Talk:Hermite constant: Difference between revisions

Bacher, Roland (2008). “A new inequality for the Hermite constants”. International Journal of Number Theory. 4 (3): 363–386. doi:10.1142/S1793042108001390.

66.177.56.143 (talk) 13:57, 14 July 2008 (UTC)[reply]

The article says that “the Hermite constant grows linearly in n as n becomes unbounded”,
but the “estimates” are clearly exponential and not linear! —Erel Segal (talk) 12:29, 19 July 2015 (UTC)[reply]

I’m confused. A fundamental region of a hexagonal lattice is a 120 degree rhombus. When the area of such a rhombus is 1, the resulting side lengths are not 2/sqrt(3). Instead, the length is sqrt(2/sqrt(3)). Am I incorrect in thinking that the shortest distance between 2 points in the lattice is supposed to be 2/sqrt(3)? BagLuke (talk) 20:02, 7 October 2023 (UTC)[reply]

David Eppstein Mind taking a look? Don’t want my image to be incorrect, and want to know how to correct it if it is. BagLuke (talk) 21:23, 9 February 2025 (UTC)[reply]

The minimum length is ${\displaystyle \lambda _{2}}$. The Hermite constant ${\displaystyle \gamma _{2}}$ is the square of that length. The rational number 4/3 (for dimension two) in the table in the example section is the square of the Hermite constant. I think your image is ok but its caption is wrong. The caption writes that the Hermite constant is the length. It is not. It is the square of the length. —David Eppstein (talk) 21:38, 9 February 2025 (UTC)[reply]

David Eppstein, Gotcha, made the correction. Still confused as to why ${\displaystyle \lambda _{2}}$ is 2/sqrt(3), when a fundamental region of a hexagonal lattice, a 120 degree rhombus, only has area 1 when the side length (least length of a nonzero element in that lattice) is sqrt(2/sqrt(3)), not 2/sqrt(3). Such a rhombus with side length 2/sqrt(3) has area about 1.1547. I’m definitely missing something about the premise. Thanks, BagLuke (talk) 22:17, 9 February 2025 (UTC)[reply]

No no no. Your caption is still wrong. lambda_2, the side length of the fundamental region, is sqrt(2/sqrt3). But that is not the Hermite constant. The Hermite constant, gamma_2, is the square of that, 2/sqrt3. —David Eppstein (talk) 23:41, 9 February 2025 (UTC)[reply]

Ohhh, so the length of the side square is the Hermite constant, and that squared is the value in the table. Thanks! BagLuke (talk) 01:15, 10 February 2025 (UTC)[reply]

I might speculate that one squaring is to make it dimensionless as a ratio with the arbitrary choice of unit area for the domain, and the second squaring is to make it rational? But really I have no idea. I agree it’s confusing. —David Eppstein (talk) 08:53, 10 February 2025 (UTC)[reply]

The article cite the paper:
– A conjecture on Hermite constants, L Mächler, D Naccache – Cryptology ePrint Archive, 2022;
that article appears to remain unpublished after 3 years, its barely cited, its argumentation is more than limited, and the authors do not have a noticeable track record on the topic. I am questioning whether this citation is fit for a wikipedia mathematical article. ~2026-43000-1 (talk) 13:26, 20 January 2026 (UTC)[reply]

Agreed. This does not seem to be a reliable source. I removed the reference to this conjecture. Eigenbra (talk) 09:45, 21 January 2026 (UTC)[reply]