The ratchet that makes wool felt

Everything else in this field is reversible in principle. A relaxed cloth can be stretched back. A calendered yarn recovers some of its roundness. A raised nap can be sheared off and the cloth underneath is the cloth that went in.

Felting is not like that. A woollen jumper that has been through a hot wash is a different object, permanently, and no treatment returns it. That is a strong claim about a process made of nothing but rubbing, and the reason is a mechanism worth having: wool is a ratchet.

The asymmetry is in the surface, not in the motion #

A wool fibre is not a smooth cylinder. Its outer layer is made of overlapping cuticle cells — scales — arranged like roof tiles, all pointing towards the tip. Under a microscope the fibre looks serrated, and the serrations all lean the same way.

The consequence is that the friction between two wool fibres, or between a wool fibre and anything else, depends on the direction of sliding. Moving root-first, the scales lie down and the fibre slips. Moving tip-first, the scales catch. Measured values run about 0.2 to 0.25 with the scales and 0.4 to 0.6 against them, and the ratio between the two — the directional friction effect, or DFE — is the quantity everything here turns on. Wet the fibre and the gap nearly doubles.

Now agitate the fibre. Push it back and forth in a mass of other fibres, wet and warm so they can move at all. The push is symmetric: the same force each way, the same number of times.

But the displacement is not, because friction resists it unequally. Each cycle the fibre travels further root-first than it returns tip-first, and the difference is left behind. Repeat it a few hundred times and the fibre has migrated a long way in one direction, with no directional force having been applied at any point.

Why this is a ratchet and not just friction #

The word matters, because ordinary friction does not do this.

A ratchet is a device that converts undirected motion into directed motion by making one direction cheaper than the other. It needs no directed input, and it is not reversible: running it backwards is not a matter of applying the opposite force, because the opposite force meets the expensive direction.

Wool’s version has all of those properties. And it has the crucial one: the state it reaches is not one the system was pushed towards, so releasing the push does not release the state. A stretched spring returns; a ratcheted pawl does not. The fibres are tangled where they ended up, and there is no restoring force anywhere in the system pointing back towards where they started.

That is why felting is irreversible in a stronger sense than the other operations in this field. It is not that undoing it is difficult. It is that there is no direction to undo it in.

Putting a number on it #

The model here is the simplest one that has the property, and it is labelled as such: the distance a fibre moves per stroke is taken as inversely proportional to the friction resisting it, so a full cycle nets

displacement ∝ 1/μ_with − 1/μ_against

At the middle of the reported range — 0.22 with the scales and 0.48 against — the forward stroke is 4.55 units, the return is 2.08, and the net is 2.46 per cycle. Thirty-seven per cent of the total motion is rectified rather than cancelled.

That is a shape, not a distance. The units are arbitrary and the model does not predict millimetres. What it does predict is how the net migration responds to the two coefficients, and that is the part with consequences: the rectification depends on the difference of the reciprocals, which means it is far more sensitive to the small coefficient than to the large one.

A fibre that slides very easily one way and moderately the other ratchets hard. A fibre with both coefficients high ratchets weakly even if the ratio between them is large. That is not obvious from the DFE ratio alone, and it is the reason a treatment that lubricates a wool fibre uniformly can increase felting rather than reduce it.

Where the friction machinery comes from #

This site already had a friction argument, and it is worth saying how the two relate because they are different.

The capstan equation governs how hard a thread is held by a wrap around another thread. It has one coefficient and an angle, and the holding force is exponential in their product. The the compound-cloths field used it to supply the quantity the integrity criterion structurally cannot see.

The ratchet is a different use of friction: not how hard something is held, but how far it moves when pushed. The capstan cares about the magnitude of μ and the ratchet cares about the difference between two μs. A fibre with a large μ and no directionality is very well held and does not felt at all. A fibre with a small μ and strong directionality felts readily and holds nothing.

So the two quantities are independent, and wool happens to have a useful amount of both — which is a large part of why it is the fibre it is.

What was counted, and how #

ratchet() computes the two stroke distances, their difference and the fraction of the total motion that is rectified, and it asserts one thing that could fail: that the against-the-scales coefficient is at least the with-the-scales one. A pair of coefficients the other way round describes a fibre whose scales point the wrong way, and the function refuses it rather than returning a negative migration that would read as a small one.

That assertion needed a tolerance, and the reason is worth recording because it is the kind of thing that turns a correct model into a failing gate. shrinkResist closes the friction gap by a fraction, and a fraction of exactly one lands a floating-point whisker below the other coefficient — so an exact comparison refused the single case the whole treatment exists to produce. The tolerance is 10⁻¹² and it is commented.

The shrinkResist family asserts the two ends: an untreated fibre ratchets forward, and a fully treated one nets exactly zero, not approximately zero. That distinction is the model’s whole claim about what the chemistry has to do.

What the ratchet needs from the fabric #

The mechanism is a fibre mechanism and it needs the fabric to let it run, which is why the same wool felts in one cloth and not in another.

A fibre can only migrate if it can move, and its ability to move is decided by how tightly it is held: by the yarn’s twist, by the sett, and by how much the fabric is compressed during the agitation. A hard-twisted yarn in a densely set cloth felts far less than a soft woollen-spun yarn in an open one, with identical fibre.

So felting propensity is a property of the assembly and not of the material, and the two halves multiply: a felting-prone fibre in a felting-resistant construction is a stable fabric, and so is the reverse. That is the practical reason worsted suitings are stable and knitted woollens are not, and it has nothing to do with the wool.

The ratio is not enough, and the arithmetic says so exactly #

The directional friction effect is quoted throughout the wool literature as a ratio — the against-the-scales coefficient over the with-the-scales one — and treatments are assessed by how far they close it. The model above says that is the wrong statistic, and it says so in a form sharp enough to be checked.

A difference of reciprocals is homogeneous of degree minus one. Multiply both coefficients by the same factor k and the net migration is divided by k, while the ratio between them does not move at all. So

at a fixed directional friction ratio, the migration is inversely proportional to the absolute level of friction.

Two fibres at 0.22 and 0.48, and at 0.44 and 0.96, have identical DFE and the second felts at half the rate. A number that cannot tell those two apart is not the number that predicts felting.

The immediate consequence is the one this essay already asserted and can now price. A uniform lubricant multiplies the migration by the reciprocal of what it multiplies the friction by: halving both coefficients doubles the net motion per cycle, exactly. A softener applied to make a woollen handle better, if it lubricates both directions equally, has made the fabric felt twice as fast and has left every DFE measurement on it unchanged.

And the reverse. A treatment that adds to both coefficients — a resin that roughens rather than a polymer that masks — helps substantially: adding a tenth to each, from 0.22 and 0.48 to 0.32 and 0.58, cuts the migration by forty-three per cent with no change to the scales at all.

Which end of the gap to close #

The same derivative settles a question a formulator actually faces: given a fixed amount of chemistry, is it better spent raising the easy direction or lowering the hard one?

The sensitivity of a reciprocal is one over the square, so the migration responds to the with-the-scales coefficient as 1/μ² = 20.7 and to the against-the-scales one as 4.34. The small coefficient is nearly five times as influential, and the two are not symmetric in any sense at all.

Put numbers on the same intervention applied at each end. Raising the with-scales coefficient by five hundredths — 0.22 to 0.27 — cuts the migration by thirty-four per cent. Lowering the against-scales coefficient by the same five hundredths — 0.48 to 0.43 — cuts it by ten.

Closing the gap from below is worth three times closing it from above, for the same movement in a coefficient, which is a design instruction rather than an observation. And it is the opposite of the intuition the word descaling suggests: the scales are the prominent feature, so the obvious target is the resistance they cause, and the arithmetic says the money is in the direction where they do nothing.

That reading is at least consistent with what the successful treatments do. A polymer that masks the surface does not remove the scales’ contribution selectively — it puts the same layer over the whole fibre, so the two coefficients converge on a common value from both ends at once, which collects the large sensitivity as well as the small one. A treatment that only smoothed the scales’ overhang would be spending everything at the cheap end.

None of that is a prediction the model can be trusted on quantitatively, since the stroke law is a stand-in. What survives is the ordering and the exact scaling, both of which follow from the shape of the expression rather than from its calibration — and the exact scaling is the one that says a ratio is the wrong thing to report.

Where the model stops #

This is the most heavily modelled thing in this field and the qualifications are correspondingly heavy.

Displacement inversely proportional to friction is a stand-in, not a law. A fibre in a wet fibrous mass under compression is not a block sliding on a plane. Its motion is limited by entanglement with its neighbours, by its own bending stiffness, and by the compressive force at each contact, none of which is here. What the model preserves is the sign and the sensitivity, and nothing more.

Nothing about the process appears. Real felting depends on temperature, on pH — alkaline conditions swell the fibre and open the scales, which is why soap is part of the recipe — on the water content, and on the kind of mechanical action. All of that is absent.

The fibre is treated as an isolated object. Felting is a collective phenomenon: fibres migrate through one another and entangle, and the entanglement is what produces the fabric change. That collective half is the next rung’s subject and it needs percolation rather than friction.

And the scale geometry is not modelled at all. The two coefficients are inputs standing in for a surface structure that has a pitch, a height and an angle, all of which vary between wools and none of which is represented.

Why the mechanism was worth finding #

There is a version of this essay that says wool felts because its fibres have scales that hook together, and that version is wrong in a way that matters.

Hooking would be a static explanation: the fibres are rough and they catch. If that were the mechanism, then felting would be a matter of how much force is applied, and enough force in the reverse direction would undo it, and a smooth fibre would not felt while a rough one would regardless of directionality.

None of those follow from the ratchet. The ratchet needs the roughness to be directional, which is a much narrower requirement, and it predicts that a fibre with symmetric roughness will not felt however rough it is. That is testable and it is true: cotton fibres are far from smooth, they tangle readily, and cotton does not felt.

It also explains why the treatment that works is the one that works. Descaling by chlorination, or masking the scales with a thin polymer, does not make the fibre slippery — it makes it equally slippery both ways, and the ratchet stops. A treatment that merely reduced friction overall would leave the ratio intact and the mechanism running.

Why the effect needs water and warmth #

The model has two coefficients and no environment, so it is worth saying what the environment does — because a woollen jumper does not felt in a cold rinse and does in a hot wash, and nothing in the arithmetic explains that.

Water swells the fibre and lifts the scales. A wet wool fibre takes up around a third of its own mass in water, its diameter rises, and the cuticle cells lift away from the shaft. A lifted scale has a larger overhang, so the against-the-scales coefficient rises while the with-the-scales one does not — the gap widens and the ratchet works harder.

Alkali does the same thing further. Soap is alkaline, and alkaline conditions swell wool more than neutral water does, which is one reason soap felts more aggressively than a neutral detergent.

Warmth reduces the fibre’s bending stiffness, so a given mechanical action moves fibres further per cycle. That is a change in the stroke length rather than in the friction, and it multiplies the whole effect.

So the recipe for felting — hot, wet, alkaline, agitated — is four separate interventions in a two-parameter model, three of which widen the friction gap and one of which lengthens the stroke. None of them is in the arithmetic, which is why the model reports a shape rather than a rate, and why the essay quotes no time.

The mechanism has a direction and the fabric does not #

A last consequence, and it is the one that makes felting a fabric-level phenomenon rather than a fibre-level one.

Each fibre ratchets root-first. But the fibres in a yarn point in every direction — a spun yarn contains fibres laid in with roots and tips distributed at random, and a woollen-spun yarn deliberately so. So the migrations do not add up to a net motion of anything: they add up to fibres moving past one another, in all directions, which is entanglement.

That is why felting shrinks a fabric rather than moving it. The individual displacements cancel as a vector sum and accumulate as a tangle, and the tangle draws the structure together because every entangled pair is a pair that can no longer separate.

A ratchet with randomly oriented ratchets is a mixer, and a mixer applied to a fibre assembly is a felt.

The scale structure that supplies all of this is itself a fibre-growth artefact rather than a design: a wool fibre is built by a follicle laying cuticle cells down in overlapping rings as the fibre extrudes, so the overlap direction is fixed by the direction of growth. Every wool fibre from every sheep has its scales pointing the same way relative to its own root, and that consistency is what makes the effect usable — a population of fibres with randomly oriented scales would ratchet in random directions and produce nothing but noise.

Who found it, and when #

The directional friction effect was described by Martin in 1944 and measured systematically through the following decade; the recognition that it is what causes felting, rather than simple entanglement, belongs to that period. The mechanism is sometimes called the scaly-fibre ratchet and the analogy to a mechanical ratchet is old enough to be unattributable.

What came later, and matters commercially, is the treatment. Chlorination was in use before the mechanism was understood — it was found empirically, as most textile chemistry was — and the understanding arrived afterwards and explained why it worked. The Hercosett process, a chlorination followed by a cationic polymer, dates from the 1960s and is still the standard route to machine-washable wool.

That order is worth noticing. The treatment did not come from the theory. The theory explained the treatment and then told everyone what a better treatment would have to do, which is where the essay on what shrink-resist actually changes picks up.

Where the ladder goes next #

A ratchet moves fibres. It does not, on its own, make a fabric — and the interesting question is what the migrated fibre amounts to once there is enough of it.

The answer connects this field to one of the oldest arguments on the site: a milled cloth holds together for a second reason, and the second reason is the nonwoven’s, arriving inside a woven fabric that was already holding perfectly well.