What is E8 — and why did QPC run it on IBM?

A short, non-specialist guide. No PhD required. This page explains the idea behind the test, what QPC actually did on hardware, and what we do not claim.

Start here — three sentences

1. In eight dimensions, there is a famous “best way” to pack equal spheres — the E8 arrangement. Mathematicians proved it is optimal (Maryna Viazovska, Fields Medal 2022).

2. That proof did not come from trying every possibility. It came from several different checks at once (geometry, bounds, symmetry, stability) that together act like a certificate.

3. QPC’s specialty is running many parallel “contexts” in one quantum job. This test asks: can we run that same style of multi-check search on a real IBM chip — using E8 only as a known answer key, not to re-prove the theorem?

The packing problem (everyday picture)

Imagine stacking identical oranges in a crate. In 2D you might use a honeycomb pattern; in 3D, grocers use pyramids. Mathematicians ask: what is the densest possible regular stacking pattern in N dimensions?

In 8 dimensions, the record-holder is called E8 (pronounced “E eight”). It is not a single sphere — it is a whole symmetric family of arrangements described by advanced geometry. The key number: E8 achieves the maximum packing density in 8D, about 25.37% of space filled.

Analogy: E8 is like the verified world-record score on a very hard exam — we use it to check that our measuring equipment works, not to take the exam again.

Why mathematicians cared

For decades, proving “nothing beats E8 in 8D” was open. Viazovska’s breakthrough (2016) showed how several mathematical tools — linear bounds, modular forms, symmetry — lock together into one proof. That coordinated multi-tool approach is what people mean by a certificate: not brute search, but parallel evidence that closes the case.

Why this matters for QPC (not for pure math)

QPC is not a mathematics lab. We build an architecture where a hard problem is split into parallel contextures — separate objective channels — that execute together on one quantum processor, coupled by transjunctions.

Same pattern, different domain. Cerrado on this site = carbon score + biodiversity score + social score, one IBM submission. E8 test = packing check + shortest-distance check + stability check + … (eight channels), one IBM submission. The business case and the math case look unrelated — but the computing pattern is the same: multi-context certificates instead of one giant brute-force loop.

What we demonstrated: classical software scores hundreds of candidate 8D arrangements using eight certificate channels; E8 wins as expected. Then we encode that winner into 8 parallel quantum context blocks (152 qubits total) and run one auditable job on IBM Heron. That validates QPC’s architecture on a hard, structured problem — not a new theorem.

What we do not claim

What the eight “certificates” mean

Each row is one QPC contexture — a parallel check, like a different expert reviewing the same candidate arrangement.

Plain nameWhat it checksWhy it matters
Packing densityHow tightly spheres fill spaceMain score — E8 is the 8D champion
Shortest gapSmallest nonzero distance between centersBad arrangements have awkward “near collisions”
Kissing countHow many neighbours touch one sphereE8 has a famously regular 240-neighbour shell
Shell shapeAre neighbour distances even?Symmetry hint — irregular shells score worse
Cell volumeSize of the repeating unit cellNormalizes density across candidates
ConditioningNumerical stability of the geometryFilters fragile, artifact-prone candidates
Bound gapDistance from known E8 optimumZero for E8 — quick “are we at the record?” test
StabilityDoes score survive tiny perturbations?Real solutions should be robust, not lucky flukes

Classical search results

500 candidate arrangements (E8 plus random perturbations). Eight channels combined into one composite score. E8 should win — it did.

CandidateComposite scoreDensity vs E8Gap to optimum
E8 (reference) 0.868 100% 0
Best random perturbation 0.699 44% large

IBM Heron run (hardware proof)

After classical selection, the E8 reference was encoded into eight coupled context circuits and submitted as one job on IBM’s 156-qubit Heron processor.

FieldValuePlain reading
Backendibm_fezIBM Heron, 156 qubits
Layout8 × 19 = 152 qubitsEight parallel context blocks in one circuit
Job IDd8o47bjqv2lc7389jo60Auditable on IBM Quantum dashboard
Shots1024Number of repeated measurements
Runtime11.4 sWall-clock on cloud queue
Unique outcomes1024 / 1024Every shot produced a distinct bit pattern — high diversity
ICC (context coupling)0.025Low cross-context correlation — channels stay distinct while coupled

✓ Architecture check passed: one submission, eight live contextures, full measurement diversity, auditable job ID.

Where this goes next

Architecture program overview Comparable benchmarks Separability (ICC) results