Expander Decomposition with Almost Optimal Overhead

Abstract: We present the first polynomial-time algorithm for computing a near-optimal \emph{flow}-expander decomposition. Given a graph $G$ and a parameter $φ$, our algorithm removes at most a $φ\log<sup>{1+o(1)}n$ fraction of edges so that every remaining connected component is a $φ$-\emph{flow}-expander (a stronger guarantee than being a $φ$-\emph{cut}-expander). This achieves overhead $\log<sup>{1+o(1)}n$, nearly matching the $Ω(\log n)$ graph-theoretic lower bound that already holds for cut-expander decompositions, up to a $\log<sup>{o(1)}n$ factor. Prior polynomial-time algorithms required removing $O(φ\log<sup>{1.5}n)$ and $O(φ\log<sup>{2}n)$ fractions of edges to guarantee $φ$-cut-expander and $φ$-flow-expander components, respectively.

• The paper presents a polynomial-time algorithm that removes at most φm log(n)exp(√log log n) edges while ensuring every remaining component is a φ-flow-expander.
• The method combines concurrent-flow duality with spreading-metric clustering, multiscale voting, and mass-based recursion to avoid the extra logarithmic losses of sparsest-cut and flow–cut approaches.
• The result improves prior bounds to within a log^{o(1)}n factor of the Ω(log n) lower bound and extends to capacitated graphs, terminal settings, and general node-weightings.

Overview

Expander decomposition is a structural primitive in which a small fraction of edges is removed from an undirected graph so that every remaining connected component has expansion at least $\phi$. The quality of such a decomposition is measured by its overhead $\gamma$: the number of removed edges, expressed as $|C| \le \gamma \phi m$. Existentially, the problem is fully understood: removing $O(\phi m \log n)$ edges always suffices, and the hypercube shows that $\Omega(\log n)$ overhead is necessary even for cut expanders. However, all prior polynomial-time constructions lost additional polylogarithmic factors: $O(\log^{1.5} n)$ overhead for cut-expander decompositions (via the Arora–Rao–Vazirani sparsest-cut approximation) and $O(\log^2 n)$ for flow-expander decompositions (via the Leighton–Rao flow–cut gap).

The paper by Bansal, Jambulapati, and Saranurak (2602.15015) closes this gap up to lower-order factors. Its main theorem states that there is a polynomial-time algorithm that, given an undirected graph and parameter $\phi$, returns an edge set $C$ with $|C| \le \phi m \log(n)\exp(\sqrt{\log\log n})$ such that every component of $\gamma$0 is a $\gamma$1-flow-expander. This improves the state of the art by $\gamma$2 for cut expanders and $\gamma$3 for flow expanders, and it does so for the stronger flow-expansion guarantee rather than merely cut expansion. The result extends to capacitated graphs, terminal versions, and general node-weightings.

Background and definitions

A node-weighting $\gamma$4 is $\gamma$5-cut-expanding in $\gamma$6 if every nontrivial set $\gamma$7 satisfies $\gamma$8. It is $\gamma$9-flow-expanding if every $|C| \le \gamma \phi m$0-respecting demand is routable with congestion at most $|C| \le \gamma \phi m$1. Flow expansion strictly implies cut expansion; conversely, a $|C| \le \gamma \phi m$2-cut-expander is only an $|C| \le \gamma \phi m$3-flow-expander by the flow–cut gap. A key fact used throughout is that the $|C| \le \gamma \phi m$4-product demand $|C| \le \gamma \phi m$5 is, up to a factor of 2, the hardest $|C| \le \gamma \phi m$6-respecting demand: if it routes with congestion $|C| \le \gamma \phi m$7, then $|C| \le \gamma \phi m$8 is $|C| \le \gamma \phi m$9-flow-expanding.

The benchmark lower bound comes from the hypercube: isoperimetric arguments show any decomposition into $O(\phi m \log n)$0-cut-expanders must remove an $O(\phi m \log n)$1 fraction of edges, so $O(\phi m \log n)$2 overhead is the best possible target.

Why polynomial-time algorithms previously fell short

The classical cut-and-recurse procedure achieves the existential bound assuming exact sparsest cuts: find a $O(\phi m \log n)$3-sparse cut, remove it, recurse on both sides; a charging argument in which each vertex lies on the smaller side at most $O(\phi m \log n)$4 times yields $O(\phi m \log n)$5 overhead. In polynomial time this fails twice over. For cut expansion, the best sparsest-cut approximation is $O(\phi m \log n)$6, inflating each cut to $O(\phi m \log n)$7 sparsity and yielding $O(\phi m \log n)$8 overhead. For flow expansion the obstruction is more fundamental: when a graph fails to be a $O(\phi m \log n)$9-flow-expander, no $\Omega(\log n)$0-sparse cut need exist at all, so cut-and-recurse inherently pays the full flow–cut gap and gives $\Omega(\log n)$1 overhead. Improving either approximation factor is itself a major open problem, so circumventing these barriers required a different structural approach.

Algorithmic approach

The algorithm adapts the spreading-metric clustering framework developed by Bansal, Katzelnick, and Schwartz for approximating cutwidth to the expander-decomposition setting. It solves the concurrent multicommodity flow LP for the $\Omega(\log n)$2-product demand together with its dual. If the optimal congestion $\Omega(\log n)$3, the product demand already routes and the component is $\Omega(\log n)$4-flow-expanding, so the recursion stops. Otherwise LP duality supplies edge lengths $\Omega(\log n)$5 with total length at most 1 under which the expected distance between a random pair drawn from the product distribution is at least on the order of $\Omega(\log n)$6 — a spreading metric.

In an idealized setting where every vertex sees roughly the same mass $\Omega(\log n)$7 within radius $\Omega(\log n)$8, and ball masses are stable under constant-factor radius growth, one can take a packing net of size at most $\Omega(\log n)$9, run a sparse neighborhood cover clustering routine around the net terminals, and obtain clusters whose boundary cost is $O(\log^{1.5} n)$0 while each cluster has mass at most $O(\log^{1.5} n)$1, so recursive cost is only $O(\log^{1.5} n)$2. These telescope exactly to $O(\log^{1.5} n)$3.

Since no single scale works globally, the actual algorithm discretizes radii as $O(\log^{1.5} n)$4 and masses as $O(\log^{1.5} n)$5, where $O(\log^{1.5} n)$6 and the number of scales is $O(\log^{1.5} n)$7. Each vertex "votes" for the pair $O(\log^{1.5} n)$8 where its ball-mass growth curve stabilizes logarithmically; a pigeonhole argument guarantees a consensus class carrying at least $O(\log^{1.5} n)$9 of the mass. Clustering at the consensus radius then peels off substantial mass while keeping each cluster's mass bounded by $O(\log^2 n)$0, preserving the telescoping bound up to factors of $O(\log^2 n)$1.

Two technical points deserve emphasis. First, the heavy-cluster case — a vertex whose coarsest ball contains half the total mass — is handled separately via a Leighton–Rao-style sweep cut from the dense core, which finds a $O(\log^2 n)$2-sparse cut with respect to $O(\log^2 n)$3, allowing standard cut-and-recurse at $O(\log^2 n)$4 loss for that level. Second, the authors note explicitly why Seymour-style volume telescoping does not apply: the dual metric changes upon recursion into induced subgraphs, so they telescope instead using true $O(\log^2 n)$5-mass, which is invariant across recursive calls. They also concede that the covering factor of 2 in the packing-net argument is tight (the projective plane is a counterexample to improving it), which is precisely why the final overhead carries an $O(\log^2 n)$6 factor rather than $O(\log^2 n)$7.

Correctness analysis

The proof establishes three properties per recursive step. Validity: every vertex's radius and mass scales lie in $O(\log^2 n)$8, since otherwise the geometric decay of ball masses would force $O(\log^2 n)$9, contradicting integrality of $\phi$0. Expansion: whenever the recursion terminates, the product demand routes with congestion below $\phi$1, giving $\phi$2-flow-expansion in each final component. Cut size: the heavy case contributes at most $\phi$3 times the smaller side's mass; the balanced case contributes $\phi$4 in boundary edges while each cluster satisfies $\phi$5 and the clusters capture at least $\phi$6 mass. Induction on the potential $\phi$7 with $\phi$8 then telescopes cleanly in both cases, because the boundary term depends on $\phi$9 while the recursive term depends on $C$0.

An immediate consequence of the main theorem is that near-optimal flow-expander decompositions are computable in polynomial time, which strengthens the toolkit available to applications requiring flow guarantees — oblivious routing, edge-disjoint paths, all-or-nothing flow, and vertex sparsifiers — without any change to those applications' downstream analyses beyond substituting the improved overhead.

Limitations and open questions

The paper leaves several questions open. The $C$1 slack over the $C$2 lower bound is not known to be removable; whether a truly $C$3-overhead polynomial-time construction exists remains unresolved, and the tightness of the factor-2 covering argument suggests the current technique cannot directly achieve $C$4 overhead. The authors propose two conjectures motivated by their result. First, whether the well-linked decomposition underlying the $C$5-approximation for all-or-nothing flow can be built with $C$6 overhead — noting that the connection is not black-box, since well-linked decomposition interleaves concurrent-flow and maximum-throughput flow steps. Second, whether tree flow sparsifiers of quality $C$7 admit polynomial-time construction, mirroring the existential $C$8 bounds for tree cut sparsifiers against the current $C$9 algorithms. Finally, extensions to vertex-expander and directed-expander decompositions remain open.

Conclusion

This paper gives the first polynomial-time expander decompositions whose overhead matches the $|C| \le \phi m \log(n)\exp(\sqrt{\log\log n})$0 information-theoretic lower bound up to a $|C| \le \phi m \log(n)\exp(\sqrt{\log\log n})$1 factor, and it achieves this for the stronger flow-expansion guarantee. The key insight is a clustering scheme over a spreading metric that votes across discretized radius and mass scales, enabling a telescoping potential based on true node mass that survives the changing dual metrics across recursion. Beyond settling a two-decade-old question about the polynomial-time complexity of expander decomposition, the result narrows the gap between existential and algorithmic guarantees for hierarchical routing structures and provides concrete targets — well-linked decomposition, tree flow sparsification, and directed variants — where analogous improvements may now be attainable.