Overview
Computational morphogenesis generates form by simulating how living things grow: cells divide, chemicals diffuse, branches split, surfaces buckle under their own expansion. The designer writes the rules and the constraints, and the geometry is what comes out. The method draws on reaction-diffusion chemistry (Turing, 1952), branching grammars (Lindenmayer, 1968), and physical form-finding with hanging chains and soap films, and it now drives research pavilions, 3D-printed rooms, and mass-produced jewellery.Examples
- Elytra Filament Pavilion (2016), a roughly 200 square metre canopy at the V&A by Achim Menges, Moritz Dörstelmann, Jan Knippers, and Thomas Auer, each cell robotically wound in about three hours
- Digital Grotesque II (2017) by Michael Hansmeyer and Benjamin Dillenburger, a 7-tonne sandstone grotto with 1.3 billion surfaces generated by recursive subdivision
- Nervous System's Kinematics dress (2014), 2,279 hinged panels grown to fit a body scan and printed as a single piece, now at the Museum of Modern Art
- Cellular Forms (2014), sculptures produced by simulating the division of millions of cells under nutrient flow (Lomas)
- Airbus's bionic cabin partition (2016), designed with Autodesk and The Living, 45 percent lighter than the part it replaced
Growth and Form
In biology, form is the result of local rules acting over time under physical constraint. A shell, a horn, or a leaf records that process, and computational morphogenesis tries to reproduce it in software. The idea reached design from three fields: mathematical biology, developmental theory, and structural engineering.
Reaction-Diffusion
The shapes of cells, skeletons, and shells can be explained by forces and geometry rather than heredity alone, and a coordinate grid deformed in a few directions maps one fish species onto another (Thompson, 1917). Parametric modelling software now performs that transformation routinely. Animal markings arise when two chemicals, an activator and an inhibitor, diffuse at different rates and react (Turing, 1952). The resulting equations produce spots, stripes, and labyrinths from a uniform field, and a change in one diffusion rate moves the result from leopard to zebra. Confirmed in living tissue only decades later, in zebrafish stripes and the ridges of the mammalian palate, the equations are still the most common method of pattern generation in creative computation.
Branching Grammars
L-systems are string-rewriting grammars that model branching in plants and algae (Lindenmayer, 1968). A rule as simple as replacing every F with F[+F]F[-F]F, applied repeatedly and drawn as turtle graphics, yields a shrub. Extended with parameters, randomness, and sensitivity to light and neighbouring branches, the grammars reproduce the architecture of specific species (Prusinkiewicz and Lindenmayer, 1990), and they still underlie most procedural vegetation in games and film.
Physical Models
The second lineage uses physical material to do the calculation. A hanging chain settles into a catenary, and inverted it becomes a pure compression arch; weighted chain models found the vaults of the Colònia Güell crypt, and soap films, suspended nets, and wet wool were systematized as form-finding instruments at the Institute for Lightweight Structures in Stuttgart from 1964. The Munich Olympic roofs (1972) and the Mannheim Multihalle (1975), a timber gridshell spanning 60 metres, were derived from hanging models. In each case the geometry was found by letting a model settle under gravity rather than drawn in advance.
Simulation Methods
Contemporary practice combines several algorithm families, often inside one project. Each models growth in a different way.
Differential Growth
Differential growth models a surface that expands faster than the space available to it, so it buckles and folds as coral, lettuce, and cortex do. Each vertex of a mesh pushes away from its neighbours while the surface resists stretching and bending, and the folds are where the two demands collide. Floraform (2015) grew jewellery this way, and Cellular Forms (2014) simulated cell division driven by nutrient flow to produce sculptures with millions of cells. Reaction-diffusion run on a mesh does related work for surface relief and perforation.
Agents and Particles
Three local rules, separation, alignment, and cohesion, are enough to produce flocking (Reynolds, 1986). Swarm-based design lets agents deposit material along their paths, so that structure is the accumulated trace of interaction rather than a drawn object; the practices Kokkugia and Biothing built a body of work on it. Space colonization grows branches toward scattered attractor points and removes each point once reached (Runions et al., 2005), and it remains the standard method for generating leaf venation and tree crowns.
Topology Optimization
Topology optimization removes material wherever stress is low until only a load path remains (Bendsøe and Kikuchi, 1988). Dynamic relaxation, as in the Kangaroo plug-in for Grasshopper, reproduces hanging models numerically, letting a mesh settle under gravity or pressure until its forces balance. Both are now the core of commercial generative design software, which is why bone-like brackets appear so often in examples of the method.
Buildings and Objects
Morphogenetic design entered architecture in the mid-2000s and has been tested mainly through research pavilions, built almost yearly since 2010 at the Institute for Computational Design and Construction (ICD) in Stuttgart. The BUGA Fibre Pavilion (2019) spans 23 metres with 60 components containing about 150 kilometres of robotically wound glass and carbon fibre, its cell structure modelled on beetle wing cases. The Elytra Filament Pavilion (2016) used the same winding logic for a canopy of about 200 square metres, each cell taking roughly three hours to wind. Digital Grotesque (2013, second version 2017) applied recursive subdivision to generate ornament with more than a billion surfaces, then printed it in sand as a 7-tonne grotto. The Silk Pavilion (2013) let 6,500 silkworms complete a robotically laid scaffold, so that the last stage of growth was done by the organism itself. Grown lamps and jewellery from Nervous System are produced in the thousands, one of the few cases where morphogenetic form has reached a consumer price point. These works are part of a wider interest in biomimetic systems, but they borrow from biology at different depths. The fibre layout of a pavilion is based on the structure of a beetle's wing case, while a recursive subdivision rule has no biological reference at all.
Material Computation
A growth simulation means little until it is tied to real material. In material computation, the elastic bending of plywood or the anisotropic swelling of wood becomes part of the generative system rather than a constraint checked afterwards (Menges, 2012). The simulation only produces forms the material can take, and the material's behaviour completes the form during fabrication. This aligns the field with material intelligence and with embodied computation, where physical form performs part of the calculation. Some designers work with organisms directly rather than with models of them. 4D-printed strands fold into programmed shapes when wetted, and engineered living materials research grows components from bacteria and fungi. The distinction between simulating growth and using it is becoming less clear. An open question is how much of a biological principle actually transfers. Many morphogenetic forms resemble organisms without sharing their structural or ecological logic, so it is useful to ask which principle a project borrows, at what scale, and whether the result has been tested under load rather than judged by resemblance.