When placed in moving water, these neutrally buoyant shapes self-assemble into interesting coral-like patterns and growths.

See here for video: https://youtu.be/aCJM2R2Kdmc

Print the “coral buds” with 100% infill. The pieces are designed pre-hollowed to achieve as close to neutrally buoyant as possible with the magnets in place. In reality, the pieces are slightly denser than neutral in part because achieving perfectly neutral is next to impossible and because it allows the option of filing off a bit of material if you would like them to float more.

Each “coral bud” requires six 1/8” x 1/16” cylindrical magnets. Insert the magnets side-to-side as pictured with the polarity consistent across all surfaces and all pieces. They can be press fit with pliers and should not require glue.

The pump base is sized for a 6” (OD) glass vase from a dollar store. The pump is a 3V submersible pump from Adafruit: https://www.adafruit.com/product/4546 . The legs of the pump base have platforms for adding weights such as coins to keep the base and pump submerged.

While the shape of the coral buds looks simple, the design constraints were challenging and required a balancing of magnetism, buoyancy, geometry and surface topology.

The basic design constraints were that the pieces had to be as small as possible, require as few magnets as possible, and be printable in one piece with minimal supports.

The underlying geometry is Laves graph, achieved by centering each pair of magnets on three faces of a rhombic dodecahedron.

Using the rhombic dodecahedron as the final shape might seem like the easier option, but this results in a high level of “mutations” in which the pieces join but not in the correct geometric alignment. The end result of this is that the overall geometry breaks down and they do not form nice structures.

Adding the saddle-shaped contours to the faces reduces the frequency of mutations. The saddle shaped contours allow for strong connections when pieces meet in the proper alignment but makes weaker connections when the pieces join in incorrect alignment (“mutations”). With the pieces in motion, “mutations” either break apart on their own or are knocked out of place by other pieces.

I would love to see the results of different approaches to this design challenge. I also invite any suggestions for improving on this design.