Good morning Bruce. Let me verify the checkable pieces before commenting.
Ran codeFailed to run code
typescriptCopy
async function main() { // 1) Monte Carlo: second moment of uniform octahedron |x|+|y|+|z|<=1 let N = 400000, sum = 0, inCount = 0; for (let i = 0; i < N; i++) { const x = Math.random()*2-1, y = Math.random()*2-1, z = Math.random()*2-1; const s = Math.abs(x)+Math.abs(y)+Math.abs(z); if (s <= 1) { inCount++; sum += x*x+y*y+z*z; } } const meanR2 = sum/inCount; // expect 3/10 const intR2 = meanR2 * (4/3); // expect 2/5 // 2) Ratio map from (3,4) let a = 3, b = 4; const rs: number[] = []; for (let k = 0; k < 8; k++) { rs.push(b/a); const t = a + b; a = b; b = t; } const phi = (1+Math.sqrt(5))/2; return { mc_meanR2: meanR2, expect_3_10: 0.3, mc_intR2: intR2, expect_2_5: 0.4, ratioIterates: rs, phi, twoPhiOver5: 2*phi/5 };}
Result
textCopy
run_typescript exceeded maxProgramOperations (1024).
Ran code
typescriptCopy
async function main() { // Closed-form check of the second moment instead of Monte Carlo (budget limits). // Density of X on octahedron: f(x) = (3/2)(1-|x|)^2. E[X^2] = 3*int_0^1 x^2(1-x)^2 dx = 3/30 = 1/10. const integral = (() => { let s = 0; const n = 10000; for (let i=0;i<n;i++){ const x=(i+0.5)/n; s += x*x*(1-x)*(1-x); } return s/n; })(); const EX2 = 3*integral; // expect 1/10 const meanR2 = 3*EX2; // expect 3/10 const intR2 = meanR2*(4/3); // expect 2/5 let a = 3, b = 4; const rs: number[] = []; for (let k = 0; k < 8; k++) { rs.push(b/a); const t=a+b; a=b; b=t; } const phi=(1+Math.sqrt(5))/2; return { EX2, meanR2, intR2, ratioIterates: rs, phi, twoPhiOver5: 2*phi/5 };}
Result
textCopy
{"EX2":0.09999999999999956,"meanR2":0.2999999999999987,"intR2":0.39999999999999825,"ratioIterates":[1.3333333333333333,1.75,1.5714285714285714,1.6363636363636365,1.6111111111111112,1.6206896551724137,1.6170212765957446,1.618421052631579],"phi":1.618033988749895,"twoPhiOver5":0.6472135954999579}
Thought for 1s
Write the review. Structure: what checks out, what’s shaky, verdict. Cite knowledge only if used — memories had nothing relevant, so no knowledge link.
Good morning Bruce. I worked through the document rather than just reacting to it. Here’s my honest read.
✓ What checks out
- Second moment [P]: verified independently. The density of one coordinate on the octahedron is (3/2)(1−|x|)², giving ⟨x²⟩ = 1/10, ⟨r²⟩ = 3/10, ∫r²dV = (4/3)(3/10) = 2/5. Correct, and the remark that “the 5 is L⁵ dimension, not a pentagon” is a good deflationary move.
- Plate geometry: P₁ ∩ O is indeed the regular hexagon through the six edge-midpoints, side √2/2, and the 6-triangle fan with apex at the center is right. I also checked the cell bookkeeping with sign patterns of the four plate normals: exactly 14 bounded regions (one degenerate sign combination), splitting as 6 corner cells + 8 tetrahedra, and each fan sector does border a corner cell on one side of the plate and a tetrahedron on the other. So the inventory is consistent.
- Ratio map (Lemmas 2.2–2.3): r ↦ 1 + 1/r, unique positive fixed point φ, |derivative| = 1/φ² < 1, iterates 4/3 → 7/4 → 11/7 → 18/11 → … all correct.
- D₃ stabilizer conclusion (order 6, 120° rotation survives, 60° fails) is right for an alternately 2-colored 6-sector fan.
⚠ What I’d push back on
- Definition 2.1 smuggles in the Fibonacci addition. True alternation is just the swap N_A → N_B, N_B → N_A, which has no φ at all. The φ appears only because the second line was written as N_B(k+1) = N_A(k)+N_B(k). That’s Fibonacci recurrence, chosen, not derived. This is the single biggest gap: φ is baked into the setup, so Corollary 2.4 doesn’t “emerge” — it’s assumed. The page should either justify why interface alternation generates addition (a combinatorial reason cells on one side count both sides’ predecessor populations), or downgrade Lemma 2.2 to a definition.
- Corollary 1.4’s 4/(4+6) = 2/5. The remark honestly separates it from the second moment, but it’s still a coincidence of small integers doing load-bearing work. Ratios like 4/10 = 2/5 will match almost any target constant with a little freedom. I’d mark this [O], not [P].
- Lemma 1.2’s reflection claim looks backwards. For an alternating 6-sector coloring, a reflection through a vertex-to-vertex axis maps an A sector onto a B sector (I believe it fails), while a reflection through the axis joining opposite sector centers preserves the coloring. The D₃ conclusion survives, but the stated classification of which reflections survive appears swapped. Worth a corrective pass.
- Theorem 3.1 is not yet a theorem. The candidate set {duplication, general addition, alternation} is hand-picked (the falsification list admits this), “duplication has no continuum limit” is asserted, and (H) — “has a unique scale-invariant positive attracting fixed point” — is satisfied by a huge family of symmetric local maps, so C+S+H doesn’t uniquely select the Fibonacci-winning alternation. I’d relabel this [D] or [O].
- Corollary 3.2 (2φ/5) is the weakest link. “Different symmetries at different levels, so they multiply” is an assertion, not a derivation. Multiplying a static orbit ratio by a dynamical attractor constant has no established mathematical license, and 2φ/5 ≈ 0.6472 doesn’t yet name a specific, measured muon observable. As written, the falsification condition can’t actually be triggered, which makes it a placeholder rather than a prediction.
My overall take
The framing discipline is genuinely good: status keys, a remark that self-corrects a prior page, falsification conditions, and the acknowledgment that the whole theorem shrinks if the two load-bearing pieces fail. The load-bearing piece that actually holds is the static geometry (fan, hexagon, D₃, inventory); the load-bearing piece that doesn’t yet hold is Lemma 2.3 as a derivation, because its input (Definition 2.1) contains the conclusion. I’d save the 2φ/5 claim for a separate hypothesis page rather than letting it ride on Corollary 3.2.
If you want, I can do a clean re-derivation page of the fan geometry and cell inventory with the coordinate-level details filled in, or try to find a non-arbitrary reason the alternation should obey a Fibonacci-type addition instead of a pure swap.Sources
7:35am
I have been watching your “Thinking” and it is an education unto itself. There is hidden complexity within the simplicity of the hexarchon and octahedron! So, yes, please do a clean re-derivation page of the fan geometry and cell inventory. Thinking about your statement, “I’d save the 2φ/5 claim for a separate hypothesis page rather than letting it ride on Corollary 3.2.”
7:45am
Thought for 10min 14sThought for 10min 14s
Bruce wants a “clean re-derivation page”. This is a document artifact — a canvas. Load canvas skill first.
Plan: verify the geometry carefully (fan hexagon, cell inventory 6+8, volumes, orbits) and write a mathematical derivation page. Deliver as a text/markdown canvas. Could also do the 2φ/5 hypothesis page? He said “thinking about your statement” — ambiguous. I could include a brief separate hypothesis section/page. Maybe create one canvas with the re-derivation, and ask? Better: do the re-derivation page as asked, and offer the hypothesis page after — or add a short appendix noting 2φ/5 as separate. Keep scope: main deliverable = re-derivation page; briefly note hypothesis split at end of the page maybe. I’ll do the re-derivation page and include a short final section “What this page does not claim” pointing to the hypothesis page, then offer to draft it.
Let me verify facts before writing:
- O = {|x|+|y|+|z|≤1}, R=1, edge √2, V = 4/3. ✓ (V = 4/3·a³? octahedron volume = √2/3 a³ = √2/3·2√2 = 4/3 ✓.)
- Plates P1..P4: x±y±z=0 with even number of minus… listed: x+y+z=0, x+y−z=0, x−y+z=0, −x+y+z=0. Normals n_i = sign vectors with product… (1,1,1),(1,1,−1),(1,−1,1),(−1,1,1): these have an even number of negatives? (1,1,1): 0 negs; (1,1,−1): 1 neg… hmm (1,1,−1) has one minus. The four are: (1,1,1),(1,1,−1),(1,−1,1),(−1,1,1) = sign vectors with at most one minus. These are the normals of facets of… they point toward 4 alternate vertices of a cube → vertices of a regular tetrahedron. So the 4 plates are the symmetry planes of the tetrahedron {n·x=1}? Each plate is perpendicular to a facet direction of the dual tetrahedron, i.e., each plate is the plane through 3 midpoints of opposite edges of… Actually each plate is parallel to 3 of the cube… Note each plate passes through the centers of 6 of O’s 12 edges? P1 (x+y+z=0) passes through the 6 edge-midpoints connecting a positive to a negative axis: 6 midpoints, and those 6 edges are the edges of O NOT incident… O’s edges connect axis points of different coordinates; sign combinations: edges between (+i) and (−j), i≠j: these come in two types:
- Both coordinates… all 6 edges from +i to −j (i≠j): (1,0,0)-(0,−1,0): midpoint (1/2,−1/2,0): sum=0 ✓ on P1. There are 3×2? Edges connecting + to −: for each pair of distinct axes with opposite sign: 3 (axis pairs) × 2 (sign combos? (+i,−j) and (−i,+j) for each unordered axis pair i<j: 3 pairs × 2 = 6 ✓.) All 6 have midpoint sum 0 ✓. So P1 ∩ O = hexagon through the 6 opposite-sign edge midpoints. ✓
- Other 6 edges connect (+i, +j) and (−i, −j): midpoints (1/2,1/2,0) etc., on P4-type planes? P4: −x+y+z=0: contains (0,1/2,1/2): −0+1/2+1/2=1≠0. Hmm. P4 contains midpoints of edges between −x and +y?? (−1,0,0)-(0,1,0) midpoint (−1/2,1/2,0): −(−1/2)+1/2+0 = 1 ≠ 0. Hmm! Which edges are on P4 (−x+y+z=0)? Midpoint (a,b,c)/2 with −a+b+c=0 where (a,b,c) axis values ±1 with distinct axes: need −a+b+c=0: options: a=−1,b=−1? −(−1)+… a=−1, b+c=… wait −a+b+c=0 → a = b+c, b,c∈{±1}, b+c ∈ {2,0,−2}: a = b+c: a=1 requires b=c=1: edge (1,0,0)–(0,1,0): but wait a=1,b=1,c must… a = b+c = 2 impossible with distinct… hold on: a is the coordinate ON the axis of x, b,c coordinates of the other axis point: edge between (a,0,0) and (0,b,0): midpoint (a/2, b/2, 0): P4 value = −a/2 + b/2 = (b−a)/2 = 0 → a=b: edge between (1,0,0)-(0,1,0): mid (1/2,1/2,0): P4 value = −1/2+1/2 = 0 ✓! I miscalculated before (forgot the x). So P4 passes through midpoints of same-sign edges with c=0: (±1,±1,0) with a=b: (1/2,1/2,0) ✓ and (−1/2,−1/2,0) ✓, plus for the (y,z)-plane edges: midpoint (0,b/2,c/2): −0+b+c = 0 → b=−c: (0,1/2,−1/2),(0,−1/2,1/2) ✓, and (x,z) edges: (−a/2+c/2) = 0 → a=c: (1/2,0,1/2),(−1/2,0,−1/2). Total 6 ✓. So each plate passes through exactly 6 edge-midpoints: P1 through the 6 “opposite-sign” mids, P2,P3,P4 through same-sign pairs… P2 (x+y−z=0): x+y=z: (1/2,0,1/2)? a=1,c=1: (1/2)… midpoint of (1,0,0)-(0,0,1): (1/2,0,1/2): 1/2+0−1/2=0 ✓. Yes.
So the 4 plates collectively pass through all 12 edge midpoints, each midpoint on exactly 2 plates? 4 plates × 6 mids = 24 = 12×2 ✓. Each edge-midpoint lies on exactly 2 plates: e.g. (1/2,1/2,0) on P1? sum = 1 ≠ 0 no; on P2: 1/2+1/2−0 = 1 ≠0; P3: x−y+z = 0 ✓; P4: 0 ✓. Yes 2 plates.
Also: the three coordinate planes x=0,y=0,z=0 each contain… not part of construction.
- Hexagon of P1: vertices the 6 opposite-sign mids; side √2/2 (computed earlier: distance between (1/2,0,−1/2) and (0,1/2,−1/2) = √(1/4+1/4) = √2/2 ✓); circumradius √2/2; regular hexagon ✓. Inradius = √3/2·√2/2 = √6/4.
- Cell decomposition by the 4 plates ∩ O: 14 regions. Formal count: four linear forms h1..h4 with h4 = h1 − h2 − h3 (= −x+y+z: check: h1−h2−h3 = (x+y+z)−(x+y−z)−(x−y+z) = x+y+z−x−y+z−x+y−z = −x+y+z ✓). Sign patterns (s1,s2,s3) free, s4 = sign(h1−h2−h3) forced iff h1 and h2+h3 have same sign and |h1|>|h2+h3|… forced only in two cases: (s1,s2,s3)=(+,−,−) → h2+h3 <0, h1>0 → h4 = h1−(neg) >0 forced: pattern (+,−,−,+). And (−,+,+) → h4<0 forced: (−,+,+,−). Wait earlier I derived (−,+,+) forces s4=− ✓ and (+,−,−) forces +. Hmm earlier I wrote “(−,−,−) forces s4=−”: check (−,−,−): h4 = h1−h2−h3 = neg+|h2|+|h3|… h1<0, −h2>0, −h3>0: sign ambiguous! Earlier I wrongly said forced. Correct: with h1<0, h2<0, h3<0: h4 = h1 − h2 − h3 = negative + positive + positive: ambiguous. So forced cases: (+,−,−): h4 = h1 + |h2| + |h3| > 0 forced ✓. (−,+,+): h4 = −|h1| − h2… h4 = h1 − h2 − h3 = −|h1| − |h2| − |h3| < 0 forced ✓. Other 6 patterns: s4 free. Total regions = 8 patterns: (−,−,−): 2 wait (−,−,−) is among the 6 free → 2; total = 2 forced patterns (1 each) + 6×2 = 14 regions ✓.
Antipodal: O and plates centrally symmetric → regions in antipodal pairs: 7 pairs. Region types by sign pattern (up to global sign):
All-positive-type: (s1,s2,s3,s4) = (+,+,+,+): contains direction (1,1,1)/√3? h’s at (1,1,1): (3,1,1,1) all >0 ✓. This is the cone {x<y+z, y<x+z, z<x+z, 0<sum}?? = {h1>0: x+y+z>0; h2>0: x+y>z; h3: x+z>y; h4: y+z>x} = positive hull of… the cone around (1,1,1) = cone spanned by (1,0,0)? x<y+z… The cone = {triangle-inequality region} = cone over the medial triangle of facet x+y+z=1: cone ∩ O = tetrahedron conv(0, (1/2,1/2,0),(1/2,0,1/2),(0,1/2,1/2)) — wait: is the cone actually equal to the positive hull of those three points? Positive hull of e1+e2, e1+e3, e2+e3 (scaled): spanned vectors u=(1,1,0),v=(1,0,1),w=(0,1,1) with nonneg coefficients: point = αu+βv+γw: x=α+β, y=α+γ, z=β+γconditions x<y+z ⟺ α+β < α+γ+β+γ ⟺ 0<2γ ✓; similarly. And h1 = 2(α+β+γ)>0 ✓. So cone = poshull{u,v,w} ✓. So 8 regions of “tet type”? Which sign patterns are tet-type: patterns that are “single-sign cones” around cube-corner directions ±(1,1,1)±: direction where all h’s same sign: 2 (all+around (1,1,1), all−). Other tet-type patterns: (+,+,−,−): at direction? Solve: want interior direction of pattern (+,+,−,−): region = {x+y+z>0, x+y−z>0, −(x−y+z)>0 → y>x+z, −(−x+y+z)>0 → x>y+z} → x>y+z and y>x+z: adding: x+y > x+y+2z → z<0. Example direction (1,1,−1)! Check h at (1,1,−1): h1 = 1, h2 = 3, h3 = −1, h4 = −1 ✓. So (+,+,−,−) is the cone around (1,1,−1) = positive hull of e1+e2−… cone around (1,1,−1) = poshull{(1,0,0)? } tetrahedral cone spanned by (1,0,0),(0,1,0),(0,0,−1)?? check: pos cone of e1,e2,−e3: contains (1,1,−1) ✓; condition: {x,y≥0, z≤0}∩… no wait that box-cone isn’t {h signs}. Cone around (1,1,−1) bounded by the 4 h=0 planes: these are the planes through pairs of O-face… hmm, the cone = positive hull of the three axis rays of the adjacent O vertices: the cone containing (1,1,−1) and bounded by 4 planes is trihedral? A cone in R³ bounded by 4 halfspaces through origin is generally a tetrahedral cone bounded by 3 facets (4th redundant) or has 4 facets (still pointed, 3 extreme rays… a pointed cone in 3D has ≥3 facets; with 4 halfspaces, if bounded (pointed) with 4 facets: possible (like a cone over a quadrilateral? no pointed + 4 facets = 4 extreme rays possible). Simplest: the cone around (1,1,−1): intersection conditions h1>0,h2>0,h3<0,h4<0: h3<0: y > x+z; h4<0: x > y+z → z < min(x−y, y−x)·… x>y+z ∧ y>x+z → adding 0 > 2z → z<0 ✓ consistent. This cone = poshull of rays: intersection lines: h3=0∧h4=0: x−y+z=0, −x+y+z=0 → z=0… wait: subtract: x−y+z −(−x+y+z) = 2x−2y = 0 → x=y; then z = y−x = 0: line {(t,t,0)}: t>0 ✓ extreme ray (1,1,0). h3=0∧h1=0: x−y+z=0, x+y+z=0 → x=z=0… add: 2x+2z=0 → x=−z; then −z−y+z=0 → y=0: line {(t,0,−t)}: ray (1,0,−1). h4=0∧h1=0: −x+y+z=0 & x+y+z=0 → x=y line… subtract: 2x−2y = 0 → x = y; then 2x+z=0 → z=−2x: ray (1,1,−2)?? Hmm that’s h1=h4=0 line: (t,t,−2t): check in cone: h2 = t+t+2t = 4t>0 ✓, h3 = t−t−2t = −2t <0 ✓: ray (1,1,−2). And h2=0? h2>0 with h1>0: h2=0∧h3<0… h2=0: x+y=z; with z<0 → x+y<0: and h1>0 requires… h1 = x+y+z = 2z <0 ✗ contradict: so h2=0 doesn’t bound this cone: cone has 3 facets (h1=0? no wait: facets are h3=0, h4=0, and h1=0 or h2=0: from above extreme rays: (1,1,0) [h3=h4=0], (1,0,−1) [h3=h1=0], (1,1,−2) [h1=h4=0]: three rays → cone = poshull{(1,1,0),(1,0,−1),(1,1,−2)}: a trihedral cone (3 facets: h1, h3, h4; h2 redundant ✓ since within cone h2 = sum of others >0). So region (+,+,−,−)∩O = tetrahedron conv(0, (1,1,0)/norm… clipped by O: the cone ∩ O: O boundary points: O ∩ cone: vertices at O: ray (1,1,0): hits O at (1/2,1/2,0) ✓; ray (1,0,−1) → (1/2,0,−1/2) ✓; ray (1,1,−2): length |1|+|1|+|2| = 4 → hits O at (1/4,1/4,−1/2) — an interior point of an O FACE! (face x+y−z=… check (1/4,1/4,−1/2): x+y−z = 1: lies on O’s face ✓.) So region = tetrahedron conv(0, (1/2,1/2,0),(1/2,0,−1/2),(1/4,1/4,−1/2))! Volume: det[(1/2,1/2,0),(1/2,0,−1/2),(1/4,1/4,−1/2)]: = 1/2·(0·(−1/2) − (−1/2)(1/4)) − 1/2·((1/2)(−1/2) − 0) + 0 = 1/2·(1/8) − 1/2·(−1/4) = 1/16 + 1/8 = 3/16 → vol = 1/6·3/16 = 1/32.
Hmm! So the 8 “tetrahedra” per the doc… the doc’s inventory says 8 tetrahedra. My decomposition: 14 regions: 8 of “one-side” type? Let’s recount types: patterns and their cones:
Each sign pattern (of 14 regions, 7 antipodal pairs) corresponds to cone around one of the 8 cube-corner directions (±1,±1,±1)? 8 slices around the 8 cube corners + 6 remaining = 14. Cube-corner directions: (1,1,1): all h >0 ✓; (1,1,−1): (+,+,−,−) ✓; (1,−1,1): h=(1,−1,3,−1): (+,−,+,−) ✓; (−1,1,1): (1,−1,−1,3): (+,−,−,+) ✓… wait that’s forced pattern (+,−,−,+) ✓ consistent (forced patterns are cube-corner directions on coordinate… (−1,1,1)·h: h1=1>0,h2=−1,h3=−1,h4=3: (+,−,−,+) ✓ and yes this is the physically forced case? earlier said (+,−,−) s3 free forces s4 = +: matches ✓.) So 8 cube-corner cones: patterns around directions (±1,±1,±1): each is one of the 14. Then 6 more regions = around the 6 coordinate-axis directions! Direction (1,0,0): h = (1,1,1,−1): pattern (+,+,+,−): the CORNER regions ✓ (region containing O’s vertex (1,0,0)).
So the 14 regions = 8 cones over cube-corner dirs (whose O∩ are tetrahedra? the (1,1,1) cone ∩ O = tet conv(0 + medial facet triangle) ✓ vol 1/24; the (1,1,−1) cone ∩ O = tet conv(0, (1/2,1/2,0),(1/2,0,−1/2),(1/4,1/4,−1/2)) vol 1/32; by symmetry the 6 non-(1,1,1)-type cube corners: dirs with one minus: (1,1,−1),(1,−1,1),(−1,1,1) and antipodes: 6 such: each ∩ O = tet of vol 1/32) and 6 axis cones (corner regions): corner region (+,+,+,−)∩O: shape: cone bounded by h1,h2,h3>0, h4<0 around (1,0,0): facets: extreme rays: h1=0∧h3=0: computed → (1/2,0,−1/2)-direction (1,0,−1); h1=0∧h2=0: x+y=0, x+y−z=0 → z=0: line (t,−t,0): ray (1,−1,0)? check sign: h3 = 1+1+0 = 2>0 ✓ h4 = −1−1 = −2 ✓: ray (1,−1,0). h2=0∧h4=0: shown (1,0,1)? h2=h4=0: x=z… wait earlier: h2=0,h4=0 → y=0, x… h2: x+y−z=0, h4: −x+y+z=0: add: 2y = 0 → y=0, then x=z: ray (1,0,1) ✓. h3=0∧h4=0 → (1,1,0) ✓ computed earlier. h1=0∧h4=0 → (0,1,−1)? h1=h4=0: x=0, z=−y: ray (0,1,−1): check region: h2 = 1+1=2>0 ✓ h3 = 0−1−1 = −2 <0 ✗ not in region. So extreme rays: from facets {h1,h3},{h1,h2},{h2,h4},{h3,h4}: rays (1,0,−1),(1,−1,0),(1,0,1),(1,1,0). 4 rays → region = cone over quadrilateral! Cone facet planes: between rays (1,0,−1),(1,−1,0): facets h1; between (1,−1,0),(1,1,0): plane h4? (1,1,0): h4 = −1+1 = 0 ✓ (1,−1,0): h4 = −1−1 = −2 ≠ 0. Hmm. Let me systematically: region = {h1>0,h2>0,h3>0,h4<0} around (1,0,0). Facet planes: each extreme ray = intersection of 2 constraint planes. Rays found: R1=(1,0,−1): h1=0,h3=0. R2=(1,−1,0): h1=0,h2=0. R3=(1,0,1): h2=0,h4=0. R4=(1,1,0): h3=0,h4=0. Each plane appears twice? h1: R1,R2. h2: R2,R3. h3: R1,R4. h4: R3,R4 ✓ each exactly twice ✓ cone over quadrilateral with vertices R1,R2,R3,R4 in cyclic order? Order: R1 (h1=h3=0) adjacent to R2 (shares h1) and R4 (shares h3); R2 adjacent R1(h1), R3(h2); R3 adjacent R2(h2), R4(h4); R4 adjacent R3(h4), R1(h3). Cyclic: R1→R2→R3→R4→R1. So region cone over quad R1R2R3R4. ∩ O: the cone rays exit O at: R1 → (1/2,0,−1/2) ✓ (on O edge), R2 → (1/2,−1/2,0) ✓, R3 → (1/2,0,1/2) ✓, R4 → (1/2,1/2,0) ✓ — all four on the plane x=1/2 square. Plus O vertex (1,0,0): is (1,0,0) inside the cone? h at (1,0,0) = (1,1,1,−1) all strict ✓. And its boundary: region ∩ O = pyramid conv((1,0,0), square)? But wait, region also extends toward origin? The cone {h-cone} ∩ O: points of cone with small norm: e.g. ε(1,0,0)-ish ✓ all of cone ∩ O. Does cone contain points with x < 1/2? Cone over rays all with x-coordinate positive: any point of the cone = positive combo of the four rays → x>0, but can be small: (0.1, 0.05, −0.05) ∈ cone? h = (0.1, 0.2, 0.2, −0.1)? h1 = 0.1+0.05−0.05 = 0.1 ✓ >0, h2 = 0.1+0.05+0.05 = 0.2 ✓, h3 = 0.1−0.05−0.05 = 0 ✓ boundary. OK. Cone ∩ O: this is the cone (apex origin!) over quad… no wait: cone = poshull of the 4 rays, apex at ORIGIN. ∩ O: since O is convex and contains origin… region ∩ O = O ∩ cone. O ⊂ {|x|+|y|+|z| ≤ 1}. The region ∩ O = cases: for each ray r, the segment from 0 to O-boundary ∩ ray. Region ∩ O = {λ r_i combos}: it’s O∩cone, where O is convex containing 0: O∩cone = union over… not simply conv(apex…). O∩cone: the extreme points: (1,0,0) (O vertex, inside cone), the 4 ray∩O points, and origin (0,0,0)? Is origin ∈ region? h_i(0) = all 0: origin is a boundary point of ALL 14 regions (all plates meet there). So region ∩ O = full-dimensional but origin on its boundary? Origin is in the closure of region but is it a vertex of ∩O? Yes: near origin inside region cone = the cone itself (small points): so origin is the apex-type vertex, with the local cone structure = the 4-ray cone: so origin is a vertex of the region with 4 edges (the rays). So region shape: vertices (0,0,0),(1,0,0),(1/2,0,±1/2),(1/2,±1/2,0): six vertices. Faces: 4 side faces (on plates h1=0: quad? face on plate h1=0: bounded by rays… the facet of the region lying in h1=0: cone facet between R1,R2: triangle-ish region: points with h1=0 in region closure: {(x,y,z): x+y+z=0, h2,h3>0, h4<0, |·|≤1}: vertices on this facet: rays R1, R2 exit points (1/2,0,−1/2),(1/2,−1/2,0), and origin: triangle ✓. Face h2=0: between R2,R3: triangle (0,0,0)? wait R2,R3 and origin: triangle conv(0,(1/2,−1/2,0),(1/2,0,1/2)) ✓. Similarly h3, h4. Then faces on O: (1,0,0) and the 4 mids: the 4 triangles conv((1,0,0), mid_i, mid_{i+1}): these lie on the four O faces at vertex (1,0,0) ✓. Total faces: 4 plate triangles + 4 O triangles = 8 faces, 6 vertices, 12 edges ✓ a (non-regular) octahedron! Combinatorially an octahedron ✓. So “6 corner octahedra” is combinatorially right — but they’re NOT regular octahedra: vertex (0,0,0) has 4-face conical structure… wait octahedron vertices should each have 4 faces ✓ (both (0,0,0) and (1,0,0) have 4 triangles). It’s combinatorially an octahedron but not regular (edges: from (1,0,0) to mids: √2/2… distance (1,0,0)-(1/2,1/2,0) = √(1/4+1/4) = √2/2; from origin to mids: √2/2; but edges of square: (1/2,1/2,0)-(1/2,0,1/2) = √2/2 ✓ ALL edges √2/2?! (0,0,0)-(1/2,1/2,0) = √2/2 ✓; (1,0,0)-(1/2,1/2,0): √(1/4+1/4)=√2/2 ✓. Wait is (1,0,0) adjacent to all four mids? Edges: (1,0,0)-mid pair triangles: yes each mid connects to (1,0,0) ✓. And mids form 4-cycle with edges √2/2 ✓. Origin connects to all 4 mids ✓. So octahedron with ALL 12 edges = √2/2 → it IS a regular octahedron (edge √2/2, i.e., a half-scale octahedron)! Volume of regular octa edge s: √2/3 s³ = √2/3·(√2/2)³ = √2/3 · √2·(1/2)… (√2)³ = 2√2; (2√2)/8 = √2/4; ·√2/3 = 2/12 = 1/6. Volume 1/6 each. But wait 6 × 1/6 = 1! Plus 8 tets: 2 of type (1,1,1)-cone vol 1/24 and 6 of vol 1/32: 2/24 + 6/32 = 1/12 + 3/16 = 4/48 + 9/48 = 13/48. Hmm total = 1 + 13/48 = 61/48 > 4/3 = 64/48?! Overshoot! Contradiction. So somewhere wrong: corner octahedron volume: check disjointness: corner octa = conv(origin? no wait: does the corner region actually contain origin? The points near origin within the region cone: e.g. 0.01·(1,0,0): yes. But hold on — the eight tets also have origin as vertex, and each pair of adjacent regions shares a plate face containing… So origin is shared by all 14 regions: fine, no overlap issue. But wait: my “corner octahedron” conv(0,(1,0,0),4 mids): does it really equal region∩O? Region∩O = cone∩O where cone = poshull{R1..R4} (all of which have x>0 region… hmm but also the cone contains (1,0,0)-direction: is (1,0,0) ∈ poshull{R_i}? (1,0,0) = a·R1 + b·R2 + c·R3 + d·R4 with a,b,c,d ≥ 0: y: −b+d = 0 → b=d; z: −a+c = 0 → a=c; x: a+b+c+d = 2a+2b = 1. Take a=b=1/4 ✓. Yes (1,0,0) in cone interior ✓. So region∩O ⊇ {tetrahedra from origin}: conv(0, R_i∩O points) ⊆ region∩O, plus O-vertex part: region∩O = conv(0, (1,0,0), mids)? Check: volume of that = 1/6. Then 14 region volumes: 6×(1/6) + 2×(1/24) + 6×(1/32) = 1 + 1/12 + 3/16… = 1 + 0.0833 + 0.1875 = 1.2708 = 61/48 ≈ 1.27 vs O = 4/3 ≈ 1.333. UNDERSHOOT now (earlier overshoot was my arithmetic: 6×1/6 = 1, +13/48: 48/48+13/48 = 61/48 < 64/48). Missing 3/48 = 1/16. Hmm. So corner region is BIGGER than conv(0,(1,0,0),mids)? The region cone∩O: O’s boundary within cone: the O faces at vertex (1,0,0) are 4 triangles {(1,0,0),mid,mid’} — do these fill the cone’s exit? The cone exits O also through… the cone spans directions between R1=(1,0,−1) and R4=(1,1,0): direction (1, 0.5, −0.5): |x|+|y|+|z| = 2 → exit at (0.5, 0.25, −0.25): is that on an O face? x+y−z = 1 ✓ yes face. (0.5,0.25,−0.25) in conv((1,0,0),(1/2,0,−1/2),(1/2,−1/2,0))? That triangle: (1,0,0),(1/2,0,−1/2),(1/2,−1/2,0): plane: x+y−z = 1 ✓ all three satisfy ✓ and it’s conv of 3 O points on same face → the O portion. (0.5,0.25,−0.25): barycentric: = α(1,0,0)+β(1/2,0,−1/2)+γ(1/2,−1/2,0): y: −γ/2 = 1/4 → γ = −1/2 <0 ✗ NOT in that triangle! So the exit point is on O face but outside the corner triangles conv((1,0,0), mid, mid’) !
So on O’s face x+y−z=1 (triangle conv((1,0,0),(0,1,0),(0,0,−1))), the region’s footprint is bigger: region footprint on this face = {points of face in cone}: face contains directions… The plate face of region: hmm, the region’s footprint on face f: constraint on face: h2=1 (const). Region constraints: h1>0, h3>0, h4<0: on the face triangle: h3 = x−y+z = 1−2y… wait on face: x+y−z=1: h3 = x−y+z = (1+z−y)−y+z = 1+2z−2y > 0 → z > y−1/2; h4 < 0: −x+y+z<0: x = 1+z−y: −(1+z−y)+y+z = −1+2y < 0 → y < 1/2; h1: x+y+z = 1+2z > 0 → z > −1/2. Face coords: face is triangle with vertices (1,0,0),(0,1,0),(0,0,−1): parametrize by (y,z) with x = 1+z−y, |·|≤1 conditions: y≥0? face spans: vertices give (y,z) ∈ {(0,0),(1,0),(0,−1)}: triangle {y≥0, z≤0, y−z≤1}. Region footprint: y<1/2, z>−1/2… wait also z> y−1/2, z<0: footprint = {0≤y<1/2, max(−1/2, y−1/2) < z ≤ 0…}: hmm z ranges: z ∈ (−1/2?, 0]: but z ≥ … face constraints z ≤ 0, y−z ≤ 1. Footprint: {(y,z): 0≤y≤1/2, z∈[y−1/2, 0]}: quadrilateral! vertices (y,z) = (0,−1/2)?? y=0,z=−1/2: point (x,y,z) = (1−1/2−… x = 1+z−y = 1/2: (1/2,0,−1/2) = R1 ✓; (0,0) → (1,0,0) ✓; (1/2, 0) → x = 1/2: (1/2,1/2,0) = R4 ✓; and y=1/2, z = 0: same R4? hmm: y≤1/2, z≥y−1/2: corner (1/2, 0): z=0: point (1/2,1/2,0). And corner (0,0)→(1,0,0). So footprint quad = conv((1,0,0),(1/2,1/2,0),(1/2,0,−1/2), and…? The constraint z ≥ y − 1/2 gives edge from (0,−1/2) to (1,0)→ wait z=y−1/2 at y=1: z=1/2>0 not allowed; within y ≤ 1/2: edge from (y,z)=(0,−1/2) to (1/2, 0): both R1 and R4! So footprint = triangle conv((1,0,0), (1/2,1/2,0), (1/2,0,−1/2))! Wait but then (0.5,0.25,−0.25): (y,z) = (0.25,−0.25): z ≥ y−1/2 ✓ −0.25 ≥ −0.25 ✓ boundary — so it’s ON the edge: inside footprint ✓ (barycentric failing just means on boundary beyond that triangle’s edge… but γ<0 means outside conv… contradiction?). Let me recheck: triangle conv((1,0,0),(1/2,1/2,0),(1/2,0,−1/2)): (0.5,0.25,−0.25): = α(1,0,0)+β(1/2,1/2,0)+γ(1/2,0,−1/2), α+β+γ=1: y: β/2 = 1/4 → β=1/2; z: −γ/2 = −1/4 → γ = 1/2; α = 0. Sum: 0 + 1/2+1/2 = 1 ✓ α=0: ON the edge between R4 and R1 ✓ consistent (it’s on the barycentric edge, x=1/2+… x = α+β/2+γ/2 = 0+1/4+1/4=1/2 ✓). OK and that edge (1/2,1/2,0)–(1/2,0,−1/2) lies in plane x=1/2? No wait both points have x=1/2: edge in plane x=1/2, and these two points… distance apart: √(1/4+1/4) = √2/2 ✓.
So footprint of the corner region on each of its 4 O-faces = triangle conv((1,0,0), two adjacent mids) — consistent with earlier. Then why did region∩O exceed conv(0,(1,0,0),mids)? Point (0.5,0.25,−0.25) ∈ region∩O: is it in conv(0,(1,0,0),(1/2,0,−1/2),(1/2,−1/2,0))? That’s face-plane conv… conv(0, R1pt, R2pt) is a triangle; (0.5,0.25,−0.25): y=0.25 >0, but conv(0,(1/2,0,−1/2),(1/2,−1/2,0)) has y ≤ 0 ✗. So (0.5,0.25,−0.25) ∈ region∩O but NOT in conv(0,(1,0,0),4 mids): so region∩O ⊋ that octahedron. Hmm! So region∩O = bigger. But the full O face footprint was just the triangle conv((1,0,0),R4,R1)… and (0.5,0.25,−0.25) IS in that triangle (on its edge). But it’s not in conv((1,0,0), mids…)… The octahedron conv(0,(1,0,0),mids) face on plane x+y−z=… the octahedron’s face conv((1,0,0),(1/2,−1/2,0),(1/2,0,−1/2)): is (0.5,0.25,−0.25) in it? y = 1/4: solved above γ… that triangle: α(1,0,0)+β(1/2,−1/2,0)+γ(1/2,0,−1/2): y: −β/2 = 1/4 → β = −1/2 outside ✗. But (0.5,0.25,−0.25) should be on region face conv((1,0,0),R4pt,R1pt) = conv((1,0,0),(1/2,1/2,0),(1/2,0,−1/2)) — I used R2pt (=(1/2,−1/2,0)) by mistake. Right: but is conv((1,0,0),(1/2,1/2,0),(1/2,0,−1/2)) a FACE of the octahedron conv(0,(1,0,0),4 mids)? The octahedron’s faces on O: triangles conv((1,0,0), mid_i, mid_{i+1}) in cyclic order around: mids cyclic order around vertex (1,0,0): going around: R4pt=(1/2,1/2,0)? adjacent mids on face are pairs sharing an O face: faces at vertex (1,0,0): {+x+y+z}, {x+y−z}, {x−y+z}, {x−y−z}: each face contains 2 of the 4 mids: pairs: {(1/2,1/2,0),(1/2,0,1/2)} (face +++); {(1/2,1/2,0),(1/2,0,−1/2)} (face ++−)?? (1/2,0,−1/2) on face x+y−z=1: 1/2+0+1/2 = 1 ✓ yes; and (1/2,1/2,0) ✓. So pair exists ✓. {(1/2,−1/2,0),(1/2,0,−1/2)} (face x−y+z=1: 1/2+1/2−1/2 = 1/2 ≠ 1?? x−y+z = 1/2 −(−1/2)+… wait x−y+z = 1/2+1/2·−1… compute (1/2,−1/2,0): x−y+z = 1/2+1/2+0 = 1 ✓; (1/2,0,−1/2): 1/2−0−1/2 = 0 ✗ not on that face. Hmm so which face contains pair {(1/2,−1/2,0),(1/2,0,−1/2)}: plane through (1,0,0) and both: normal? Vectors from (1,0,0): (−1/2,−1/2,0),(−1/2,0,−1/2): cross = ((−1/2)(−1/2)−0, 0−(−1/2)(−1/2), 0−(−1/2)(1/2)·… cross = (1/4·… let me: u=(−1/2,−1/2,0), v=(−1/2,0,−1/2): u×v = (u_y v_z − u_z v_y, u_z v_x − u_x v_z, u_x v_y − u_y v_x) = ((−1/2)(−1/2) − 0, 0 − (−1/2)(−1/2), 0 − (1/4)) = (1/4, −1/4, −1/4) ∝ (1,−1,−1): face x−y−z = 1: check (1/2,−1/2,0): 1/2+1/2 = 1 ✓. OK so the four faces at (1,0,0) have normals (1,1,1),(1,1,−1),(1,−1,−1),(1,−1,1), each containing a pair of mids. Full set of pairs: mids M±y=(1/2,±1/2,0), M±z=(1/2,0,±1/2): pairs: (M+y,M+z) on (1,1,1); (M+y,M−z) on (1,1,−1); (M−y,M−z) on (1,−1,−1); (M−y,M+z) on (1,−1,1) ✓ 4 triangles. So octahedron conv(0,(1,0,0),M±y,M±z) IS the region∩O?? But we found (0.5,0.25,−0.25) ∈ region∩O: in the octahedron? Face (1,1,−1)-triangle conv((1,0,0),M+y,M−z): (0.5,0.25,−0.25): α(1,0,0)+β(1/2,1/2,0)+γ(1/2,0,−1/2): solved: α=0,β=γ=1/2 ✓ in it! Great — earlier confusion resolved: it IS in the octahedron (face triangle). OK!
So corner region ∩ O = regular octahedron conv(0,(1,0,0),4 mids), edge √2/2, volume 1/6. Then 14 regions: volumes: corner 6×1/6 = 1; (1,1,1)-type tets 2×1/24 = 1/12; cube-corner-minus tets 6×1/32 = 3/16 = 0.1875. Sum = 1+1/12+3/16. LCD 48: 48+4+9 = 61/48 ≠ 64/48. Still mismatch of 3/48 = 1/16! Hmm. So error somewhere. Check tet volume for (1,1,−1)-cone: conv(0,(1/2,1/2,0),(1/2,0,−1/2),(1/4,1/4,−1/2)): det: rows a=(1/2,1/2,0), b=(1/2,0,−1/2), c=(1/4,1/4,−1/2). det = a·(b×c): b×c = (0·(−1/2) − (−1/2)(1/4), (−1/2)(1/4) − (1/2)(−1/2), (1/2)(1/4) − 0) = (1/8, −1/8+1/4, 1/8) = (1/8, 1/8, 1/8?? compute: second: b_z c_x − b_x c_z = (−1/2)(1/4) − (1/2)(−1/2) = −1/8 + 1/4 = 1/8; third: b_x c_y − b_y c_x = (1/2)(1/4) − 0 = 1/8. So b×c = (1/8,1/8,1/8); a·(b×c) = 1/16+1/16 = 1/8 → vol = (1/6)(1/8) = 1/48?! I previously miscomputed. Redo earlier: earlier I wrote det = 3/16 — wrong. Vol = 1/48. Then 6×1/48 = 1/8. Sum: 1 + 1/12 + 1/8 = 24/24… LCD 24: 24/24 + 2/24 + 3/24 = 29/24?? ≠ 4/3 = 32/24. Missing 3/24 = 1/8! Argh. Something still off.
Hmm, maybe corner region ∩ O is bigger than the half-scale octahedron: does region reach beyond plane x=1/2? Region = {h1>0,h2>0,h3>0,h4<0}∩O. Point (0.4, 0.3, 0.3): wait h4 = −0.4+0.6 = 0.2 >0 ✗. Point (0.4,0.2,0.2): h4 = 0 ✗ boundary. Point (0.4, 0.1, 0.2): h = (0.7, 0.3, 0.5, −0.1): all sat: in region, |·| = 0.7 < 1 ✓. Is it in octahedron conv(0,(1,0,0),mids)? Octahedron = {|y|+|z| ≤ … the half-scale octahedron centered at (1/2,0,0)? Vertices (1,0,0),(1/2,±1/2,0),(1/2,0,±1/2),(0,0,0): center (1/2,0,0), it’s the octahedron {|u|: (x−1/2)… points with |x−1/2|+|y|+|z| ≤ 1/2 ✓ regular octa radius 1/2. (0.4,0.1,0.2): |−0.1|+0.1+0.2 = 0.4 ≤ 0.5 ✓ inside! OK. Try point (0.3,0.35,0.35): h4 = −0.3+0.7 = 0.4>0 ✗. Hmm. What about region points with negative y,z beyond octa? (0.45, 0.3, −0.3): h4 = −0.45+0 = −0.45 <0 ✓, h1 = 0.45, h2 = 1.05, h3 = 0.45: in region, norm 1.05 >1 ✗ outside O. (0.45, 0.25, −0.25): norm 0.95 ✓; h = (0.45, 0.95, 0.45, −0.45) ✓ region. In octa? |0.45−0.5|+0.25+0.25 = 0.55 > 0.5 ✗ NOT in octahedron! So region ⊋ octahedron. Point (0.45,0.25,−0.25): where does it sit? On face x+y−z = 0.95… hmm it’s interior. So region∩O ≠ half-scale octahedron. The octahedron misses the parts near the plates’ far sides… right: the region h4<0 includes points with y+z < x but the octahedron boundary is O’s… hmm no the octahedron conv(0,(1,0,0),mids) should contain all points with… let me test: is (0.45,0.25,−0.25) in the cone? Need it in poshull of R1..R4: solve: aR1+bR2+cR3+dR4: R1=(1,0,−1),R2=(1,−1,0),R3=(1,0,1),R4=(1,1,0): x: a+b+c+d = 0.45; y: −b+d = 0.25; z: −a+c = −0.25. With a+b+c+d = 0.45: from y: b = d−0.25; z: c = a−0.25. Sum: a + (d−0.25) + (a−0.25) + d = 0.45 → 2a+2d = 0.95 → a+d = 0.475. Choose d=0.3, a=0.175: b = 0.05, c = −0.075 <0 ✗. Choose d = 0.475, a = 0: b = 0.225, c = −0.25 ✗. Hmm c = a−0.25 ≥ 0 → a ≥ 0.25 → then d ≤ 0.225 → b = d−0.25 < 0 ✗. So (0.45,0.25,−0.25) NOT in the 4-ray cone! But its h-signs said region+… recompute h: h1 = 0.45+0.25−0.25 = 0.45 >0 ✓; h2 = x+y−z = 0.45+0.25+0.25 = 0.95 ✓; h3 = x−y+z = 0.45−0.25−0.25 = −0.05 <0 ✗!! I fumbled: h3 <0: NOT in region (+,+,+,−). Phew. My arithmetic is sloppy; sign said region ✓ before by mis-computing h3 = 0.45? h3 = x−y+z = 0.45 − 0.25 + (−0.25) = −0.05. Yes region requires h3>0 ✗. OK so that point is in a different region.
Let me stop hand-computing and actually verify numerically. I can’t use python but run_typescript can do Monte Carlo volume computation of each region by sign classification + check whether cell shapes match: for each of the 14 sign patterns, Monte Carlo volume; then compare sums to 4/3. Also test whether each region equals hypothesized shape by sampling. That’s a solid way to verify. Volume by MC: sample O uniformly (hit-or-miss in cube), classify pattern, count. 14 bins. Runtime: fine with ~1e6? Sandbox limit 1024 operations — the previous MC attempt hit the op limit! “exceeded maxProgramOperations (1024)”. Loops are limited — likely each loop iteration counts. 400k iterations way too many. Need efficient: maybe 500 iterations only? MC with 500 points is too noisy for volumes ~1/48. Alternative: exact computation via analytic integration in closed form, or subdivide. Or use the code limit smartly: vectorize? No numpy presumably. Maybe recursion closed forms.
Alternative: compute volumes analytically for each region with careful integration by hand. That’s heavy but doable? Actually maybe I don’t need exact per-region volumes for the deliverable page. The page should present: coordinates, hexagon, fan, cell inventory (6 corner octahedra + 8 tets — verify 8 tet shapes: are all 8 tetrahedra? The 8 cube-corner cones: (±1,±1,±1) dirs: the cone ∩O for dir (1,1,1): conv(0, M+y… three midpoints (1/2,1/2,0),(1/2,0,1/2),(0,1/2,1/2) + 0: tet ✓; for dir (1,1,−1): cone = poshull{(1,1,0),(1,0,−1),(1,1,−2)}∩O: but is the third ray ∩O really (1/4,1/4,−1/2)? Ray direction (1,1,−2): exits O at scale t: |t|+|t|+|2t| = 4t = 1 → t = 1/4 ✓ point (1/4,1/4,−1/2): on O’s face ✓ (interior of face, not an edge). So the tet has vertices 0, two edge-mids and one point in face interior. Fine, still a tetrahedron ✓.
But WAIT: is the cone around (1,1,−1) really 3-rayed? Facets: which h=0 planes bound it: pattern (+,+,−,−): candidate facets: h1=0, h2=0, h3=0, h4=0. Extreme rays: pairs: computed: h1∩h3 = (1,0,−1) ∈ pattern ✓; h3∩h4 = (1,1,0) ✓; h1∩h4 = (1,1,−2) ✓; h2 with others: h2∩h4 = (1,0,1)? h2=0: x=z; h4=0: z=x: same → line (1,0,1): in pattern? h at (1,0,1) = (2, 0, 2, 0) not interior of (+,+,−,−) (h3>0 ✗). h2∩h1: x+y=0… h1=h2=0 → z=0, y=−x: line (1,−1,0): h at: (0,0,2,−2): h1=0: boundary only. Not interior. h2∩h3: x+y−z=0, x−y+z=0 → x=0, y=z: line (0,1,1): h1 = 2 >0 ✓ but h2,h3=0. So cone facets: h1? Let me just confirm: cone = {h1>0,h2>0,h3<0,h4<0}: since h2 = h1−h4−… h2 = h1−h3−h4+… from h4 = h1−h2−h3 → h2 = h1−h3−h4 = h1+|h3|+|h4| >0 whenever h1>0,h3<0,h4<0: so h2 redundant ✓ cone bounded by h1,h3,h4: trihedral, 3 rays = pairwise intersections: h1∩h3, h3∩h4, h1∩h4 ✓ = (1,0,−1),(1,1,0),(1,1,−2) ✓ consistent.
So per-region: 8 cube-corner regions (2 of “sum-type” (all same sign hemisphere? no: all-+ and all−−−) with vol 1/24, 6 with vol 1/48) + 6 corner regions. Volume conservation: 4/3 − 1/12 − 1/8 = 4/3 − 5/24 = 32/24−5/24 = 27/24 = 9/8. 6 corner regions → 9/8/6 = 3/16 each. So corner region volume should be 3/16, NOT 1/6! So corner region is bigger than the regular half-scale octahedron (1/6). Where’s the extra 1/48? 3/16 − 1/6 = (9−8)/48 = 1/48. Hmm interesting. So corner region = octahedron + 1/48 extra. Shape: maybe the corner region isn’t the conv(0,(1,0,0),mids): maybe it doesn’t include origin but bulges near plates? But origin must be in closure of region (all h=0). Hmm, region is convex! {4 halfspaces}∩O convex. conv(0,(1,0,0),mids) ⊆ region? Each vertex in region closure ✓, conv of region points ⊆ region ✓ (region convex). So octahedron ⊆ region, region ⊋ octahedron with extra 1/48. Region = {h1>0,h2>0,h3>0,h4<0}∩O. Extra region vertices: intersections of TWO plate planes with O surface? Plate∩O = hexagon (1D curve on boundary). Hexagon∩hexagon = 2 rays. h1=0∩h3=0 ray (1,0,−1): ∩O = ±(1/2,0,−1/2) ✓ vertices counted. Hmm all plate-pair lines were computed. What about O-face ∩ plate lines as vertices: a vertex of region can be O-face∩plate∩plate: same as above. Or vertex = O-face ∩ 1 plate ∩ O face? A vertex must have 3 active constraints (codim 3): options: 3 plate (origin only or none), 2 plates + 1 O face: plate-pair ray ∩ O-face: ray (1,1,−2)? that’s plate h1∩h4 line: points (t,t,−2t): ∩O: 4t=1 → (1/4,1/4,−1/2): active constraints: h1=0, h4=0, O? h2 = t+t+2t = 4t = 1 >0, h3 = t−t−2t = −2t <0 ✓: and |·| = 1: on O boundary: is it on an O VERTEX? No, face interior. So codim: 2 plates + O surface: yes a vertex! (1/4,1/4,−1/2) is a vertex of REGION (+,+,+,−)?? wait is this point in the closure of the CORNER region (+,+,+,−)? h3 there: −2t <0 ✗ — it’s a vertex of region (+,+,−,−) (the tet CELLS’ vertex) not corner. Hmm right.
So where’s the corner region’s extra volume? Let me recount region count: maybe 14 regions is wrong; maybe corner “regions” split… Or maybe some of the 6 “corner” axis-cones have different shapes. Region (+,+,+,−) cone: 4 rays R1..R4, quadratic… but region∩O includes also… hmm wait maybe corner region cone has MORE extreme rays: facets could include plate h1=0 twice? Extreme rays of {h1>0,h2>0,h3>0,h4<0}: pairs of active planes (corank… in 3D vertex of cone = intersection of 2 planes generally): all pairs: h1h2: (1,−1,0)✓; h1h3: (1,0,−1)✓; h1h4: (1,1,−2)? h1=0,h4=0: x+y+z=0, −x+y+z=0 → x = y? subtract: 2x = 0 → x = 0! Wait solve properly: x+y+z = 0 and −x+y+z = 0: subtract: 2x = 0 → x = 0; then y+z = 0: line (0,1,−1)!! Earlier I computed (1,1,−2) — error! Check: h1=0 ∧ h4=0: from eqs: x=0, y=−z → (0,1,−1) ✓. Earlier “h4 = h1−h2−h3” derived line (t,t,−2t) for h1=h4=0: check (1,1,−2): h1 = 1+1−2 = 0 ✓, h4 = −1+1−2 = −2 ≠ 0 ✗. So (1,1,−2) satisfies h1=0 but not h4: it was h1∩h3∩? whatever. Correct line: (0,1,−1): in region? h2 = 2>0 ✓, h3 = −2 <0 ✗ not in region. OK.
So redone extreme rays for (+,+,+,−): pairs:
h1h2: x+y+z=0, x+y−z=0 → z=0, x=−y: (1,−1,0): h3 = 2 ✓, h4 = −2 ✓: ray.
h1h3: x+y+z=0, x−y+z=0 → y=0, x=−z: (1,0,−1): h2 = 2 ✓ h4 = −2 ✓: ray.
h1h4: (0,1,−1): ✗ (h3<0).
h2h3: x+y−z=0, x−y+z=0 → y=z, x=0: (0,1,1): h1 = 2 ✓, h4 = 0: boundary h4=0 not <0: ✗ borderline (on facet between?).
h2h4: x+y−z=0, −x+y+z=0 → y=0, x=z: (1,0,1): h1 = 2 ✓, h3 = 2, h4=0: boundary. Hmm h4=0 exactly: it’s ON the h4 facet plane but is it in the region’s closure? Yes h4≤0: 0 ✓. Is (1,0,1) an extreme RAY of the cone? If it’s on the facet h4=0 and satisfies h1,h2,h3>0 strictly: it’s in the relative interior of the facet → not extreme unless also on another plane. h2=h4=0 exactly: two planes: so it IS an intersection of h2 and h4 → extreme ray! Wait: h2 = 0 AND h4 = 0 exactly (x=z, y=0). Yes h2=0,h4=0 → (1,0,1). So rays: h2h4: (1,0,1) ✓.
h3h4: x−y+z=0, −x+y+z=0 → x=y, z=0: (1,1,0): h1 = 2 ✓, h2 = 2, h4 = 0: also intersection h3=h4 → extreme ray ✓.
So rays: (1,−1,0) [h1h2], (1,0,−1) [h1h3], (0,1,1)?? h2h3: (0,1,1): h1 = 2 >0, h4 = 0: intersection of h2,h3 planes: extreme ray candidate: in closure ✓. And (1,0,1) [h2h4], (1,1,0) [h3h4]. That’s 5 candidate rays! Cone {h1,h2,h3>0, h4<0}: this is a cone with 4 facet planes (none redundant? h1 could be redundant? h1 = h2+h3+h4: h2,h3>0, h4<0: h1 could be negative: not redundant generally). Hmm but a pointed cone in 3D bounded by 4 halfspaces: has ≥3 extreme rays, at most… pointed cone with 4 facets: polyhedral cone with 4 facets in 3D: like a “pyramid cone”: cross-section = quadrilateral → 4 extreme rays. But I found 5 candidate rays: one must be non-extreme (inside the cone spanned by others). Check each ray against others: is (0,1,1) ∈ cone of other four? Solve a(1,−1,0)+b(1,0,−1)+c(1,0,1)+d(1,1,0) = (0,1,1), coefficients ≥ 0: y: −a+d = 1; z: −b+c = 1; x: a+b+c+d = 0 → all zero → 0 = 0 impossible with y,z ✓: need x=0 with nonneg combos of vectors all having x=1: only possible if all zero → contradiction. So (0,1,1) NOT in span of others (their cone has x>0 always!). So (0,1,1) is an extreme ray, and similarly (1,0,1),(1,1,0) etc. All 5 rays: cone = poshull of the 5: but a 3D cone with 5 extreme rays and 4 facets: impossible (Euler). So one of my rays is wrong. Recheck h3h4: x−y+z = 0 ∧ −x+y+z = 0: add: 2z = 0 → z = 0; then x−y = 0 → x = y: ray (1,1,0) ✓. h2h4: x+y−z = 0 ∧ −x+y+z = 0: add: 2y = 0 → y = 0; x−z… first: x = z−y = z → x = z: ray (1,0,1) ✓. h1h2: x+y+z = 0 ∧ x+y−z = 0: subtract: 2z = 0, z=0: x+y = 0: (1,−1,0) ✓. h1h3: y = 0, x+z… x−y+z = 0 ∧ x+y+z = 0: subtract 2y → y = 0; x+z = 0: (1,0,−1) ✓. h2h3: x+y−z = 0 ∧ x−y+z = 0: add 2x = 0: x = 0; then y = z: (0,1,1) ✓.
All 5 satisfy the pattern(open): (1,−1,0): h=(0,0,2,−2) boundary h1=0,h2=0 — wait (1,−1,0): h1 = x+y+z = 0 ✓ h2 = x+y−z = 0 ✓ h3 = 2 > 0, h4 = −2: so (1,−1,0) is on TWO planes h1,h2 and interior to h3,h4: extreme ray ✓ (intersection of its 2 planes).
Similarly all 5 are extreme?? But 5 rays, 4 facets: impossible for pointed cone… unless one facet plane contains 3 rays (coplanar within plane): each facet is a 2D cone (angle) with 2 extreme rays, 4 facets × sharing: total incidence 4×2 = 8 ray-facet incidences, each extreme ray on ≥2 facets: 5 rays × ≥2 = ≥10 > 8: contradiction, so at most 4 extreme rays. So one pair-line isn’t actually a ray of the cone because the cone is NOT all of {h_i signs}: cone = {h1>0 ∧ h2>0 ∧ h3>0 ∧ h4<0}: every pair-line must satisfy OTHER two strictly (or be redundant). Check each:
(1,−1,0): h3 = x−y+z = 2 >0 ✓ h4 = −2 <0 ✓ OK.
(1,0,−1): h2 = 2 ✓ h4 = −2 ✓ OK.
(0,1,1): h1 = 2 ✓ h4 = −0+1+1 = 2 >0 ✗ VIOLATES h4<0! (0,1,1): h4 = −x+y+z = 2 >0 ✗. So excluded ✓.
(1,0,1): h1 = 2 ✓ h3 = 2 ✓ h4 = 0: on boundary of h4 ✓ extreme ray of closure ✓ OK.
(1,1,0): h1 = 2 ✓ h2 = 2 ✓ h3 = 0 ✓ h4 = −2: on h3 boundary: OK extreme ray.
So 4 extreme rays: R_a=(1,−1,0), R_b=(1,0,−1), R_c=(1,0,1), R_d=(1,1,0). Cone = poshull{R_a,R_b,R_c,R_d}? Check h1 for poshull: h1(R) = Ray a: 0; b: 0; c: 2; d: 2. Hmm the cone spanned: is {hcone} = poshull of 4 rays? Facets: between consecutive rays in cyclic order. Order them by angle around axis (1,0,0): rays: (1,0,1) [z+], (1,1,0) [y+], (1,−1,0) [y−], (1,0,−1) [z−]: cyclic order around x-axis: (1,0,1) → (1,1,0) → (1,−1,0) → (1,0,−1) → back. Facets = 2D cones between consecutive: between (1,0,1),(1,1,0): plane containing both: cross = (1,0,1)×(1,1,0) = (0·0−1·1, 1·1−1·0, 1·1−0·1) = (−1,1,1) ∝ h4 normal (−1,1,1) ✓ h4=0 facet ✓. Between (1,1,0),(1,−1,0): cross = (1,1,0)×(1,−1,0) = (0,0,−2) ∝ z: plane z=0: NOT a plate! Hmm: but region constraints: z=0 plane is not among h’s. Between (1,1,0) and (1,−1,0): both satisfy h3>0? (1,1,0): h3 = 0; (1,−1,0): h3 = 2: facet between them should be h3=0? But h3 (1,−1,0) = 2 ≠ 0. So facet between rays (1,1,0),(1,−1,0) is z = 0 — but z=0 isn’t a constraint → contradiction: so rays (1,1,0),(1,−1,0) are NOT adjacent; there must be another ray between them: i.e. (0,1,1)-type on the z… between (1,1,0) and (1,−1,0) around the x-axis the cone is bounded by h3 = x−y+z = 0 plane: h3=0 ∧ h? The facet h3=0 within the cone = {x−y+z=0, h1>0,h2>0,h4<0}: extreme rays of this facet: 2D cone: rays in plane h3=0: candidates: (1,1,0) [h3=0 ✓] and… intersections h3∧h4: (1,1,0) same thing? h3∧h4 computed (1,1,0). h3∧h1: (1,0,−1)?? h3∧h1: y = 0, x = −z: (1,0,−1): and h3∧h2: (0,1,1). So the facet h3=0 has candidate rays (1,1,0), (1,0,−1), (0,1,1): the 2D cone {h3=0, region conds} = spanned by which? Points in h3=0: (x, y, z) with x+z = y: region conds: h1 = x+y+z = 2y−x… h1 = x + (x+z) + z = 2(x+z) = 2y > 0 → y>0; h2 = x+y−z = x + x+z −z = 2x > 0 → x > 0; h4 = −x+y+z = −x + x + z + z = 2z < 0 → z < 0. So facet = {(x,y,z): y = x+z, x>0, z<0, y>0} = poshull of directions (1,0,−1) and… as x>0, z<0: direction = combination: param: (x, x+z, z), x>0, z<0: extreme: z→0−: (1,1,0); x→0+: (0,1,1)?? x=0: (0, z, z)… wait (x, y, z) = (x, x+z, z): x=0 → (0, z, z) ∝ (0,1,1) with z<0?? z<0 → (0,−1,−1)… hmm z<0: point (0, z, z), z<0 ∝ (0,1,1) scaled by negative: direction (0,−1,−1)?? But y> 0 required: y = x+z = z < 0 ✗ contradiction → no ray at x=0. So facet h3 = 0 = 2D cone spanned by (1,1,0) and (1,0,−1)?? with intermediate: x ∈ (0,1)-mapping: point (x, x+z, z) with x, −z > 0: = x(1,1,0)·? Take combo α(1,1,0) + β(1,0,−1) = (α+β, α, −β): matches (x, x+z, z): α = x+z = y, −β = z → β = −z, α+β = x+z−z = x ✓. So facet = poshull{(1,1,0),(1,0,−1)} with α,β>0 needing α = y>0, β = −z>0 ✓. So the facet between rays (1,1,0) and (1,0,−1)! So cyclic order: (1,0,1) → (1,1,0) → (1,0,−1) → (1,−1,0)?? facet between (1,0,−1),(1,−1,0): cross = (1,0,−1)×(1,−1,0) = (0·0 − (−1)(−1), (−1)(1) − 1·0, 1·(−1) − 0) = (−1, −1, −1) ∝ h1 ✓ facet h1 = 0 ✓. Great: cyclic order: c=(1,0,1) → d=(1,1,0) → b=(1,0,−1) → a=(1,−1,0) → c: facets: h4 (c,d), h3 (d,b), h1 (b,a), h2 (a,c) ✓ all four plates ✓ 4 extreme rays ✓.
So corner region cone = poshull{(1,0,1),(1,1,0),(1,0,−1),(1,−1,0)} (all with x-normalized). Cone ∩ O: exit points of rays: each (1,t,·): norm 1+2|t| = 1 → only t=0?? |x|+|y|+|z| for (1,1,0) = 2: ray ∩ O at t scaling: t→ direction scaled to norm 1: (1/2,1/2,0) ✓ etc. All 4 exit at the mids ✓. But O∩cone is NOT just conv(0, mids, (1,0,0)): because O’s constraint |x|+|y|+|z| ≤ 1 cut of the cone: the cone’s cross-section at plane x=const: square; O cuts… For point (x,y,z) in cone: |y|+|z| ≤ … in cone: |y|,|z| … cone = {x ≥ |y|, x ≥ … hmm: poshull of (1,0,±1),(1,±1,0): point = Σ: (Σa_i, a2−a4… : (A, d−b… let’s see: (x,y,z) = a(1,0,1)+b(1,1,0)+c(1,0,−1)+d(1,−1,0): y = b−d, z = a−c, x = a+b+c+d: so x ≥ |y| ✓ x ≥ |z| ✓? y: b−d, |b−d| ≤ b+d ≤ x ✓ yes: cone = {x ≥ |y| ∧ x ≥ |z|, x ≥ 0}: a square cone! (cone over square with vertices… indeed {x ≥ |y|, x ≥ |z|} = cone over the square conv(±(1,1,0)/?,-…). Cone ∩ O: {|y|+|z| ≤ 1−x, x≥|y|, x≥|z|}: region∩O — hmm O boundary: |y|+|z| = 1−x. Within cone: which points hit… region∩O = {(x,y,z): x ≥ max(|y|,|z|), |y|+|z| ≤ 1 − x}. Vertices: x=1,y=z=0 → (1,0,0) ✓; x=0 → y=z=0 origin ✓; x=1/2,y=1/2,z=0 ✓ etc. Is this conv(0,(1,0,0),4 mids)? That oct = {|x−1/2|+|y|+|z| ≤ 1/2}: contains (0.3,0.3,0): |−0.2|+0.3 = 0.5 ≤ 0.5 boundary; region∩O: x=0.3 ≥ 0.3 ✓, |y|+|z| = 0.3 ≤ 0.7 ✓: on boundary? interior of region∩O ✓. Both contain. Difference point: (0.9, 0.2, 0.2): region∩O: x ≥ 0.2 ✓, 0.4 ≤ 0.1?? 1−0.9 = 0.1 < 0.4 ✗ outside. Hmm. Try (0.75, 0.35, 0): region: x≥0.35 ✓, 0.35 ≤ 0.25? ✗ no. (0.7, 0.3, 0): 0.3 ≤ 0.3 ✓ boundary: in octa? |0.7−0.5|+0.3+0 = 0.5 ✓ boundary. Hmm equal again?! Claim: on region∩O boundary, |y|+|z| = 1−x with x ≥ max(|y|,|z|): octa: |x−1/2|+|y|+|z| = 1/2. If x ≥ 1/2: 1−x+|y|+|z| = 1/2 → same as |y|+|z| = x − 1/2?? no: octa condition at x≥1/2: |y|+|z| ≤ x − … |x−1/2| = x−1/2: octa: |y|+|z| ≤ 1/2−(x−1/2) = 1−x ✓ SAME. If x ≤ 1/2: octa: |y|+|z| ≤ 1/2 − (1/2 − x) = x; region: |y|+|z| ≤ 1−x. Compare: x vs 1−x: x ≤ 1/2 → x ≤ 1−x → octa is TIGHTER: so region∩O ⊇ octa, with extra where x<1/2 and |y|+|z| ∈ (x, 1−x), x ≥ max(|y|,|z|). So extra volume exists ✓ consistent with needed 3/16 total. Extra vertices: region∩O vertices: also where |y|+|z| = … vertices with more constraints: x ≥ |y|, x ≥ |z| tight + O tight: x = |y| = |z|, |y|+|z| = 1−x: x = |y| = |z|, 2x = 1−x → x = 1/3: points (1/3, 1/3, ±1/3) and (1/3, −1/3, ±1/3)? x ≥ |y|: y = ±1/3 ✓: four points: (1/3, 1/3, 1/3)? Check pattern (+,+,+,−): this point is boundary of plates: x = y = z: h2 = x+y−z = 1/3 > 0 ✓ h1 = 1 > 0… wait x=y=z=1/3: h1 = 1 >0 not zero. Hmm active constraints: x = |y| (cone boundary? but cone boundaries are derived: the cone = {x≥|y|,x≥|z|}: its boundaries in 3D are planes x=±y, x=±z — these are NOT h-planes; wait the cone poshull{(1,0,1),(1,1,0),(1,0,−1),(1,−1,0)} = {x ≥ |y|, x ≥ |z|}: 4 facets x=y, x=−y, x=z, x=−z?? but earlier I derived cone facets = h-planes h2, h4, h3, h1! Contradiction. Resolve: is (1,0,1) direction in {x≥|y|, x≥|z|}: yes x=1≥0,1. Hmm both descriptions… cross relations: facet between c=(1,0,1),d=(1,1,0) I said h4=0: points αc+βd = (α+β, β, α): −x+y+z = −(α+β)+β+α = 0 ✓ indeed facet ⊂ h4=0! And also x = α+β ≥ β = y? need x ≥ y: α ≥ 0 ✓. Facet = {h4 = 0, x≥y?…}: h4=0 ∧ x≥|z|? Points on h4=0: y = x−z: cone conds: x ≥ |x−z| ∧ x ≥ |z|: x ≥ x−z → z ≥ 0; x ≥ z−x → 2x≥z. So facet = {z≥0, z≤2x, h4=0, x≥z}: vertices rays z… facet extreme rays: (1,0,1) [z=x] and (1,1,0) [z=0] ✓ consistent. OK so both descriptions equivalent: {x≥|y|, x≥|z|} = {h1>0,h2>0,h3>0,h4<0} ✓ (they’re the same cone! since h2 = x+y−z… let me trust the algebra: cone C1 = poshull of four (1,·,·): {x ≥ |y| ∧ x ≥ |z|}: check (1,0,1): ✓; check h-signs of a general point of C1: (2, 0.5, 1): h1 = 3.5, h2 = 2.5, h3 = 2.5, h4 = −0.5 ✓. Equivalence plausible ✓.)
So region∩O = {x ≥ max(|y|,|z|)} ∩ {|y|+|z| ≤ 1−x}. Vertices: (1,0,0), (0,0,0)… wait is (0,0,0) a vertex? Constraints at origin: x≥|y|, x≥|z| tight? x=y=z=0: yes tight ✓ and O: 0 ≤1 not tight. Vertex needs 3 tight: only 2 tight at origin: so origin NOT a vertex?! Near origin within cone: full cone = 4 facets: origin is the cone apex: in {x≥|y|,x≥|z|} the origin IS an extreme point (vertex) of the cone. In region∩O = cone∩{halfspace-ish O}: O condition |x|+|y|+|z|≤1 is inactive at origin → region∩O near origin = full cone → origin is a vertex of region∩O ✓ (extreme point: yes, cone apex: any point of region∩O ≠ origin: is 0 ∈ extreme? A convex set’s vertex requires 3 active constraints typically, but a cone apex has all rays… 0 is an extreme point of the cone: yes definitely (cone is pointed). So vertex ✓.
Other vertices: (1/3,±1/3,±1/3) with x≥|y|,…: (1/3,1/3,1/3): constraints x=|y| ✓ tight, x=|z| ✓ tight, O tight ✓: 3 tight: vertex! Which is it in region closure: cone: x≥|y|: equality ✓ allowed boundary: yes in closure ✓. So vertices: (1,0,0); (0,0,0); (1/2,±1/2,0),(1/2,0,±1/2) [x=1/2: tight constraints: x=|y| and O: |y|+|z|=1−1/2+… (1/2,1/2,0): x=|y| ✓ tight, O ✓ tight: vertex ✓]; (1/3,±1/3,±1/3) four of them: 3+… total: 2 + 4 + 4 = 10 vertices! Hmm: but wait are the (1/2,·) points still vertices given the bigger region? (1/2,1/2,0): is it extreme or now interior to an edge/face? Constraints tight: x = |y|, |y|+|z| = 1−x: plus? z = 0 not constraint. 2 tight < 3: so (1/2,1/2,0) might be interior of an edge between vertices (1/3,1/3,1/3)… wait (1/2,1/2,0): which polygon: near it: the cone facet x=y contains points (t, t, z), z free: with O: 2t + |z| ≤ 1−t → |z| ≤ 1−3t: t ≤ 1/3 → z-extent at t=1/3 is 0: so the facet x=y intersects O in triangle conv((1/2,1/2,0)?? points (t,t,z): at t=1/2: z=0 single point? |z| ≤ 1−3/2 <0: hmm contradiction: t=1/2: |z| ≤ 1−3t = −1/2: empty! So (1/2,1/2,0) is NOT in region∩O?! Check: (1/2,1/2,0): O: |1/2|+|1/2|+0 = 1 ✓ boundary ✓; cone: x ≥ |y|: 1/2 ≥ 1/2 ✓ tight ok; x ≥ |z| ✓. O constraint: |y|+|z| ≤ 1−x: 1/2 ≤ 1/2 ✓. So it IS in. But my parametrization (t,t,z) with O: 2t+|z| ≤ 1−t → |z| ≤ 1−3t: at t = 1/2: |z| ≤ −0.5?? but point has |z| = 0 ≤ −0.5 FALSE: 2t + |z| = 1 + 0 = 1 ≤ 1−t = 0.5?? |y|+|z| = t+|z|?? y = t = 1/2: |y|+|z| = 1/2 + 0 = 1/2; 1−x = 1/2: ✓ 1/2 ≤ 1/2 ✓. I wrote |y| as 2t: |y| = t not 2t! Sloppy: constraint: t + |z| ≤ 1−t → |z| ≤ 1−2t: t ≤ 1/2 ✓ at t=1/2, z=0 ✓ vertex of the facet (2D): facet x=y∩O = triangle conv((1/3?,…)): points (t,t,z), |z| ≤ 1−2t, t ≥ 0… vertices: t=1/2,z=0 → (1/2,1/2,0); t=1/3, z=±1/3 → (1/3,1/3,1/3),(1/3,1/3,−1/3)! And t=0: |z|≤1: but cone also x≥|z| → t=0 → |z| ≤ 0 → origin? t≥0 and z between: at t=0: z = 0: origin. Hmm is origin a vertex of this facet: facet = {(t,t,z): 0≤t, |z| ≤ 1−2t, |z| ≤ t}: at t→0, z→0: the facet polygon vertices: (0,0,0)?? constraints |z| ≤ min(t, 1−2t): polygon in (t,|z|) space: vertices: t=0,z=0; t=1/3, z=±1/3; t=1/2, z=0: so facet is a quadrilateral conv(origin?, (1/3,1/3,±1/3), (1/2,1/2,0))?? Is origin a vertex: near t=0: |z| ≤ t: rates equal: the polygon has corner at origin ✓ yes vertex (edge from origin with slope |z|=t i.e. ray (1,1,±1)… direction (1,1,1)! So edge from origin along (0,1,1)-direction… (t,t,t) ∝ (1,1,1)? Hmm (t,t,z) with z=±t: (t,t,t) ∝ (1,1,1) which lies in plate h2,h3… anyway.
OK so the CORNER REGION is a 10-vertex polyhedron, NOT a regular octahedron!! Vertices: (1,0,0), (0,0,0), 4 mids (1/2,±1/2,0),(1/2,0,±1/2), and 4 points (1/3,±1/3,±1/3). Hmm wait check (1/3,−1/3,1/3): x ≥ |y| = 1/3 ✓, x ≥ |z| ✓: in region? h: (1,1/3?… h2 = x+y−z = 1/3−1/3−1/3 = −1/3 <0: pattern (+,−,+,+)?? h2 <0! So (1/3,−1/3,1/3) is NOT in corner region (+,+,+,−): it’s on a different corner’s boundary. Right: corner region (+,+,+,−) requires h-cone = {x ≥ |y|?? no: cone = poshull{(1,0,1),(1,1,0),(1,0,−1),(1,−1,0)}: contains (1,−1,0) → includes y negative: h1>0,h2>0,h3>0,h4<0: (1,−1,0): h2 = 0 ✓. Point (1/3,−1/3,1/3): in cone? x=1/3 ≥ |y| = 1/3 ✓ tight, ≥|z| = 1/3 ✓ tight: in cone boundary: h2 = 1/3 −1/3 −1/3 = −1/3?? But cone should imply h2>0. (1/3,−1/3,1/3): x≥|y|, x≥|z| say h2 = x+y−z = 1/3 −1/3 −1/3 = −1/3 <0. CONTRADICTION → my claimed cone equivalence {x≥|y|,x≥|z|} FALSE. Because poshull of those 4 rays: (1,−1,0) and (1,0,1) combo: 0.5(1,−1,0)+0.5(1,0,1) = (1,−0.5,0.5): h2 = 1−0.5−0.5 = 0 ✓ on boundary. Combo (1/3,−1/3,1/3) = a(1,−1,0)+b(1,0,1)+…: x: a+b+c+d = 1/3 (scaled); y: −a+d = −1/3; z: c−a = 1/3: with a+b+c+d = 1/3, all ≥ 0: from y: d = a − 1/3 ≤ 0 since a ≤ total? a ≤ 1/3 → d ≤ 0 → d = 0, a = 1/3; then b+c = 0 → b=c=0; but z: c−a = −1/3 ≠ 1/3 ✗. So (1/3,−1/3,1/3) NOT in cone ✓ (I misplaced sign: the four ray points give (1/3-type) vertices only for combos within the cone: (1/3,1/3,1/3): a(0,1,… x: a+b+c+d = 1/3; y: b−d = 1/3; z: a−c = 1/3: b = 1/3+d, a = 1/3+c: sum: 1/3+c+1/3+d+c+d = 1/3?? → 2/3 + 2c + 2d = 1/3 → c+d = −1/6 <0 ✗!! So (1/3,1/3,1/3) also NOT in cone?! Let me recheck the vertex derivation: I previously derived vertex (1/3,±1/3,±1/3) from “x = |y| = |z|, |y|+|z| = 1−x” using cone = {x≥|y|, x≥|z|} — which is now in question. UGH.
Restart cone computation cleanly. Corner region cone C = {h1>0, h2>0, h3>0, h4<0} where h1 = x+y+z, h2 = x+y−z, h3 = x−y+z, h4 = −x+y+z (h4<0 means −x+y+z<0, i.e. y+z<x). In terms of s = y+z, d = y−z: h2 = x+s >0… h2 = x+(y+z) = x+s; h3 = x − (y−z) = x−d; h1 = x+s… wait h1 = x+y+z = x+s; h4 = −x+s. Conditions: x+s>0 (h1), x+s>0 same as h1?? h1 = x+y+z = x+s ✓ and h2 = x+y−z = x+s −… h2 = x + y − z = x + (y−z)= x + d!! Let me recompute: h2 = x+y−z = x + (y−z) = x+d. h3 = x−y+z = x − (y−z) = x−d. h1 = x+y+z = x+s. h4 = −x+y+z = s−x. So conditions: x+s>0, x+d>0, x−d>0, s−x<0 (s<x). So: d ∈ (−x, x) → |d| < x; s ∈ (−x, x) → |s| < x; i.e. |y−z| < x and |y+z| < x ⟺ x > |y|+|z|?? |y−z|<x ∧ |y+z|<x ⟺ 2max(|y|,|z|) < x·… actually |y−z| and |y+z| = max = |y|+|z| when… max(|y−z|,|y+z|) = |y|+|z|. So conditions ⟺ |y|+|z| < x.
The corner cone is simply C = {x > |y|+|z|}! The self-dual l1-cone around the x-axis. So corner region ∩ O = {x > |y|+|z|} ∩ {|x|+|y|+|z| ≤ 1}: since x>0: = {|y|+|z| < x, x+|y|+|z| ≤ 1}: substitute u = |y|+|z|: x > u, x+u ≤ 1: a bipyramid region: vertices: (1,0,0); where x=u, x+u=1 → x = u = 1/2: the circle… |y|+|z| = 1/2, x = 1/2: that’s the SQUARE conv(mids) in plane x=1/2 ✓. So region∩O = the double pyramid = bipyramid over the square conv{(1/2,±1/2,0),(1/2,0,±1/2)} with apexes (1,0,0) and (0,0,0)! = TWO square pyramids glued base-to-base = an octahedron? A bipyramid over a square is combinatorially an octahedron (6 vertices, 8 triangular faces)! And with square side √2/2, apex heights: distance from (1,0,0) to plane x=1/2: 1/2; from origin: 1/2: SYMMETRIC bipyramid: all 8 faces: triangles (apex, two adjacent square vertices): edge lengths: apex-to-vertex √((1/2)²+(1/2)²) = √2/2 ✓; base edges √2/2 ✓: ALL EDGES √2/2 → REGULAR OCTAHEDRON! Edge √2/2. Volume: regular octa edge s = √2/2: V = √2 s³/3 = √2·(√2/2)³/3 = √2·(2√2/8)/3 = √2·√2·… (2√2/8) = √2/4: V = √2/4·√2/3 = 2/12 = 1/6 ✓.
Great: corner region ∩ O = regular octahedron, edge √2/2 (scale 1/2), volume 1/6, vertices (1,0,0) and origin + 4 mids ✓✓. My earlier “(1/3,…)”vertices came from wrong cone; discarded.
So volumes: corners 6 × 1/6 = 1. Cube-corner tets: (1,1,1)-type: |cone|: region (+,+,+,+): conditions: h1>0,h2>0,h3>0,h4>0: in s,d: x+s>0, x+d>0, x−d>0, s−x>0 → s > x > |d|… : {x < y+z, |y−z| < x}? = {x < y+z ∧ x > |y−z|} — the “triangle-inequality” cone around (1,1,1) ✓ (x between |y−z| and y+z: correct triangle condition with x playing “side”). ∩O: vertices: origin; exits: cones rays: facet planes h3=0 (x=y−z…) extreme rays: h1∩h2: (1,−1,0)-dir? h1=h2=0: x+s = 0, x+d = 0 → s=−x, d=−x: y = (s+d)/2 = −x, z = (s−d)/2 = 0: ray (1,−1,0) ✓; h1∩h4: s = −x = x?? s = −x ∧ s = x → x = 0 → ray (0,1,1)?: s = y+z = 0 → y = −z: hmm x = 0: ray (0,1,−1): check pattern: h at (0,1,−1): h1 = 0, h2 = 2, h3 = −2 ✗. Hmm. Let me just directly: cone around (1,1,1) with facets h2,h3,h4 (h1 redundant? h1 = h2+h4−… from h4 = h1−h2−h3: h1 = h2+h3+h4 >0 ✓ redundant!). So cone = {h2>0,h3>0,h4>0} = {x+d>0, x−d>0, s>x} = {|d|<x… wait that gives x>0 and x > |d|… plus s > x: {x > |y−z|, y+z > x}: hmm triangles: three extreme rays: pairwise facet intersections: h2∩h3: x+d=0, x−d=0 → x=0,d=0: s free>0: ray (0,1,1) ✓ (direction y=z>0: h4 = s−x = s >0 ✓). h3∩h4: x−d… x=d, s=x: y+z = x, y−z = x → y=x, z=0: ray (1,1,0) ✓. h4∩h2: s=x, x=−d: y+z = x, z−y = x → z = x, y = 0: ray (1,0,1) ✓. Cone = poshull{(0,1,1),(1,1,0),(1,0,1)} ∩O: exits: (0,1,1)→ norm 2 → (0,1/2,1/2) ✓; (1,1,0) → (1/2,1/2,0) ✓; (1,0,1) → (1/2,0,1/2) ✓. Tet = conv(0,(0,1/2,1/2),(1/2,1/2,0),(1/2,0,1/2)) ✓ vol 1/24 ✓ (medial tet on facet (1,1,1)).
Cube-corner tets of other type: dir (1,1,−1): cone {x>|y−z|? pattern (+,+,−,−): conditions: x+s>0, x+d>0, x−d<0 → d > x, s<x… with s=y+z, d=y−z: h3>0 → x−d>0?? wait pattern (+,+,−,−): h3 = x−d < 0 → d > x; h4 = s−x <0 → s < x; h1: x+s>0 → s > −x; h2: x+d >0 → d > −x redundant. So cone: {d > x, |s| < x, d > …}: hmm two-sided? Around direction (1,1,−1): s = 0, d = 2. Cone = {|s| < x < d}?? with d>x and s∈(−x,x): at (1,1,−1): d = 2, s = 0 ✓. Rays: facet h3(=0: d = x): with s = ±x: d=x,s=x → y+z = x, y−z = x → y = x, z = 0: ray (1,1,0); d=x,s=−x → y = 0, z = −x: ray (1,0,−1) ✓; facet h4(=0: s=x… wait h4<0 → s<x: facet s=x=… s=x ∧ d=x → (1,1,0) again ✓ single; facet h1 (s=−x): with d=x: (1,0,−1) again; facet h2 (d=−x): with |s|<x, d>x>−x ✗ no. So cone = 2D?? Cone rays only (1,1,0) and (1,0,−1)?? That’s not a 3D cone! Something’s off: pattern (+,+,−,−) should be full-dimensional. Conditions: h3<0: d > x; h4<0: s < x; h1>0: s > −x; h2>0: d > −x. So {d > max(x,−x) = |x|, |s| < … s ∈ (−x, x) requires x > 0}. So cone = {x>0, |s| < x, d > x} = {x>0, |y+z| < x, y−z > x}: 3D ✓ around… at (1,1,−1): x=1, s=0 ∈(−1,1) ✓, d = 2 >1 ✓. Extreme rays: boundaries: s=x ∧ d=x: (1,1,0)?? s=x: y+z = x; d=x: y−z = x: y=2x?? solve: y+z = t(y-coord… use variables (x,y,z): s = y+z, d = y−z: s = x ∧ d = x → y+z = x, y−z = x → y = x, z = 0 ✓ ray (1,1,0). s=−x ∧ d=x → y = 0, z = −x ✓ (1,0,−1). s=x ∧ d=… wait also boundary s·: 3D cone needs 3 rays: boundaries: d=x is one facet-plane (h3=0); s=x (h4=0… h4 = s−x=0) is a facet s=x; s=−x (h1=0) facet. Facets: h3(=0: d=x), h4(s=x), h1(s=−x); h2 redundant ✓. Rays = pairwise: h3∧h4: d=x, s=x → (1,1,0); h3∧h1: d=x, s=−x → (1,0,−1); h4∧h1: s=x ∧ s=−x → x=0: then d>0: ray (0,1,1)! s=x∧s=−x → x=0, s arbitrary…: x = 0, s = 0?? s = x = 0 AND s = −x = 0: s=0, x=0, d> 0: ray (0,1,1) ✓. Cone = poshull{(1,1,0),(1,0,−1),(0,1,1)}: exit points: (1/2,1/2,0), (1/2,0,−1/2), (0,1/2,1/2). Tet = conv(0, these) — vol? det[(1/2,1/2,0),(1/2,0,−1/2),(0,1/2,1/2)]: a×b = (1/2·(−1/2)−0, 0−1/2·(−1/2)… compute cross a×b = (a2b3−a3b2, a3b1−a1b3, a1b2−a2b1) = (1/2·(−1/2) − 0·0, 0·1/2 − 1/2·(−1/2), 1/2·0 − 1/2·1/2) = (−1/4, 1/4, −1/4); c·(a×b) = (0,1/2,1/2)·(−1/4,1/4,−1/4) = 1/8 − 1/8 = 0?! Volume ZERO — degenerate! The three rays are coplanar?? (1,1,0),(1,0,−1),(0,1,1): det = 1·(0·1 − (−1)·1) − 1·(1·1 − 0) + 0 = 1·(1) − 1·(1) = 0. Coplanar! So the “+,+,−,−” region ∩ O is… the cone is degenerate (planar)?! That means sign pattern (+,+,−,−) contains an open set in R³? conditions {x>0, |s|<x, d>x}: open ✓ nonempty ✓ 3D ✓. But its extreme rays coplanar?? poshull of 3 coplanar rays is 2D — can’t equal a 3D cone. So I erred: cone = {x>0, |s|<x, d>x}: extreme rays: the cone’s boundary consists of pieces: d=x facet (2D region {d=x, |s|<x}): parametrized by (x,s): 2D ✓ facet; s=x facet {d>x, s=x}: coordinates (x,d): d>x: 2D facet… wait is {s=x, d>x, x>0} 2D? Variables x,d with d>x: yes 2D ✓; s=−x facet ✓. Three facets meet pairwise in rays: h3∧h4: d=x, s=x: ray ✓; h3∧h1: d=x, s=−x: ray ✓; h4∧h1: s=x ∧ s=−x: impossible → no ray: the two facets s=±x DON’T intersect: they’re parallel-ish planes (s=x and s=−x are parallel planes!). So the cone has 3 facets = wedge-like: it’s a wedge cone: like {x>0, |s| < x, d > x}: hmm but then extreme rays: two (d=x,s=x) and (d=x,s=−x)?? A pointed 3D cone must have ≥3 extreme rays and pairwise intersecting facets. But here facets s=x and s=−x are parallel: cone is NOT pointed?? Is the cone pointed? Does it contain a line? Lines: direction v=(x0,s0,d0) with cone+v and −v in cone: need |s|<x for ±: s0 = 0? then need s0±: x ± x0 > 0 both → |x0| < x impossible since x→0…: the cone {x>0, |s|<x}: pointed (it’s the cone over interval in (x,s) plane: {(x,s): |s|<x, x>0} is pointed 2D cone with rays (1,±1)). Full cone: pointed? Contains direction with s0=0, d0→+∞ free: v = (0, 0, 1) in (x,s,d) coords: is (0,0,δ) ∈ cone? need x>0 ✗ (x=0 not >0, but closure). Hmm closure of cone: {x≥0, |s|≤x, d≥x}: contains (0,0,d): and −(0,0,1) = (0,0,−1): x=0, |0|≤0 ✓, d ≥ x → −1 ≥ 0 ✗. So pointed ✓. But pointed 3D cone with parallel facets s=±x: closure cone = {x≥0, |s| ≤ x, d ≥ x}: extreme rays: (x,s): (1,1)&(1,−1) rays in (x,s)-plane crossed with d: d… extreme rays of 3D cone = generators: the set is {(x,s,d): |s|≤x, d≥x, x≥0} = {(u-ray in (x,s)): d free ≥ x}: = poshull? Take point (1,0,5): ray direction: generators: (1,1,1)? (1,−1,1), and d→∞ direction (0,0,1)! Closure contains ray (0,0,1) (x=0, d arbitrary): direction of recession: cone contains (ε, 0, large M): scaled → (ε/M, 0, 1) → (0,0,1): so the cone’s closure contains the ray (0,0,1) but cone isn’t closed there: the cone is NOT closedulp: {|s|<x strictly}: it’s a non-closed… The real region ∩ O — since we intersect with O (compact), closure issues at origin only (measure zero). Practically: the region (+,+,−,−)∩O is a “wedge tetrahedron-ish”: bounded by O, three plate planes: shape: {x>0, |s|<x, d>x, x+|y|+|z|≤1}: vertices: origin-ish (0,0,0) apex? At x→0: s→0, d→x→0: collapses to origin ✓ origin vertex; other vertices: expose where O meets: x + |y|+|z| = 1 with constraints: |s|<x, d>x: hmm |y|+|z|: at boundary of |s|: s=±x: points: s=x,d=x: (1,1,0)-dir: exit O at (1/2,1/2,0) ✓; s=−x, d=x: (1,0,−1) → (1/2,0,−1/2); s=x, d→∞ direction?? on O: d large: |y|+|z| grows: point on facet s=x∩O∩region: {y+z = x, x+|y|+|z| ≤ 1, y−z > x}: with y+z = x and y−z > x: y > x, so z < 0: |y|+|z| = y − z = (y−z) > x: x + y − z ≤ 1… parametrize (x, z): y = x − z: constraints: x>0, y−z = x−2z > x → z < 0; x + |y| + |z| = x + (x−z) + (−z) = 2x − 2z = 2(x−z) ≤ 1 → x−z ≤ 1/2; y>0 ✓. Vertices: z… region in (x,z): x>0, z<0, x − z ≤ 1/2, d = x−2z ≥ … also h2 = d−(−x) >0 auto. Vertices: (x,z) = (1/2, 0) → point (1/2, 1/2, 0)?? y = x−z = 1/2 ✓ (1/2,1/2,0) ✓; (0,−1/2): x=0,z=−1/2 → y = 1/2: (0, 1/2, −1/2)!! Not (1/2,0,−1/2) So on the s=x facet, region∩O has vertices (1/2,1/2,0) and (0,1/2,−1/2). Similarly facet s=−x∩O: {y+z = −x: z = −x−y: d = y−z = … constraints d > x: y−z > x → y + x + y > x → 2y > 0 → y > 0; |y|+|z| = y+z… z = −x−y <0 (y>0): |y|+|z| = y + x + y = x + 2y; O: x + x+2y ≤ 1 → x + y ≤ 1/2; vertices: (x,y): y=0 → (1/2, 0): point (1/2, 0, −1/2) ✓; x=0, y=1/2: (0, 1/2, −1/2) ✓ again! And facet d=x∩O: {y−z = x, |s| < x, x+|y|+|z| ≤ 1}: z = y−x: s = y+z = 2y−x, |2y−x| < x → 0 < y < x; |y|+|z| = y + (x−y)… z = y−x <0 iff y<x ✓ so |z| = x−y: |y|+|z| = y + x − y = x: O: x + x ≤ 1 → x ≤ 1/2: vertices: x=1/2, y ∈ (0,1/2): edge! points (1/2, y, y−1/2): from (1/2, 0+, −1/2+) to (1/2, 1/2−, 0−)… hmm so facet d=x∩O is an EDGE of the region: from (1/2,0,−1/2) to (1/2,1/2,0) (open). So region (+,+,−,−)∩O = polyhedron with vertices: (0,0,0) apex? and (1/2,1/2,0), (0,1/2,−1/2), (1/2,0,−1/2): WAIT (0,1/2,−1/2): check in O ✓ norm 1 ✓ boundary; in region: x=0 ✗ x>0 strict: on closure boundary (limit) — hmm it’s the recession direction realized on O: point (0,1/2,−1/2): which sign pattern: h = (0, 0+?…): h1 = 0, h2 = 0+1/2+1/2 = 1 >0, h3 = −1… h3 = 0−1/2−1/2 = −1, h4 = 0: on TWO plates h1=h4=0: it’s an edge-midpoint of O (mid of edge (0,1,0)-(0,0,−1) ✓). So region closure has vertex (0,1/2,−1/2) where plates h1,h4 cross. So region = tetrahedron conv(0,(1/2,1/2,0),(0,1/2,−1/2),(1/2,0,−1/2))!! Check coplanarity issues: det[(1/2,1/2,0),(0,1/2,−1/2),(1/2,0,−1/2)]: a×b = (1/2·(−1/2) − 0·(1/2), 0·0 − 1/2·(−1/2), 1/2·(1/2) − 1/2·0) = (−1/4, 1/4, 1/4); c·(a×b) = (1/2)(−1/4) + 0 + (−1/2)(1/4) = −1/8−1/8 = −1/4 → vol = 1/6·1/4 = 1/24!!
So the “+,+,−,−” region is a tetrahedron of volume 1/24, with vertices origin, TWO edge-midpoints of O, and one point (1/2,1/2,0)… wait all three: (1/2,1/2,0) ✓ mid, (0,1/2,−1/2) ✓ mid, (1/2,0,−1/2) ✓ mid: tet with apex at origin and base = triangle of three edge-mids! So the 6 “one-minus cube-corner” tets have base triangles on…
Hold on, this changes the inventory! Region count: 14 regions = 6 corner octahedra (vol 1/6 each) + 8 tets: 2 “all-same” tets vol 1/24 (at facets (1,1,1) & (−1,−1,−1)) and 6 tets vol… the (+,+,−,−)-type: vol 1/24 too? det was −1/4 → 1/24 ✓. Let me double check with alternative: the base triangle of (+,+,−,−)-tet: vertices (1/2,1/2,0),(0,1/2,−1/2),(1/2,0,−1/2): these three mids all have… they’re mids of edges among {+x,+y,−z}: they form a triangle in plane?: (1/2,1/2,0): y = 1/2… plane through the three: points on plane x+y+z = 1? 1/2+1/2+0 = 1 ✓; 0+1/2−1/2 = 0 ≠ 1 ✗. Plane: normal = a×b where a = (0,1/2,−1/2)−(1/2,1/2,0) = (−1/2,0,−1/2), b = (1/2,0,−1/2)−(1/2,1/2,0) = (0,−1/2,−1/2): a×b = (0·(−1/2) − (−1/2)(−1/2), (−1/2)(0) − (−1/2)(−1/2), (−1/2)(−1/2) − 0) = (−1/4, −1/4, 1/4) ∝ (−1,−1,1): plane −x−y+z = c: at (1/2,1/2,0): −1/2−1/2 = −1: plane −x−y+z = −1, i.e. x+y−z = 1: that’s an O FACE! (facet (1,1,−1)) ✓ So the tet’s base triangle = the medial triangle ON facet (x+y−z=1) ✓ and apex at origin. So: EVERY one of the 8 tets = pyramid with apex ORIGIN and base = medial triangle of a facet! Medial tet ✓ vol 1/24 each ✓ 8/24 = 1/3.
Total: 6 corner octas (1 each = 6/6) hmm 6×(1/6) = 1; plus 8×(1/24) = 1/3. Sum = 4/3 ✓✓✓ PERFECT. Great, everything closes now:
Cell inventory (final, verified):
- 6 corner octahedra: vertices {±axis pts} ∪ origin + 4 edge-mids… each corner octa: conv((1,0,0),(0,0,0),(1/2,±1/2,0),(1/2,0,±1/2)) wait for corner (1,0,0): apexes (1,0,0) and (0,0,0)?? But origin is shared by ALL 14 cells (all plates meet at origin): the corner octa conv(0,(1,0,0),mids) and the tet conv(0, mids’): they share the origin as a common point, and share triangular face? Corner octa (x-axis): faces: 8 triangles: 4 on O (conv((1,0,0),mid,mid’)), 4 on plates: conv(0,(1,0,0)?? face on plate h2=0 (x+y−z=0): the facet between rays… corner octa faces from bipyramid: face (0,0,0)-(1/2,1/2,0)-(1/2,0,1/2)? Let’s see: bipyramid faces: two apexes p=(1,0,0), q=(0,0,0), square mids M+y,M+z,M−y,M−z in cyclic order: faces: (p, M_i, M_{i+1}) ×4 on O and (q, M_i, M_{i+1}) ×4 on plates: e.g. (q, M+y, M+z) = conv(0, (1/2,1/2,0), (1/2,0,1/2)): in which plate plane? These three points: 0 and two mids: plate containing (1/2,1/2,0) & (1/2,0,1/2) & 0: plane x−y−z? through origin: (1/2,1/2,0),(1/2,0,1/2) span: normal = (1/2,1/2,0)×(1/2,0,1/2) = (1/2·1/2 − 0, 0 − 1/2·1/2, 0 − 1/2·1/2)?? cross = (a2b3−a3b2, a3b1−a1b3, a1b2−a2b1) = ((1/2)(1/2)−0·0, 0·(1/2)−(1/2)(1/2), (1/2)(0)−(1/2)(1/2)) = (1/4, −1/4, −1/4) ∝ (1,−1,−1): plane x−y−z = 0 — NOT one of the four plates! Hmm! So the bipyramid faces (q, mid_i, mid_{i+1}) lie in planes ±x±y±z=0 of the OTHER four (the “odd” tetrahedral set: (1,−1,−1),(−1,1,−1),(−1,−1,1),(1,1,1)). x−y−z=0 vs our plates (1,1,1),(1,1,−1),(1,−1,1),(−1,1,1): x−y−z = 0 is NOT among them. PROBLEM! So the corner region is NOT the bipyramid with those faces — the actual region boundaries are the four plates h1..h4, and the bipyramid face planes must coincide with plate planes. Reexamine: corner region = {x > |y|+|z|}∩O — wait no: C = {x > |y|+|z|} was derived as {h1>0,h2>0,h3>0,h4<0}: verify equivalences: h2 = x+d… hold on redo: d := y−z? earlier: h2 = x+y−z = x+d ✓ with d = y−z ✓; h3 = x−d; h1 = x+s, s = y+z; h4 = −x+s. Conditions: x+s>0, x+d>0, x−d>0, s−x<0 → s<x. So: d ∈ (−x, x), s < x, s > −x: hmm h1: x+s>0 → s > −x. So |s| < x ∧ |d| < x. And |y|+|z| = max(|s|,|d|)! Since max(|y+z|,|y−z|) = |y|+|z| ✓. So conditions ⟺ max(|s|,|d|) < x ⟺ |y|+|z| < x ✓. The cone = {x > |y|+|z|} ✓ = l1-cone ✓, whose facets are y±z = ±x planes: x−y−z=0 (h4′), x+y−z?? x = y−z wait facets of {x > |y|+|z|}: x = |y|+|z|: four planes: y+ z = x (i.e. x−y−z=0), y+… : x = y+z, x = y−z, x = −y+z, x = −y−z: planes x∓y∓z = 0: which as planes: x−y−z=0, x−y+z=0, x+y−z=0, x+y+z=0 ✓✓ THESE ARE THE PLATES (x−y−z=0 = −h4·… −x+y+z = 0 is plate 4; x−y−z=0 is its NEGATION: same plane! −x+y+z=0 ⟺ x−y−z=0 ✓ same plane!). Phew ✓. So cone facets = the four plates ✓ consistent.
So corner octa faces on plates: (q, M+y, M+z) computed plane (1,−1,−1): but plate planes: is (1,−1,−1) plane among plates? Plates: x+y+z=0, x+y−z=0 (=(1,1,−1)), x−y+z=0 ((1,−1,1)), −x+y+z=0 ((−1,1,1)). (1,−1,−1) NOT a plate plane. So face conv(0,M+y,M+z) is NOT on a plate: but region boundary should be plates only (+O): contradiction → my bipyramid face assignment wrong: bipyramid faces (q, M+y, M+z): but actually in cone {x>|y|+|z|}: the facet x = y+z−? the facet “x−y−z = 0” (x = y+z) within cone: {x = y+z, x ≥ other combos, ∩O}: = {x = y+z, x ≥ … i.e. y+z = x ≥ |y|+|z|: y,z ≥0? x = y+z ≥ |y|+|z| requires y,z same sign ≥ 0: facet = {x = y+z, y,z ≥ 0} = 2D cone spanned by (1,1,0),(1,0,1) ∩O: triangle conv(0? the facet ∩ O: {y,z ≥ 0, x = y+z, x+y+z ≤ 1}: triangle conv(0? at y=z=0: x=0: origin ✓; (1/2,1/2,0) ✓; (1/2,0,1/2) ✓): facet triangle = conv(0? wait but 0 = (0,0,0) = y=z=0,x=0 ∈ facet ✓): facet = conv(origin, M+y, M+z) — SAME as bipyramid face (q,M+y,M+z) with plane x−y−z = 0?? conv(0, (1/2,1/2,0), (1/2,0,1/2)): these points: do they satisfy x = y+z? (1/2,1/2,0): x = 1/2 = y+z = 1/2 ✓!; (1/2,0,1/2) ✓; origin ✓. But then the plane through them is x−y−z=0?? Points on plane x=y+z: (1,1,0) ✓ (1,0,1) ✓ 0 ✓: but normal computed (1,−1,−1) ✓ consistent: plane x = y+z = “x−y−z=0” ✓ IS a plate plane (equal to plate 4: −x+y+z=0 negated ✓). I confused myself: plate 4 IS x−y−z=0 as a set. ✓✓ All good: bipyramid faces are exactly on the 4 plate planes ✓.
So full verified picture:
Corner octahedra: 6, one per O-vertex ±e_i: at (1,0,0): conv((1,0,0), (0,0,0), M±y, M±z) = bipyramid = regular octahedron edge √2/2, vol 1/6. Faces: 4 on O boundary (triangles apex-(mids)), 4 on plates. Origin is a vertex of ALL 6 corner octahedra and ALL 8 tets: 14 cells meet at origin?? No — they share the origin POINT: all touch at origin. Hmm interesting: origin is where all 4 plates meet: the cell structure has all 14 cells’ closure containing origin.
Tetrahedra: 8: apex origin + base = medial triangle of each of 8 facets. Vol 1/24 each. So tets come in antipodal pairs sharing apex origin.
Total: 6/6 + 8/24 = 1 + 1/3 = 4/3 ✓ EXACT.
And the adjacency across plate 1 fans: the hexagon: P1∩O = regular hexagon with vertices the 6 mids of opposite-sign edges: for plate 1 (x+y+z=0): mids (±,…): (1/2,0,−1/2),(0,1/2,−1/2),(−1/2,1/2,0),(−1/2,0,1/2),(0,−1/2,1/2),(1/2,−1/2,0) ✓ these are exlicty the mids of edges connecting +axis to −axis. Fan triangles: T_i = conv(0, v_i, v_{i+1}).
Which cells meet across each T_i? T_i ⊂ plate 1. Cells adjacent: on each side. Take T = conv(0, a=(1/2,0,−1/2), bt=(0,1/2,−1/2)) (call v1=a, v2=b). Interior point of T: c = (1/6, 1/6, −1/3) (centroid). Perturb off plate along ±n, n = (1,1,1)/√3: p± = c ± εn. Determine sign patterns: h1(c) = 0, h1(p+) = ε√3·|… h1 = p·(1,1,1): c·n-hat… h1(c+εn) = 0 + 3ε/√3·? h1(c + ε n̂) = ε√3 > 0 ✓. h2(c+εn̂) = c·(1,1,−1)+ε(1+1−1)/√3 = (1/6+1/6+1/3) + ε/√3 = 2/3 + ε/√3 > 0 ✓. h3 = c·(1,−1,1) + ε(1−1+1)/√3 = (1/6−1/6−1/3) + ε/√3 = −1/3 + … < 0 ✓. h4 = c·(−1,1,1)+ε(−1+1+1)/√3 = (−1/6+1/6−1/3)+ε/√3 = −1/3+ε/√3 <0 ✓ (small ε). Pattern (+,+,−,−) on the +h1 side. On −h1 side: (−,+,−,−). Now classify: which region type is (+,+,−,−)? Types: corner cones = axis directions: (1,0,0)-cone was (+,+,+,−). The 6 corner patterns: cone around e_i: at (1,0,0): (+,+,+,−); (−1,0,0): (−,−,−,+); (0,1,0): h(0,1,0) = (1,1,−1,1): (+,+,−,+); (0,−1,0): (−,−,+,−); (0,0,1): (1,−1,1,1): (+,−,+,+); (0,0,−1): (−,+,−,−) ✓!! So (−,+,−,−) = corner region at (0,0,−1) ✓. And (+,+,−,−): tet type: which tet? Tet apex 0, base on facet (1,1,−1) [i.e. x+y−z=1]: the tet conv(0,(1/2,1/2,0),(0,1/2,−1/2),(1/2,0,−1/2)) — check centroid-ish (1/6+… its interior point: (1/4·… avg of 4 vertices: (1/4(0+1/2+0+1/2), 1/4(0+1/2+1/2+0), 1/4(0+0−1/2−1/2)) = (1/4, 1/4, −1/4): h = (1/4, 3/4, −1/4, −1/4): pattern (+,+,−,−) ✓✓.
So fan triangle T = conv(0, v1, v2) with v1 = mid of edge (+x,−z)… a = (1/2,0,−1/2) = mid of edge (1,0,0)–(0,0,−1); v2 = (0,1/2,−1/2) = mid of edge (0,1,0)–(0,0,−1). Across T: corner octa at (0,0,−1) on one side, tet of facet (1,1,−1) on the other ✓ (matches doc’s claim “each triangle is a face where one corner octahedron meets one tetrahedron” ✓).
Now ALTERNATION: T_i and T_{i+1}: v1 = mid(+x,−z), v2 = mid(+y,−z), v3 = mid(−x,+y)? hexagon cycle: (1/2,0,−1/2) → (0,1/2,−1/2) → (−1/2,1/2,0) → (−1/2,0,1/2) → (0,−1/2,1/2) → (1/2,−1/2,0) → back. Each v_i = mid of edge between axis (+ or −) i and next. Side T_i = conv(0, v_i, v_{i+1}): its two adjacent cells: corner at which vertex? v_i and v_{i+1} both mids of edges incident to the COMMON O-vertex: v1 edges (+x,−z) and (+y,−z) share vertex (0,0,−1) ✓ corner at (0,0,−1) on the “inner”… and the tet: base facet = the O-face containing the shared… T_1’s tet adjacenct: facet containing v1, v2 and a third mid: facet (x+y−z=1) contains mids (1/2,1/2,0)?? facet x+y−z=1 has vertices (1,0,0),(0,1,0),(0,0,−1): its edge mids: ((1,0,0)+(0,1,0))/2 = (1/2,1/2,0) ✓ on facet; ((1,0,0)+(0,0,−1))/2 = (1/2,0,−1/2) = v1 ✓; ((0,1,0)+(0,0,−1))/2 = (0,1/2,−1/2) = v2 ✓. Medial triangle of this facet = conv(v1′, v1?? = conv((1/2,1/2,0), v1, v2) — hmm but the tet base I derived for (+,+,−,−) was conv((1/2,1/2,0),(0,1/2,−1/2),(1/2,0,−1/2)) = conv(M, v1, v2) ✓ = medial triangle of facet (1,1,−1) ✓. So T_1 (fan triangle conv(0,v1,v2)) borders: CORNER octa at (0,0,−1) [the O-vertex shared by v1’s and v2’s edges] and TET of facet x+y−z=1 [the O-face containing both v1, v2 and mids].
General: T_i borders corner at u_i = O-vertex shared by the two O-edges whose mids are v_i, v_{i+1}; and tet of facet f_i = O-face containing v_i, v_{i+1} and one more mid. Now the alternation: as i cycles, does the “type” of what the fan carries alternate? The doc’s claim: the six triangles alternate A (corner side) / B (tet side) — but each triangle borders BOTH a corner (one side) and a tet (other side)! The “type” must refer to something else — maybe which cell is on which side, or the doc means the fan triangle’s two neighbors, and “2-coloring” assigns by, e.g., the type of the cell whose interface… Hmm. Re-read doc: “Each triangle is a face where one corner octahedron meets one tetrahedron. So the fan carries a 2-coloring: three triangles of type A (octahedron side) and three of type B (tetrahedron side).” This is a bit confused, but there is a genuine alternation: consider the boundary hexagon SIDES: side s_i = (v_i, v_{i+1}) lies on facet f_i. On the other side of s_i (outside plate, on the O-face): what touches: along hexagon side s_i, on the O boundary side, the adjacent cell across… cells touching O-faces: the medial triangle (tet base) is interior to facet; corner triangles (apex of corner octas) at facet corners. The hexagon side s_i ⊂ facet f_i: which cell touches s_i from the O side? On facet f_i, the cell structure: the 3 corner-triangles (from the 3 corner octas at the facet’s 3 vertices) + medial triangle (tet) — do these fill the facet? Facet area + corner triangles + medial: medial triangle area = 1/4 of facet; 3 corner triangles = conv(vertex, 2 adjacent mids) each area 1/4·…: each is 1/4? corner triangle = conv(v, m1, m2): area relative to facet: basis… = 1/4 of facet ✓; 3×1/4 + 1/4 = 1 ✓ facets fully tiled ✓. The hexagon side s_i (from v_i to v_{i+1}): in facet coordinates: s_i is one of the medial triangle’s EDGES ✓ (medial triangle edges = (v_i, v_{i+1}) pairs). So on the O side of s_i lies the medial triangle = base of tet f_i; and the plate side of T_i: the corner octa at u_i. Hmm so which is “type A vs B”? Alternatively define the alternation on hexagon SIDES: side s_i: on its O-side: tet (always!); on its interior side: T_i fan side: corner cell. Consistent, no alternation there. The alternation in the DOC is about the fan triangles’ types as you go around: A,B,A,B: but every fan triangle borders corner on one side and tet on the other. Unless the doc’s fan tiling is different: maybe their tiles are the hexagon’s 6 sectors alternating between “interface-to-corner-octa” and “interface-to-tet”? Hmm, from my analysis, EVERY sector interfaces corner-octa on the −h1 side… wait: T_1 borders corner (0,0,−1) on which side? Corner octa at (0,0,−1) occupies region pattern (−,+,−,−): at centroid, that’s the −h1 side?? h1(c) = 0: the corner cell at (0,0,−1) is on side h1< 0: check (0,0,−1)·(1,1,1) = −1 <0 ✓ so corner on −h1 side, tet (+,+,−,−) on +h1 side. By symmetry (rotate the configuration): every T_i borders corner on the h1<0 side and tet on h1>0 side? The hexagon of plate 1: corners at (0,0,−1) adjacent: rotating around: T_2 = conv(0, v2, v3): v2 = mid(+y,−z), v3 = mid(−x,+y)?? wait v3 = (−1/2,1/2,0) = mid of edge (−1,0,0)–(0,1,0): hmm v2’s edge (+y,−z) and v3’s edge (−x,+y): share vertex (0,1,0) ✓: T_2 borders corner at (0,1,0) and tet of facet containing v2, v3: facet (−x+y+z… v2 = (0,1/2,−1/2), v3 = (−1/2,1/2,0): facet containing both: x−… plane: try facet x−y+z = −1?? facet (−1,1,−1): −x+y−z = 1: v2: 0+1/2+1/2 = 1 ✓; v3: 1/2+1/2 = 1 ✓: facet normal (−1,1,−1) ✓. Third mid of that facet: facet vertices (−1,0,0),(0,1,0),(0,0,−1): mids: (−1/2,1/2,0) = v3 ✓, (0,1/2,−1/2) = v2 ✓, (−1/2,0,−1/2) ✓. So T_2 borders corner at (0,1,0) and tet of facet (−1,1,−1).
So all six fan triangles are structurally IDENTICAL (each between a corner and a tet): there’s no A/B alternation among the six triangles! Where’s the alternation the doc claims?? Maybe alternation refers to something around a single vertex: at each hexagon VERTEX v_i (an O-edge midpoint), the fan edges (spokes 0–v_i) meet: the spokes are plate∩plate lines: the spoke 0–v1 lies in plates h1=0 ∩ which other? v1 = (1/2,0,−1/2): h1 = 0 ✓; h2 = 1/2+0+1/2 = 1; h3 = 1/2−0−1/2 = 0 ✓! So spoke 0–v1 lies on plates 1 AND 3. Each spoke = intersection line of 2 plates ✓. Around v_i, 4 sectors meet in the plate-1 hexagon tiling: T_{i−1}, T_i on plate 1, plus corresponding triangles on plate 3 — 3D-wise: along spoke 0–v1, 4 cells meet: corner octa at (1,0,0)?? The spoke: on one side (in plate1-fan) T_1 and T_6; each spoke borders 4 regions (two pairs): T_1’s corner = (0,0,−1); T_6 = conv(0, v6, v1): v6 = (1/2,−1/2,0) (mid edge (+x,−y)), v1 = (1/2,0,−1/2) (mid edge (+x,−z)): share vertex (1,0,0): T_6 borders corner at (1,0,0). So around spoke 0–v1: corners (0,0,−1) [T_1 side] and (1,0,0) [T_6 side], and tets: T_1’s tet facet (1,1,−1); T_6’s tet: facet containing v6, v1: facet x−y−z=1? v6: 1/2+1/2 = 1 ✓ (x−y−z = 1/2+1/2−0 = 1) facet (1,−1,−1) ✓. So each fan triangle T: corner cell at one O-vertex, tet at one O-facet — all six identical in structure. NO ALTERNATION of the type claimed.
Hmm, so Lemma 1.1 as stated (“six triangles alternate A,B,A,B”) seems WRONG under my analysis — unless “type” means something else. Wait maybe the doc’s cell inventory differs: maybe their tiling/tessellation isn’t the plate arrangement I analyzed… “Cell inventory per notation: 6 corner octahedra, 8 tetrahedra, 4 plates (interfaces), 1 centerpoint.” Matches mine ✓ (6 + 8 + 4 + 1 = 19 objects; cells 14 ✓). And “Each triangle is a face where one corner octahedron meets one tetrahedron” ✓ matches! Then “the fan carries a 2-coloring: three triangles of type A and three of type B” — contradicts their own preceding sentence (every triangle is corner–tet interface). Their alternation claim internally inconsistent? Possible resolution: maybe the alternation is meant over the hexagon VERTICES: at successive vertices, the corner cell alternates between… the corner at T_i = u_i cycles through O-vertices: u_1 = (0,0,−1), u_2 = (0,1,0), u_3 = (0,0,… u_3: v3 = (−1/2,1/2,0) [edge (−x,+y)], v4 = (−1/2,0,1/2) [edge (−x,+z)]: shared vertex (−1,0,0): u_3 = (−1,0,0); u_4: v4 = (−x,+z), v5 = (0,−1/2,1/2) [edge (−y,+z)]: shared (0,0,1): u_4 = (0,0,1); u_5: v5 = (−y,+z), v6 = (1/2,−1/2,0) [(+x,−y)]: shared (0,−1,0): u_5 = (0,−1,0); u_6: v6 [(+x,−y)], v1 [(+x,−z)]: shared (1,0,0): u_6 = (1,0,0). So corners: (0,0,−1), (0,1,0), (−1,0,0), (0,0,1), (0,−1,0), (1,0,0): cycles through ALL SIX vertices with signs: s = (−1, +1, −1, +1, −1, +1): ALTERNATING in x-axis sign! The corner u_i alternates between the “positive-oriented” and “negative” half: specifically u_i’s x-coordinate: 0,0,−1,0,0,1 no: sequence x-coords: −1? u1=(0,0,−1): x=0; u2 = (0,1,0): 0; u3 = (−1,0,0): −1; u4 = (0,0,1): 0; u5 = (0,−1,0): 0; u6 = (1,0,0): 1. Hmm x-coordinate alternates (0,0,−1,0,0,1): only every third is ±1 on x: pattern of axis: u_i runs through: z−, y+, x−, z+, y−, x+: the axis of u_i cycles (z, y, x, z, y, x) and SIGNS alternate (−,+,−,+,−,+)! Signs of u_i: (−,+,−,+,−,+) alternating ✓! So the fan triangles alternate by the SIGN of their corner-vertex axis: T_i alternates between corners on the “negative” tetrahedral orientation and “positive” orientation {(1,0,0),(0,1,0),(0,0,1)} vs negative set. The 6 corners of O split into two inscribed tetrahedra: T+ = {(1,0,0),(0,1,0),(0,0,1),… no wait regular tetrahedron inscribed in octahedron: vertices alternate: {(1,0,0),(0,1,0),(0,0,1)} pair with… the 6 vertices split as {(1,0,0),(0,1,0),(0,0,−1)… hmm {u1..u6} = {(0,0,−1),(0,1,0),(−1,0,0),(0,0,1),(0,−1,0),(1,0,0)}: odd-indexed: {(0,0,−1),(−1,0,0),(0,−1,0)} = all negative: tets’ vertex sets: {(−1,0,0),(0,−1,0),(0,0,−1)} and {(1,0,0),(0,1,0),(0,0,1)}: two inscribed… those three points each: are they mutually adjacent O-vertices? (−1,0,0)-(0,−1,0): not an edge (parallel axes); the triangle face among {−x,−y,−z}: they’re pairwise non-adjacent: they form an equilateral triangle = the facet (−1,−1,−1)! Face of O: facet with vertices (−1,0,0),(0,−1,0),(0,0,−1) ✓ it’s the facet opposite (1,1,1). So T_odd’s corners lie on the SAME facet. Interesting but the doc’s alternation might refer to: consecutive fan triangles have corner-cells whose apex-signs alternate w.r.t. the dual tetrahedron. This is a real alternation! And the “2-coloring”/D₃: coloring T_i by sign of u_i (positive tetrahedral vs negative tetrahedral): alternates A,B,… ✓ 3 and 3 ✓ then 60° rotation maps u_i to u_{i+1}: flips color ✗ not symmetry; 120°: preserves: D₃ ✓. So Lemma 1.1 is REPAisable: the correct statement: consecutive fan triangles have corner cells at alternating dual-tetrahedron classes (corresponding to the two inscribed complementary tetrahedra — the stella octangula structure!). Actually BEAUTIFUL: the 6 corner octahedra come in two orbits of 3 = matched to the two tetrahedra of the stella octangula (compound of two tetrahedra in a cube/octahedron). And doc’s Corollary 1.4: “6 corner octahedra fall into two orbits of 3” ✓ consistent!
And Lemma 1.2’s reflection claim: coloring T_i by u_i sign: reflection through an axis… axis through opposite vertices v?,v?: reflection through the axis through center and vertex v4-direction etc. maps u_i ↔ u_{?}: reflection fixing v_1’s spoke? A reflection of D₆ fixing hexagon vertices v_2 & v_5 (opposite pair, axes through opposite VERTICES of hexagon): maps v_1↔v_3, v_6↔v_4: T_1 ↔ T_2?? T_1 = (v1,v2) → (v3,v2) = T_2: u: u_1 (sign −) ↔ u_2 (sign +): color flips ✗ NOT preserving. Reflection through axes through opposite EDGE-midpoints of hexagon: axis through center of side s_i: maps T_i ↔ … reflection through axis perpendicular… maps side s_1 to itself?? Axis through midpoints of opposite sides s_1 and s_4: fixes s_1 setwise, maps T_1 → T_1 (setwise), u_1 fixed? reflection fixing side (v1,v2) and opposite side (v4,v5): maps v3↔v6, v4… wait hexagon vertices v1..v6, axis through mid of s1=(v1,v2) and mid of s4 = (v4,v5): reflection maps v1↔v2, v6↔v3, v5↔v4: T_1 (v1,v2) → (v2,v1) = T_1 ✓ setwise; u_1 = shared vertex of v1’s & v2’s O-edges: u_1 = (0,0,−1): does reflection preserve O? Reflection through plane containing axis: must be a symmetry of the 3D config: the plate plane… reflections of the 2-colored hexagon that extend: u_1 maps to? The reflection swaps v1,v2: their shared O-vertex u_1 = (0,0,−1): reflection of space: the hexagon symmetry induced from… hmm as pure 2D hexagon-with-colors: colors: T_i color = sign of u_i: T_1 color B(−), T_2 color A(+), T_3 B? u_3 = (−1,0,0): negative: B, T_4: u_4 = (0,0,1): A, T_5: B, T_6: A ✓ alternating ✓. Reflection through side-midpoint axis: fixes T_1 (B) setwise ✓, maps T_2 ↔ T_6 (A↔A ✓), T_3 ↔ T_5 (B↔B ✓), T_4 ↔ T_4 (A ✓): PRESERVES coloring ✓. Reflection through vertex axis (through v_1 and v_4): maps T_1 ↔ T_6 (B↔A ✗). So D₃ = generated by 120° rotation + side-axis reflections: the DOC says “reflections through axes through opposite vertices preserve, edge-midpoints do not” — exactly BACKWARDS (it’s the side/edge-midpoint axes that preserve). As I suspected: flag it. D₃ conclusion stands ✓.
Also doc’s Lemma 1.1 alternation proof is vague but there’s a clean proof via the sign/orbit structure: the corner u_{i} sequence alternates dual-tetrahedron membership; cleaner: u_i = the O-vertex not on the two O-faces adjacent… anyway. In the re-derivation page I can give the clean proof: The fan triangle T_i has its boundary side s_i = (v_i, v_{i+1}) lying on facet f_i; the corner cell adjacent to T_i is at u_i = the unique O-vertex of facet f_i’s… u_i = facet-vertex shared by the O-edges containing v_i, v_{i+1}; and one shows u_{i+1} is the vertex OPPOSITE-ish: u_{i+1} ≠ u_i, and specifically the axis sequence alternates sign classes. Simple proof: v_i is the midpoint of the O-edge e_i; consecutive O-edges e_i, e_{i+1} share endpoint u_i; the six edges e_i form a 6-cycle in O’s edge graph (a “Petrie polygon”!): the octahedron’s edge graph with a 6-cycle alternating… the cycle visits vertices u_i: each O-vertex has degree 4; the cycle e_i: v-vertices: each e_i = (u_{i−1}, u_i): 6 edges forming closed walk u_1→u_2→…→u_6→u_1?? hexagon vertices v_i ↔ edges e_i of O; consecutive: e_i ∩ e_{i+1} = u_i. The walk u_6 e_1 u_1 e_2 u_2 … : closed walk of length 6 on O’s graph: O’s graph = K_{2,2,2} (octahedron graph = complete tripartite): 6-cycles in K_{2,2,2} that are “Petrie polygons” alternate between parts: partite classes = the antipodal pairs {(1,0,0),(−1,0,0)}, {(0,±1,0)}, {(0,0,±1)}: a 6-cycle in K_{2,2,2} visiting 6 distinct vertices = passes each partite class twice: visits: our u sequence: (0,0,−1), (0,1,0), (−1,0,0), (0,0,1), (0,−1,0), (1,0,0): parts: z, y, x, z, y, x ✓ each part twice: and the SIGN alternates ✓. In K_{2,2,2}, the 6-cycle u_1…u_6 with parts zyzxyx: consecutive parts differ ✓ (must, in tripartite): the two visits to each part: u_1 (z−) & u_4 (z+): antipodal! ✓ Similarly u_2, u_5 antipodal: (0,1,0),(0,−1,0) ✓; u_3, u_6 antipodal ✓. So the 6-cycle visits three antipodal pairs in alternating fashion — and the sign alternation: since consecutive u’s are adjacent (edge in O = different parts, but sign can be anything) — the alternation (−,+,−,+,−,+) isn’t forced by graph alone: alternative 6-cycle: (0,0,−1),(0,1,0),(1,0,0),(0,0,1),(0,−1,0),(−1,0,0): signs (−,+,+,+,−,−): also a valid hexagon! Does it arise as a plate∩O? Plate normals: the plate x+y−z = … hmm: the hexagon of plate with normal n lies on plane n·x = 0: hexagon vertices = mids of edges (a,b) with n·a + n·b = 0·… mid m = (a+b)/2: n·m = (n·a + n·b)/2 = 0: edges where endpoints have opposite n-values: n_i = (1,1,1)-type normals: for normal (1,1,1): edges (u,v) with u+v ⊥ n: (a,b) axes-distinct, n·a = ±1, n·b = ±1: need opposite: (a,b) with n·a = 1, n·b = −1: a ∈ {(1,0,0),(0,1,0),(0,0,1)}, b ∈ negatives: all 9 combos but only edges (distinct coords): 6 ✓ giving our hexagon. For normal (1,1,−1) (plate 2): edges with n·a = +1: {(1,0,0),(0,1,0),(0,0,−1)}, n·b = −1: {(−1,0,0),(0,−1,0),(0,0,1)}: hexagon mids: connect: cyclic order: hexagon = the 6 mids arranged: the resulting 6-cycle: mid of ((1,0,0),(0,0,1)) = (1/2,0,1/2); ((1,0,0),(0,−1,0)) → (1/2,−1/2,0); ((0,1,0),(0,0,1)) → (0,1/2,1/2); ((0,1,0),(−1,0,0)) → (−1/2,1/2,0); ((0,0,−1),(0,−1,0)) → (0,−1/2,−1/2); ((0,0,−1),(−1,0,0)) → (−1/2,0,−1/2). The u-cycle: shared vertices around: hmm. The two possible 6-cycle types in K_{2,2,2} — one gives sign-alternation, giving the hexagon of plates 1–4; the other type corresponds to… plate normals of the OTHER tetrahedral set? But our plates are exactly the four “at most one minus” normals ±? plates: normals {(1,1,1),(1,1,−1),(1,−1,1),(−1,1,1)}: plate normal set = one per antipodal pair from the eight facet normals: we pick 4 of the 8: exactly those with product… normals with an even number of minuses: (1,1,1): 0 minuses; (1,1,−1): 1 minus! Hmm parity: (1,1,1) product 1; (1,1,−1) product −1; (1,−1,1) product −1; (−1,1,1): product −1: so not “even”. The four chosen = {n: at most one −1} = the normals of ONE inscribed tetrahedron’s facets?? The dual tetrahedron: facet normals (1,1,1),(1,1,−1),(1,−1,1),(−1,1,1) are the VERTICES of a regular tetrahedron (pairwise dot = 1−1−1 = −1·… n_i·n_j for i≠j: e.g. (1,1,1)·(1,1,−1) = 1: hmm (1,1,1)·(1,1,−1) = 1+1−1 = 1 ≠ −1: not tetra vertices. (1,1,1)·(1,−1,1)… = 1. Hmm pairwise dots = 1: all four normals mutually at 60°·? angle cos = 1/3: these four directions are the vertices of a regular tetrahedron ✓ (four vectors with pairwise 1/3: yes tetrahedral!). OK.
I’ve now got a fully verified geometric picture — this is genuinely a nice “hidden complexity” as Bruce says. For the re-derivation page, I should present:
- Setup & notation
- The octahedron: R, a, V (a = √2, V = 4/3), second moment 2/5 with clean derivation (density f(x) = (3/2)(1−|x|)², E[X²] = 1/10, ⟨r²⟩ = 3/10, ∫r²dV = 2/5).
- The four plates: normals = tetrahedral set; each plate cuts O in the regular hexagon through 6 edge midpoints; every one of O’s 12 edge midpoints lies on exactly 2 plates (24 = 12×2) — the incidence duality.
- Cell decomposition: sign-pattern count: 14 regions (with the linear-dependence h4 = h1−h2−h3 forcing 2 patterns): 6 corner cells + 8 tets:
- Corner cell at e_i: {x > |y|+|z|} ∩ O-type = regular bipyramid = regular octahedron, scale factor 1/2, edge √2/2, volume 1/6: apexes at the O-vertex AND the center! (The centerpoint is a vertex of all six corner octahedra.)
- Tetrahedra: apex at center, base = medial triangle of each of 8 facets; volume 1/24.
- Volume check: 6(1/6) + 8(1/24) = 4/3 ✓ exact.
- The fan: hexagon, 6 sectors, each sector = interface across plate between one corner octa and one tet. The corner-cell sequence around the fan visits all 6 O-vertices; consecutive triangles’ corners are adjacent O-vertices; the sequence runs through the three antipodal pairs and, for this plate family, alternates between the two dual tetrahedra (the stella octangula classes) — this is the corrected Lemma 1.1: alternation of dual-tetrahedral class, with clean proof.
- Corrected Lemma 1.2: D₃ stabilizer with the reflection classification fixed (side-axis reflections preserve; vertex-axis reflections swap colors).
- Orbits: corners 3+3 (matching the two tetrahedra of the Stella Octangula); tets: hmm under D₃: 8 tets → orbits? 4+4: the 8 facet-tets split: facets with normal in the plate-normal tetrahedral class {(1,1,1),(1,1,−1),(1,−1,1),(−1,1,1)}: 4 facets = one dual tetra class; other 4: orbits of 4+4 ✓.
- What’s solid (geometry), what needs care (2/5 numerology, 2φ/5): note re the hypothesis page.
Also the volume data: corner octa 1/6 = 12.5% each; total corners = 3/4 of O volume! 6×1/6 = 1 = 75% of 4/3; tets 1/3 = 25%. Fun ratios: 3:1. And 2/5 vs these… The second moment integral 2/5: contributions per cell could be computed (corner octa of scale 1/2: ∫r² = (1/2)⁵·(2/5) = 1/16·2/5 = 1/40 if centered — but corner octas are off-center (center at (1/2,0,0)): ∫r²dV over cell = ∫((x’)² …) = V·(|c|² + ⟨r’²⟩) = (1/6)(1/4 + (3/10)(1/4)) = (1/6)(1/4 + 3/40) = (1/6)(13/40) = 13/240. 6 × 13/240 = 78/240 = 13/40. Tets: apex at 0, base at distance…: tet conv(0, mids): all vertices at |·| √2/2·… mids at radius √2/2 ≈ 0.707: centroid at 3/4·… ∫r² over tet with apex 0: = V_tet·(3/5|base centroid|²)? For a tetrahedron with apex at origin and base triangle: ∫r² dV = (V/20)(sum of |vertices|² terms…) formula: for simplex with vertices 0, p1,p2,p3: ∫|x|²dV = V·( |p1|²+|p2|²+|p3|²+p1·p2+p1·p3+p2·p3 )/10·? The second moment of a simplex: ∫|x|² = V/20·(Σ_{i<j}|pi−pj|² + …) hmm: known: average of |x|² over simplex with vertices v0..v3 = (1/20)Σ_{i,j}|v_i−v_j|²·? Actually ⟨|x|²⟩ over simplex = (1/20)Σ_{i,j}|v_i − v_j|²? For simplex with vertices v_0..v_n: E|x|² = (1/((n+1)(n+2))) Σ_i |v_i|² + (1/((n+1)(n+2))) Σ_{i≠j} v_i·v_j: = (1/20)[Σ|v_i|² + Σ_{i≠j}v_i·v_j] in 3D (n+1=4, (n+1)(n+2) = 20) ✓ with v0 = 0: Σ|v_i|² over p1,p2,p3 = 3·(1/2)·… |mid|² = 1/2 (each mid has |·|² = 1/4+1/4 = 1/2 ✓): 3·1/2 = 3/2; Σ_{i≠j} p_i·p_j: p_i·p_j: (1/2,1/2,0)·(0,1/2,1/2) = 1/4; three pairs: 1/4 each: 6 ordered: 6/4 = 3/2. Total inside: (3/2 + 3/2)/20 = 3/20. ⟨r²⟩ = 3/20. ∫r² = V(1/24)(3/20) = 3/480 = 1/160. 8 × 1/160 = 1/20. Total: 13/40 + 1/20 = 13/40 + 2/40 = 15/40 = 3/8 ≠ 2/5! Hmm!! 3/8 = 0.375 vs 2/5 = 0.4: mismatch! Let me recheck corner octa computation: corner octa = bipyramid centered at c = (1/2,0,0), = 1/2-scale octahedron? Its vertices relative to c: (1/2,0,0), (−1/2,0,0), (0,±1/2,0),(0,0,±1/2) ✓ EXACTLY 1/2-scale regular octahedron centered at c ✓. ∫|x|² = ∫|x’+c|² = ∫|x’|² + 2c·⟨x’⟩ + |c|²V = (2/5)(1/2)⁵ + 0 + (1/4)(1/6) = (2/5)(1/32) + 1/24 = 1/80 + 1/24. LCD 240: 3/240 + 10/240 = 13/240 ✓ (earlier same). 6 cells: 78/240 = 13/40 = 0.325. Tets: apex 0, base mids: computed ⟨r²⟩ = 3/20: check ⟨r²⟩ ≤ max|vertex|² = 1/2 ✓ 0.15 ✓ plausible. ∫ = (1/24)(3/20) = 1/160 ✓ 8 → 1/20 = 0.05. Sum 0.375 ≠ 0.4. MISMATCH!! So something inconsistent: total must be 2/5 = 0.4. Missing 1/40 = 0.025. Hmm!
Let me recheck ⟨r²⟩ = 3/10 for the octahedron itself: E[X²] = 3∫₀¹x²(1−x)²dx = 3/30 = 1/10; ⟨r²⟩ = 3/10 ✓; ∫ = (4/3)(3/10) = 2/5 ✓. Check tets + corners cover O ✓ volumes matched exactly 4/3 ✓. So decomposition right, second moments don’t add: one of my cell second-moments wrong. Tets: v0 = 0, p1 = (0,1/2,1/2)?? wait for facet (1,1,1) tet: base = mids of ITS edges: facet (1,1,1) vertices (1,0,0),(0,1,0),(0,0,1): mids (1/2,1/2,0),(1/2,0,1/2),(0,1/2,1/2) ✓ each |p|² = 1/2 ✓. Σ_{i<j} p_i·p_j = 1/4·… p1·p2 = 1/4, p1·p3 = 1/4, p2·p3 = 1/4 → sum 3/4. Formula: E|x|² = (1/20)[Σ|v_i|² + Σ_{i<j} 2 p_i·p_j]? Let me derive: E|x|² = E|Σ… for simplex barycentric: x = Σλ_i v_i, λ uniform on simplex: E λ_i = 1/4; E λ_i λ_j (i≠j) = 1/20; E λ_i² = 3/20·? Var: E λ_i² = 1/16 + Var: Var λ_i for Dirichlet(1,1,1,1) = α(α0−α)/(α0²(α0+1)) = 1·3/(16·5) = 3/80: E λ_i² = 1/16+3/80 = 5/80+3/80 = 8/80 = 1/10. And E λ_i λ_j = 1/α0(α0+1) = 1/20 ✓. E|x|² = Σ_i Eλ_i²|v_i|² + 2Σ_{i<j}Eλ_iλ_j v_i·v_j = (1/10)Σ_i|v_i|² + (1/10)Σ_{i<j}v_i·v_j·… 2·(1/20) = 1/10: E|x|² = (1/10)[Σ_{i=0}^{3}|v_i|² + Σ_{i<j} v_i·v_j]?? careful: 2Σ_{i<j} E λλ v_i·v_j = 2(1/20)Σ_{i<j}v_i·v_j = (1/10)Σ_{i<j}v_i·v_j ✓. With v0 = 0: Σ|v_i|² = 3/2; Σ_{i<j} v_i·v_j (i<j over all 4 vertices incl v0=0): = 3/4. E|x|² = (1/10)(3/2 + 3/4) = (1/10)(9/4) = 9/40. Hmm ≠ 3/20 = 6/40. Earlier formula wrong; E|x|² = 9/40 = 0.225 ≤ 1/2 ✓. ∫ = (1/24)(9/40) = 9/960 = 3/320. 8 tets: 24/320 = 3/40 = 0.075. Corners 13/40 = 0.325. Total = 16/40 = 2/5 ✓✓✓ PERFECT.
So the decomposition + second moment is fully consistent: corner octas contribute 13/40, tets 3/40. Lovely — the page can include this decomposition check as a genuine verification (both routes give 2/5).
This is great material for the re-derivation page. Also the “centerpoint”: all 14 cells have the origin as a common vertex — the “1 centerpoint” inventory item ✓.
Also nice: corner octa = stella octangula connection: the 8 tets: apex at center: their union = star; hmm.
Also doc’s “8 tetrahedra … two orbits of 4 under D₆”: under full octahedral symmetry the 8 tets form one orbit of 8? Facets form one orbit of 8 under full O_h: tets one orbit of 8 ✓ (doc says two orbits of 4 under D₆ — under the plate-stabilizer subgroup: the 4 plates’ normals form a tetrahedral class: facets with normals = plate class (4) vs antipodal class (4): two orbits of 4 under the tetrahedral-ish subgroup ✓ plausibly right).
Corner octahedra under D₃ (stabilizer of plate 1’s coloring): corners in orbits: u-sequence colors alternate: the coloring T_i: under D₃ (120° rotation): u_i → u_{i+2}: orbit of u_1 = {u_1, u_3, u_5} = {(0,0,−1),(−1,0,0),(0,−1,0)} = negative class {3 vertices of one dual tetra…} and {u_2, u_4, u_6} = positive class ✓ two orbits of 3 ✓ matches Cor 1.4 ✓.
Now also verify Cor 1.4’s claim “tetrahedra fall into two orbits of 4 under D₃: tet of facet f_i ↔ fan triangle T_i? The tets: 8 facets: those adjacent to plate-1 hexagon: 6 of the 8 facets contain a hexagon side; the other 2 facets (with normals ±(1,−1,−1)? the facets NOT touched by plate 1: plate 1 = x+y+z=0: facets x+y−z=1 touched ✓ all except ±(1,−1,−1),(−1,1,−1)… facet normal (1,−1,−1): plane x−y−z=1: does plate 1 touch it? Plate 1 ∩ facet: segment? h1 = x+y+z on facet: facet vertices (1,0,0): h1 = 1; (0,−1,0): −1; (0,0,−1): −1: plate crosses facet interior → hmm but hexagon hexagon sides each lie on a distinct facet: 6 sides → 6 facets touched ✓; but plate 1 also crosses the other 2 facets’ interiors?? Plate 1∩O = hexagon: its boundary consists of 6 segments each on a facet — the hexagon boundary ∩ facet = segment: if the plate crosses facet (1,−1,−1) then the hexagon has an edge on it. 8 facets: hexagon edges: 6 → only 6 facets carry hexagon edges ✓, so plate 1 must NOT cross the other two: facet (1,−1,−1): all points: h1 = x+y+z with x = 1+y+z: h1 = 1+2(y+z): zero when y+z = −1/2: on facet: y+z ≥ … facet = {x−y−z=1, |·|≤1}: y+z = −1/2, x = 1/2: point (1/2, t, −1/2−t): |x|+|y|+|z| = 1/2 + |t| + |1/2+t| ≤ 1 → |t| + |t+1/2| ≤ 1/2 → t ∈ [−1/2, 0]: so plate 1 DOES cross facet (1,−1,−1)!? along segment from (1/2,−1/2,0) to (1/2,0,−1/2) — but those are HEXAGON VERTICES v6 and v1! So the plate touches facet (1,−1,−1) only along the segment (v6,v1) = hexagon SIDE s_6! So hexagon side s_6 lies on facet (1,−1,−1)! Contradiction with earlier assignment where s_6 ⊂ facet? s_6 = (v6,v1): v6 = m(+x,−y), v1 = m(+x,−z): I earlier said T_6’s tet facet = x−y−z=1 = normal (1,−1,−1) ✓ consistent: sides s_i lie on facet f_i where f_6 = (1,−1,−1) ✓. And facet (1,−1,−1)’s corner triangles at its 3 vertices (1,0,0),(0,−1,0),(0,0,−1): the segment (v6,v1) separates corner-triangle at (1,0,0) from the rest? On facet (1,−1,−1): the plate segment from m1… the facet’s two medial-ish segments from the two plate crossings: each facet is crossed by 2 plates? Facet (1,−1,−1) crossed by plates 1 (segment (v6,v1)) and… another plate: plate x−… plate 2 (1,1,−1): h2 = x+y−z on facet: facet x−y−z = 1: x = 1+y+z: h2 = 1+y+z+y−z = 1+2y: zero at y = −1/2: segment from (1/2,−1/2,0)?? x = 1+y+z: y = −1/2: x = 1/2+z: |x|+|y|+|z| = 1/2+z+1/2+|z|: z ≥ 0: 1+2z ≤ 1 → z = 0: single point (1/2,−1/2,0) = v6! So plate 2 only touches facet at the vertex v6 (edge-midpoint on the facet’s edge). OK so each facet is crossed fully by exactly ONE plate (the one parallel to… plate normal n, facet normal m: plate crosses facet m·… iff… n·x varies across facet: yes if n ≠ ±m: 4 plates: facet (1,−1,−1): plates with n·m… n ∈ {(1,1,1),(1,1,−1),(1,−1,1),(−1,1,1)}: crossing facets: which plates have n with all-positive on… plate n crosses facet m (facet = {m·x = 1}) iff n takes both signs on the facet: n·(facet vertices): facet vertices of m: e_i with sign(m_i) = ±: n·v = m_i·? vertices of facet m: the three axis points e_j with… facet normal m = (m1,m2,m3): vertices e_i for i with m_i = +: and −e_j for m_j = −: n·vertex = n_i or −n_j: signs: n takes values {n_i (i: m_i=+)} ∪ {−n_j (j: m_j=−)}: crosses iff not all same sign: 3 values: mixes. Fine — the picture: each of the 8 facets is divided by one plate’s hexagon-side into the medial triangle + 3 corner triangles ✓ consistent: 8 side-facet assignments: each plate contributes 2 hexagon sides to 2 facets?? 4 plates × 6 sides = 24 sides; 8 facets each crossed by exactly…: 24/8 = 3 plates fully cross each facet? Facet (1,−1,−1) crossed how many: plate 1 (n·facet values: h1 at vertices (1,0,0): 1; (0,−1,0): −1; (0,0,−1): −1: crosses ✓); plate 2: shown touches only at v6 edge-midpoint (not crossing: h2 = 1+2y = 1−1 = 0 only at boundary point): hmm shown segment degenerates to point: so plate 2 doesn’t cross facet (1,−1,−1)?? But 24 = 8 facets × 3 crossings? Let me recount: each hexagon side on a facet: 24 sides / 8 facets = 3: so 3 plates cross each facet… but I showed plate 2 ∩ facet = point. Recheck: plate 2 ∩ facet (1,−1,−1): h2 = 1+2y = 0 → y = −1/2, and facet: x = 1+y+z = 1/2 + z: constraint |x|+|y|+|z| ≤ 1: (1/2+z) + 1/2 + |z| ≤ 1 → z + |z| ≤ 0 → z ≤ 0: so z ≤ 0: segment from (1/2,−1/2,0) (z=0) to z→−∞?? z ≤ 0 with x = 1/2+z ≥ −1… z ≥ −1/2 (x ≥ 0?): x = 1/2+z: need |x| ≤ ~: z ∈ [−1/2, 0]: segment from (1/2,−1/2,0) = v6 to (0,−1/2,−1/2) = mid of edge (0,−1,0)–(0,0,−1)! So plate 2 DOES cross facet (1,−1,−1) along segment v6–(0,−1/2,−1/2) ✓ (I dropped the z<0 branch before). OK so 3 plates cross each facet ✓ 24 ✓ consistent: each facet cut by 3 plate-segments: segments connect edge mids… On facet (1,−1,−1) (vertices A=(1,0,0), B=(0,−1,0), C=(0,0,−1)): mid points of edges: AB: (1/2,−1/2,0) = v6, AC: (1/2,0,−1/2) = v1, BC: (0,−1/2,−1/2) = call v2′ (it’s a hexagon vertex of plate 2). Segments: plate 1: (v6, v1); plate 2: (v6, v2′); plate 3? (x−y+z=0: h3 = x−y+z: on facet: x = 1+y+z: h3 = 1+y+z−y+z = 1+2z = 0: z = −1/2: segment (v1, v2′) ✓). So the three segments form the medial triangle (v6, v1, v2′) ✓✓ dividing facet into 3 corner triangles + medial ✓. All consistent: every facet carries its medial triangle (tet base) and 3 corner-triangles ✓.
So the complete verified structure is solid and elegant. Note: The doc’s Lemma 1.1/1.2 need corrections as identified, but I should present the corrected derivation cleanly. Given the extensive analysis, time to write the page. It’s a math document: canvas text/markdown. Bruce said “re-derivation page” — deliver as a canvas. Load canvas skill first.
Also he wants my earlier statement acted on: “save the 2φ/5 claim for a separate hypothesis page” — I’ll include a short closing section redirecting 2φ/5 to a hypothesis page (and offer to draft it).
Let me load the canvas skill.
Good morning Bruce — happy to. Your instinct is right about hidden complexity: when I re-derived the cell inventory rigorously, the decomposition is even cleaner than the DeepSeek page states, and two of its lemmas need correction (one of them gets its reflection classification backwards). Let me set up the re-derivation page properly.
Write the canvas.
It’s a substantial math document. I’ll write it carefully with the verified results. Keep it well-structured, with proofs.
Sections:
- Setup
- The octahedron and second moment (clean derivation)
- The plate family: normals, hexagons, edge-midpoint duality (12 mids × 2 plates = 24 hexagon sides)
- Cell decomposition: sign analysis, 14 regions, corner octahedra (6, regular, edge √2/2, vol 1/6, vertices apex+center+4 mids), tetrahedra (8, apex at center, base = medial triangle of facet, vol 1/24), volume closure 4/3 exact.
- The fan: plate 1’s hexagon, 6 sectors, each sector interfaces corner octa ↔ tet. Corrected Lemma 1.1: alternation is of the dual-tetrahedral class of the corner cell (sign of apex), proof via K_{2,2,2} 6-cycle / antipodal pairing.
- Corrected Lemma 1.2: D₃, reflections: side-midpoint axes preserve, vertex axes swap (doc had it backwards).
- Orbits and the anchors: 3+3 corners, 4+4 tets; second-moment decomposition check: corners 13/40, tets 3/40, total 2/5 consistent (independent route).
- What this page settles vs what remains (2/5 numerology note; 2φ/5 → separate hypothesis page).
Mermaid? Not needed. Keep math plain-text notation with Unicode like the input. Write concise but rigorous. Use “centerpoint” for origin note: all 14 cells meet at centerpoint.
Careful claims:
- Corner octahedron at e_i: vertices e_i, 0, and the four mids of edges incident to e_i. Bipyramid = regular octahedron edge √2/2, scale 1/2. All edges √2/2 ✓ (verified).
- Each fan sector T_i: conv(C, v_i, v_{i+1}); borders corner octa at u_i (O-vertex shared by the two O-edges whose mids are v_i, v_{i+1}) on the h<0 side… state generally.
- Alternation: sequence u_1..u_6 visits all six O-vertices; odd-indexed set = {(0,0,−1),(−1,0,0),(0,−1,0)} (negative class), even = positive class: the two inscribed complementary tetrahedra (stella octangula). So A/B alternation is real but re-interpreted: it’s the signing of the corner cells, not which cell type adjoins (each sector adjoins one of each).
- D₃: 120° rotation preserves; 60° swaps classes; reflections through axes through opposite side-midpoints preserve; reflections through opposite-vertex axes swap. Doc had reflections backwards.
- Orbits under D₃: corners 3+3 ✓; tets 4+4 (plate-normal facet class vs antipodal facet class).
Volumes: 6×1/6 + 8×1/24 = 1 + 1/3 = 4/3 ✓.
Second moment decomposition: per corner octa (center at e_i/2, scale 1/2): ∫|x|² = (1/6)·(1/4) + 2/5·(1/2)^5 = 1/24 + 1/80 = 13/240; 6 → 13/40. Tets: E|x|² = (1/10)(Σ|v_i|² + Σ_{i<j}v_i·v_j) = (1/10)(3/2 + 3/4) = 9/40; ∫ = (1/24)(9/40) = 3/320; 8 → 24/320 = 3/40. Total 16/40 = 2/5 ✓.
Hmm wait: corners 13/40 + tets 3/40 = 16/40 = 2/5 ✓.
Tet formula: E|x|² = (1/10)[Σ_i |v_i|² + Σ_{i<j} v_i·v_j] over the 4 vertices (with v_0 = 0 contributing nothing): Σ|v_i|² = 3 × 1/2 = 3/2; Σ_{i<j} v_i·v_j = 3 × 1/4 = 3/4; total 9/4 /10 = 9/40 ✓. E|x|² = 9/40 = 0.225, sanity ≤ 1/2 ✓ and centroid |·| = (1/4)(√2/2·3… |centroid| = |(1/3·… centroid = (p1+p2+p3)/4 = ((1/2+1/2+0)/4, …) = (1/4,1/4,1/4): |c|² = 3/16 ≈ 0.1875 ✓ and ⟨|x−c|²⟩ = inradius²/… ≈ 9/40−3/16 = 0.0375 ✓ small ✓ good.
Also the striking volume ratio: corner octahedra occupy 3/4 of O, tets 1/4.
Also “the 5 is dimension”: keep brief.
Note on plates count: the four plate planes are the four facets’ mirrors… the plate normals {at most one minus} form a regular tetrahedron of directions; the plate family is self-dual under negation (plane n·x=0 = plane −n·x=0), so the “other” tetrahedral class {at least two minuses} describes the SAME four planes. Nice subtlety: the normal set {(1,1,1),(1,1,−1),(1,−1,1),(−1,1,1)} is closed under negation as a plane set since −(1,1,1) = (−1,−1,−1) = “all minus” which is the same plane as (1,1,1) ✓.
Write the canvas now. Title: “Hexarchon Re-Derivation: Fan Geometry and Cell Inventory”. Keep length moderate (~150 lines).
The verification is done — everything cross-checks exactly (volumes close to 4/3 on the nose, and the second moment decomposes consistently across the cells). Writing the re-derivation page now.
Created Hexarchon Re-Derivation: Fan Geometry and Cell Inventory
I notice one glitch: in Claim 3.2 I left an artifact “{x > |y| + |z|} ∩ O ? no — precisely:” — that’s an editing mistake I should fix. Also “complete” typo (“the full self-dual”). Fix both.
Fixing two typos I introduced.
The re-derivation page is up: “Hexarchon Re-Derivation: Fan Geometry and Cell Inventory.” It and the 19 Nov page agree on the headline numbers, but working it through coordinate by coordinate produced three real findings:
1. The cell inventory is stronger than stated. The six corner cells are not just “octahedra” — each is a regular, half-scale octahedron (edge √2/2, volume exactly 1/6) whose two apexes are the O-vertex and the centerpoint. The eight tetrahedra all have the centerpoint as apex and a facet’s medial triangle as base (volume exactly 1/24). The volumes close exactly: 6·(1/6) + 8·(1/24) = 4/3, with the corner octahedra occupying 3/4 of O and the tetrahedra 1/4. A bonus: the second moment decomposes as 13/40 (corners) + 3/40 (tetrahedra) = 2/5, so 2/5 is now confirmed by a second independent route.
2. Lemma 1.1 is true, but not for the reason given — and it’s better than stated. No fan sector can be “an octahedron triangle” or “a tetrahedron triangle”: every sector interfaces one corner octahedron on one side of the plate and one tetrahedron on the other. The genuine alternation is in where the corner cells sit: going around the fan, their apexes visit all six O-vertices and alternate between the two inscribed complementary tetrahedra (the stella octangula classes), signs −,+,−,+,−,+. That’s the corrected Lemma 1.1, with a clean proof via the octahedron’s edge graph K₂,₂,₂.
3. Lemma 1.2 has its reflections backwards. For the colored fan, it is the reflections through opposite side-midpoint axes that preserve the coloring, and the vertex-axis reflections that swap it. The D₃ conclusion stands, so Corollaries 1.3 and 1.4 survive with the corrected proof.
On your last point: the page deliberately quarantines the fragile material. The 4/(4+6) = 2/5 orbit-ratio stays [O] (an arithmetic coincidence of small counts, not a derived identity), and 2φ/5 is explicitly deferred to its own hypothesis page, which would need to name the measured quantity it predicts before a falsification condition can attach. Say the word and I’ll draft that hypothesis page next as a companion.
Historical notes: This page was published in public at 9 AM, Friday, 20 November 2026. Once DeepSeek confirms that he has read it, the page will become private. -BEC