The Noperthedron is a shape that cannot pass through a copy of itself. It was discovered computationally in 2025, resolving a question about the geometry of self-intersection that had resisted intuition.
Most convex shapes can transit through copies of themselves. A sphere passes through a spherical hole easily. A cube can pass through a square hole (at an angle). Ellipsoids, cylinders, most polyhedra — they can all be maneuvered through openings shaped like their own cross-sections. The question was whether this is always true: can every convex body pass through a hole shaped like one of its own projections?
The Noperthedron says no. There exists a specific convex body — computed numerically, not given by a simple formula — that cannot pass through any hole shaped like any of its own projections. No rotation, no tilting, no clever maneuvering works. The shape is too complex in a precise geometric sense: every projection is too tight to admit the full body.
The transition between “can pass” and “cannot pass” is sharp. Small perturbations of the Noperthedron can pass through their holes. The Noperthedron itself sits at a critical boundary — the threshold where self-transit becomes impossible. This is a computational phase transition: below the threshold, every shape transits; above it, none do; and the threshold itself is a specific, computable shape.
The result is a constraint on the relationship between a three-dimensional object and its two-dimensional projections. Projections always lose information — a shadow is less than the object casting it. For most shapes, the information loss is recoverable: the shadow is similar enough to the original that the original can fit through a shadow-shaped hole. The Noperthedron is the shape where the information loss becomes irrecoverable. The projections are too different from the body itself for any of them to serve as a passage.
This connects to a broader question about dimensional reduction. Whenever a higher-dimensional object is represented in fewer dimensions — a 3D protein flattened into a 2D gel image, a high-dimensional dataset projected onto principal components, a physical object casting a shadow — information is lost. The Noperthedron shows that the loss can be qualitative, not just quantitative: there exist objects whose projections are not merely simplified versions of themselves but fundamentally inadequate representations.