Morphogenesis

In 1952, between his famed code-breaking work during World War II and his tragic death in 1954, Alan Turing published a paper titled "The Chemical Basis of Morphogenesis." It proposed a stunningly simple mechanism for how patterns form in nature: take two chemicals — an "activator" that promotes its own production and an "inhibitor" that suppresses it — let them diffuse at different rates, and patterns spontaneously emerge from an initially uniform mixture. Stripes, spots, spirals, labyrinths. The mathematics predicted them all.¹

What makes Turing's insight so powerful isn't that it explains one pattern. It's that the same mechanism keeps appearing everywhere, at every scale, in systems that have almost nothing in common except the abstract mathematical structure of activator-inhibitor dynamics. Leopard spots. Zebra stripes. The arrangement of bacterial colonies. Windswept ripples in sand. The spacing of human settlements across a landscape. Psychedelic visual hallucinations. Even, it turns out, the arrangement of atoms in a crystal only a nanometer wide.¹

Turing Patterns at the Atomic Scale

Several years ago, Aharon Kapitulnik's lab at Stanford was trying to grow a thin layer of bismuth crystal on a metallic surface. Instead of forming a uniform sheet, the crystal became a patchwork of uneven growth. In areas where the crystal was one atom thick, striking stripes appeared — small parallel ridges filling irregular patches, with neighboring patches oriented at different angles. On a trip to Paris in 2017, Kapitulnik showed the images to theorist Yuki Fuseya. "This is like a zebra," Fuseya said. And if the stripes were really like a zebra's, they could be a Turing pattern.¹

The suggestion was surprising because Turing's original mechanism requires diffusion — molecules randomly spreading out while reacting with each other. Bismuth atoms don't diffuse. They're locked into a crystal lattice. But the forces at play are structurally analogous. Bismuth atoms want to fit into particular spots on the metal substrate, but those spots are closer together than the atoms find comfortable. The sheet buckles. The vertical displacement of atoms acts as the activator; the in-plane strain acts as the inhibitor. The "morphogens" here aren't molecules at all — they're mechanical displacements. And yet when Fuseya ran Turing's equations with these variables, the simulated pattern was "an amazing match" to the real stripes.¹

What I find most striking about this result is the scale. Each bismuth stripe is about 1 nanometer wide — ten millionths the width of a human hair. Chemical and biological Turing patterns typically form at scales a million times larger. The fact that the same mathematical structure governs pattern formation from atomic crystals to animal hides suggests that Turing patterns aren't really about chemistry or biology at all. They're about a much more fundamental property of systems with competing short-range activation and long-range inhibition. The substrate doesn't matter. The math does.

And the bismuth has one more trick: Turing patterns are self-healing. When part of a pattern is disrupted, it grows back. Fuseya's simulations showed the bismuth crystal mending itself, and the way neighboring stripe regions grew together in the real experiments suggested the same. An inorganic crystal repairing its own surface pattern isn't something you'd expect. But if the pattern is truly an expression of the underlying energy landscape — the stable configuration of competing forces — then it should be as robust as any thermodynamic equilibrium.

Biofilm Architecture: Shape From Cells

Turing patterns are one route to biological form. Peter Yunker's biophysics lab at Georgia Tech has been investigating another: the emergent architecture of bacterial biofilms, which arrive at complex three-dimensional shapes through a mechanism that's less about reaction-diffusion and more about geometry, growth, and nutrient competition.²

A biofilm starts as a flat layer of cells on a surface. As cells divide, the colony grows outward and slightly upward. Then something interesting happens. Cells in the center get starved — the outer cells gobble up oxygen and nutrients before they can diffuse inward. Lars Dietrich at Columbia found that even biofilms just 50 micrometers deep develop internal oxygen gradients steep enough to trigger completely different gene expression in interior versus exterior cells. The center collapses. The edges keep expanding. The result is a complex wrinkled topography — folds and ridges resembling a neocortex — that emerges entirely from the geometry of how cells push against each other and compete for resources.²

Yunker, trained in soft matter physics, treats biofilms like colloids — suspensions of particles whose collective behavior emerges from simple pairwise interactions. Two cells can't occupy the same space (repulsive force). Cells are covered with sticky proteins that fasten them to neighbors (attractive force). If attraction dominates, cells aggregate into biofilms. But unlike colloids, biofilms grow — and they have to balance horizontal versus vertical expansion. It's not unlike urban planning: Houston sprawls horizontally because land is cheap; Queens builds vertically because it's hemmed in by water. Biofilms face the same trade-off. Stickier cells grow more vertically but spread less. A single parameter — the contact angle between the biofilm's lip and its substrate — controls nearly the entire architecture.²

When Dietrich changed the growth medium — different sugar source, less protein, more salt — entirely different topographies emerged from the same bacterial species. Same genome, same cell-level rules, wildly different collective form. This is Cellular Automata-style emergence at its most vivid: local rules (cell division, diffusion, stickiness) producing global structure with no central coordinator. The collective acquires properties — shape, internal differentiation, metabolic resilience — that no individual cell possesses.

The Multicellularity Connection

What makes biofilm morphogenesis particularly interesting is that it may recapitulate life's earliest steps toward multicellularity. Biofilms exist at the liminal boundary between unicellular and multicellular life. They're aggregations of single cells that grow into unified life forms with emergent properties, yet can split back into component cells under duress. They're somehow both unicellular and multicellular — and simultaneously neither.²

The biophysical rules generating their complex shapes — cells pushing on neighbors, competing for resources, secreting extracellular matrix — may be the same rules that drove the first transitions from single-celled to multicellular organisms. As Stuart Newman at New York Medical College puts it, "the characteristics of multicellularity are very strongly influenced by physics." You don't need a genome encoding for multicellular body plans. You need cells that stick to each other, compete for nutrients, and divide. Physics handles the rest.²

This connects to a question that's been bugging me about Convergent Evolution: why does multicellularity evolve independently so many times (at least 25 times across the tree of life)? The biofilm work suggests an answer. If the transition from single cells to multicellular collectives is primarily a physical phenomenon — driven by geometry, diffusion, and adhesion rather than by specific genetic innovations — then it should happen readily whenever the conditions are right. The physics is universal; only the genetic details differ.

Turing's Reach

Since the bismuth paper was published, Fuseya reports hearing from scientists identifying Turing patterns in their own materials. "Experimentalists already see that pattern, but they never realized that this is the same mechanism as in tropical fish." The list of Turing pattern systems now includes chemical reactions, biological development, ecological distributions, crystal growth, and hallucination geometry. Andrew Krause at Oxford notes that scientists have modeled Turing patterns using predator-prey interactions, whole cells as variables, and now mechanical displacements.¹

I think the reason Turing patterns keep turning up isn't because nature is following Turing's equations. It's because the conditions that produce them — local positive feedback competing with longer-range negative feedback — are absurdly common. Any system where "more of X here promotes more X nearby, but suppresses X further away" will tend to form patterns. This is true whether X is a chemical concentration, a mechanical strain, a population density, or a neuronal firing rate. The patterns are an attractor in the space of possible dynamics, not a specific mechanism. That's why they're so universal, and why Turing's 1952 paper — written as a pure mathematical exercise about hypothetical chemicals — turned out to describe everything from fish scales to crystal growth.