Projection studies

Turn a solid, fold its shadow, and measure the match.

An orthographic camera keeps the horizontal and vertical coordinates and drops depth. The silhouette is the filled shadow. A transparent wireframe draws every recorded edge, so front and back lines can share the same spot on screen.

Rounded chiral cube

A handed solid can still cast a shadow that nearly matches across a fold.

Infinitely many angles give the cube a near-mirror outline. The exact contour count remains open.

Five green cube silhouettes at 0, 41.52, 80, 105 and 130 degrees, each with a dashed vertical fold line.
Five silhouettes from one camera path. Every outline in this family matches its reflection within 0.000001 mm in the rounded model.

Why the fold survives a turn

Keep a half-turn axis upright and look perpendicular to it. The half-turn keeps height and reverses both left and right and front and back. The camera drops the front-to-back reversal, leaving a left-to-right reflection.

What the measurement means

Measure the distance from each reflected point to the nearest original point, and check both directions. The largest distance is the mirror gap. Here the eight core corners match their half-turned partners to within 0.00000060 mm.

The near-mirror +X silhouette alongside the transparent 51-edge drawing from a different angle.
The +X outline folds within 0.000001 mm. The angled view brings all 51 recorded edges within 0.0078 mm of their mirror image.
Camera directions and the exact wireframe proof

The +X sightline points from the solid toward the camera along (1, 0, 0). Its screen up is approximately (0, 0.920469, −0.390816). Record other angles by turning +X around this upright axis while staying perpendicular to it.

The all-edge near-match uses camera direction (0.34777182, −0.93060115, −0.11417640) and screen up (0, 0.12177787, −0.99255738). Both fold lines pass through the screen projection of (17.51735887, −0.00000003, −0.00000040) mm. At +X, a CAD seam in the all-edge drawing misses its reflection by about 0.104 mm.

A circular arc viewed at an angle lies on an ellipse. A mirror of the whole curve must also mirror the ellipse center. Eight center groups supply at least one genuine ellipse each at every angle. The exact certificate checks all 764 ways to pair separated centers and all 28 directions where center pairs can overlap, including their small neighborhoods. It excludes a perfect fold of the literal 51-edge STEP wireframe at every angle.

The exact zero-gap question for the outer contour alone remains open here. The STEP file declares a connectivity uncertainty of 0.01 mm. The reported gaps describe the recorded geometry.

Download the exact wireframe verifier (Python)

Five differently oriented transparent cube drawings, each with its viewing vector underneath.
Five random orthographic wireframes at one drawing scale. Projection sheet (PDF)
The minimum-area silhouette beside the larger +X and angled silhouettes, all drawn to one scale.
The smallest shadow encloses 573.95 mm². The sightline follows a short core edge, about 41.52° from +X.
How the cube area minimum covers all angles

The rounded cube is modeled by a convex core with a 0.5 mm radius offset. The projection is a polygon expanded by a disk. Its area is A + rP + πr², where A and P are the core shadow’s area and perimeter. At the minimum they are 525.0523 mm² and 96.2241 mm.

The six planar faces divide the sphere of viewing directions into 32 regions with fixed contributing faces and outline edges. Supporting bounds on area and projected edge lengths cover every region. The minimizing sightline is approximately ±(0.748730168, 0.259062087, 0.610155694). Either sign gives the same area.

Five rows pair a tilted transparent cube model with its flattened drawing. One edge is orange in every row.
Follow the same orange edge through five rotations. As it points toward the viewer, its projected length shrinks. Open the animated rotation (GIF)
A 76-second lesson follows the parallel camera rays, the left-to-right fold and the edge-center test.
Lesson in text

Parallel rays transfer points to a flat screen and remove depth. At +X the cube folds along the projected vertical axis. Drawing every CAD edge reveals a seam gap in that same view.

Under projection, a circular edge becomes an ellipse unless viewed edge-on. A fold has to mirror the centers of those ellipses. Checking 764 separated pairings and 28 overlap directions rules out exact equality for the wireframe. An angled view brings all 51 curves nearly into alignment.

Read The Shadows of a Chiral Cube (PDF)

Double Bowl

A triangulated model reveals which details survive the same fold.

Five near-mirror outlines were checked. The exact contour count across all angles remains open.

Five Double Bowl silhouettes from 0, 12, 30, 45 and 110 degree turns, each with its area and mirror gap.
Five turns about a fitted axis give five shadows. Their gaps about the tested fold are each below 0.0072 length unit.

Compare the edge sets

The 60 broad-face sides match their half-turned partners to within 0.000072 length unit. The complete 2,709-edge triangle grid has gaps of about 1.77 to 2.47 units in the five views. A separate turn near +X reduced the grid gap to about 1.33 units.

Read the model’s scale

The STL records 1,806 triangles and 905 vertices on a closed surface. Its coordinate format leaves physical units unspecified, so lengths and areas are given in file units and squared file units.

A Double Bowl outline and triangle grid share a vertical fold, with an orange segment exposing a gap in the grid.
The 0° outline nearly folds onto itself. On the grid, the orange segment measures a missed match of about 2.39 units.
A magnified overlay of the original teal outline and its dashed orange reflection, showing a tiny separation.
At the circled location the reflected contour misses the original by about 0.00501 unit. Use the scale bar to read the magnification.
Five camera turns and what the exact test covers

The screen-up axis is approximately a = (0.68472164, 0.44462567, −0.57746367). Project +Z onto the plane perpendicular to a and normalize to obtain the starting sightline n0. The directions are n0 cos θ + (n0 × a) sin θ at 0°, 12°, 30°, 45° and 110°. The screen fold passes through the projected volume centroid.

At each angle, the convex hull of the shadow has a single longest side. A mirror must preserve that side, so only its perpendicular bisector can serve as the fold. An exact calculation sends a hull vertex outside the hull for each candidate fold. All five tested outlines therefore have a small, measurable asymmetry. Exact symmetry at some other camera angle remains open for this mesh.

Independent triangle-union calculations agree on the five areas to five decimal places. The outline gap estimates sample the boundary at most 0.0006 unit apart and add half that spacing to bound the continuous gap. Three approximate half-turn axes support a looser all-angle-on-each-axis-family bound of 0.025 unit.

Download the five-view mirror verifier (Python)

Five shaded Double Bowl projections showing the broad pentagonal regions from different directions.
Five randomly sampled opaque camera views at the same scale, with at least 27° of separation between their viewing axes.
A nearly vertical sightline through the Double Bowl mesh and the resulting smallest measured silhouette.
The global minimum rounds to 615.7 square units. The best measured mesh view gives 615.70694 square units close to the Z axis.
Why the Double Bowl minimum is bounded

The recorded surface has shallow inward deviations. An inscribed convex solid gives an all-direction lower bound for the mesh shadow. Its supporting planes reduce the lower-bound calculation to 921,381 sightlines from plane-normal pairs. A projected convex outer hull at a tested direction gives a separate upper bound.

The global area lies between 615.689655 and 615.707095 square units. The best tested mesh direction is approximately (0.00000002, −0.000257887, −0.999999967), or its opposite. The exact minimizing direction remains unresolved.

Download the minimum-area verifier (Python)

Five rows show the Double Bowl in space on the left and its transparent triangle grid on the right.
Match the turning solid to the camera’s drawing at each stop. Open the animated rotation (GIF)
A 57-second lesson shows the difference between folding a shadow and folding the triangle grid.
Lesson in text

One sample of five directions gives five profiles. The orange line divides two triangles within one flat face. As the camera turns, its projection changes with the other mesh lines.

At the 12° stop the outline misses its reflection by about 0.005 unit. The full grid misses by about 2.08 units. The smallest shadow area rounds to 615.7 square units on the file’s coordinate scale.

Read the Double Bowl Orthographic Study (PDF)

Other solids

Apply the same tests to three different shapes.

Pentagonal icositetrahedron

In the [001] direction of this standard chiral solid, the outline and complete wireframe agree with their reflection across the marked fold. The visible interior edges give a different fold result. Rotating the drawings by 90°, 120° and 180° reproduces the projected patterns along [001], [111] and [110], respectively.

Eight-faced STEP model

Eight planar faces join through rounded transitions. Its 39 line edges and 45 circular arcs give 12 circle-center groups. The certificate checks 140,152 possible pairings and the neighborhoods of 66 overlap directions, ruling out an exact mirror of the complete wireframe. Its silhouette remains a separate question.

Two [001] pentagonal icositetrahedron drawings show visible edges and all edges against the same fold axis.
Compare where visible-only and transparent edges fall on each side of the fold. Try the interactive projections
Three pentagonal icositetrahedron views along [001], [111] and [110] illustrate their different rotational symmetries.
These three representative axes show different orders of rotational symmetry. A fold asks whether a drawing agrees with its reflection.
Five transparent orthographic wireframes of the rounded eight-face STEP model at a common scale.
Five sampled views of the eight-faced model, including front and back edges. Projection sheet (PDF)

Constructed asymmetric polyhedron

This separate convex example has 10 vertices and 16 triangular faces. Its extreme shadow areas are about 12,368.5 mm² and 27,348.8 mm². Drag in the viewer, choose an extreme, or use the arrow keys while a control is focused. The area readout uses the convex surface formula.

Open the polyhedron viewer

The constructed asymmetric polyhedron, a faceted solid with uneven edges and faces.
The constructed solid in three dimensions. Open the rotating view (GIF)
Five orthographic views of the constructed polyhedron from different directions.
Random viewing directions produce a range of different silhouettes.
The minimum-area silhouette of the constructed asymmetric polyhedron.
The minimum-area view of the constructed convex example.
The maximum-area silhouette of the constructed asymmetric polyhedron.
The maximum-area view, drawn with its area scale.
Two silhouettes of the constructed polyhedron overlaid for a projection containment comparison.
Superimpose the two projected outlines to see how their shapes fit together.

Study library

Read the reports, replay the lessons, and reproduce the checks.