REVIEW 5 minor 18 references
Geometric Bounds for the Mean Gauge and the Mean Width in Isotropic Position
T0 review · 0 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash
Pith's one-line read This paper proves deterministic geometric bounds $M(K) \leq C \log(n)/\sqrt{n}$ and $M^*(K) \leq C\sqrt{n}\log(n)$ for origin-symmetric convex bodies in isotropic position, so $M(K)M^*(K) \leq C\log^2 n$.
desk verdict A careful deterministic proof of near-optimal M-M* bounds that trades a logarithmic factor for a clean geometric argument; the technique is the contribution, not the record. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
Centroid bodies are defined by $h_{Z_p(K)}(u)=(E|\langle X,u\rangle|^p)^{1/p}$, and Laplace bodies by $L_p(K)=p\{\Lambda_K\le p\}^{\circ}$, where $\Lambda_K$ is the logarithmic Laplace transform of the uniform measure; the two families are comparable by a known theorem, which is what makes the Laplace-body route possible. The main mechanism is dyadic aggregation: at each scale $p$ a curvature estimate holds only outside a subspace of codimension $O(p)$, and by summing over dyadic scales with weights and applying the min-max principle and the layer-cake formula, the paper extracts global spectral information, namely an inverse-trace bound. Legendre duality turns upper or lower curvature bounds on one body into the complementary bounds on its polar, and the spherical Laplacian identity converts the inverse-trace bound into the desired spherical $L^2$ estimate.
What would settle it
Compute, for an explicit family of isotropic log-concave measures, the centered small-ball probability $\nu(\varepsilon\sqrt{r}\,B_2^r)$ for all $\varepsilon>0$; if any measure exceeds $(C\varepsilon)^r$ with the universal constant required in the proof, then the codimension control in the weighted-covariance lemma and in the tilted-covariance estimate breaks down, and the theorems' conclusions would not follow. A direct disproof would be a sequence of isotropic origin-symmetric bodies with $M(K)M^*(K)$ growing faster than $C\log^2 n$.
Extended reading notes
Core claim
The central claim is that both the isotropic mean gauge and the isotropic mean width can be bounded by deterministic geometry at the orders $\log n/\sqrt{n}$ and $\sqrt{n}\log n$. For $M(K)$, the proof forms a body whose support function squares as $|u|^2$ plus the sum over dyadic $p$ of centroid-body moment terms; a weighted-covariance lemma shows that each dyadic summand has Hessian curvature of order $p$ outside a subspace of codimension $O(p)$, and a dyadic eigenvalue-counting argument converts this into a logarithmic inverse-trace bound. Legendre duality and the spherical Laplacian identity then transfer that bound to the spherical average of the gauge. For $M^*(K)$, the same machine is run on the Laplace bodies associated to the logarithmic Laplace transform; a spectral estimate for tilted covariance matrices supplies the curvature, the dyadic aggregation yields an inverse-trace bound of order $n^2\log^2 n$ on the dual gauge, and the spherical integration gives the mean-width estimate. The arguments are stated for origin-symmetric bodies and extended to centered non-symmetric bodies.
Load-bearing premise
The load-bearing assumption is the dimension-free small-ball estimate that every isotropic log-concave measure in $\mathbb{R}^r$ gives mass at most $(C\varepsilon)^r$ to any ball of radius $\varepsilon\sqrt{r}$; if that estimate failed, the codimension-$O(p)$ curvature lemmas would fail and the bounds would degrade by a factor depending on isotropic constants.
Editorial extensions
If this is right
- Every origin-symmetric isotropic body satisfies $M(K)M^*(K) \leq C\log^2 n$, within one logarithmic factor of the optimal order.
- The mean-width bound $M^*(K) \leq C\sqrt{n}\log(n)$ improves the previous mean-width estimate by one logarithmic factor, while avoiding the heavier functional-analytic ingredients of earlier approaches.
- The mean-gauge argument reaches the optimal order $\log n/\sqrt{n}$ by a purely Euclidean route, without stochastic localization, martingale inequalities, or heat flow.
- Both estimates extend to centered non-symmetric bodies with only minor modifications to the one-sided centroid bodies or the Laplace-body construction.
- Without the slicing theorem, the arguments yield bounds with an extra factor equal to the supremum of the isotropic constants of the relevant marginals, so the method does not give an independent proof of slicing.
Reading between the lines
- The dyadic-aggregation template should apply to other rotationally averaged functionals of log-concave measures whenever one can prove a codimension-$O(p)$ curvature estimate at each scale; $M$ and $M^*$ are the two natural test cases.
- Because the small-ball estimate is the only high-dimensional input, any sharpening of the small-ball bound for centered balls would automatically improve constants in these theorems without changing the geometric argument.
- The Laplace-body formulation suggests that mean-width control is essentially a spectral statement about tilted covariance matrices, so the same estimates may transfer to families of tilted measures and to variance bounds for log-concave measures.
- A natural check is whether the dyadic layer-cake step is what loses the extra logarithm in the $MM^*$ product, by optimizing the aggregation weights or replacing the dyadic sum with a continuous one.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper proves deterministic geometric bounds for the mean gauge and the mean width of origin-symmetric convex bodies in probabilistic isotropic position. The two main results, Theorem 2.1 and Theorem 4.1, give M(K) ≤ C log(n)/√n and M*(K) ≤ C√n log(n), hence M(K)M*(K) ≤ C log² n. The first proof uses a quadratic aggregate of dyadic centroid bodies; the second uses a weighted aggregate of Laplace bodies. The only high-dimensional input is the dimension-free small-ball estimate (1), a consequence of the slicing theorem. Remark 4.3 records explicitly what the arguments give without slicing, replacing absolute constants by the hereditary quantities L_M(K) and L_{M*}(K).
Significance. The estimates themselves are not optimal: the mean-gauge bound was already known with the same order and the mean-width bound was recently improved to optimal up to constants. The value of the paper lies in the method. The proofs avoid stochastic localization, heat flow, and the Milman–Pisier theorem, and they isolate precisely where the slicing theorem enters. The curvature formulas (9) and (20), the spectral small-ball arguments, and the inverse-trace estimates are derived rather than asserted. Remark 4.3 is particularly valuable because it makes the dependence on lower-dimensional isotropic constants fully explicit. The extension to non-symmetric bodies in Remark 4.4 is a useful addition.
minor comments (5)
- [§2, after Eq. (10)] The passage from the per-scale bound #{i : λ_i(H(u)) < 1 + cp} ≤ Cp to the uniform counting estimate N_u(t) ≤ Ct for all t ≥ 1 is compressed into a single sentence. Please spell out the dyadic matching for t near 1 and for t of order n, since this step is load-bearing for the inverse-trace estimate (8).
- [Theorems 2.1 and 4.1] The Hessian computations assume C² smoothness of the auxiliary bodies R, L_p(K), and S, while the hypotheses only give convexity. Please add a standard approximation or regularization remark, or state explicitly that the Hessian identities are understood through approximation, so that the curvature arguments are fully justified at the level of the original body.
- [Remark 4.3] The displayed estimates in Remark 4.3 are labeled (9.1)–(9.3), but the numbering in the main text stops at (28). Please renumber these displays to avoid confusion.
- [§3, after Eq. (13)] The inclusions K ⊆ C Z_n(K) ⊆ C(n/p) Z_p(K) are invoked as 'standard' without proof or reference. A citation to [7] or a one-line explanation would make the derivation of (14) easier to verify.
- [§2, proof of Theorem 2.1] The phrase 'the trivial bound by n for t ≳ n' is terse. Please make explicit that for t of order n one uses N_u(t) ≤ n ≤ Ct, and that the remaining range of t is covered by choosing a dyadic scale p comparable to t.
Circularity Check
No significant circularity: the proofs use external slicing and Laplace-centroid theorems, and the paper explicitly tracks the non-slicing factors.
full rationale
The derivations in Theorems 2.1 and 4.1 are self-contained conditional on two stated external inputs: the dimension-free small-ball estimate (1), cited to the slicing theorem (Klartag-Lehec), and the Laplace-centroid equivalence L_p(K) ≃ Z_p(K), cited to Klartag-Milman [13]. Neither input is supplied by the present paper, and neither input is the target M(K) or M*(K) bound. In the weighted-covariance and tilted-covariance arguments, (1) is applied only to isotropic log-concave marginals or whitened exponential tilts, and the resulting codimension bounds r ≤ Cp and r ≤ 2p are used to count eigenvalues, not to fit constants. The dyadic aggregation, Legendre duality, and spherical-Laplacian steps convert these spectral bounds into the stated inequalities. Remark 4.3 explicitly records the L_M(K), L_M*(K), and L_n factors that would appear without (1), showing that the slicing input is not smuggled in as the conclusion. The only self-citation is the background reference [7] to the authors' own textbook, used for standard facts and not load-bearing. No equation is defined in terms of the target estimate, and no fitted quantity is renamed as a prediction.
Assumptions & free parameters
assumptions (5)
- domain assumption Slicing theorem: L_n ≤ C, yielding the small-ball estimate (1).
- domain assumption Klartag-Milman Laplace-centroid comparison (13).
- standard math Log-concavity of uniform measures on convex bodies and their projections and exponential tilts.
- standard math Logarithmic Berwald / Borell inequality (W ≥ c m^(p-2)).
- standard math Legendre duality, min-max principle, and spherical Laplacian identities for convex homogeneous functions.
Cite this review
Pith. "Pith review of Geometric Bounds for the Mean Gauge and the Mean Width in Isotropic Position." pith.science (2026). https://pith.science/paper/IK4KXDQI
@misc{pith2026260807216,
author = {Pith},
title = {Pith review of: Geometric Bounds for the Mean Gauge and the Mean Width in Isotropic Position},
year = {2026},
howpublished = {\url{https://pith.science/paper/IK4KXDQI}},
note = {Machine review of arXiv:2608.07216}
}
abstract
Let $K \subset \mathbb{R}^n$ be an origin-symmetric convex body and assume that its uniform probability measure is isotropic in the probabilistic normalization, namely \[ \int_K x \otimes x \, d\mu_K(x) = \mathrm{Id}_n. \] We give deterministic geometric proofs of \[ M(K) \leq C \frac{\log(n)}{\sqrt{n}} \qquad \text{and} \qquad M^*(K) \leq C \sqrt{n} \, \log(n), \] where \[ M(K) = \int_{\mathbb{S}^{n-1}} \|\theta\|_K \, d\sigma(\theta), \qquad M^*(K) = \int_{\mathbb{S}^{n-1}} h_K(\theta) \, d\sigma(\theta). \] Combining both estimates yields \[ M(K) M^*(K) \leq C \log^2(n). \] The first proof uses a quadratic aggregate of dyadic centroid bodies. The second uses the analogous weighted aggregate of the Laplace bodies $p\{\Lambda_K \leq p\}^{\circ}$, which are equivalent to the centroid bodies by the work of Klartag and E. Milman. In both cases, curvature at each dyadic scale outside a subspace of codimension $O(p)$ leads, via the min--max principle, Legendre duality, and the spherical Laplacian, to the required estimate. The only high-dimensional input is the dimension-free small-ball consequence of the slicing theorem.
Reference graph
Works this paper leans on
-
[1]
Ball,Logarithmically concave functions and sections of convex sets in Rn, Studia Math.88(1988), no
K. Ball,Logarithmically concave functions and sections of convex sets in Rn, Studia Math.88(1988), no. 1, 69–84
work page 1988
-
[2]
Bizeul,The slicing conjecture via small ball estimates, arXiv:2501.06854, 2025
P. Bizeul,The slicing conjecture via small ball estimates, arXiv:2501.06854, 2025
arXiv 2025
-
[3]
Optimal $MM^*$ bounds for convex bodies
P. Bizeul,OptimalM M ∗ bounds for convex bodies, arXiv:2607.29458, 2026
work page Pith review arXiv 2026
-
[4]
P. Bizeul and B. Klartag,Distances between non-symmetric convex bodies: optimal bounds up to polylog, arXiv:2510.20511, 2025
arXiv 2025
-
[5]
Borell,Convex measures on locally convex spaces, Ark
C. Borell,Convex measures on locally convex spaces, Ark. Mat.12(1974), 239–252
work page 1974
-
[6]
Bourgain,On high-dimensional maximal functions associated to convex bodies, Amer
J. Bourgain,On high-dimensional maximal functions associated to convex bodies, Amer. J. Math.108(1986), no. 6, 1467–1476. 16
work page 1986
-
[7]
S. Brazitikos, A. Giannopoulos, P. Valettas and B.-H. Vritsiou,Geometry of Isotropic Convex Bodies, Mathematical Surveys and Monographs, vol. 196, American Mathemat- ical Society, Providence, RI, 2014
work page 2014
-
[8]
N. Dafnis and G. Paouris,Small ball probability estimates, ψ2-behavior and the hyper- plane conjecture, J. Funct. Anal.258(2010), no. 6, 1933–1964
work page 2010
Show all 18 references
-
[9]
Giannopoulos and E
A. Giannopoulos and E. Milman, M -estimates for isotropic convex bodies and their Lq-centroid bodies, in:Geometric Aspects of Functional Analysis: Israel Seminar (GAF A) 2011–2013, Lecture Notes in Math.2116, Springer, Cham, 2014, 159–182
2011
-
[10]
Giannopoulos, P
A. Giannopoulos, P. Stavrakakis, A. Tsolomitis and B.-H. Vritsiou,Geometry of the Lq-centroid bodies of an isotropic log-concave measure, Trans. Amer. Math. Soc.367 (2015), no. 7, 4569–4593
2015
-
[11]
Guan,A note on Bourgain ’s slicing problem, arXiv:2412.09075, 2024
Q. Guan,A note on Bourgain ’s slicing problem, arXiv:2412.09075, 2024
2024 arXiv
-
[12]
Klartag and J
B. Klartag and J. Lehec,Affirmative resolution of Bourgain ’s slicing problem using Guan ’s bound, Geom. Funct. Anal.35(2025), 1147–1168
2025
-
[13]
Klartag and E
B. Klartag and E. Milman,Centroid bodies and the logarithmic Laplace transform—a unified approach, J. Funct. Anal.262(2012), no. 1, 10–34
2012
-
[14]
Letwin,The KLS constant isO(log 1/4 n), arXiv:2607.24164, 2026
B. Letwin,The KLS constant isO(log 1/4 n), arXiv:2607.24164, 2026
2026 arXiv
-
[15]
Milman,On the mean-width of isotropic convex bodies and their associated Lp- centroid bodies, Int
E. Milman,On the mean-width of isotropic convex bodies and their associated Lp- centroid bodies, Int. Math. Res. Not. IMRN (2015), no. 11, 3408–3423
2015
-
[16]
Paouris,Small ball probability estimates for log-concave measures, Trans
G. Paouris,Small ball probability estimates for log-concave measures, Trans. Amer. Math. Soc.364(2012), no. 1, 287–308
2012
-
[17]
Paouris and R
G. Paouris and R. Pathak,Optimal mean width and metric entropy estimates for convex bodies, arXiv:2607.29522, 2026
2026 arXiv
-
[18]
Vritsiou,Regular ellipsoids and a Blaschke–Santal´ o-type inequality for projections of non-symmetric convex bodies, J
B.-H. Vritsiou,Regular ellipsoids and a Blaschke–Santal´ o-type inequality for projections of non-symmetric convex bodies, J. Funct. Anal.286(2024), no. 11, Paper No. 110414. 17
2024
Reviewed August 10, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.