It is readily said that nature prefers certain numbers. The three of the triangle, the six of the honeycomb cell, the twelve of the fullerene recur often enough for one to suspect an intention, a signature or a hidden law. The observation that feeds this impression is correct. The interpretation is not. The numbers associated with natural architectures do not all have the same status, and one and the same numerical value may appear for reasons entirely foreign to one another. The number does not suffice to characterise a structure. Only the mechanism that produces it allows its significance to be understood.
Some numbers correspond to the elementary operations of spatial construction. Others appear as general solutions to precisely defined geometric problems. Some are imposed by global topological constraints. Others result from local configurations stabilised by matter, chemistry, development or evolution. A final category emerges when dynamic processes select a spatial scale whose ratio to the size of the system determines a variable number of bands, spots or repetitions. This distinction prevents real regularities from being turned into a disguised numerology. Two structures may display the same number of elements without sharing the same origin, the same function or the same constraints.
Moreover, not all these numbers recur. The nine of axonemes, the seven of certain membrane channels or the eight of cubic coordinations appear only where particular conditions have stabilised them. This rarity is part of the demonstration. If every value recurred, one would indeed have to seek a reason for that recurrence. It is because some recur and others do not that the explanation must be sought in the constraints, and not in the numbers themselves.
This reading leads beyond an inventory of numerical regularities toward an examination of the mechanisms that produce them. The following sections distinguish geometric, topological, material and dynamic constraints. The final part draws the consequences for the origin of our knowledge.
1. The elementary operations of spatial construction
The numbers 1, 2, 3 and 4 correspond to four elementary thresholds: individuating a unit, establishing a relation, closing a contour and closing a volume. They do not constitute universal optima, but the minimal operations from which a geometric representation can be built.
1: the individuated unit. Every architecture begins with a unit distinguished from the rest: particle, molecule, cell, organism or building element. The number 1 does not yet describe a relation or a form. It designates the element from which an organisation becomes possible. Before counting bonds, neighbours or symmetries, one must identify what is being brought into relation. The 1 must not be confused with the geometric point. A point is an object without dimension, whereas 1 expresses a quantity. The comparison remains useful nonetheless, since both occupy an elementary position in the construction of a representation.
2: the minimal relation. Two units introduce a first relation: distance, direction, bond, polarity, complementarity or opposition. Two distinct points determine a line, and the segment joining them is the shortest path between them in Euclidean space. The 2 also appears in systems founded on pairing, such as the two complementary strands of DNA, certain protein dimers or bilateral symmetry. These examples do not proceed from a single numerical law. They share only one elementary operation: linking, pairing, opposing or separating two units.
3: the closure of a contour. Three non-collinear points define a plane and form the first closed polygon. When the length of its sides is maintained, the triangle cannot change shape without deformation of one of its elements. It thus constitutes the first rigid structure in the plane. This property explains its importance in trusses, frameworks, reticulated structures and certain bony or biological networks. The triangle corresponds to the minimal threshold at which an articulated contour becomes geometrically stable.
4: the closure of a volume. Four non-coplanar points make it possible to form a tetrahedron, the simplest polyhedron. When the length of its edges is maintained, it constitutes an elementary rigid cell of three-dimensional space. Tetrahedral coordination also plays a major role in chemistry and in materials, notably in the organisation of sp³-hybridised carbon, of diamond, of silica tetrahedra and of many molecular complexes. The 4 does not have a single status: it corresponds to the minimal closure of a volume, to a fundamental coordination of matter and, in other contexts, to the orthogonality of the square.
2. The general geometric results
Some numbers appear as remarkable answers to precisely defined mathematical problems. Their validity is general within the framework of their assumptions, but it must not be extended to all systems indiscriminately.
3: minimal rigidity in the plane. The triangle is the first rigid polygon when the length of its sides is preserved. An articulated quadrilateral can be deformed into a rhombus without altering the length of its sides, whereas the triangle has no such freedom. The number 3 corresponds to the minimal threshold of rigid closure of a contour in the plane. This property must not be confused with the static determinacy of a complete structure, which also depends on the number of members, joints and degrees of freedom.
6: the economy of boundary in the plane. The regular hexagonal tiling minimises the total length of boundaries when a plane is divided into regions of equal area. The hexagon tiles the plane without leaving gaps, shares its sides with neighbouring cells and offers a greater economy of perimeter than the square or the equilateral triangle for the same area. This result is established by the honeycomb theorem. It helps us understand the appearance of networks close to the hexagon in bee cells, foams, certain cell layers and various systems dominated by interfacial tensions. Natural structures do not, however, always produce perfectly regular hexagons. Growth, local irregularities, mechanical constraints and boundary conditions often modify their geometry. The theorem defines an optimum under precise conditions, not a form obligatory for every natural structure.
12: maximal neighbourhood in three-dimensional space. Twelve identical spheres can simultaneously touch a central sphere of the same size without overlapping. A thirteenth sphere cannot be added. The maximal contact number of identical spheres in three-dimensional space equals 12. This coordination appears in face-centred cubic and hexagonal close packings. Each sphere there has twelve immediate neighbours and the whole reaches the maximum possible density for identical spheres. The 12 here expresses a limit of neighbourhood and compactness. It does not constitute a general law of all organisation in three dimensions.
The numbers 3, 6 and 12 do not form a sequence governing all natural forms. They answer three different problems: minimal rigidity, economy of boundary and maximal neighbourhood. These results remain tied to precise conditions: preserved lengths, regions of equal area, identical and rigid spheres.
3. Topological constraints and combinatorial curvature
The 5, the 6 and the 7 can be brought together under a single mechanism: their contribution to the curvature of a trivalent network, that is, a network in which three faces meet at every vertex. This condition is essential. Without it, the number of sides of the faces does not suffice to determine either the total curvature or the number of pentagons required for closure.
In a closed trivalent network, Euler's relation leads to a general rule. If fₙ denotes the number of faces with n sides, the sum of the contributions of all the faces must satisfy the following relation:
| Σ (6 − n) fₙ = 12 Trivalent network of spherical topology |
|---|
This relation can be expressed as a combinatorial curvature. Each face with n sides makes a contribution equivalent to 60 multiplied by 6 minus n degrees. This contribution does not mean that the real geometric curvature is necessarily concentrated at the centre of each face. It distributes over the faces a balance whose sum equals the total angular deficit borne by the vertices. For a closed surface of spherical topology, the sum reaches 720 degrees.
| Contribution of a face with n sides: 60 × (6 − n) degrees |
|---|
6: null contribution and planar extension. A hexagonal face makes a null contribution, since 60 multiplied by 6 minus 6 equals 0. In a regular trivalent network, three hexagonal angles of 120 degrees meet around a vertex and total 360 degrees. The network can extend locally without curvature. This neutrality completes the other status of the 6: the same hexagon that adds no combinatorial curvature also achieves the economy of boundary of a tiling into cells of equal area. One and the same numerical value here intervenes in two distinct mechanisms.
5: positive contribution and closure. A pentagonal face makes a combinatorial contribution of 60 degrees. Introducing a pentagon into a hexagonal network adds one positive unit of curvature. This rule is independent of the local arrangement of the faces and holds for the dodecahedron as much as for C₆₀, C₇₀ or the other fullerenes.
| Pentagon: 60 × (6 − 5) = +60° |
|---|
In a spherical trivalent network composed solely of pentagons and hexagons, the hexagons contribute nothing to the balance. The surface must accumulate 720 degrees, while each pentagon supplies 60. Exactly twelve pentagons are required, whatever the number of hexagons.
| 12 pentagons × 60° = 720° |
|---|
The number 12 is not here a primary property or a mysterious preference of nature. It results from three conditions: a spherical topology, a trivalent network and faces limited to pentagons and hexagons. The same result can be obtained directly from Euler's relation. Euler expresses the global constraint of connectivity, while combinatorial curvature shows how each type of face takes part in the balance required for closure.
The relation between the dodecahedron and the fullerenes follows directly. The dodecahedron is the minimal solution: twelve pentagons and no hexagons. C₆₀ adds twenty hexagons, C₇₀ twenty-five, without ever modifying the balance, since the hexagon contributes nothing. The fullerenes are not a family of another kind. They are dodecahedra whose twelve pentagons have been moved apart by the insertion of neutral faces. The number twelve does not vary, because nothing that was added counts.
A neighbouring principle intervenes in icosahedral viral capsids, which comprise twelve pentameric positions associated with the vertices of the icosahedron. The comparison remains limited to capsids answering to this organisation and cannot be extended to all viral architectures.
7: negative contribution and compensation. A heptagonal face makes a combinatorial contribution of minus 60 degrees. Its regular interior angle equals 900 divided by 7 degrees, that is, about 128.57 degrees. It is not, however, necessary to imagine three regular heptagons assembled in the Euclidean plane. The per-face contribution directly supplies the general relation: the heptagon introduces one unit of negative combinatorial curvature into a trivalent network.
| Heptagon: 60 × (6 − 7) = −60° |
|---|
In a spherical network composed of pentagons, hexagons and heptagons, the hexagons remain neutral. Each heptagon removes the equivalent of 60 degrees from the positive balance and requires an additional pentagon to preserve the same spherical topology.
| f₅ − f₇ = 12 Each additional heptagon requires an additional pentagon |
|---|
This compensation makes it possible to understand the pentagon–heptagon pairs present in certain network defects, in the necks and junctions of nanotubes, or in structures of opposite curvature. A Stone–Wales transformation can convert four hexagons into two pentagons and two heptagons. The combinatorial balance remains unchanged, but the curvature is redistributed locally.
Heptameric biological assemblies also exist, such as the channel formed by pannexin 1. Their seven-subunit organisation nevertheless arises from another mechanism, founded on molecular interactions, on the diameter of the channel and on its functions of passage or selectivity. Their common number does not suffice to establish a kinship with the curvature of polygonal networks.
The 5, the 6 and the 7 thus form three consecutive regimes of one and the same geometric accounting: positive contribution for the 5, null contribution for the 6, negative contribution for the 7. This accounting does not, however, serve only to classify faces. Restricted to a single type of polygon, it ceases to be a balance and becomes an equation whose solutions are finite in number.
4. One constraint, three polyhedra
When a closed trivalent network comprises only one type of face, the preceding accounting reduces to a single term. The number of faces then ceases to be free and can be read directly off the number of sides.
| fₙ = 12 / (6 − n) Closed trivalent network all of whose faces have n sides |
|---|
The type of polygon then fixes the size of the polyhedron. Only three values provide a solution. Three sides give four faces, that is, the tetrahedron. Four sides give six faces, that is, the cube. Five sides give twelve faces, that is, the dodecahedron. Six sides admit no finite solution, since a null contribution can never reach 720 degrees: the hexagonal network does not close and extends indefinitely in the plane.
These three polyhedra are exactly the trivalent Platonic solids, and there are no others. The octahedron and the icosahedron are absent from this list because four and five faces meet at each of their vertices. Their absence confirms that trivalence is not a formal precaution but the condition on which the result depends.
The reach of this reduction extends beyond the case of the regular polyhedra. The 3, the 4 and the 5 were presented above under distinct statuses: elementary closure of a contour, contextual orthogonality of the square, positive curvature of the pentagon. They appear here as the only values of n compatible with a trivalent closure, and the numbers 4, 6 and 12 they engender are the sizes of the three corresponding structures. A single relation thus produces four values that the preceding sections had to treat separately. No inverse generalisation is thereby permitted: the same value may issue elsewhere from an unrelated mechanism, and the 12 of the contact number between spheres provides the immediate example.
It is not the numbers that are primary. They are the numerical traces of the constraints that make an architecture possible, without those constraints themselves being the last word of the explanation. The tetrahedron, the cube and the dodecahedron are not three figures retained for their elegance or their antiquity. They are the only three possible answers to a single question: how to close a surface with identical faces three of which meet at every vertex. The 4, the 6 and the 12 are what that question leaves behind.
5. Other local and contextual solutions
4: orthogonality and modularity. In the plane, the square organises two perpendicular directions. This orthogonality facilitates alignment, measurement, modular repetition, cutting and construction. An articulated quadrilateral is not rigid, but this weakness disappears when its angles are locked or a diagonal is added. The square tiling has a greater perimeter than the hexagonal tiling for cells of equal area. It may nevertheless become preferable when orthogonality, standardisation, accessibility or ease of fabrication count for more than maximal economy of boundary. The 4 intervenes in several mechanisms: closure of the tetrahedron, tetrahedral coordination and orthogonal organisation of the plane.
8: cubic coordinations and particular symmetries. The number 8 appears notably when an element is surrounded by neighbours placed at the vertices of a cube. It intervenes in certain crystalline coordinations, molecular configurations, protein structures and quasicrystalline symmetries. These occurrences result from local constraints and do not confer on the number 8 a general role comparable to that of the hexagon in the division of the plane or to that of the number 12 in close packings.
9: the organisation of motile axonemes. The motile cilia and flagella of many eukaryotes display a so-called 9 + 2 architecture. Nine peripheral doublets of microtubules surround a central pair. The molecular motors of dynein produce sliding between the doublets, subsequently converted into curvature and movement. This organisation is very widely conserved through eukaryote evolution and constitutes a particularly robust biological solution. Nothing shows, however, that nine represents a unique geometric optimum or a necessary consequence of three-dimensional space. It is a stabilised evolutionary architecture, not a general invariant.
10: an organisation often derived from the 5. The number 10 frequently appears as a doubling of pentameric symmetry. Five axes may each produce two elements, or two series of five structures may be associated. The 10 does not always possess an autonomous geometric principle. Its role must be analysed from the precise organisation in which it appears.
6. Numbers produced by a dynamic scale
Some numerical regularities are imposed neither by a topological closure nor by a geometric optimum independent of the size of the system. They appear when a dynamic process selects a characteristic spatial scale.
In a reaction–diffusion system, the reaction rates and the diffusion coefficients can destabilise an initially homogeneous state and produce a periodic pattern. The mechanism then selects a preferred wavelength, that is, a characteristic distance between the bands, spots or other elements of the pattern. The visible number of repetitions depends on the ratio between that wavelength and the size of the domain in which the pattern develops.
Number of repetitions ≈ size of the domain ÷ selected wavelength
This number can vary when the domain grows, when its limits change or when the dynamic parameters are modified. Two organisms governed by a related mechanism may thus display a different number of bands or spots without the generating rule having changed. The patterns described by Turing models and subsequently developed by the biology of morphogenesis belong to this dynamic status.
This variability contrasts with the case of the fullerenes. Adding hexagons can increase the size of the network considerably without modifying the number of pentagons required for its closure. Enlarging the domain can increase the number of bands, whereas enlarging a fullerene by inserting hexagons leaves its twelve pentagons rigorously unchanged. Two numbers observed in nature may possess opposite rigidities: one varies with the scale of the system, the other remains invariant under its enlargement.
7. One value, several mechanisms
Numbers do not form a hierarchy running from the universal to the anecdotal. Their status depends on the mechanism that produces them. Some correspond to the elementary operations of any spatial construction. Others are the general solutions of precise geometric problems. Others are imposed by the closure or the global topology of a structure. Others again describe local architectures stabilised by the properties of matter, by molecular mechanisms, by development or by evolutionary history. Finally, some result from the ratio between a scale selected by a dynamic and the contingent size of the domain in which it unfolds.
One and the same number may belong to several categories. The 3 corresponds to the minimal closure of a contour and to the elementary rigidity of the triangle. The 4 intervenes in the closure of the first volume, in tetrahedral coordination and in the orthogonal organisation of the plane. The 6 intervenes both in the economy of boundary and in the absence of combinatorial curvature of a trivalent hexagonal network. The 12 appears as the maximal contact number between spheres and as a topological consequence of the closure of a spherical trivalent network. Numerical coincidence never constitutes an identity of mechanism.
8. One constraint, several numbers
One and the same numerical value may result from different mechanisms, but the converse relation is equally true: one and the same general constraint may produce different values according to the nature of the system and according to what one chooses to count.
Spherical topology provides a particularly clear example. In a trivalent polygonal network composed of pentagons and hexagons, Euler's relation imposes twelve pentagons. In a nematic field tangent to a sphere, that same Euler characteristic imposes a total topological charge equal to 2. A frequent configuration distributes this charge among four defects of positive charge equal to one half (+1/2 defects, typically arranged at the vertices of an inscribed tetrahedron).
4 × 1/2 = 2
The sphere universally imposes neither the number 12, nor the number 4, nor the number 2. It imposes a topological constraint. The value observed then depends on the object counted, on its rules of connection and on the way the constraint is distributed among its elements.
The two directions of the argument complete one another. One value may conceal several causes, and one cause may engender several values. In neither case does the number alone allow the mechanism that produced it to be reconstructed.
9. What these numbers reveal about the origin of our knowledge
This classification shows that our geometric and technical knowledge is not born in an abstract void. In the real we observe units, relations, closures, coordinations, tilings, curvatures and packings. We compare these organisations, we measure their properties and we look for the constraints that make them possible. Mathematics isolates these regularities and turns them into demonstrable relations. Science verifies the conditions under which they apply. Technique transposes them into objects, buildings, materials and devices.
We did not invent the rigidity of the triangle, the economy of boundary of the hexagon, the tetrahedral coordination of carbon or the maximal number of contacts between identical spheres. These possibilities already belong to the real. Human invention often consists in recognising them, abstracting them, recomposing them and applying them in another context. Nature does not hand us ready-made theorems or plans. It exposes structures, behaviours and constraints. We build the concepts that allow them to be understood, and then the techniques that allow them to be reused.
Our inventions do not draw on a mysterious list of numbers. They draw on a diversity of mechanisms already at work in the real. The numerical value is never the final explanation. It is not the numbers that are primary: they are the visible traces of constraints, relations, dynamics and histories that remain to be identified. Those constraints are not themselves necessarily primary. They merely mark the level at which the inquiry provisionally stops, before it can be pushed back toward the structure of space, of matter and of the laws that make them possible. It is by moving from the trace to the mechanism, and then from the mechanism to its own conditions, that our knowledge advances.