Zomeable 4-polytopes

duoprism_4_q snap+signature census

Per-q distinct shape signatures on the (a,b,0,0) and (a,b,c,d) kernel planes at multiple rngs. Probe: ongoing_work/probes/duoprism_4q_snap_sig.py.

Signature = robust 5-decimal SHA-256 (same as tools/dedup_corpus_by_shape.py).

(a,b,0,0) plane — square plane only

q rng=2 sigs rng=3 sigs rng=4 sigs rng=4 snap rng=4 align
5 0 0 0 0 3280
6 1 3 3 104 3280
7 0 0 0 0 0
8 0 0 0 0 3280
9 0 0 0 0 0
10 2 2 5 160 3280
11 0 0 0 0 0
12 0 0 0 0 0

Finite-probe saturation for q=6 and q=10 (rng=4..8)

rng q=6 sigs q=6 snap q=10 sigs q=10 snap
4 3 104 5 160
5 3 160 5 236
6 3 224 5 328
7 3 304 5 444
8 3 384 5 576

Both q=6 and q=10 saturate at small finite-probe signature counts and remain there as rng grows. At rng=8 the snap-count is 3.6–3.7× the rng=4 count, yet the distinct-probe signature count is unchanged. The SHA-256 hash sets are bit-identical across all 5 rngs for both q values.

Superseding note: this finite-probe saturation is not a boundedness proof. Later exact constructions found inf families for {4}x{6} and {4}x{10}; see duoprism_4_6_inf_family_resolved.md and duoprism_4_10_inf_family_resolved.md.

(a,b,c,d) plane — full 4D directions

q rng=2 sigs rng=3 sigs rng=2 snap rng=3 snap rng=3 align
5 0 0 0 0 2400
6 1 3 24 64 2512
7 0 0 0 0 1200
8 0 0 0 0 2400
9 0 0 0 0 1200
10 2 2 40 84 2400
11 0 0 0 0 1200
12 0 0 0 0 1200

Reading the table

A q-value with snap=0 at a high rng confirms that no kernel direction in that range can produce a zomeable projection.

Headline findings

The eight q values cleanly partition into three classes:

class q values (a,b,0,0) align (a,b,c,d) align snap n_shapes
zomeable 6, 10 yes yes yes finite probe saturates (3, 5); exact inf families
align-only 5, 8 yes yes no 0
non-aligned (square-plane) 7, 9, 11, 12 no yes no 0

Interpretation:

Why q=12 fails despite being “zome-compatible”

The dodecagon does have axes appearing in zome, but inside the 4D duoprism the edge set’s projected directions don’t match the 4-fold square pattern under any 4D rotation accessible to _try_align. Conclusion: q=12 is not zomeable as a duoprism_4_q.

Cross-reference with prismatic_manifest

q manifest n_shapes (rng=4) census ab00 rng=4 census abcd rng=3
5 0 0 0
6 3 3 3
7 0 0 0
8 0 0 0
9 0 0 0
10 5 5 2
11 0 0 0
12 0 0 0

All entries agree — except (q=10, abcd rng=3) which undercounts because the abcd-plane scan at rng=3 can’t reach the wider 1D range that the ab00-plane scan at rng=4 explores in the square plane.

Output JSON files (one per q × plane × rng)

ongoing_work/probe_4q_sigs_q{5..12}_ab00_rng{2,3,4}.json
ongoing_work/probe_4q_sigs_q{5..12}_abcd_rng{2,3}.json

Each contains the full sig list, n_aligned, n_snapped, n_errors, and elapsed_s.

Reproducing

python ongoing_work/probes/duoprism_4q_snap_sig.py --q 10 --plane ab00 --rngs 4,5,6
python ongoing_work/probes/duoprism_4q_census_driver.py