Three Numbers Can Draw Almost Every Seashell Ever Made

Walk along any beach and you are surrounded by a manufacturing miracle. Cone snails that taper to a lethal point. Nautiluses scrolled into a logarithmic perfection. Limpets splayed flat as satellite dishes. Tusk shells like miniature elephant tusks. Each form is consistent across every individual of a species, across oceans and decades, regardless of water temperature, food availability, or growth rate. Evolution has somehow encoded extraordinarily precise three-dimensional geometry into animals that have no architects, no measuring instruments, and no ability to see their own shells as they build them.

How? That question has animated biologists and mathematicians for more than a century, and a new paper from Caltech's Division of Engineering and Applied Science may finally answer it. According to Liu and Bhattacharya (2026), the shapes of nearly all known molluscan shells — spanning every major class of the phylum — can be reproduced with essentially three parameters: a single number governing scale, a vector governing rotation, and the curve of the shell's embryonic starting edge. Not dozens of genes. Not a biochemical orchestration of staggering complexity. Three numbers and a curve.

The Science

The intellectual lineage of this work begins with D'Arcy Wentworth Thompson, the Scottish biologist who in 1917 published On Growth and Form, a book-length argument that biological shapes are constrained and explained by physical and mathematical laws. Thompson noticed that shell apertures — the growing mouth of the shell — seem to retain their shape as the shell expands, tracing a path that blends rotation with dilation (scaling). That simple observation has been the seed for many subsequent mathematical treatments of shell morphology.

Liu and Bhattacharya (2026) take this intuition and formalize it rigorously using the machinery of Lie groups. A Lie group — named for the nineteenth-century Norwegian mathematician Sophus Lie — is a mathematical object that describes continuous, smooth transformations. Think of all possible rotations of a sphere: they can be combined, reversed, and composed, and they vary continuously. That set of transformations forms a Lie group. What the Caltech team recognized is that the accretion of a molluscan shell — each day's new layer deposited at the mantle edge — is precisely this kind of continuous, composable transformation. Yesterday's growing edge is rotated, scaled, and set down again to become today's edge, which is then rotated and scaled to become tomorrow's.

The key insight — and the thing that makes this more than a geometric curiosity — is that only local information is needed. The shell does not need a global blueprint stored anywhere. The mantle cells at the growing edge only need to respond to the geometry immediately around them: how curved is the edge right here, right now? From that local rule, applied repeatedly over months or years, the entire global form of the shell emerges.

Formally, the authors express the growth front $\mathbf{y}(t,x)$ — the position of the growing edge parameterized by arc length $x$ at time $t$ — as evolving under:

$$\frac{\partial \mathbf{y}(t,x)}{\partial t}=\mathbf{A}\mathbf{y}(t,x)$$

where $\mathbf{A}=\log \lambda \mathbf{I}+\mathbf{W}$ is the Lie algebra element encoding both scaling (through the scalar \lambda$) and rotation (through the skew-symmetric tensor ${\bf W} = {\bf w} \times, which encodes the rotation axis ${\bf w}$). Solving this equation gives the full shell surface:

$$\mathbf{y}(t,x)={\lambda }^t\mathbf{R}(t){\mathbf{y}}_0(x)$$

where ${\mathbf{y}}_0(x)$ is the initial protoconch edge and $\mathbf{R}(t)=\exp (t\mathbf{W})$ is the accumulated rotation. The full shell is then the orbit of the conformal Lie group $\mathcal{G}=\{{\lambda }^t\mathbf{R}(t),t\in \mathbb{R}\}$ acting on that initial seed curve. In plain language: specify the starting edge, the rate at which each new whorl is bigger than the last, and the axis around which it turns — and you get the complete shell, automatically, without ever programming what the finished shape should look like.

The protoconch — the embryonic first chamber that every molluscan shell begins with — is the biological anchor for ${\mathbf{y}}_0$. In most species (homeostrophic shells, where the larval and adult whorls share the same axis), the protoconch edge simply serves as the starting curve. In heterostrophic species, where the larval shell switches axis at metamorphosis, the mantle is reconstructed at that stage, resetting the initial condition.

What They Found

The framework reproduces an astonishing range of shell architectures. The single scaling parameter $\lambda$ — which governs how much larger each new whorl is compared to the last — turns out to be the deepest organizer of shell diversity

Scaling Parameter λ Across Molluscan Families

Label	Value
Turritellidae (Gastropoda)	1.033
Buccinidae (Gastropoda)	1.055
Conidae (Gastropoda)	1.07
Nautilidae (Cephalopoda)	1.18
Patellidae (Gastropoda)	1.45
Dentaliidae (Scaphopoda)	1.02

λ governs how much larger each new whorl is relative to the last. Larger λ values generally correspond to more rapidly flaring or flattening shells.

. Planispiral shells like the nautilus, which expand in a flat coil, require $\lambda$ values that keep successive whorls nearly the same size in the direction perpendicular to coiling. Tall helical shells like Turritella use a modest $\lambda \approx 1.033$. Broadly flaring shells like the conch family use larger values. Flat limpets use very large $\lambda$ combined with a near-zero rotation, so the shell essentially expands outward in one shot rather than coiling. Tusk shells (Dentaliidae), which are open at both ends, require a $\lambda$ combined with a rotation vector pointing nearly parallel to the shell axis.

What makes this especially compelling is how $\lambda$ maps onto evolutionary history. When Liu and Bhattacharya (2026) overlay their $\lambda$ values onto the molluscan phylogenetic tree — a tree reconstructed from genomic data by Chen et al. (2025) — they find a striking correspondence. Related classes tend to cluster in $\lambda$ space. Gastropods, bivalves, scaphopods, and cephalopods each occupy recognizable regions. The parameter is not random noise; it carries phylogenetic signal

Shell Shape Parameters by Major Molluscan Class

Label	Value
Gastropoda	65
Cephalopoda	85
Bivalvia	75
Scaphopoda	30
Polyplacophora	50

Axes are normalised to the maximum observed value across classes. Cephalopods have high scaling and rotation; Scaphopods have low rotation and low scaling.

.

The robustness finding is perhaps the most counterintuitive result. Because the growth law is purely local — depending only on the current geometry of the edge, not on any global template or timer — the rate of growth can vary wildly without changing the final shape. A shell growing quickly in summer and slowly in winter will still converge to the same form as one growing at a constant pace, because each increment follows the same local rule regardless of how much time has passed between increments. The animal doesn't need to know how big it is, or how big it will become. It only needs to respond to its own edge geometry.

This elegant decoupling explains a longstanding puzzle: why shell shapes are so consistent across individuals in very different environments. It's not that the animal is monitoring and correcting its growth trajectory. It's that the trajectory is an attractor of the local rule, and any reasonable growth rate leads to the same attractor.

Scaling Parameter λ Across Molluscan Families

Label	Value
Turritellidae (Gastropoda)	1.033
Buccinidae (Gastropoda)	1.055
Conidae (Gastropoda)	1.07
Nautilidae (Cephalopoda)	1.18
Patellidae (Gastropoda)	1.45
Dentaliidae (Scaphopoda)	1.02

λ governs how much larger each new whorl is relative to the last. Larger λ values generally correspond to more rapidly flaring or flattening shells.

Why This Changes Things

The implications ripple outward in several directions simultaneously.

For evolutionary biology, the finding suggests that the vast morphological diversity of molluscan shells — more than 100,000 living species — arises not from 100,000 distinct genetic programs but from continuous variation in a small parameter space. A slight shift in the biochemistry that determines $\lambda$ at the mantle edge, or a small change in the initial protoconch geometry, is sufficient to move a lineage from a tightly coiled snail to a flat limpet to a tusk. This is a powerful economy of means. Evolution does not need to redesign the whole growth program to explore a new shell morphology; it only needs to tune a few local parameters. The morphological distance between a nautilus and a cone snail is not nearly as large, in parameter space, as it appears on a beach.

For developmental biology and morphogenetics, the paper challenges the dominant paradigm of "bottom-up" explanations — the idea that to understand a biological form, you must trace every relevant gene, protein, and signaling pathway. Liu and Bhattacharya (2026) show that a top-down mathematical framework can capture the geometric essentials of shell growth with remarkable completeness, without knowing anything specific about the underlying biochemistry. This is not anti-biology; it is a complement to it. Just as thermodynamics explains the behavior of gases without tracking individual molecules, this Lie group framework explains shell form without tracking individual mantle cells.

The connection to earlier work is worth dwelling on. Raup (1966) famously parameterized shell shapes with four numbers — whorl expansion rate, distance from the axis, translation rate, and shape of the generating curve — and showed that real shells occupy only a small corner of the resulting four-dimensional morphospace. The Liu–Bhattacharya framework is both more and less than Raup's: it uses fewer parameters (three versus four), but more importantly it derives those parameters from a growth law rather than simply fitting them to observed shapes. The parameters are not just descriptive; they are mechanistic.

The comparison with Moulton et al. (2014) — who modeled growing fronts that translate, rotate, scale, and twist — is also instructive. Liu and Bhattacharya (2026) begin from a similar kinematic starting point but arrive at a more constrained and more powerful representation by requiring that the growth law be purely local. The locality constraint is not an assumption of convenience; it is a biological claim backed by experimental evidence that shell growth and mantle growth can be uncoupled, that neurosecretory activity at the mantle margin can adjust growth locally, and that shell formation appears to be a self-organizing process driven by the shell's own geometry.

For materials science and engineering, the Lie group framework potentially offers a blueprint — or rather, an anti-blueprint — for designing complex self-organizing structures. If you want to engineer a structure with a precise three-dimensional geometry, the conventional approach is to specify that geometry globally and fabricate it directly. The shell framework suggests an alternative: specify a local growth rule, and let the global shape emerge. This is especially attractive for soft robotics, programmable matter, and any application where the structure needs to grow or adapt in an environment that cannot be fully controlled in advance. The paper notes that shells achieve "hierarchical microstructures and robust growth dynamics" through this approach — and that the same principles might inform the design of engineering systems capable of autonomous self-organization.

Shell Shape Parameters by Major Molluscan Class

Label	Value
Gastropoda	65
Cephalopoda	85
Bivalvia	75
Scaphopoda	30
Polyplacophora	50

Axes are normalised to the maximum observed value across classes. Cephalopods have high scaling and rotation; Scaphopods have low rotation and low scaling.

What's Next

Like any elegant framework, this one raises as many questions as it answers — which is, arguably, the mark of a productive theory.

The most immediate open question is the biochemical one: what exactly sets $\lambda$ and $\mathbf{w}$ at the cellular level? The paper deliberately sidesteps this, arguing for a top-down mathematical language first and a bottom-up mechanistic explanation later. But connecting the Lie algebra parameters to specific molecular actors at the mantle edge — the neuropeptides, the ion channels, the crystal nucleation sites — is the obvious and necessary next step. Li et al. (2024) have shown that neurosecretory activity at the mantle margin affects shell secretion, and Schoeppler et al. (2019) have demonstrated that shell microstructures arise through classical crystal growth mechanisms. The bridge between those findings and the Lie group parameters is a rich territory for future work.

There are also shells that push at the edges of the framework. Some highly irregular shells — those of oysters, for instance, which conform to the surfaces they settle on — do not obviously fit a simple Lie group trajectory. The paper focuses on the "nearly all known" shell shapes, but characterizing the exceptions, and understanding what breaks the local-rule paradigm in those cases, would sharpen the theory considerably.

The phylogenetic mapping is suggestive rather than rigorous at this stage. The correspondence between $\lambda$ values and the molluscan family tree is visually compelling (Liu & Bhattacharya, 2026), but a quantitative phylogenetic analysis — testing whether $\lambda$ evolves in a manner consistent with Brownian motion or an Ornstein-Uhlenbeck process on the tree, for example — would put the evolutionary claim on firmer statistical ground.

And then there is the engineering frontier. The authors gesture toward "a new approach to engineering complex structures," but realizing that promise will require translating the mathematical framework into physical fabrication strategies: 3D printing protocols that apply local curvature rules, soft material systems that self-actuate according to geometric feedback, or synthetic biological systems in which engineered cells are programmed with local growth responses. None of that is easy. But the conceptual scaffolding is now in place.

Ultimately, what Liu and Bhattacharya (2026) have achieved is a kind of grammar of shells — a compact, principled description of how local rules compose into global beauty. The nautilus doesn't know it's building a nautilus. The cone snail has no image of a cone stored anywhere in its nervous system. Each only knows the geometry of its own edge, right now, and what to do next. From that ignorance, repeated faithfully over a lifetime, comes a shape so precise and so consistent that humans have been collecting and marveling at it for tens of thousands of years. The mathematics, it turns out, is simple. The simplicity is the miracle.