degrade unique-neighbor counts.

4. Analytical Frameworks: Hypercontractivity and Beyond

The extension of small-set expansion analysis to high-dimensional expanders has leveraged new analytic tools to replace Fourier and product-structure arguments:

• Hypercontractivity via approximate Efron-Stein decomposition: On $\varepsilon$-HDX complexes, hypercontractivity inequalities are proven for global functions (insensitive to small restrictions), leading directly to small-set expansion bounds for indicator functions of small, combinatorially “global” sets (Gur et al., 2021).
• Expansion for global and double-balanced sets: Functions with low influences in all links (global, double-balanced) expand nearly optimally under noise operators, and their indicator sets exhibit strong boundary expansion comparable to the classical Bonami–Beckner–KKL paradigm.
• Spectral-vs-coboundary decoupling: In the improved strong expansion regime (Kaufman et al., 11 Dec 2025), the proof partitions into cases using spectral properties of links (Cheeger’s inequality) for “cocycle-like” functions and combinatorial fat machinery when localized density is high.

5. Structural Consequences and Applications

Small set expansion in high dimensional expanders is a linchpin for several structural and algorithmic properties:

• Cosystolic expansion: Unique-neighbor-like ($\delta_1$) small set expansion is strictly strong enough to yield quantitative cosystolic expansion over arbitrary abelian groups; this in turn underlies the construction of quantum LDPC codes of linear distance and constant rate, as well as explicit CSP lower bounds hard for sum-of-squares algorithms (Kaufman et al., 2022). The explicit dependence $\delta\approx \epsilon^{2^d}/\mathrm{poly}(d,q)$, $\gamma\approx\epsilon^{2^d}$, links small set expansion parameters to cosystolic expansion constants.
• Locally testable codes: Strong small set expansion translates into tight soundness and larger testability windows for locally testable codes, as more errors are detected and corrected in a broader regime of corruptions (Kaufman et al., 11 Dec 2025).
• Quantum codes: Simpler, more robust cosystolic expansion regimes underpin modern explicit quantum LDPC code constructions, improving minimum distance and list-decoding radius.

6. Open Problems, Limitations, and Future Directions

Despite significant advances, several challenges remain:

• Optimal parameter dependence: The 2^d dependence in group-independent cosystolic expansion and related parameters is suboptimal; reducing it to polynomial in d would represent a major improvement (Kaufman et al., 2022).
• Nonabelian extension: Group-independence of cosystolic expansion has been shown only up to $k\leq1$ for nonabelian groups. Generalizing to all $k<d-1$ remains open.
• Beyond Ramanujan complexes: The search for new families of complexes (not based on the Ramanujan/Lubotzky-Samuels-Vishne framework) achieving optimal δ₁-expansion and explicit expansion parameters is active.
• Unified pseudorandomness paradigms: An “orthogonal decomposition” or substitute for high-dimensional pseudorandomness remains elusive. Double-balance is a local proxy, but further refinements may yield stronger or more unified expansion criteria (Kaufman et al., 2022).
• Tight hypercontractivity and expansion constants: Whether the thresholds and constants in globalness and small-set expansion can be improved (e.g., in (Gur et al., 2021)) is unresolved. Improvements in the hypercube would transfer to HDX settings.

7. Connections to Broader Theories

Small set expansion in high dimensional expanders forms a nexus point between combinatorial topology, spectral graph theory, group cohomology, randomized pseudo-randomness, and coding theory. It unifies traditional graph expansion (e.g., via unique-neighbor and vertex expansion), hypercontractivity for functions on product spaces, and the study of hard instances in computational complexity (notably, construction of hard CSPs and sum-of-squares lower bounds). The interplay between explicit combinatorial constructions, spectral and analytic techniques, and their cohomological/topological consequences underpins ongoing research and applications spanning mathematics and theoretical computer science.