Small-Set Vertex Expanders

• Small-set vertex expanders are graphs where every small set of vertices has an external neighborhood at least t times its size, ensuring strong local connectivity.
• They are characterized by the (α,t)-expander condition and related formulations, which influence graph minors, rapid neighborhood growth, and efficient routing paths.
• Their study bridges combinatorial, spectral, and algorithmic methods, leading to explicit constructions with applications in coding theory, distributed computing, and optimization.

Searching arXiv for recent and foundational papers on small-set vertex expanders. Small-set vertex expanders are graphs in which every sufficiently small vertex set has a large external neighborhood relative to its size. A standard formalization used in recent work is the $(\alpha,t)$-expander condition: a graph $G$ with $n$ vertices is an $(\alpha,t)$-expander if for every $X\subseteq V(G)$ with $|X|\le \alpha n/t$, one has $|N(X)|\ge t|X|$ (Krivelevich et al., 10 Mar 2025). Related formulations appear in edge-expansion, conductance, and bipartite-neighborhood settings, and much of the literature studies algorithmic, extremal, probabilistic, and coding-theoretic consequences of these local expansion guarantees. In bounded-degree graphs, small-set edge expansion and small-set vertex expansion are tightly linked up to constant factors, so algorithms and structural theorems stated for edge expansion often translate directly to vertex expansion (Bansal et al., 2011).

1. Definitions and equivalent viewpoints

For a graph $G=(V,E)$ and a set $S\subseteq V$, the external neighborhood is

$$N(S)=\{v\in V\setminus S:\exists u\in S,\ (u,v)\in E\}.$$

A common vertex-expansion ratio is

$G$0

A graph is a small-set vertex expander at scale $G$1 if all sets $G$2 with $G$3 satisfy $G$4 for some $G$5. A more parameterized version, used explicitly in "Minors in small-set expanders" (Krivelevich et al., 10 Mar 2025), is the $G$6-expander condition: $G$7 Here the expansion factor is $G$8, while the small-set regime is controlled by $G$9 (Krivelevich et al., 10 Mar 2025).

A distinct but closely related formalism appears in left-regular bipartite graphs $n$0. For $n$1, one studies the left-to-right neighborhood $n$2, and defines the $n$3-th expansion constant by

$n$4

Such a graph is a $n$5-small-set-expander if every subset $n$6 with $n$7 has at least $n$8 neighbors (Bogart et al., 22 Jun 2026). In that setting, the literature also studies unique-neighbor expansion and lossless expansion, where the neighborhood size is close to the maximum possible $n$9 in a degree-$(\alpha,t)$0 graph (Hsieh et al., 2023, Hsieh et al., 21 Apr 2025, Bogart et al., 22 Jun 2026).

Another viewpoint comes from edge expansion. For an undirected graph $(\alpha,t)$1, one may define

$(\alpha,t)$2

where $(\alpha,t)$3 is the number or weight of edges crossing $(\alpha,t)$4. In bounded-degree graphs with degrees in $(\alpha,t)$5, the inequalities

$(\alpha,t)$6

imply

$(\alpha,t)$7

up to constant factors (Bansal et al., 2011). This is why algorithmic results for small-set edge expansion can often be interpreted as results about finding weakly expanding vertex sets or certifying their absence in bounded-degree graphs (Bansal et al., 2011).

2. Parameter regimes and basic structural consequences

The $(\alpha,t)$8-expander definition emphasizes a regime in which $(\alpha,t)$9 is fixed and $X\subseteq V(G)$0 may grow with $X\subseteq V(G)$1. As $X\subseteq V(G)$2 increases, the admissible set sizes $X\subseteq V(G)$3 shrink, but the required expansion factor increases (Krivelevich et al., 10 Mar 2025). This local nature distinguishes small-set expansion from global expansion up to $X\subseteq V(G)$4, and it permits behaviors unavailable in standard global notions.

One consequence is rapid neighborhood growth. If $X\subseteq V(G)$5 is an $X\subseteq V(G)$6-expander and $X\subseteq V(G)$7 denotes the radius-$X\subseteq V(G)$8 ball around a set $X\subseteq V(G)$9, then

$|X|\le \alpha n/t$0

so balls grow exponentially until they hit size $|X|\le \alpha n/t$1 (Krivelevich et al., 10 Mar 2025). This yields a diameter bound for connected $|X|\le \alpha n/t$2-expanders: $|X|\le \alpha n/t$3 (Krivelevich et al., 10 Mar 2025). A plausible implication is that large $|X|\le \alpha n/t$4 forces short routing paths even though expansion is guaranteed only for relatively small sets.

A separate distributed-computing formalism studies $|X|\le \alpha n/t$5-vertex expanders, defined by

$|X|\le \alpha n/t$6

In that framework, there are no $|X|\le \alpha n/t$7-expanders for $|X|\le \alpha n/t$8, while if $|X|\le \alpha n/t$9 then the graph has diameter at most $|N(X)|\ge t|X|$0 (Cruciani et al., 25 Aug 2025). The paper stresses that the regime $|N(X)|\ge t|X|$1 is possible only because expansion is required only up to $|N(X)|\ge t|X|$2, not up to $|N(X)|\ge t|X|$3 (Cruciani et al., 25 Aug 2025). This suggests that small-set vertex expansion can support sublogarithmic or even constant-diameter behavior under local rather than global hypotheses.

In the left-regular bipartite setting, optimal local expansion is constrained combinatorially. If $|N(X)|\ge t|X|$4 has left degree $|N(X)|\ge t|X|$5 and $|N(X)|\ge t|X|$6, then

$|N(X)|\ge t|X|$7

More generally, if $|N(X)|\ge t|X|$8, then for every integer $|N(X)|\ge t|X|$9,

$G=(V,E)$0

(Bogart et al., 22 Jun 2026). The quantity $G=(V,E)$1 is therefore a natural benchmark for optimal small-set expansion in such graphs (Bogart et al., 22 Jun 2026).

3. Extremal and structural theory

A central recent theme is that strong local vertex expansion forces rich minor structure. "Minors in small-set expanders" (Krivelevich et al., 10 Mar 2025) proves that for every fixed $G=(V,E)$2 there exist $G=(V,E)$3 and $G=(V,E)$4 such that every $G=(V,E)$5-expander on $G=(V,E)$6 vertices with $G=(V,E)$7 contains a complete minor of order

$G=(V,E)$8

The same paper proves that for every fixed $G=(V,E)$9 there exist $S\subseteq V$0 and $S\subseteq V$1 such that every $S\subseteq V$2-expander with $S\subseteq V$3 and $S\subseteq V$4 is $S\subseteq V$5-minor-universal for

$S\subseteq V$6

meaning it contains every graph $S\subseteq V$7 with at most $S\subseteq V$8 vertices and at most $S\subseteq V$9 edges as a minor (Krivelevich et al., 10 Mar 2025).

These lower bounds are complemented by an upper bound valid for arbitrary graphs with average degree $$N(S)=\{v\in V\setminus S:\exists u\in S,\ (u,v)\in E\}.$$0: there exists a graph with at most

$$N(S)=\{v\in V\setminus S:\exists u\in S,\ (u,v)\in E\}.$$1

vertices and edges that is not a minor of $$N(S)=\{v\in V\setminus S:\exists u\in S,\ (u,v)\in E\}.$$2 (Krivelevich et al., 10 Mar 2025). The paper therefore concludes that when $$N(S)=\{v\in V\setminus S:\exists u\in S,\ (u,v)\in E\}.$$3 for a constant $$N(S)=\{v\in V\setminus S:\exists u\in S,\ (u,v)\in E\}.$$4, the minor-universality of $$N(S)=\{v\in V\setminus S:\exists u\in S,\ (u,v)\in E\}.$$5-expanders is optimal up to constant factors (Krivelevich et al., 10 Mar 2025). In its words, such small-set expanders are optimal graphs with the average degree $$N(S)=\{v\in V\setminus S:\exists u\in S,\ (u,v)\in E\}.$$6 from the point of view of minor-universality (Krivelevich et al., 10 Mar 2025).

The proofs rely on structural consequences of local expansion. One lemma yields a large induced subgraph $$N(S)=\{v\in V\setminus S:\exists u\in S,\ (u,v)\in E\}.$$7 that remains a $$N(S)=\{v\in V\setminus S:\exists u\in S,\ (u,v)\in E\}.$$8-expander and satisfies a robust connectivity property: after deleting a small set $$N(S)=\{v\in V\setminus S:\exists u\in S,\ (u,v)\in E\}.$$9, any partition $G$00 with $G$01 and $G$02 both sufficiently large still has an edge between $G$03 and $G$04 (Krivelevich et al., 10 Mar 2025). Another lemma shows that in a connected $G$05-expander, a connected set can be found that intersects many prescribed subsets $G$06, with size bounded by

$G$07

when each $G$08 and $G$09 (Krivelevich et al., 10 Mar 2025). These facts support iterative construction of branch sets for clique minors and sparse minors.

A different structural characterization arises in bipartite optimal expanders. "Optimal Small Set Expanders and Their Codes" (Bogart et al., 22 Jun 2026) proves that for a left-regular bipartite graph of degree $G$10, $G$11-optimality,

$G$12

is equivalent to the girth condition

$G$13

(Bogart et al., 22 Jun 2026). The forward implication uses the fact that the induced subgraph on $G$14 is a forest whenever $G$15, yielding

$G$16

(Bogart et al., 22 Jun 2026). The reverse implication uses a shortest cycle of length $G$17 to produce a set $G$18 with $G$19, contradicting $G$20-optimality (Bogart et al., 22 Jun 2026). This gives an exact combinatorial description of optimal small-set expansion in terms of local tree-likeness.

4. Random, spectral, and explicit constructions

Random regular graphs have long been known to be good expanders. "Local Improvement Gives Better Expanders" (Lampis, 2012) improves Bollobás’s lower bounds on the edge expansion of random $G$21-regular graphs by observing that any low-expansion set can be transformed by local improvement into a locally optimal one, and such sets are much rarer (Lampis, 2012). The paper studies edge expansion

$G$22

where $G$23 is the number of cut edges, and proves improved lower bounds for $G$24 (Lampis, 2012). Since in a $G$25-regular graph

$G$26

these lower bounds imply corresponding lower bounds on vertex expansion up to a factor $G$27 (Lampis, 2012). A plausible implication is that random regular graphs are stronger small-set vertex expanders than earlier union-bound analyses could certify.

Spectral expanders also provide natural examples of small-set vertex expanders. "Minors in small-set expanders" (Krivelevich et al., 10 Mar 2025) states that if $G$28 is an $G$29-graph with $G$30, then $G$31 is a $G$32-expander (Krivelevich et al., 10 Mar 2025). This directly converts a spectral gap into local vertex expansion with expansion factor $G$33. The same paper positions random regular graphs, binomial random graphs $G$34 with $G$35, $G$36-out random graphs, and spectral expanders $G$37 with $G$38 as typical sources of small-set expanders (Krivelevich et al., 10 Mar 2025).

Several papers address explicit constructions with stronger guarantees. "Explicit two-sided unique-neighbor expanders" (Hsieh et al., 2023) gives the first explicit two-sided construction of imbalanced unique-neighbor expanders. For large enough degrees $G$39, it constructs strongly explicit $G$40-biregular graphs in which small sets on both sides have $G$41 unique neighbors, and subsets of size up to

$G$42

expand losslessly (Hsieh et al., 2023). The analysis uses a tripartite line product and a sharp characterization of induced subgraphs in biregular spectral expanders via the non-backtracking matrix and a generalized Moore bound (Hsieh et al., 2023).

"Explicit Lossless Vertex Expanders" (Hsieh et al., 21 Apr 2025) gives the first explicit constant-degree lossless vertex expanders. For any $G$43 and sufficiently large $G$44, it constructs explicit $G$45-regular graphs where every small set $G$46 has

$G$47

which implies

$G$48

for the unique-neighbor set $G$49 (Hsieh et al., 21 Apr 2025). The construction extends to biregular bipartite graphs of any constant imbalance, gives two-sided expansion on both sides, and admits a free group action, yielding new families of quantum LDPC codes of Lin and M. Hsieh with a linear time decoding algorithm (Hsieh et al., 21 Apr 2025).

At the bipartite local-optimality scale, "Optimal Small Set Expanders and Their Codes" (Bogart et al., 22 Jun 2026) proves existence of $G$50-optimal expanders for every $G$51 by random selection and cycle-removal from families of $G$52-optimal graphs (Bogart et al., 22 Jun 2026). It also proves that an $G$53-optimal expander must satisfy

$G$54

showing that $G$55-optimality forces the number of left vertices to grow exponentially in $G$56 for fixed degree $G$57 (Bogart et al., 22 Jun 2026).

5. Algorithmic approximation and testing

Small-set vertex expansion is algorithmically subtle. A direct approximation result for vertex boundary minimization appears in "New Approximation Bounds for Small-Set Vertex Expansion" (Ghoshal et al., 2023). For a graph $G$58 and $G$59, the $G$60-Small-Set Vertex Expansion value is

$G$61

For $G$62, this is simply $G$63 (Ghoshal et al., 2023). The paper gives a randomized algorithm, running in time $G$64, that outputs a set $G$65 of size $G$66 with vertex expansion at most

$G$67

where $G$68 is the maximum degree (Ghoshal et al., 2023). The method uses the basic SDP relaxation augmented with $G$69 rounds of the Lasserre/SoS hierarchy and a Gaussian hyperedge rounding lemma (Ghoshal et al., 2023). The paper also proves integrality gaps showing that the $G$70 term reflects genuine limitations of the relaxation (Ghoshal et al., 2023).

An earlier algorithmic route goes through hypergraph and edge-expansion reductions. "Approximation Algorithms for Hypergraph Small Set Expansion and Small Set Vertex Expansion" (Louis et al., 2014) gives an $G$71 approximation algorithm for hypergraph small-set expansion, and via reductions obtains an $G$72 approximation for Small Set Vertex Expansion (Louis et al., 2014). It also gives an algorithm that finds a set with vertex expansion

$G$73

up to the polylogarithmic factors detailed in the paper (Louis et al., 2014). The analysis introduces hypergraph orthogonal separators and $G$74–$G$75 separators, extending the graph-orthogonal-separator paradigm to hypergraphs (Louis et al., 2014).

The edge-expansion literature remains highly relevant because bounded-degree edge and vertex expansion are comparable. "^^^^10^^^^" (Bansal et al., 2011) gives a polynomial-time randomized algorithm that, for a parameter $G$76, returns a set $G$77 of size $G$78 and edge expansion

$G$79

in general graphs, and constant-factor bicriteria approximations in minor-free and bounded-genus graphs (Bansal et al., 2011). In bounded-degree graphs, these algorithms can be viewed as finding sets with poor vertex expansion or certifying their absence (Bansal et al., 2011).

Property testing provides a sublinear-time perspective. "Testing Small Set Expansion in General Graphs" (Li et al., 2012) studies conductance-based $G$80-expansion and gives testers in the adjacency-list and rotation-map models. The two-sided error tester in the adjacency-list model distinguishes $G$81-expanders from graphs $G$82-far from any $G$83-expander with

$G$84

in time $G$85 (Li et al., 2012). In bounded-degree graphs, such testers can be interpreted as testers for small-set vertex expansion because conductance and vertex expansion are comparable up to constants (Li et al., 2012).

For the bipartite version, "Minimizing the Union: Tight Approximations for Small Set Bipartite Vertex Expansion" (Chlamtáč et al., 2016) studies the problem of finding $G$86 of size $G$87 minimizing $G$88 in a bipartite graph $G$89. The paper proves an $G$90-approximation algorithm for SSBVE for any constant $G$91, where $G$92, and shows matching hardness under an extension of the Dense vs Random conjecture to hypergraphs (Chlamtáč et al., 2016). It also gives a $G$93-bicriteria approximation for the more general SSVE problem on arbitrary graphs (Chlamtáč et al., 2016).

6. Expansion in partitions, geometry, and optimization

Small-set expansion governs more global partitioning tasks. In "Min-Max Graph Partitioning and Small Set Expansion" (Bansal et al., 2011), the main partitioning theorem states that Min–Max $G$94-Partitioning admits a bicriteria approximation

$G$95

and the proof uses weighted small-set expansion and weighted $G$96-unbalanced cut as the central primitives (Bansal et al., 2011). The algorithm repeatedly finds small low-boundary pieces and aggregates them into a global partition, so the existence or absence of weakly expanding small sets directly controls the achievable partition quality (Bansal et al., 2011). In bounded-degree graphs, the same story can be rephrased in vertex-expansion language.

Small-set expansion also appears in geometric rounding and spectral partitioning. "Rounding via Low Dimensional Embeddings" (Braverman et al., 2022) works with $G$97 small-set edge expanders and shows that several square-root losses in classical rounding can be avoided. One result states that if a regular graph is an $G$98 small-set expander and contains a cut of fractional size at least $G$99, then one can find a cut of size at least

$n$00

in polynomial time (Braverman et al., 2022). The paper’s main idea is to project a high-dimensional SDP solution into low dimension while roughly preserving $n$01 distances, then use low-dimensional geometry and small-set expansion to quantize the vector solution without incurring the usual square-root degradation (Braverman et al., 2022). In bounded-degree graphs, a plausible implication is that strong small-set vertex expansion should support analogous vertex-boundary rounding schemes, since the underlying obstruction is concentration of geometry on small poorly connected sets.

A distinct geometric-group-theoretic perspective appears in "Remarks on partitions into expanders" (Vigolo, 2020). There the basic objects are $n$02-F sets, namely nonempty vertex subsets $n$03 with $n$04 and $n$05, and the main theorem states that if a finite graph $n$06 has no $n$07-small $n$08-F sets, then it can be partitioned into

$n$09

where each $n$10 is $n$11-big and a $n$12-expander for

$n$13

(Vigolo, 2020). Since the work is edge-based, in bounded-degree settings this also yields partitions into large vertex-expanding pieces (Vigolo, 2020). The same paper proves that the existence of such partitions is a quasi-isometry invariant (Vigolo, 2020).

7. Unique neighbors, lossless expansion, and codes

A major branch of the subject concerns stronger neighborhood guarantees than mere linear expansion. In a graph $n$14, a vertex is a unique neighbor of a set $n$15 if it has exactly one neighbor in $n$16. Unique-neighbor expansion is stronger than ordinary vertex expansion, because it forbids excessive collisions in the neighborhood map (Hsieh et al., 2023).

"Explicit two-sided unique-neighbor expanders" (Hsieh et al., 2023) studies bipartite $n$17-biregular graphs $n$18 in which small sets on both sides have many unique neighbors. It gives the first strongly explicit infinite family of two-sided imbalanced unique-neighbor expanders, and additionally guarantees that subsets of size up to

$n$19

expand losslessly on both sides (Hsieh et al., 2023). The constructions are obtained from a tripartite line product of a large tripartite spectral expander and a constant-size unique-neighbor gadget (Hsieh et al., 2023).

"Explicit Lossless Vertex Expanders" (Hsieh et al., 21 Apr 2025) sharpens this by giving explicit constant-degree lossless vertex expanders. In a $n$20-regular graph, lossless means that every small set $n$21 has

$n$22

with $n$23 as $n$24 in a family (Hsieh et al., 21 Apr 2025). The paper proves this for explicit constant-degree graphs and for biregular bipartite graphs of any constant imbalance (Hsieh et al., 21 Apr 2025). It further notes that the lossless condition implies at least $n$25 unique neighbors (Hsieh et al., 21 Apr 2025). These graphs admit a free group action and therefore realize new families of quantum LDPC codes of Lin and M. Hsieh with a linear time decoding algorithm (Hsieh et al., 21 Apr 2025).

"Optimal Small Set Expanders and Their Codes" (Bogart et al., 22 Jun 2026) connects local expansion constants to coding properties in another way. Given a left-regular bipartite graph $n$26 of degree $n$27, it defines a binary linear code $n$28 by parity checks indexed by $n$29. The paper recalls the Sipser–Spielman criterion that if $n$30, then the minimum distance is at least $n$31, and if $n$32, then the standard bit-flipping algorithm corrects up to $n$33 errors in linear time (Bogart et al., 22 Jun 2026). Using transfer bounds,

$n$34

in $n$35-optimal expanders (Bogart et al., 22 Jun 2026), it derives decoding and cryptographic guarantees for expander-code-based key exchange protocols (Bogart et al., 22 Jun 2026).

8. Small-set expansion in complexity and optimization theory

Small-set expansion, especially in its edge form, is deeply connected to the Unique Games Conjecture and the Small-Set Expansion Hypothesis. "Testing Small Set Expansion in General Graphs" (Li et al., 2012) explicitly situates small-set expansion within the Unique Games and locally testable codes literature (Li et al., 2012). "New Approximation Bounds for Small-Set Vertex Expansion" (Ghoshal et al., 2023) similarly emphasizes the role of SSVE in the Strong Unique Games problem and in reductions between graph and hypergraph partitioning (Ghoshal et al., 2023).

A more direct SoS-based perspective appears in "Playing Unique Games on Certified Small-Set Expanders" (Bafna et al., 2020). That paper works with edge expansion and certifiable hypercontractivity, but it proves that if low-degree sum-of-squares proofs certify good small-set expansion of the underlying constraint graph, then one can solve affine Unique Games instances on that graph in polynomial time (Bafna et al., 2020). As corollaries, it gives the first polynomial-time algorithms for Unique Games on the noisy hypercube, the short code, and the Johnson graph (Bafna et al., 2020). The same paper also treats graphs that are not literal small-set expanders but whose non-expanding small sets are characterized by low-degree SoS proofs, notably the Johnson graph via restricted subcubes (Bafna et al., 2020). A plausible implication is that one fruitful notion of a “usable” small-set vertex expander in complexity theory is not only combinatorial expansion, but expansion or non-expansion that is recognizable inside a low-degree proof system.

This complexity-theoretic perspective intersects with approximation barriers. For bipartite small-set vertex expansion, "Minimizing the Union" (Chlamtáč et al., 2016) shows that the $n$36-approximation is tight under a hypergraph extension of the Dense vs Random conjecture, and it provides a matching $n$37 Sherali–Adams integrality gap and an even worse SDP gap (Chlamtáč et al., 2016). For general SSVE, "New Approximation Bounds for Small-Set Vertex Expansion" (Ghoshal et al., 2023) proves SSEH-based hardness showing that even approximations depending arbitrarily on the degree can remain hard (Ghoshal et al., 2023). These results indicate that the algorithmic theory of small-set vertex expanders inherits much of the subtlety, and many of the barriers, already familiar from edge-based small-set expansion.

9. Dynamic and distributed consequences

Small-set vertex expansion also has dynamic consequences for information spread. "Towards Constant Time Multi-Call Rumor Spreading on Small-Set Expanders" (Cruciani et al., 25 Aug 2025) studies the $n$38-PUSH&PULL rumor-spreading protocol on $n$39-vertex expanders, where

$n$40

For $n$41 and $n$42, the paper proves a diameter bound $n$43 (Cruciani et al., 25 Aug 2025). It then shows that $n$44-PUSH&PULL completes in

$n$45

rounds with high probability, and in the common regime simplifies this to

$n$46

(Cruciani et al., 25 Aug 2025). The paper also proves a lower bound

$n$47

(Cruciani et al., 25 Aug 2025). This sharpens the contrast with conductance and edge expansion, where the best possible dependence on $n$48 remains logarithmic (Cruciani et al., 25 Aug 2025). The result identifies small-set vertex expansion as a graph property that is particularly well suited for fast information dissemination.

10. Conceptual synthesis

Several distinct notions in the literature fall under the label “small-set vertex expander.” One is the direct neighborhood condition $n$49 for $n$50 (Krivelevich et al., 10 Mar 2025). Another is the minimization problem of finding a size-$n$51 set with minimum vertex boundary (Ghoshal et al., 2023). A third is the bipartite-neighborhood formalism $n$52 for left subsets of size at most $n$53 (Bogart et al., 22 Jun 2026). Still others arise from unique-neighbor or lossless requirements (Hsieh et al., 2023, Hsieh et al., 21 Apr 2025). These formulations are not identical, but they share the same governing principle: small subsets should force many external contacts, preferably with low collision.

A recurring theme is that local vertex expansion simultaneously behaves as a combinatorial, geometric, and algorithmic object. Combinatorially, it controls minors, diameter, and tree-like local structure (Krivelevich et al., 10 Mar 2025, Bogart et al., 22 Jun 2026). Geometrically, it supports quantization and low-dimensional rounding without square-root losses (Braverman et al., 2022). Algorithmically, it underlies approximation schemes, bicriteria partitioning routines, and sublinear testing procedures (Bansal et al., 2011, Li et al., 2012, Louis et al., 2014, Ghoshal et al., 2023). In coding theory and pseudorandom constructions, stronger variants such as unique-neighbor and lossless expansion are the key combinatorial drivers of decoding and robustness (Hsieh et al., 2023, Hsieh et al., 21 Apr 2025, Bogart et al., 22 Jun 2026).

A common misconception is that edge expansion and vertex expansion are interchangeable in complete generality. They are not: several papers explicitly stress that edge and vertex expansion are incomparable in general (Krivelevich et al., 10 Mar 2025, Hsieh et al., 2023). However, in bounded-degree graphs, the two notions are tightly linked by the inequalities relating crossing edges and distinct boundary vertices (Bansal et al., 2011, Lampis, 2012, Li et al., 2012). This is why many of the deepest algorithmic advances still enter through edge expansion, then translate to vertex expansion in the sparse regime most relevant to expanders.

Another misconception is that strong local expansion must be a purely asymptotic or nonconstructive property. Recent work shows otherwise. There are now explicit constant-degree lossless vertex expanders (Hsieh et al., 21 Apr 2025), explicit two-sided unique-neighbor expanders (Hsieh et al., 2023), and explicit or probabilistic constructions of $n$54-optimal bipartite expanders for every $n$55 (Bogart et al., 22 Jun 2026). These developments indicate that the subject has moved well beyond probabilistic existence and now supports finely structured constructions with algebraic symmetry and coding applications.

Small-set vertex expanders therefore occupy a central position at the intersection of graph minors, pseudorandomness, semidefinite and SoS optimization, coding theory, and distributed algorithms. Their modern theory is best viewed not as a single definition, but as a network of equivalent and near-equivalent local expansion principles whose precise form depends on whether the goal is structural analysis, algorithm design, hardness, or explicit construction.