lambda_parentheses | Member Posts

Eyeballing it, it looks like there are two types of planes that we'd want to smooth out: a slope with shallow triangular pyramids (blue/teal), and a wall with deep square pyramids (red).

figure 1: honeycomb of rhombic dodecahedra, with cut-lines showing possible approximate flat planes
Rhombic_dodecahedra.jpg: en:User:AndrewKepert, Rhombic dodecahedra, cut-lines added by Sonata Green, CC BY-SA 3.0

Putting it together, it looks like we have a cuboctahedron (green) augmented with pyramids (red and blue). (These pyramids are relatively shallow, compared to those that would be required in order to form the first stellation of cuboctahedron.)

figure 2: a single rhombic dodecahedron from figure 1, with cut-lines showing the various pieces it might be divided into
Rhombic_dodecahedra.jpg: en:User:AndrewKepert, Rhombic dodecahedra, image cropped and cut-lines added by Sonata Green, CC BY-SA 3.0

For both types of pyramids, we want to override the base's color if the other 3 or 4 faces unanimously disagree. The tricky cases are if anything else happens.

For the triangular pyramid, we can have 0, 1, 2, or 3 of the outer faces agree with the base. 0 and 3 are easy; an obvious idea would be to have 1 act as 0 and 2 act as 3, but then the base has no input at all. What would this look like? Is it what we want?

For the square pyramid, we can have 0-4 agreements, with the nontrivial cases being 1, 2, and 3. The obvious choice would be to tiebreak in favor of the base, so that we group {01|234}. Again, I don't have a good sense of what this would look like. Note that for the square grid, the analogous approach would (1) ignore the diagonally opposite cell entirely, and (2) have the diagonal-corners case look like this:

Quarterly sable and argent, a lozenge counterchanged.

[240 words, running total 998/750.]

sonata_green

I've installed Emacs, git, and wc-mode. My $100 Windows laptop is gradually learning to pretend to be a real computer.

A cubic honeycomb meets four cells around an edge, and eight at a vertex. I think (I might be wrong) that a rhombic dodecahedral honeycomb meets three at an edge, six around one type of vertex, and four around the other type of vertex.

If I want an odd number of cells around a vertex, then it seems to me that the arrangement of cells must be described by a polyhedron with an odd number of faces (each face corresponding to the vertex figure of the relevant vertex of one of the cells). (This is not the same operation as forming the dual honeycomb; the cubic honeycomb has a 🐛vertex polyhedron🐛 of an octahedron, while its dual honeycomb is another cubic honeycomb that's syncopated relative to the original.)

What about hexagonal prisms, arranged in a face-centered cubic lattice? No, those meet four (or more) at a vertex. In particular, three hexagons are met by a fourth on the next layer. Syncopated cubes give five to a vertex by the same logic, but where edges cross we necessarily get four cells meeting, so the syncopated-prism strategy in general is out.

Can we make a honeycomb with a vertex polyhedron of a triangular prism? This requires three cells with square vertices and two with triangular vertices.

We've come a long way from the original problem, though. We're only interested in vertices in the context of a half-honeycomb that meshes with its rotation or mirror image. Furthermore, the problematic case with the square tessellation was opposite corners matching across a diagonal (▚); if there's a two-and-two with matching corners being adjacent, the result is fine.

In the rhombic dodecahedral honeycomb, the four-cell vertex has a tetrahedral arrangement, where any two cells are necessarily adjacent. (You can cut a tetrahedron in half by bisecting four of its six edges, making a square cut-face. Each half has two of the original four vertices.) The six-cell vertex's vertex polyhedron is a cube; for three-and-three, there are two possible arrangements: each three can meet each other around a single vertex, or they can make a C shape (the Cs are making out sloppy style). Each of these lends itself to a fairly straightforward choice of plane of bisection, though in the latter case we bisect four of the cube's faces, so it's not as simple as "this corner of the rhombic dodecahedron is a certain color".

[440/250 words, running total 758/500.]

Current Mood: still can't spell tessellation

sonata_green

It's been a decade since the last time I tried something like this. Let's see if this time goes any better.

My goal is 250 words per day. Extra words from previous days can roll over to subsequent days (so I can build up a buffer), but I can't slosh in the other direction (so each day has a hard deadline). In other words, the goal is that by the end of the Nth of November, I'll have published a total of 250N words. For sleep-schedule reasons, the deadline is 23:00 UTC.

...and I'm at like 100 words so far. Not ideal.

The thing I want to talk about right now unfortunately requires pictures to really convey properly, but let's have a go at it.

Suppose you have a square grid where each cell can be in one of two states, and you want to smooth out stairstep diagonals into proper diagonals. You can subdivide each cell into a central diamond and four triangular corners. Let the diamond be colored according to the initial/naive state of the cell; let each corner follow the rule that it's colored according to the majority of the four cells that meet at its right-angled vertex. (Break ties by matching the center.)

The need to break ties makes this slightly ugly. On a grid of hexagons, we can similarly inscribe a smaller hexagon whose vertices are the midpoints of the edges of the cell. This lets us smooth out lines with no special cases.

How do we generalize this to three dimensions? On a cube, we have two different types of diagonals: the "wide staircase" where we cut off an edge, and the "Q*bert staircase" where we cut off a corner. If we pick one or the other, it's not too horrible, but ideally we want to handle both cases.

The 3D analogue of the hexagon might be the rhombic dodecahedron?

[318 words.]

S	M	T	W	T	F	S
						1
2	3	4	5	6	7	8
9	10	11	12	13	14	15
16	17	18	19	20	21	22
23	24	25	26	27	28	29
30

Lambda's Accounts

Member Posts

Possible progress on rhodod smoothing, with images

More smoothed-diagonal honeycomb thoughts

Let's try this again

Profile

November 2025

Page Summary

Style Credit

Expand Cut Tags