A hemisphere costs one full turn

A flat piece of cloth cannot cover a sphere. That is Gauss’s theorem and this site takes it as given — the theorem belongs to the geometry of surfaces and not to cloth, and nothing here re-derives it. What the theorem hands a pattern cutter is a number: how much angle has to be removed from a flat piece before it will lie on a curved form.

For a spherical cap of half-angle φ the total curvature is 2π(1 − cos φ). Read that as a deficit the flat pattern must supply and the answer for a hemisphere is 2π: one full turn, 360 degrees, whatever the radius.

A hat for a child and a dome over a stadium need the same total dart angle. The area differs by a factor of a million and the angle does not differ at all.

Why the size drops out, and what that means at a cutting table #

The independence from radius is the least intuitive part of the result and the most useful.

Curvature is an angle per unit area — the units are inverse area — so a cap’s total curvature is a pure number. A sphere of twice the radius has a quarter of the curvature at each point and four times the area, and the product is unchanged. A hemisphere is a hemisphere.

So the dart angle is a property of the shape and not of the garment’s size. Scaling a pattern up leaves the darts’ angles alone and only lengthens them, which is why grading a pattern across a size range is mostly a matter of lengths — and why a shape change, however small, is a different pattern rather than a scaled one.

It also gives a check any pattern cutter can apply. Add up the angles of every dart, seam-shaping and gore in a pattern for a hemispherical form; the total must be 360°, or the piece is not a hemisphere. Two darts of 180° would do it and would be unwearable; twenty of eighteen degrees would do it and would look like a swimming cap. The total is fixed and the distribution is the design.

The fabric can supply some of it by shearing #

A sheet of paper has no way of absorbing a curvature deficit and must be cut or creased. A woven cloth does: it is a trellis, and shearing it changes the angle between its two thread systems without stretching anything at all.

That is the mechanism this site has computed since its foundation, and it is why a fabric lies on a sphere at all. Laid over a curved form by the fishnet construction, a cloth’s cells close where the surface has positive curvature and open where it has negative — the sign of the shear following the sign of the Gaussian curvature, which this site asserts rather than describes.

But the shear runs out. The threads jam against their neighbours at the locking angle, which is arcsin(d/p) of included angle and therefore a function of the cover alone — and past it the cloth cannot lie on the surface however it is persuaded. It buckles instead.

So the sphere divides into two zones with a computable boundary: inside some cap the cloth drapes with shear alone, and beyond it a dart is the only option left.

The numbers, and where they put the trades #

At a cover of 0.3 or 0.5 — an open plain weave, a loose muslin — the net reaches past the equator before anything jams, so a hemisphere needs no dart at all. That is not a modelling artefact: a loose cloth really can be pulled over a ball, which is what a tailor is doing when they ease a sleeve head or shrink a shoulder with an iron.

At a cover of 0.7 the cloth locks at 46° and drapes to 82°, leaving 53° to be removed. At 0.8 it leaves 100°. At 0.9 — a densely set gabardine, a coated cloth, a laminated fabric — it leaves 148°, and the pattern is doing most of the work.

Read down that column and the practices of three trades fall out of one quantity.

Tailoring works with cloths in the middle of the range and uses both mechanisms: darts for the bulk of the deficit and easing, shrinking and stretching for the rest. A tailor’s iron is a device for exploiting the shear.

Composite draping works with fabrics that have very little shear left — a preform at high fibre content is nearly jammed — so a near-net-shape preform is darted, gored or specially woven rather than persuaded. The wrinkle map this site computes for a preform is exactly this arithmetic with the dart option removed.

Coated fabrics and membranes have almost no shear at all, because the coating carries it, so they are cut into panels whose boundaries are computed. That is the same trade as the previous rung’s compensation, and it is the extreme of this column.

A dart and a seam are the same operation #

There is a piece of vocabulary hiding a unification here, and pattern cutters use both words without usually saying they are one thing.

A dart removes a wedge from the interior of a piece and sews the cut edges together. A shaped seam — a princess line, a raglan, a gore in a hat or a sail — joins two pieces whose edges are not the same curve, so that the join itself takes up angle. The second is a dart that has been walked out to the edge of the piece and turned into a seam.

For the accounting above they are identical: both remove angular deficit, and the total each removes is the difference between the angles their edges subtend. So the 360° a hemisphere demands can be paid in darts, in gores, in shaped seams, or in any mixture — and a six-gore cap and a cap with six darts are removing the same total in different places.

That is why the total is the useful check and the distribution is the craft. A hat made of six gores has its seams doing the work and no interior darts; a bra cup has darts and a seam; a tailored shoulder has a seam, two darts and a great deal of easing. Add the angles up and the answer is fixed by the shape of the head, the body or the sail.

Knits and felts pay in a different currency #

The two mechanisms above — cut it out, or shear it away — are what a woven cloth has. Two other constructions on this site pay the same debt differently, and putting them beside each other is the clearest statement of why construction rather than material decides what a fabric can do.

A knit accommodates curvature by distorting its loops. A loop is not a rigid cell: it can lengthen, narrow, and rob length from its neighbours, so a knitted fabric absorbs a curvature deficit continuously and over a wide range — which is why a knitted hat needs no darts, why a knit recovers where a woven does not, and why the whole shaping vocabulary of knitting is increases and decreases rather than darts. The debt is paid by the loop’s geometry rather than by an angle removed.

A felt has no threads to shear or cut. Its fibres are entangled and its coherence comes from the milling ratchet — and needs water to run at all — so it can be moulded: wetted, pressed over a form and dried, with the fibre network rearranging locally. A felt hat is not cut and darted at all — it is blocked — and that is the oldest hat-making technique there is.

Three constructions, one geometric requirement, three different answers to it. The requirement does not care: 360° for a hemisphere in every case. What differs is where the angle goes, and it goes to the mechanism the construction happens to have.

What was counted, and how #

The curvature integral is quoted and used, not derived. It is 2π(1 − cos φ) for a spherical cap and this site’s arithmetic reads it off; the theorem behind it belongs to the geometry of surfaces, and the boundary is recorded in the fleet’s own claim registry rather than left implicit.

What is computed here is the split. The drape is drape.js’s pin-jointed net, laid over the sphere by the fishnet construction, and every segment is asserted to be exactly one pitch — so the shear field is solved rather than fitted. Then the first cell whose shear passes the locking angle fixes the polar angle at which the cloth gives up, and the deficit inside that cap is what the shear absorbed.

Three assertions guard the column: a more closely set cloth drapes over less of the sphere, it needs more dart, and the total the cap demands is the same in every row to within a part in 10⁹. The third is the one that would catch a curvature integral quietly picking up a radius, which is the error the headline result is most vulnerable to.

How many gores, and why a gore beats a panel #

The total is fixed and the distribution is the design, and one distribution is common enough to be worth costing: cutting the form into gores or panels and letting the seams carry all of it. That is how a balloon, a globe, a parachute and a hat crown are made, and the question a maker actually asks is not how much angle but how many pieces.

The answer is a sagitta. A flat piece cannot follow a curved surface, so it stands off it, and how far is the geometry of a chord. A gore spanning an azimuth of 2π/g on a sphere of radius R has, across its width, a chord subtending 2π/g at the centre, so the flat piece departs from the sphere by

R(1 − cos(π ÷ g)),

which for a dozen gores is 3.4 per cent of the radius and for twenty-four is 0.9. On a hat crown of a hundred millimetres’ radius that is three and a half millimetres and nine tenths of one — the first visible as facets, the second not.

Inverting it gives the piece count directly. For small angles the cosine expands and the deviation is π²R/2g², so

g = π √(R ÷ 2t)

for a tolerance t: twenty-three gores to hold a hundred-millimetre crown to a millimetre, thirty-two to hold it to half. The count goes as the square root of the tolerance, so halving the allowed facet costs only forty per cent more seams — which is why gored construction stays practical as it is made smoother, and why nobody stops at four gores except deliberately.

Now the comparison that makes gores interesting rather than merely arithmetical. A panel — a piece that is short in both directions, like a football’s — approximates a cap of half-angle α and stands off by Rα²/2, and N panels each cover a solid angle of 4π/N, which makes α² = 4/N and the deviation 2R/N. So a panelled sphere’s smoothness improves as 1/N and a gored sphere’s as 1/g².

Equate them and N = 0.41g². Twenty-two gores hold a sphere as smoothly as 196 panels would.

That is a large factor and it has a plain cause. A gore is long and thin, so it has to approximate the sphere in one direction only — along its length it can simply bend, which costs nothing, because bending without stretching is free. A panel has to approximate in both directions at once and can bend in neither without leaving the surface. A near-developable strip is a much better piece than a compact patch, and the whole advantage is that one of its two curvatures has been made small enough to bend away.

Which explains a division of practice that otherwise looks like tradition. Balloons, parachutes, globes, spinnakers and hat crowns are gored, because smoothness is what they are for. Footballs and pressure vessels are panelled, because a closed shape sewn from long tapered pieces has all its seams meeting at two poles — a weak, thick, crowded joint — and a game ball would rather have twelve small faults spread evenly than one large one at each end.

So the piece count is not chosen for smoothness alone; it is chosen against where the seams end up. The arithmetic above says what smoothness costs, and the choice between the two shapes of piece is made on the seams.

What the picture cannot show #

Every figure in this essay draws a net or a bar, and neither of them is the thing being claimed.

The claim is about an integral — a total curvature — and a total is not a local quantity that can be drawn at a point. The drape figures show where the shear is large and where it is small; they cannot show that the sum of the deficits over the whole cap is 2π, because that number is not visible anywhere on the surface. The bars show how the total divides between shear and dart; they cannot show why the total is what it is.

So this rung is unusual on a site whose rule is that an essay exists because a figure explains something better than prose can. Here the figure explains the division and the prose carries the theorem, and saying which is which is the honest arrangement. A figure purporting to show Gauss–Bonnet on a cap would be a decoration, and drawing one would also be taking ground this fleet has assigned elsewhere.

What the figures do carry, and prose could not, is the pin-jointed net’s own consistency: every segment one pitch, the shear signed by the curvature, the cells visibly closing towards the pole. That is a solved field rather than an illustration, and it is what makes the split between shear and dart a computation instead of a story.

Where the model stops #

The cap is a sphere and a body is not. Real forms are compound: a shoulder is closer to a saddle, a bust to a cap, a hip to something with both signs of curvature in it. The total curvature of a compound form is still the integral of its curvature, so the accounting survives; what does not survive is the clean split into one cap with one radius.

The fishnet is one construction among several. A cloth laid over a form can be positioned in many ways, and the fishnet — two threads laid along geodesics from a chosen point, the rest constructed from them — is the classical one and gives a solution rather than the solution. A different starting point gives a different shear field and a different locking radius. The direction of every result here survives that; the exact cap angle does not.

A dart is not a pure angle. Removing a wedge changes the piece’s boundary, adds a seam that has thickness and stiffness, and moves fabric that has already been shaped. Two patterns with the same total dart angle can fit quite differently, which is the whole craft of pattern cutting and none of it is here.

And the locking angle is a geometric jam, not a mechanical one. Real threads compress, coatings resist, and a cloth pushed past its jam wrinkles gradually rather than stopping dead. The boundary computed above is where the kinematics runs out, and a real fabric’s usable range ends a little before it.

The check has a practical form worth stating for a reader who cuts patterns rather than code. Sum the angles, and compare with the shape. A cap needs 2π(1 − cos φ); a full sphere needs 4π, which is 720° and is why a ball is made in panels rather than shaped from one piece; a cylinder needs nothing at all. If a pattern’s darts, gores and shaped seams do not add to the figure the form demands, the pattern is not that form — and the discrepancy is either ease, which is deliberate, or an error, which is not.

Who found it, and when #

The theorem is Gauss’s, from 1827, and the global form is Gauss–Bonnet. Neither is this site’s and neither is re-derived here: cartographic-projection.com holds them in this fleet, and the licence under which cloth over a curved form may use them is recorded in the registry rather than assumed.

The application is much older than the theorem. Darts, gores and gussets are medieval tailoring, and the empirical fact that a hat needs a certain amount taken out of it whatever its size was known to everybody who made hats. What the theorem adds is that the amount is exactly one turn and that no fabric, cut or eased, changes it — which converts a craft rule into an accounting identity, and gives a pattern cutter a total to check against.

The fishnet construction for laying a net over a surface is from the composites literature of the 1980s and 1990s, where the question was whether a preform could be draped rather than how a garment should be cut. It is the same construction either way, and the two trades appear not to know they share it.

Where the ladder goes next #

Cutting a pattern leaves edges, and edges have to be joined. A seam is a row of stitches, and what holds the cloth beside it is not the stitches at all: it is friction at the crossings, accumulating multiplicatively along the thread. The next rung computes it with the capstan the fancy weaves brought, and finds that a satin gives a seam almost nothing to hold with.