The thin shell conjecture (variance of $|X|^2$)
Description of constant
Let $X$ be a random vector in $\mathbb R^n$ with an isotropic log-concave distribution (i.e. $X$ has a log-concave density, $\mathbb E X=0$, and $\mathrm{Cov}(X)=\mathrm{Id}$). Since $X$ is isotropic, $\mathbb E|X|^2 = n$.
We define $C_{20a}$ to be the smallest constant such that
\[\mathrm{Var}(|X|^2) \;=\; \mathbb E\bigl(|X|^2-n\bigr)^2 \ \le\ C_{20a}\,n\]for every dimension $n$ and every isotropic log-concave $X$ in $\mathbb R^n$.
Equivalently,
\[C_{20a}=\sup_{n\ge 1}\ \sup_{X}\ \frac{\mathrm{Var}(|X|^2)}{n},\]where the inner supremum is over isotropic log-concave $X$ in $\mathbb R^n$.
This “variance” formulation implies the more common thin-shell estimate
\[\mathbb E\bigl(|X|-\sqrt{n}\bigr)^2 \ \le\ \frac{1}{n}\,\mathbb E\bigl(|X|^2-n\bigr)^2 \ \le\ C_{20a},\]so boundedness of $C_{20a}$ means that $\lvert X\rvert$ concentrates in a shell of constant width around $\sqrt n$.
Known upper bounds
Historically, results were often phrased in terms of the (dimension-dependent) thin-shell width
\[\sigma_n^2 := \sup_X \mathbb E\bigl(|X|-\sqrt n\bigr)^2,\]where the supremum is over isotropic log-concave $X$ in $\mathbb R^n$. Any bound $\sigma_n \le f(n)$ is evidence toward (and is closely related to) boundedness of $C_{20a}$.
| Bound | Reference | Comments |
|---|---|---|
| $\sigma_n \le O\left(\sqrt{\frac{n}{\log n}}\right)$ | [K2007a] | First nontrivial bound. |
| $\sigma_n \le O\left(n^{2/5+o(1)}\right)$ | [K2007b] | Improvement via power-law CLT methods. |
| $\sigma_n \le O(n^{3/8})$ | [Fle2010] | Further improvement. |
| $\sigma_n \le O(n^{1/3})$ | [GM2011] | “Thin-shell / large deviation interpolation” bound. |
| $\sigma_n \le O(n^{1/4})$ | [LV2017] | Via Eldan’s stochastic localization. |
| $\sigma_n \le \exp\bigl((\log n)^{1/2+o(1)}\bigr)=n^{o(1)}$ | [Che2021] | First subpolynomial bound (via near-constant KLS). |
| $\sigma_n \le O(\log^4 n)$ | [KL2022] | Polylog bound. |
| $\sigma_n \le O(\log^{2.23\ldots} n)$ | [JLV2022] | Improves the polylog exponent. |
| $\sigma_n \le O(\sqrt{\log n})$ | [K2023] | Further improvement. |
| $\sigma_n \le O(\log\log n)$ | [Gua2024] | Based on a $\log\log n$ KLS bound. |
| $\sigma_n \le O(1)$ (and in fact $\mathrm{Var}(\lvert X\rvert^2)\le C n$) | [KL2025] | Affirmative resolution of the thin shell conjecture. The universal constant is not optimized (and is not made explicit). |
Known lower bounds
| Bound | Reference | Comments |
|---|---|---|
| $0$ | Trivial | By definition. |
| $4/5 = 0.8$ | [KL2025] | Achieved by the cube (for the variance formulation). |
| $2$ | [KL2025] | Achieved by the standard Gaussian: if $X\sim N(0,\mathrm{Id})$ then $\mathrm{Var}(\lvert X\rvert^2)=2n$, so $C_{20a}\ge 2$. |
Additional comments and links
- Thin shell implies slicing (hyperplane) bounds (see [EK2011]). Thus [KL2025] gives (independently) slicing-type control, complementing the direct slicing resolution [KL2024].
- Many of the improvements toward thin shell proceeded via progress on the KLS isoperimetric constant; see [Che2021], [KL2022], [JLV2022], [K2023], [Gua2024].
References
- [ABP2003] Anttila, M.; Ball, K.; Perissinaki, I. The central limit problem for convex bodies. Trans. Amer. Math. Soc. 355 (2003), no. 12, 4723–4735.
- [BK2003] Bobkov, S. G.; Koldobsky, A. On the central limit property of convex bodies. In: Geometric aspects of functional analysis (2001–02), Lecture Notes in Math. 1807, Springer (2003), 44–52.
- [Che2021] Chen, Y. An almost constant lower bound of the isoperimetric coefficient in the KLS conjecture. Geom. Funct. Anal. 31 (2021), no. 1, 34–61.
- [EK2011] Eldan, R.; Klartag, B. Approximately Gaussian marginals and the hyperplane conjecture. In: Concentration, functional inequalities and isoperimetry, Contemp. Math. 545, Amer. Math. Soc. (2011), 55–68.
- [Fle2010] Fleury, B. Concentration in a thin Euclidean shell for log-concave measures. J. Funct. Anal. 259 (2010), no. 4, 832–841.
- [GM2011] Guédon, O.; Milman, E. Interpolating thin-shell and sharp large-deviation estimates for isotropic log-concave measures. Geom. Funct. Anal. 21 (2011), no. 5, 1043–1068.
- [Gua2024] Guan, Q. A note on Bourgain’s slicing problem. Preprint (2024). arXiv:2412.09075
- [JLV2022] Jambulapati, A.; Lee, Y. T.; Vempala, S. S. A slightly improved bound for the KLS constant. Preprint (2022). arXiv:2208.11644
- [K2007a] Klartag, B. A central limit theorem for convex sets. Invent. Math. 168 (2007), no. 1, 91–131.
- [K2007b] Klartag, B. Power-law estimates for the central limit theorem for convex sets. J. Funct. Anal. 245 (2007), no. 1, 284–310.
- [K2023] Klartag, B. Logarithmic bounds for isoperimetry and slices of convex sets. Ars Inveniendi Analytica, Paper No. 4 (2023), 17pp.
- [KL2022] Klartag, B.; Lehec, J. Bourgain’s slicing problem and KLS isoperimetry up to polylog. Geom. Funct. Anal. 32 (2022), no. 5, 1134–1159.
- [KL2024] Klartag, B.; Lehec, J. Affirmative resolution of Bourgain’s slicing problem using Guan’s bound. Preprint (2024). arXiv:2412.15044
- [KL2025] Klartag, B.; Lehec, J. Thin-shell bounds via parallel coupling. Preprint (2025). arXiv:2507.15495
- [LV2017] Lee, Y. T.; Vempala, S. Eldan’s stochastic localization and the KLS hyperplane conjecture: an improved lower bound for expansion. In: FOCS 2017, 998–1007.
- [Pao2006] Paouris, G. Concentration of mass on convex bodies. Geom. Funct. Anal. 16 (2006), no. 5, 1021–1049.
Acknowledgements
Prepared with ChatGPT 5.2 Pro.