Families Construction: Recursive Flexible Polyhedra

• Families Construction is a recursive method that generates flexible polyhedra by replacing a cap in a Bricard octahedron with a quadrilateral band.
• The process produces both genus-0 composites with tetrahedral caps and genus-1 tori by systematically closing and extending quadrilateral strips while maintaining flexion.
• It employs five cap-based subtypes with specific metric and angular conditions, enabling the creation of small examples like decahedra and hendecahedra and larger infinite families.

“Families construction” in the sense developed by Gerald Nelson denotes a recursive method for generating composite flexible polyhedra from Bricard octahedra by replacing a cap with a quadrilateral band and then closing the new boundary by another cap of the same Bricard type. The resulting objects are flexible polyhedra of genus $0$ and $1$, with non-constant dihedral angles under flexion, self-intersections, and indefinite size; in the type III setting the method also yields genus-$0$ examples with any number of faces $n>9$, including a decahedron with seven vertexes and a hendecahedron (Nelson, 2010).

1. Bricard octahedra as the generating units

A Bricard octahedron is a flexible octahedron with $8$ triangular faces and $6$ vertices, discovered by Bricard in 1897. It is self-intersecting and non-convex. The three classical types are: type I, with axial symmetry; type II, with planar symmetry; and type III, with two distinct planar positions where all vertices lie in a plane. Nelson refines this into five cap-based sub-types because the recursive construction is governed by constraints at a single cap, namely a four-face neighborhood of a vertex $X_0$ with faces

$$X_0A_0B_0,\quad X_0B_0C_0,\quad X_0C_0D_0,\quad X_0D_0A_0.$$

The five sub-types are determined by metric or angular conditions at that cap. In subtype I–OEE, opposite edges on the base cycle are equal: $|A_0B_0|=|C_0D_0|,\qquad |B_0C_0|=|D_0A_0|.$ In II–AEE, adjacent equalities are imposed: $|A_0B_0|=|B_0C_0|,\qquad |C_0D_0|=|D_0A_0|.$ In II–OEE, the base cycle satisfies the same opposite-edge equalities as I–OEE, while the cap edges satisfy

$1$0

In III–OAE, opposite face angles at $1$1 are equal: $1$2 In III–OAS, opposite face angles at $1$3 are supplementary. For the type III sub-types, the essential data is angular rather than metric, and Bricard’s equations govern deformation through opposite-equal or opposite-supplementary angle conditions.

This cap-based refinement is the organizing principle of the entire construction. Each of the five sub-types admits an extension operation that produces another Bricard octahedron of the same type, so flexibility is preserved throughout the recursion (Nelson, 2010).

2. Extension as a recursive operation

The extension begins with a flexible Bricard octahedron $1$4 with caps at $1$5 and $1$6. To extend at the cap $1$7, one prolongs the four edges

$1$8

beyond the base vertices by a common scale factor $1$9, producing new vertices $0$0 such that

$0$1

and similarly for the other three edges. Each old base edge then determines a quadrilateral: $0$2 The four triangular faces at $0$3,

$0$4

are removed, leaving an open flexible polyhedral surface consisting of the cap at $0$5 and the four new quadrilaterals.

The essential step is that the scaling factor is chosen so that the new outer cycle $0$6 closes as the base of another cap, producing a second Bricard octahedron $0$7 of the same sub-type. Repeating this at $0$8 yields an infinite recursive family. In this sense, extending a Bricard octahedron means replacing one cap by four quadrilateral faces so that the new boundary again supports a Bricard cap.

The family is parametrized by the initial edge-length or angular data of $0$9, together with extension parameters chosen at each stage. For the edge-controlled sub-types I–OEE, II–AEE, and II–OEE, the initial octahedron is determined by six independent edge-length parameters. For III–OAE and III–OAS, the initial octahedron requires five independent parameters, completed by Bricard-type angular relations. At each extension stage there is additional freedom: up to two independent extensions in I–OEE and II–AEE, always two nonzero parameters in II–OEE, and up to three independent extensions in III–OAE and III–OAS. Special III–OAE cases allow three of the four extension parameters to be zero (Nelson, 2010).

3. Genus-$n>9$0 and genus-$n>9$1 families

The generic genus-$n>9$2 construction has the combinatorics of two tetrahedral caps joined by a chain of quadrilateral bands. After $n>9$3 extensions, if one closes the outer boundary by a cap at $n>9$4 instead of extending further, the result is a closed flexible polyhedron with

$n>9$5

faces and

$n>9$6

vertices. All vertices have index $n>9$7, and Euler’s formula gives genus $n>9$8. Nelson proves that for every integer $n>9$9 there exist flexible closed polyhedra of this form, so the genus-$8$0 family is infinite in combinatorial size as well as in metric parameter space.

The genus-$8$1 construction closes the quadrilateral tube back onto itself. A basic example uses three Bricard octahedra $8$2, arranged so that $8$3 and $8$4 are similar, while terminal caps are extended until a fourth octahedral closure condition is met. After removing the original caps, the remaining surface is a torus with $8$5 quadrilateral faces. The general theorem states that for every integer $8$6 there exist flexible tori with

$8$7

vertices and

$8$8

quadrilateral faces.

In the toroidal case, the construction may be described as taking $8$9 identical Bricard octahedra in a cylindrical or toroidal arrangement with $6$0 layers of smaller octahedra. Because the overlapping octahedra flex compatibly, their intersection vertices remain coincident through the motion. Retaining only the quadrilateral strips between those intersections produces genus-$6$1 flexible polyhedra. This suggests that the family concept is simultaneously metric and combinatorial: fixed incidence structure, continuously variable realization subject to Bricard constraints (Nelson, 2010).

4. Type III constructions and the smallest examples

The most delicate part of the theory concerns the III–OAE family, which yields the smallest known examples in Nelson’s framework. The flexible decahedron is built from two intermediate type III octahedra $6$2 and $6$3, together with a degenerate extension at $6$4. The construction uses five independent parameters on two adjacent faces of $6$5, assigns OAE and OAS status to selected vertices, and determines the remaining face angles through Bricard’s equations.

A central calculation uses the law of sines and Bricard’s angular relations to derive

$6$6

and

$6$7

Eliminating $6$8 gives

$6$9

Solving this quadratic determines the missing cap angle and hence the entire octahedron $X_0$0. The same pattern of relations is then used to construct $X_0$1.

The decahedron arises when three extension lengths at $X_0$2 are effectively zero and only $X_0$3 is extended to a nontrivial point $X_0$4. The final polyhedron consists of the cap at $X_0$5, two triangles from the degenerate extension, and the cap at $X_0$6, giving

$X_0$7

faces. The paper states explicitly that this is a decahedron with seven vertexes. The hendecahedron is obtained by a related III–OAE construction in which only two extension lengths are zero; the first extension contributes two triangles and one quadrilateral, so the total number of faces is $X_0$8.

These examples are the base cases for a broader theorem: there exist flexible genus-$X_0$9 polyhedra with any number of faces $$X_0A_0B_0,\quad X_0B_0C_0,\quad X_0C_0D_0,\quad X_0D_0A_0.$$0. The decahedron and hendecahedron provide the initial cases $$X_0A_0B_0,\quad X_0B_0C_0,\quad X_0C_0D_0,\quad X_0D_0A_0.$$1 and $$X_0A_0B_0,\quad X_0B_0C_0,\quad X_0C_0D_0,\quad X_0D_0A_0.$$2, and the generic four-quadrilateral extension adds $$X_0A_0B_0,\quad X_0B_0C_0,\quad X_0C_0D_0,\quad X_0D_0A_0.$$3 faces at a time thereafter. Because the underlying octahedra are of type III, these polyhedra also inherit the property of having two planar positions (Nelson, 2010).

5. Flexibility, dihedral variation, and self-intersection

The flexion is described by a single dihedral angle, typically

$$X_0A_0B_0,\quad X_0B_0C_0,\quad X_0C_0D_0,\quad X_0D_0A_0.$$4

All coordinates and all other dihedral angles become functions of $$X_0A_0B_0,\quad X_0B_0C_0,\quad X_0C_0D_0,\quad X_0D_0A_0.$$5. Since every stage of the recursion preserves the Bricard type and cap relations, the entire composite polyhedron moves coherently under variation of this single parameter.

The non-constancy of dihedral angles is inherited from the constituent Bricard octahedra. Face shapes remain fixed, but dihedral angles vary continuously during flexion. The quadrilateral bands do not introduce extra rigidity because they are formed from compatible pieces of successive Bricard octahedra. For this reason, the genus-$$X_0A_0B_0,\quad X_0B_0C_0,\quad X_0C_0D_0,\quad X_0D_0A_0.$$6 and genus-$$X_0A_0B_0,\quad X_0B_0C_0,\quad X_0C_0D_0,\quad X_0D_0A_0.$$7 composites retain non-constant dihedral angles under flexion.

All polyhedra in the construction are non-convex and self-intersecting. Nelson’s definition explicitly allows closed surfaces in $$X_0A_0B_0,\quad X_0B_0C_0,\quad X_0C_0D_0,\quad X_0D_0A_0.$$8 with intersecting faces, and the paper states that all closed flexible polyhedra described there have zero oriented volume. Self-intersection is therefore not an incidental by-product but an intrinsic geometric feature inherited from Bricard octahedra and preserved by the extension process. This places the constructions within the classical immersed, rather than embedded, flexible-polyhedron tradition (Nelson, 2010).

6. Families, variations, and mathematical context

The family structure has two levels. First, each subtype admits continuous metric variation in the initial octahedron and in the extension parameters. Second, the recursive process permits arbitrarily many bands, producing objects of indefinite size. For genus $$X_0A_0B_0,\quad X_0B_0C_0,\quad X_0C_0D_0,\quad X_0D_0A_0.$$9, the generic family has two caps and a chain of quadrilateral bands. For genus $|A_0B_0|=|C_0D_0|,\qquad |B_0C_0|=|D_0A_0|.$0, the family has an annular or toroidal quadrilateral pattern. Nelson also identifies a sixth family of more exotic variations, including nonuniform extensions, switching sub-types between stages in the compatible pairs I–OEE ↔ II–AEE and II–OEE ↔ III–OAE, and allowing fewer than four nonzero extensions at a stage.

The work belongs to the line of research initiated by Bricard’s 1897 classification of flexible octahedra and connected by the paper to Bennett’s 1912 and Goldberg’s 1943 work on flexible prismatic linkages, Connelly’s 1977 counterexample to the rigidity conjecture, and results of Stachel and Alexandrov in 2002 and 2009 on the Dehn invariant and type III flexibility. Within that lineage, Nelson’s contribution is a systematic extension scheme based entirely on Bricard octahedra rather than on embedded flexible spheres.

Several features are explicitly identified as novel. The first is the recursive extension scheme itself, which generates infinite families of genus-$|A_0B_0|=|C_0D_0|,\qquad |B_0C_0|=|D_0A_0|.$1 composites and genus-$|A_0B_0|=|C_0D_0|,\qquad |B_0C_0|=|D_0A_0|.$2 tori from any of the three classical Bricard types, refined into five cap-based sub-types. The second is the cap-based sub-classification, which is tailored to the constructive mechanism rather than to a general configuration-space taxonomy. The third is the explicit production of small flexible polyhedra, especially the decahedron with 7 vertices and the hendecahedron with 8 vertices. The fourth is the construction of flexible genus-$|A_0B_0|=|C_0D_0|,\qquad |B_0C_0|=|D_0A_0|.$3 tori built purely from Bricard octahedra.

The paper limits itself to vertex index $|A_0B_0|=|C_0D_0|,\qquad |B_0C_0|=|D_0A_0|.$4, self-intersecting polyhedra, and families whose flexibility is directly inherited from Bricard octahedra. A plausible implication is that the method functions less as a search for isolated examples than as a mechanism for propagating a local flexion law through a recursively assembled global combinatorial structure. In that sense, families construction is the controlled extension of Bricard caps into chains and loops of flexible polyhedral pieces (Nelson, 2010).