A sheet that grows cannot lie flat

A leaf of kale is not flat and never was. Neither is the frilled edge of a lettuce, the rim of a chanterelle, or the wrinkled surface of a growing brain. The usual explanation is mechanical: the tissue is soft, it is under compression, and soft things under compression buckle.

That explanation is true and it is not the reason. The reason is available before any material has been named, and it is the same theorem this site has been leaning on since its first phase.

Growth is a change of distances #

The thing to notice about a growing sheet is that growth is not a force applied to it. It is a change in what the sheet is.

Take two material points on a young leaf — two particular cells — and measure the distance between them along the surface. Let the leaf grow. Measure again. If every part of the leaf grew by the same factor, every such distance has been multiplied by that factor and nothing else has changed: the leaf is a scaled copy, and a scaled copy of a flat thing is flat.

If different parts grew by different amounts, the distances have changed by different factors, and the collection of all of them is a new object. It is called a metric — the rule that says how far apart any two nearby points are — and it belongs to the sheet rather than to the space the sheet is sitting in.

The whole of this essay is one consequence of that. A metric is not a shape, but it decides something about every shape that could carry it.

What a metric decides #

Gauss’s Theorema Egregium says that the Gauss curvature of a surface can be computed from the metric alone. It is the theorem cartographic-projection.com derives and the one this site has used since its expansion phase without re-deriving: curvature is intrinsic, so a creature living inside the surface and measuring only distances can work it out without ever seeing the surface from outside.

That is normally read in one direction — bend a sheet without stretching it and the curvature cannot change, which is why a flat sheet of paper can be rolled into a cylinder and not into a sphere.

Read it the other way and it says something about growth. A grown sheet has a new metric, so it has a new curvature, and it has that curvature whether or not anything in the world is prepared to accommodate it. A plane has curvature zero everywhere. If the grown metric’s curvature is not zero, no arrangement of the sheet in a plane can carry it — not a clever one, not a stiff one, not one held down.

The arithmetic, which is short #

For a disc growing isotropically — each little patch expanding by the same factor in every direction, with the factor varying only with distance from the centre — the metric can be written down in one line. If Ω® is the linear growth factor at radius r on the original sheet, the grown metric is Ω² times the flat one, and the curvature it carries is K = −(1/Ω²) Δ ln Ω, where Δ is the ordinary flat Laplacian and the logarithm is the part worth staring at.

Two readings follow immediately and both are the substance of the field.

If Ω is constant, its logarithm is constant, the Laplacian of a constant is zero, and K is zero everywhere. Uniform growth is flat, exactly, and that is the first panel above.

If Ω is not constant, K need not be zero — and the sign of K is the sign of −Δ ln Ω, which is a statement about where the growth is rather than how much of it there is. That is the whole of the next rung and it is worth its own essay.

The growth solver behind these figures computes the second derivative on a radial grid, which is a numerical operation and therefore a thing that can be quietly wrong. So it is checked against four closed forms it never uses: uniform growth, which must give exactly zero; the growth field of a spherical cap, which must give 1/ρ² at every radius; its hyperbolic counterpart, which must give −1/ρ²; and the quadratic profile, whose curvature at the centre is exactly −4a by a series expansion that shares no code with the grid.

Uniform is not the only flat case #

The formula deserves one more reading, because it says something stronger than the first panel does and the stronger statement is easy to miss.

Flatness is $K = 0$, which is $\Delta \ln \Omega = 0$. That is Laplace’s equation, and its solutions are the harmonic functions — a very much larger set than the constants.

So a sheet can grow wildly unevenly and stay perfectly flat, provided the logarithm of its growth factor is harmonic. Grow it as $\Omega = e^{bx}$, exponentially along one direction and not at all across it: the logarithm is $bx$, its Laplacian is zero, and the curvature is zero at every point of the sheet however large $b$ is.

That is not a technicality. It is a growth field in which one edge of the sheet has expanded by a factor of a hundred and the opposite edge not at all, and the result lies flat on a table.

Why the radial setting hides it #

The essay’s own figures never show such a case, and the reason is the restriction they are drawn under rather than an oversight.

The model is radially symmetric: $\Omega$ depends on $r$ alone. The harmonic functions with that symmetry are $A + B\ln r$, so the flat growth fields are $\Omega = C r^{B}$ — and every one of them with $B \ne 0$ is singular at the origin, which a disc containing its own centre cannot have.

So on a full disc with radial growth, the only flat case is the constant one, and the first panel is right within the setting it is drawn in. Remove either restriction — the radial symmetry, or the requirement that the sheet include its centre — and the flat cases become a whole function space.

That matters for reading the field’s conclusion. Things that grow unevenly cannot stay flat is true of radially symmetric growth on a disc and false in general, and the correct statement is that a sheet leaves the plane exactly when the logarithm of its growth is not harmonic.

It also sharpens what the measurement on a real leaf would have to establish. Marking two points and watching the distance is enough to show the growth is uneven; it is not enough to show the sheet must leave the plane, because unevenness is not the condition. What has to be shown is a non-zero second difference in the logarithm — which is a harder measurement and the honest one.

Why this is not the mechanical story #

The mechanical account of a ruffled leaf is that growth puts the rim in compression and compressed thin things buckle. That is a good account and it answers a different question.

It answers how — what the sheet does, on what timescale, into what shape, with what wavelength. The geometric account answers whether, and it answers it with no material constants at all. There is no modulus in the computation above, no thickness, no rate. A sheet a thousand times stiffer with the same growth field has exactly the same curvature and exactly the same impossibility.

This is worth being careful about, because the two accounts are easy to blur and the blurring runs in one direction. It is tempting to say the sheet buckles because it is soft. It buckles because it is a sheet with a non-zero curvature and there is nowhere flat for it to be. Softness decides what happens next.

The division of labour is real and this site respects it: what the material then does is elasticity, which is illustrated-physics.com’s and not folding’s, and how a section carries the resulting moment is structural-engineering-statics.com’s. Nothing in this field derives a bending energy, and where the geometry runs out the essays say so.

The negative case is the informative one #

The strongest thing about a geometric argument is what it forbids, and the reason to trust this one is that it forbids something checkable.

If a leaf’s growth field were uniform, it would be flat, and no amount of softness would ruffle it. That is not a hedge: the first figure is a computation of exactly that case and the curvature comes out zero to the last digit the grid can express. A flat growing sheet is not a sheet that happens to have avoided buckling; it is one whose metric has nothing to buckle about.

So the phenomenon is not “growing things wrinkle”. It is “things that grow unevenly cannot stay flat”, and the unevenness is measurable in principle on any real leaf by marking two points and watching the distance between them.

That distinction has a practical edge. It says which measurement would settle a disagreement about a particular organ, and the measurement is not a mechanical one. Nobody needs the tissue’s stiffness, its thickness or its rate of growth to predict whether it has to leave the plane; what is needed is a map of how much each part of it grew relative to its neighbours, which is a kinematic quantity obtainable by marking the surface and photographing it twice.

The same reasoning cuts the other way and is worth stating plainly, because it is the honest limit of the argument. A sheet observed to be flat has not been shown to have grown uniformly — it may have grown unevenly and then been held, trimmed, or torn. Geometry forbids a flat state for a non-uniform growth field only if the sheet is intact and unstressed, and a leaf pressed in a book is neither.

Where folding comes in, and it is not where it looks #

The reason this belongs on a site about folding is not that leaves fold, although some do. It is that folding and growth are the same geometry approached from opposite ends.

Folding keeps the metric and gives up the flatness. A crease pattern never changes a single distance measured along the paper — that is what “no stretching” means, and it is the constraint everything on this site follows from. What a fold changes is where the sheet sits, and the curvature stays zero everywhere except at the crease lines and the vertices.

Growth changes the metric and has flatness taken away from it. Nothing about the surface’s position is chosen; the distances change, and the consequences arrive whether or not anything wanted them.

Read side by side, that is a statement about who is doing the work. A folder decides the shape and inherits the metric. A leaf decides the metric and inherits the shape. The intermediate case, where the curvature is concentrated onto lines instead of spread over an area, is exactly what a crease is — and putting the two on one page is the fourth rung of this ladder.

The surfaces a flat sheet can reach, and the ones it cannot #

There is a second way to see the same wall, and it is one this site already had.

A surface a flat sheet can be bent into without stretching is called developable, and the classification is short: planes, cylinders, cones, and the surfaces swept by the tangent lines of a space curve. What every one of them has in common is zero Gauss curvature at every point, which is precisely the property the Theorema Egregium says bending cannot change. Everything else — every sphere, every saddle, every surface that curves two ways at one point — is outside what an unstretched sheet can reach.

So the folding side of this site and the growth side are asking the same question with the roles swapped. Folding starts with zero curvature and asks which shapes are reachable; the answer is the developables, plus whatever creases and vertices add. Growth starts with a shape requirement — a prescribed curvature — and asks whether a flat sheet can supply it; the answer is no, unless the growth was uniform.

The reason a leaf is more interesting than a map is that a leaf is allowed to leave the plane. A cartographer has to produce something flat and therefore has to pay for the curvature in distortion, which is the whole subject of another site in this fleet. A leaf pays for it in shape, which costs the leaf nothing and is often the point of the exercise — a ruffled edge has more surface exposed to the light than a flat one of the same span.

That asymmetry also explains why the two disciplines find different things hard. The map problem has no solution and the work is in choosing which error to accept. The leaf problem has too many solutions and the work is in finding out which one is taken, which is where the geometry hands over to the material.

What the picture cannot show #

Three things, and the third is the one that matters most.

The figures plot a curvature against a radius. They do not plot the shape the sheet takes, because that is a different and much harder problem — finding an isometric embedding of a given metric in three dimensions, which may have many solutions or none that are smooth. Nothing here solves one, and no figure in this field draws a surface as though it had been solved for.

The growth modelled is isotropic and radial: every patch grows the same amount in every direction, and the amount depends only on distance from the centre. Real tissue does neither. It grows more along some directions than others, along directions the tissue itself lays down, and on a schedule that responds to the shape already produced. The consequence is that a real leaf’s metric is a richer object than Ω®, and the part that survives the simplification is the part that is geometry rather than biology.

And nothing here was measured on a plant. That is the rule this whole field runs under and it is stricter than it needs to be: a biology figure on this site draws a fold or a metric this repository computed, or it is not drawn. What the essays supply is the organism that motivated the geometry, named in prose, cited where it is somebody’s measurement, and never presented as though a generator had established it.

The idealisation, named #

Every essay here states the thing its figure is pretending, and this one has an unusually clean version.

The sheet is a surface. It has no thickness, so there is no distinction between the growth of its top and the growth of its bottom, and a real leaf’s most common trick — growing one face faster than the other — is invisible to the model entirely. That mechanism produces curvature too, by a completely different route, and it is what makes a bimetallic strip bend and a drying pine cone open.

The model is worth having anyway, because the two mechanisms are separable. Differential growth in the plane forces curvature by the argument above, with no reference to thickness. Differential growth through the thickness forces curvature by bending, with no reference to the metric. A real organ usually does both, and untangling them on a specimen is a research problem rather than an observation.

Where this ladder goes next #

Three directions, and they are genuinely different questions.

The immediate one is the sign: given a growth field, does the sheet dome or does it ruffle? That turns out to be decided by where the growth is concentrated rather than by how much of it there is, and one dial takes the sheet through both.

Then the question of what the sheet does with the curvature it has been given. A metric fixes the curvature; it does not fix the shape, and a fixed excess of length admits a whole family of ways to spend it. Choosing among them is where geometry stops.

And then the comparison this site is unusually placed to make, which is between a crease and a growth field carrying the same curvature. The surprising part is not that they can be matched. It is that the matched pair are completely different surfaces — one flat everywhere except at a single point that a walker could miss entirely, the other curved at every point with nothing anywhere that is special — and the total, the only number the two share, tells a reader nothing about which they are looking at.