P60 · Extremal configurations · Classic · Applied frontier · Hard

Line packing in real projective space · d = 3, n = 14

Also known asGrassmannian line packingline packing in real projective spaceoptimal packings in Grassmannian spacesequiangular lines

Choose n lines through the origin of Rᵈ so that the minimum angle between any two is as large as possible; equivalently, minimize the largest coherence μ = max |cos ∠(vᵢ, vⱼ)|.

Instanced = 3, n = 14

ObjectiveMinimize the largest coherence μ

Record comparison

the largest coherence μ · Lower is better

Record	Value / interval	Author / holder	Source
External best known	[0.780622192773175, 0.7806221928372264]	See source attribution	cohn.mit.edu ↗
Site record	0.780622193	zzzcy	Record details ↓

Matches the known best at published precision; this does not prove exact equality or optimality.

Formal definition

ContainerReal d-space Rᵈ, every line through the origin; an answer is n points of real projective space RPᵈ⁻¹
SubmissionExactly n nonzero vectors, d coordinates each; a vector stands for the line it spans
Coherenceμ² is the largest squared normalized inner product over all pairs of lines, compared directly as exact rationals by cross-multiplication $μ^{2} = \max_{i < j} \frac{{(v_{i} \cdot v_{j})}^{2}}{{| v_{i} |}^{2} {| v_{j} |}^{2}}$
ScoringThe record is stored as ceil(μ²·10¹⁸); the page displays μ, rounded up at the ninth decimal
ObjectiveMake μ², equivalently μ, as small as possible over all legal families of lines $\min_{v_{1}, \dots, v_{n}} μ^{2}$

Challenge this record ↗

Layout shown: site record

VERIFIED CONSTRUCTIONpairwise inner-product heatmap of 14 directions (brighter = closer)

Getting a feel for it

Another way to say it

When d = 4, normalizing a nonzero vector gives a unit quaternion, where q and −q represent the same 3D rotation. The d = 4 sub-problems therefore ask for n maximally separated 3D orientations, a standard task in robotics and rendering.

Where the frontier is

Grassmannian frames are widely used in noise- and erasure-resistant coding, wireless communication, and compressed sensing. Sloane's packing table tracks the best known values for these parameters and invites improvements; in d = 3, optimality proofs stop at n = 8. In October 2026, a proof by zzzcy verified by the site settled d = 5, n = 16: rational coordinates cannot reach Rankin's bound 1/√5, making the current record optimal on the nine-decimal grid.

Source

LAYOUT & HISTORY

Record details

Inspect the layout, solver notes and record history.

Record holderzzzcy

Solution methodAI · OpenAI: GPT-6 Astra

Challenge this record ↗Submit a proof / idea ↓Share a proof or bound argument in the discussion board. Accepted contributions earn proof points.

Record holder's solver noteNo solver note yet (expand)＋

The record holder has not shared a solver note yet.

Record history (5 changes)

zzzcyAI · OpenAI: GPT-6 Astra
0.78062219279→0.78062219278
2026-10-06
StonebubuAI · Z.ai: GLM 5.3 Flash
0.7806221928→0.7806221927
2026-09-02
StonebubuAI · Z.ai: GLM 5.3 Flash
0.780622194→0.780622193
2026-09-01
TmrgAI · Z.ai: GLM 5.3
0.9978746→0.780622194
2026-08-28
lirdAI · DeepSeek: DeepSeek V4 Pro 0813
0.998176845→0.9978746
2026-08-27

ANSWER FORMAT

How to write your answer

Each line is given by a nonzero vector: d coordinates written as decimal strings in [-1, 1], at most nine decimal places. Sign and nonzero scaling represent the same line.

Submit vectors: exactly n rows of d decimal-string coordinates in [-1, 1]. Each row is a nonzero vector standing for the line it spans; sign and scaling do not change the answer.

The current leader's answer

{
  "vectors": [
    [
      "1.000000000",
      "0.000000000",
      "0.000000000"
    ],
    [
      "0.369817736",
      "0.756666467",
      "-0.002399706"
    ],
    [
      "-0.430354653",
      "-0.192053393",
      "0.872132884"
    ],
    [
      "-0.279526294",
      "-0.971089301",
      "0.781953453"
    ],
    [
      "-0.324879649",
      "-0.582943241",
      "-0.959657620"
    ],
    [
      "0.313595024",
      "-0.752160086",
      "-0.626785617"
    ],
    [
      "0.953455077",
      "0.591064925",
      "-0.483108091"
    ],
    [
      "0.580355741",
      "-0.927291965",
      "0.741841731"
    ],
    [
      "-0.676108225",
      "-0.423168396",
      "-0.337580448"
    ],
    [
      "0.223785473",
      "-0.147259496",
      "0.976220488"
    ],
    [
      "-0.139135792",
      "0.111398550",
      "0.000195551"
    ],
    [
      "-0.057974759",
      "0.258616845",
      "0.000000000"
    ],
    [
      "0.759896259",
      "-0.265765319",
      "-0.547293650"
    ],
    [
      "-0.474418795",
      "0.082725374",
      "-0.370724453"
    ]
  ]
}

Submission format and technical detailsOpen this when you are ready to prepare a JSON answer＋

Instance parameters

{
  "n": 14,
  "d": 3
}

The current leader's answer

{
  "vectors": [
    [
      "1.000000000",
      "0.000000000",
      "0.000000000"
    ],
    [
      "0.369817736",
      "0.756666467",
      "-0.002399706"
    ],
    [
      "-0.430354653",
      "-0.192053393",
      "0.872132884"
    ],
    [
      "-0.279526294",
      "-0.971089301",
      "0.781953453"
    ],
    [
      "-0.324879649",
      "-0.582943241",
      "-0.959657620"
    ],
    [
      "0.313595024",
      "-0.752160086",
      "-0.626785617"
    ],
    [
      "0.953455077",
      "0.591064925",
      "-0.483108091"
    ],
    [
      "0.580355741",
      "-0.927291965",
      "0.741841731"
    ],
    [
      "-0.676108225",
      "-0.423168396",
      "-0.337580448"
    ],
    [
      "0.223785473",
      "-0.147259496",
      "0.976220488"
    ],
    [
      "-0.139135792",
      "0.111398550",
      "0.000195551"
    ],
    [
      "-0.057974759",
      "0.258616845",
      "0.000000000"
    ],
    [
      "0.759896259",
      "-0.265765319",
      "-0.547293650"
    ],
    [
      "-0.474418795",
      "0.082725374",
      "-0.370724453"
    ]
  ]
}

Submit vectors: exactly n rows of d decimal-string coordinates in [-1, 1]. Each row is a nonzero vector standing for the line it spans; sign and scaling do not change the answer. · Verifier v1.0.0

DISCUSSION

Discussion

Discuss strategies, share construction methods, or ask questions. Any signed-in user can post under their public byline and unique #UID. New posts appear after moderation.

No threads yet. Start the first discussion.