Where this collection's thread model now stands

This collection’s model of a thread was a curve with a bending stiffness. It minimised the integral of the square of its curvature subject to where its two ends had to be, and everything mechanical the site has computed came out of that.

A thread has two stiffnesses and it has a thickness, and the model had one of the first and none of the second.

Both were addressed in one phase. Neither reached the question it was built for. The accounting is worth more than either.

What was added #

A torsional rigidity, with its own bracket, and the observation that the bracket is identical to the bending one — so the ratio C/B is 2G/E at both ends and everywhere between.

A shear modulus table for the ten fibres the collection carries, with one row whose answer was known in advance.

Two integrals: the Gauss writhe and the Gauss linking number, each a definition rather than an approximation, each checked against arrangements whose answers are known by inspection.

And a distance measurement over the model’s own solved fabric, which had never been made.

That is a table, a ratio and three routines. It is a small addition by the standards of the ladders that preceded it, and it produced more than either of them.

What it settled #

A yarn’s two stiffnesses are in a fixed ratio, and the ratio is knowable when neither stiffness is.

A slack yarn’s snarl is a threshold and a length, and the length is bracket-free.

The trade’s folding rule is a surface-angle condition rather than a torque balance, and the surface rule reproduces all three of its numbers where the torque balance reproduces none.

A woven cloth’s threads are unlinked and a knitted fabric’s courses are linked once per wale, and the two failure modes everybody knows follow from those two integers.

A knitted fabric’s own geometry demands a flattened yarn, and the flattening follows the tightness factor.

Contact between courses does not explain the extension ceiling, which was the recorded diagnosis and is wrong.

And a woven cloth’s geometry demands almost no flattening, so the flattening a cloth’s section shows is a memory of the loom.

Seven, and none of them was the target of either ladder.

What it did not settle #

Spirality, which the torsion ladder was built for. The mechanism needs the fabric to trade twist for writhe, and the model’s course has no writhe — exactly nought, by a mirror symmetry — so the nine degrees per unit of twist factor remains fitted.

The extension ceiling, which the contact ladder was built for. Stopping the courses overlapping moves it seven per cent against a gap of two thirds.

The wale-direction curl, untouched.

And the torsional share of a loop’s energy, which needs a solve whose configuration space includes the material’s rotation rather than only the centre line’s shape.

Both ladders reached their targets and found the obstacle was somewhere else. In both cases the obstacle was the same one.

The one obstacle #

Five findings, recorded over three earlier ladders in four ladders, turn out to be one modelling decision.

The interlacing was declared to be a point where two centre lines pass one yarn diameter apart. That decision produced the fabric’s thickness, which is this collection’s most falsifiable knitted result, and it cost:

the course’s writhe, because a near miss has no handedness; the linking number between adjacent courses, because a point cannot link; the extension ceiling, because two threaded loops cannot separate and two passing ones can; the wale-direction curl, because the solve runs away to a configuration threaded neighbours would block; and the two half periods at a crest, which run within a fiftieth of a diameter because nothing is drawn between them.

One sentence, five symptoms, and it took a quantity that returns an integer to name it.

What the model can be trusted for #

A reader wants a rule, and this work supplies a clean one.

Trust any quantity that is an integral over one segment. A stitch’s bending energy, the contact force, the fabric’s thickness, its rigidity in both directions, its thermal resistance, its fibre fraction, the pressure a cuff applies. Each is decided by where a half period’s two ends are and how much yarn runs between them, and threading the curve through another one changes none of them to first order.

Trust any comparison at fixed topology. A rib against a jersey, a tight fabric against a slack one, one relaxation state against another: the omission is the same on both sides and cancels.

Distrust anything about the fabric holding together. The extension ceiling, the run, the curl about the wale axis, the spirality.

And treat every null result with suspicion. Two of the collection’s knitted results are exactly nought, and only one of them is nought for a reason a fabric shares.

Four rules, and the first two cover most of what the collection has computed.

What this work changed about how to work #

Three habits came out of it and all three are cheaper than the results.

Ask whether a model’s output can be built. Every gate on this site reads a fabric and asks whether something about it is right; not one asks whether the object is realisable. A distance calculation found a fabric occupying its own space, four ladders after the geometry was written.

Convert every threshold into a unit somebody would notice it in. A critical tension of 0.37 millinewtons is a number nobody can disagree with; 1.9 metres of the yarn’s own weight is a claim about something everybody has seen, and it settled a bracket by being disagreeable.

And ask of every null result whether it is zero for a reason the subject shares. A writhe of six parts in a million reads as a converged calculation of something negligible. It is a calculation of something forbidden.

What is owed #

The shortfall list, which is what a ladder leaves behind and what the next one reads.

The asymmetric solve. Two half periods that are not mirror images, with a closure condition between them. It stays a smooth equality-constrained problem, it would give the course a chirality, and it fixes three of the five symptoms. It is the cheapest item on the list and should be done first.

The contact solve, which is a different class of problem — a non-penetration inequality active on part of an unknown domain — and which fixes the other two. It costs fifteen per cent of a stitch’s yarn and a bend at the tightest radius in the fabric, and it should not be done for the thickness’s sake.

A knitted fabric’s shear stiffness, which nothing here has and which the spirality question needs alongside the writhe.

A torsion term in the loop’s own solve, which needs a configuration space that includes the material frame.

The set fraction, which multiplies every torque on the torsion ladder and which the snarl test would measure.

And a moisture argument to the diameter function, which would let every table on the site be computed at a stated moisture content rather than a dry basis.

Six items. One is half a day, three are a day or two each, one is a research programme, and one is a measurement rather than a calculation.

This work in one sentence each #

For a reader who wants the two ladders compressed rather than summarised.

The torsion ladder added a second stiffness, found that its bracket is the bending one exactly and its ratio therefore knowable, used the ratio to settle a snarl and to refute a folding rule, and could not reach the spirality it was built for because the model’s fabric has no chirality.

The contact ladder gave the thread a thickness, found that the fabric it had been computing with occupies its own space, read the shortfall as a prediction of a parameter the site had swept for years, ruled out the extension ceiling’s recorded diagnosis, and could not repair the model because a non-penetration constraint is a different class of problem.

And together they found that four everyday observations place a yarn at the free end of its own stiffness bracket, that five recorded shortfalls are one modelling decision, and that a woven cloth’s geometry is nearly self-consistent where a knitted one’s is not — because one crosses its threads and the other runs them alongside.

Three sentences, two ladders, and the third is the one neither would have produced alone.

What was counted, and how #

Every number in this rung is recomputed from this collection’s own functions rather than quoted from its essays, so a change anywhere would show here.

The bracket identity is checked across counts, fibres and packing factors at both ends, to a part in ten thousand million million.

The linking integral is checked against four arrangements whose answers are known by inspection, to four parts in ten thousand at four hundred segments a curve.

The overlap is checked as a band rather than as an inequality, so that a repaired model fails it — which is the signal that the ladder’s subject has been dealt with.

And the snarl threshold is checked at its degenerate case and as an exponent, because a relation that holds at one pair of points holds for any monotone function.

What a reader coming to this collection should take #

A rung that closes a ladder should say what it says to somebody who does not intend to read the rest of it, and this one has a short answer.

A knitted fabric’s geometry and forces are computed here to a standard nothing else in the literature reaches, from first principles, with every input named and every bracket carried. Thicknesses, contact forces, rigidities, warmth, extension curves.

All of them are computed on a fabric whose loops are not threaded through one another, which is invisible in every one of those numbers and fatal to a different set.

The two sets are cleanly separable by one rule: local integrals survive and holding-together does not.

And the fabric’s own arithmetic now predicts one thing everybody else has fitted: how flat its yarn has to be.

That is a fair statement of a collection eighteen phases into a subject: better than the literature on the questions it has asked, silent on a class of question it had not framed, and honest about which is which.

A note on the shape of this work #

This work took two questions rather than one and it is worth recording that the arrangement paid.

The torsion ladder needed a flattening to correct its headline ratio and would not have had one. The contact ladder needed a chirality argument to interpret its crest overlap and would not have had one. The bracket rung needed four observations and two of them came from each side.

Neither ladder alone produces its own best results, and the cross-products cost nothing.

That is an argument for pairing questions when work is planned, and it runs against the instinct — which is to go deep on one thing, because depth is what produces a ladder.

Both are right. The depth produces the machinery and the pairing produces the results, and work with only one of them gets half of what it could.

Where the model stops #

The fibres are straight and parallel to the yarn’s axis. That idealisation runs through both brackets and is inherited rather than introduced.

The section is circular in every solve, and this collection’s own contact ladder says it is not — which puts a correction of about twenty per cent on the torsion ladder’s headline ratio in the direction the loop actually bends.

Nothing is re-solved under contact. Every measurement on the contact ladder is a diagnosis of a free solve.

And the yarn is elastic in torsion for ever, which it is not: a set yarn has no torque at any twist, and setting is the trade’s whole answer to liveliness.

Two things this work got wrong on the way #

A ladder that reports only its findings is reporting half of itself, and two mistakes are worth recording because both are the kind that would be made again.

A closure was treated as free. The first attempt at a linking number joined a flat course’s two ends with a straight chord and got minus 0.037 — small enough to look like convergence, large enough to be a hundredth. A closure is part of the curve, and a quantity that must be an integer and comes back at a few hundredths is reporting its closure rather than its topology. The fix was to use a tube, which is a real fabric rather than a device.

And an empty search was read as a ceiling. The contact ceiling calculation asked for a round yarn at every extension and found nothing feasible anywhere. That is not a ceiling of nought; it is the relaxed fabric’s own overlap restated as an empty result, and recognising it took an afternoon. The fix was to ask for the flattening the fabric actually demands.

Both are the same mistake in different costumes: a result that looks like an answer and is a restatement of the question. Neither would have been caught by a tolerance, because both produced plausible numbers.

The generalisation #

This collection’s most transferable finding is not a number and it is worth ending on.

A model’s missing piece is often not in the same category as its symptom.

Both ladders were built to supply something material — a second stiffness, a thickness — and both found that what was missing was topological: an arrangement rather than a constant. A fabric of folds has all the geometry of a knitted fabric and none of its integrity, and no material constant supplies the difference.

That is a hazard specific to a collection that computes well. A model with good arithmetic and a good energy will answer every question that is about arithmetic or energy, correctly, and will be silent rather than wrong about anything that is about arrangement — so its failures look like small numbers rather than like errors.

The instrument for that is an invariant: a quantity that takes an integer value, survives every constant, and cannot be nearly right.

This collection had one — the connected-components count that decides whether a draft makes one cloth — and used it for eighteen phases without recognising it as a member of a family. It has three now, and the oldest of them is the reason the site exists.

The three instruments this work leaves behind #

Machinery outlasts results and this work leaves three pieces of it, each cheap and each reusable on questions nobody has asked yet.

The Gauss linking integral. Two closed curves in, one integer out. It works on any structure whose paths this collection can produce, and it has been applied to two of them. A rib, an interlock, a warp knit, a braid and a leno are all waiting.

The minimum-approach measurement. Two sampled curves in, a distance out. It answers “does this fit together” for any geometry the site can lay out, which is all of them, and it has been asked of two.

And the shear-modulus table, with its one row whose answer is known in advance — which is a pattern for how to build a table of constants that can be checked rather than trusted.

None of the three took long to build and all three found something on their first use. That is a reasonable expectation for an instrument in a collection this size: there is a great deal here that has never been measured, because nothing measured it.

Who found it, and when #

Kirchhoff wrote the rod equations in 1859, Greenhill the stability criterion in 1883, Gauss the linking integral in 1833, Călugăreanu the twist–writhe identity around 1960, and Euler the capstan relation in 1775. None of the physics in this work is new.

Peirce’s cloth geometry is from 1937 and Munden’s knitted constants from 1959, and both are inputs rather than results here.

What is this collection’s own is the application: putting these instruments to its own model, on its own numbers, and reporting what they said about it — including the parts that were uncomfortable, which are the parts worth having.

Where the ladder goes next #

Whoever writes the next ladder reads this list. The asymmetric solve is the first item and it is small; the flattening prediction wants a set of sections to check it against; and the snarl threshold wants a hanging loop, a balance and an afternoon.

None of those is a research programme, all of them are specified, and each closes something this work left open rather than opening something new — which is what a shortfall list is for.