Kannan–Lovasz–Simonovits (KLS) isoperimetric conjecture

Prove that there exists a universal constant C > 0 such that for every log-concave probability measure μ on R^n, the isoperimetric constant satisfies ψ_μ ≥ C inf_{halfspaces H} μ^+(H)/min{μ(H), μ(R^n \ H)}, equivalently ψ_μ ≥ C / √(||A||_{op}), where A is the covariance of μ; in particular, establish ψ_n bounded below by a universal constant for isotropic log-concave measures.

Background

KLS posits that halfspaces are essentially optimal isoperimetric sets for log-concave measures, implying dimension-free expansion, Poincaré inequalities, concentration, and fast mixing for Markov chains like the ball walk. The best current lower bound is ψ_n ≳ 1/√(log n).

Recent breakthroughs resolved related conjectures (Thin Shell and Bourgain’s slicing), and Thin Shell implies KLS up to √(log n), but a dimension-free KLS bound remains open.

References

Indeed, Kannan, Lov\u00e1sz and Simonovits conjectured in 1995 that for all log-concave measures, the isoperimetric constant is, up to a universal constant, attained by minimizing over half-spaces, like in the Gaussian case: There exists a universal constant $C>0$ such that for every log-concave measure $\mu$, the isoperimetric constant satisfies \begin{align*} \psi_\mu \geq C \inf_{H \subseteq \mathbb{R}n, H \text{ is halfspace} \frac{\mu+(S)}{\min{\mu(S), \mu(\mathbb{R}n\backslash S)}. \end{align*} Equivalently, there exists a universal constant $C>0$ such that \begin{align*} \psi_\mu \geq \frac{C}{\sqrt{|A|_\text{op}, \end{align*} where $A$ denotes the covariance matrix of the measure $\mu$, $A=\mathbb{E} \left(X- \mathbb{E}X\right)\left(X- \mathbb{E}X\right)\top$, where $E$ denotes expectation of $X\sim\mu$.

Randomstrasse101: Open Problems of 2025  (2603.29571 - Bandeira et al., 31 Mar 2026) in Conjecture, Section “The KLS Conjecture and Implications (AR)” (Entry 13)