Forty-six spherical codes in dimensions 3, 5, 6, 7 and 8 improving entries of the table of spherical codes

We give spherical codes for 46 pairs (d, N) whose maximal cosines are below the values listed in Henry Cohn's table of spherical codes (https://spherical-codes.org/, archived data set https://hdl.handle.net/1721.1/153543) on 19 November 2026: in R^3 for N = 481, in R^5 for N = 171, 191 and 220, in R^6 for N = 115, 121, 149, 289, 398, 567, 601 and 689, in R^7 for N = 396, and in R^8 for N = 173, 301 to 328 and 679 to 682. Version 1 of this record had 35 of these cells; version 2 adds 11 (R^3 N = 481; R^5 N = 220; R^6 N = 121, 149, 289, 398, 689; R^7 N = 396; R^8 N = 173, 327, 328). Each code was verified by an exact integer check of all pairs. The previous values are due to A. Jay (2026), Jay et al. (2026; A. Jay, M. Poberznik, M. Brut, A. Hemeryck, L.-H. Hoang, M. Gunde, N. Salles, G. Cruchon and M. Roux, "Equidistribution of points on the hypersphere including symmetries", preprint, https://doi.org/10.2139/ssrn.6427919), I. Vahteristo (arXiv:2610.19988), C. Smolen (2026) and A. Canli (2026), as listed per cell in CELLS.csv. The margins range from 1.6e-7 (R^6, N = 115) and 2.9e-7 (R^3, N = 481) to 6.7e-4 (R^8, N = 679). For R^8 and N = 314, 322, 323, 325, 326, 327 and 328 the table has no separate row; there the best known value is that of the next listed larger N, for example N = 314 (no separate row; best known value 0.586047842011, from N = 315), and likewise 322 and 323 from N = 324 and 325 to 328 from N = 332. Construction: one point is inserted at the deepest hole of the previous code (the table's (N-1)-point code, or our own previous code along a chain in N), and the maximal cosine is then minimised with L-BFGS and jittered restarts. The cells were chosen by a simple selector that looks for jumps in the table. This is a simple greedy method and each improvement is small. Verification: the coordinate files have 17-digit decimals; each row stands for the vector it gives, normalised to unit length. check_cell.py uses only the Python standard library: it reads the decimals exactly as integers and compares the exact squared cosines of all pairs as fractions. CELLS.csv gives the command values for every cell. The zip holds all coordinates, the search scripts, the run logs and the exact replay with its logs. The certificate does not depend on the numerical optimization. These are constructions at single cardinalities. They say nothing new about kissing numbers, and the optimal angles for these N remain open. AI use: this note comes from an AI-assisted research lab directed by the author. The constructions, the exact verification and the text were produced with Claude (Anthropic) under the author's direction, and the check was repeated by a separate session of the same AI system, so this is not independent human review. Not peer reviewed. Contact: [email protected].

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-19
DOI
https://doi.org/10.5281/zenodo.24360189
Primary Topic
Mathematical Approximation and Integration
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Forty-six spherical codes in dimensions 3, 5, 6, 7 and 8 improving entries of the table of spherical codes

Учрал Ням-Осор
Zenodo (CERN European Organization for Nuclear Research)
Mathematical Approximation and Integration
preprint

Forty-six spherical codes in dimensions 3, 5, 6, 7 and 8 improving entries of the table of spherical codes

Учрал Ням-Осор
preprint en

Abstract

We give spherical codes for 46 pairs (d, N) whose maximal cosines are below the values listed in Henry Cohn's table of spherical codes (https://spherical-codes.org/, archived data set https://hdl.handle.net/1721.1/153543) on 19 November 2026: in R^3 for N = 481, in R^5 for N = 171, 191 and 220, in R^6 for N = 115, 121, 149, 289, 398, 567, 601 and 689, in R^7 for N = 396, and in R^8 for N = 173, 301 to 328 and 679 to 682. Version 1 of this record had 35 of these cells; version 2 adds 11 (R^3 N = 481; R^5 N = 220; R^6 N = 121, 149, 289, 398, 689; R^7 N = 396; R^8 N = 173, 327, 328). Each code was verified by an exact integer check of all pairs. The previous values are due to A. Jay (2026), Jay et al. (2026; A. Jay, M. Poberznik, M. Brut, A. Hemeryck, L.-H. Hoang, M. Gunde, N. Salles, G. Cruchon and M. Roux, "Equidistribution of points on the hypersphere including symmetries", preprint, https://doi.org/10.2139/ssrn.6427919), I. Vahteristo (arXiv:2610.19988), C. Smolen (2026) and A. Canli (2026), as listed per cell in CELLS.csv. The margins range from 1.6e-7 (R^6, N = 115) and 2.9e-7 (R^3, N = 481) to 6.7e-4 (R^8, N = 679). For R^8 and N = 314, 322, 323, 325, 326, 327 and 328 the table has no separate row; there the best known value is that of the next listed larger N, for example N = 314 (no separate row; best known value 0.586047842011, from N = 315), and likewise 322 and 323 from N = 324 and 325 to 328 from N = 332. Construction: one point is inserted at the deepest hole of the previous code (the table's (N-1)-point code, or our own previous code along a chain in N), and the maximal cosine is then minimised with L-BFGS and jittered restarts. The cells were chosen by a simple selector that looks for jumps in the table. This is a simple greedy method and each improvement is small. Verification: the coordinate files have 17-digit decimals; each row stands for the vector it gives, normalised to unit length. check_cell.py uses only the Python standard library: it reads the decimals exactly as integers and compares the exact squared cosines of all pairs as fractions. CELLS.csv gives the command values for every cell. The zip holds all coordinates, the search scripts, the run logs and the exact replay with its logs. The certificate does not depend on the numerical optimization. These are constructions at single cardinalities. They say nothing new about kissing numbers, and the optimal angles for these N remain open. AI use: this note comes from an AI-assisted research lab directed by the author. The constructions, the exact verification and the text were produced with Claude (Anthropic) under the author's direction, and the check was repeated by a separate session of the same AI system, so this is not independent human review. Not peer reviewed. Contact: [email protected].

Zenodo (CERN European Organization for Nuclear Research)
Mathematical Approximation and Integration
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon