Shrink-resist is one number

The label says machine washable and the description says the scales have been smoothed or coated. Both are true and neither is the requirement.

The requirement is exact and it is a statement about a difference: a shrink-resist treatment must make the two friction coefficients equal. Not small — equal. A fibre with coefficients of 0.5 in both directions does not felt at all, and a fibre with 0.05 and 0.10 felts readily.

Why equality rather than smallness #

The ratchet nets, per cycle of agitation,

1/μ_with − 1/μ_against

and that expression is zero exactly when the two coefficients are equal. It does not become zero as they become small. It becomes larger as they become small, because the reciprocals grow: halving both coefficients doubles the net migration while leaving their ratio untouched.

So the two obvious interventions do opposite things:

• Lubricating the fibre uniformly — reducing both coefficients while preserving their ratio — makes felting worse.
• Roughening the fibre uniformly — raising both — makes it better, at the cost of a harsh handle nobody wants.
• Closing the gap — bringing the two together at any level — stops it.

Only the third is a treatment. That the first is counterproductive is not a curiosity: softeners are applied to wool for handle, and a softener that reduces both coefficients without touching their ratio is genuinely making the fabric more likely to felt. The trade knows this and formulates around it.

What the chemistry actually does #

Two routes, both of which close the gap and neither of which is well described by “gluing the scales down”.

Chlorination — the older route, and still the first stage of the standard process — oxidises the outer surface of the fibre. The scale edges, which are what catch, are degraded: softened, partially dissolved, their sharp overhangs blunted. The fibre remains scaly to look at, and the directional component of its friction largely goes, because directionality lives in the overhang rather than in the roughness.

Polymer masking — the Hercosett stage, a cationic polyamide-epichlorohydrin resin — deposits a thin film that bridges between scale edges. The film does not fill the surface in; it spans the steps, so a fibre sliding tip-first no longer meets a wall. The polymer’s own friction is what the fibre then has, in both directions equally.

The two together are the standard route to machine-washable wool and have been since the 1960s. Chlorination alone gives an incomplete closure and a harsher handle; the polymer alone does not adhere well enough to a scaly, hydrophobic surface without the oxidation to key into.

Both are described in the literature as scale modification and both are really gap closure, which is the same statement one level down.

The cost, and where it lands #

Nothing in a finishing works is free, and this treatment’s costs are all consequences of the same fact: it changes the fibre’s surface, and the surface is what the fibre does everything with.

Handle. A chlorinated fibre is different to the touch — the trade calls it a harsher or more “cotton-like” hand — and the polymer helps but does not restore it. Wool’s characteristic handle comes substantially from the scale structure, so a treatment that neutralises the scales cannot leave it untouched.

Lustre and dye behaviour. The surface decides both. Treated wool dyes at a different rate and to a different depth, which is why the treatment is placed at a specific point in the sequence rather than wherever is convenient.

Strength. Oxidation is not selective; some of it happens to the fibre rather than only to its surface, and a chlorinated wool is somewhat weaker.

And the effluent, which is the reason the industry has spent forty years looking for a replacement. Chlorination on wool produces adsorbable organic halides, and the discharge limits are the binding constraint on the process in most of the world.

What was counted, and how #

shrinkResist takes the two coefficients and a list of closures — the fraction of the gap each treatment removes — and reports the net migration at each. It asserts the two ends.

An untreated fibre ratchets forward, so the zero-closure row is strictly positive. If the model had somehow produced a non-felting untreated wool, the whole family would be describing something else.

A fully treated fibre nets exactly zero, to within 10⁻¹². This is the assertion worth having, because it is a claim about exactness rather than about size. The chemistry’s proposition is not that treated wool felts slowly; it is that the mechanism has been removed. A model that produced a small residual migration at full closure would be making a materially weaker claim, and the assertion is written so that a model doing that would fail rather than look approximately right.

That assertion also produced the one piece of numerical care in this file. Closing a gap by a fraction of exactly one lands the treated coefficient a floating-point whisker below the untreated one, and ratchet’s own guard — that the against-the-scales coefficient is at least the with-the-scales one — refused it. The guard now carries a 10⁻¹² tolerance with a comment saying why. A guard that refuses the one case the function exists to produce is worse than no guard, and it fails in the direction that looks like a genuine defect.

The treatment that does not treat the fibre #

There is a second route to machine-washable wool that this model describes just as well, and it is worth including because it makes the point about mechanism rather than material.

Blend the wool. A yarn in which wool fibres are separated by fibres that do not ratchet — polyester, nylon, treated wool — migrates far less, because a wool fibre’s ratchet needs other fibres to ratchet against and the neighbours it finds do not cooperate.

The model has nothing to say about the proportions, because it treats a single fibre. What it does say is that the mechanism is a property of the fibre-to-fibre contact rather than of the fibre alone, so anything that changes what a wool fibre meets is an intervention of the same kind as anything that changes the fibre’s surface.

That reframing is the useful part. Felting is a property of a contact, not of a material, which is why the same wool felts in one fabric and not in another, and why the fabric’s construction — how tightly the yarns are twisted, how densely they are set, how much the fibres can move at all — matters as much as the chemistry.

Why the number is a difference and not a ratio #

One detail of the model repays attention, because the obvious summary of the directional friction effect is the ratio μ_against/μ_with and the ratio is not what decides the felting rate.

The net migration goes as 1/μ_with − 1/μ_against, which is a difference of reciprocals. Two fibres with the same ratio can have very different net migrations: 0.1 and 0.2 give a net of 5, while 0.4 and 0.8 give 1.25 at exactly the same ratio of two.

So the DFE ratio, which is what the literature quotes, is a summary that loses the part that matters most. A fibre’s felting propensity depends on the absolute level as well as the ratio, and it is the low-friction fibre that ratchets hardest.

That is a testable difference between the two summaries and it points the same way as the observation that lubricating wool makes it felt more, which is a thing the ratio cannot explain and the difference of reciprocals predicts.

The index the trade should quote instead #

The essay observes that the DFE ratio loses the part that matters, and it is worth taking one step further, because the replacement is available and it reorders real fibres.

The net migration per cycle is 1/μ_w − 1/μ_a, which can be written

(μ_a − μ_w) ÷ (μ_w μ_a),

a difference over a product. Call that the felting index. It has units of one over a friction, it is zero exactly when the two coefficients are equal, and it grows without bound as either coefficient falls.

Against the ratio, it ranks fibres differently — and not in a corner case:

The ratio calls A and B equally bad and C the best of the three. The index calls A four times worse than B, and C better than B. So the two summaries agree about C being best and disagree completely about how far apart A and B are — which is the disagreement that matters, because A and B are the same fibre lubricated to two different levels.

Three things recommend the index.

It predicts the lubrication result and the ratio cannot. Halving both coefficients leaves the ratio untouched and doubles the index, which is the observation that a softener makes wool felt more — a fact the trade knows and the standard summary has no room for.

It is zero at the target. A shrink-resist treatment’s aim is an exact zero, and the index reaches it while the ratio reaches one — so a specification written on the index has its target at the origin and its tolerance as a plain upper bound, which is the easier document to write.

And it is measurable with the same apparatus. Both coefficients come from one directional-friction measurement; the index is a different arithmetic on the same two numbers and costs nothing to report beside the ratio.

The caution is that the index is not dimensionless and the ratio is. Comparing indices across fibre pairs with very different absolute frictions is comparing quantities with a unit in them, which is exactly the property that makes the index informative here and would make it awkward anywhere the absolute level did not matter. Here the absolute level is the point, so the dimensionality is a feature — and the ratio’s dimensionlessness, which is what recommended it, is precisely what discards the information.

Where the model stops #

The gap closure is a single number and a real treatment is not. Chlorination and polymer masking change the two coefficients by different amounts and in different ways, and neither reduces the against-the-scales figure to exactly the with-the-scales one. Real treated wool has a residual DFE and felts very slowly rather than not at all, which is why machine-washable wool still carries a care label.

Nothing here models the treatment’s uniformity. A partially treated fibre with treated and untreated regions is not the same as a uniformly half-treated one, and a real process leaves a distribution rather than a value.

And the whole model is single-fibre. Everything above about blends and construction is reasoning around the model rather than from it.

Why this is the clearest case in the field #

Every other essay here computes a quantity that lands somewhere on a scale and asks how far. This one has a target and the target is a zero, which makes it the easiest claim in the field to state and to check.

That is worth something beyond wool. A mechanism with a clean off-switch is a mechanism that has been correctly identified, because a wrong identification almost never produces one. If felting were entanglement of a rough surface, there would be no single quantity whose vanishing stops it; there would be a gradual reduction as roughness fell, with no particular point of interest. The existence of an exact zero at μ_with = μ_against is evidence for the ratchet rather than a consequence of assuming it.

The same test can be run against the site’s other frictional claims and it is instructive that they fail it. The capstan comparison of a leno and a plain weave has no setting of μ at which the two grips become equal, because the inequality is between two angles and μ cancels. That is a different and equally strong shape — a result with no free parameter — and the two together are the pattern this field’s arguments try to reach: either the parameter cancels, or there is an exact value at which the effect switches off.

What the treatment does to the rest of the fabric #

A treatment aimed at one coefficient reaches everything the surface decides, and the list is worth setting out because it explains why treated wool is a different product rather than the same product made safe.

Felting is what makes a woollen cloth in the first place. A melton, a loden, a billiard cloth and a tennis ball’s covering all depend on the mechanism this treatment removes, so shrink-resist wool cannot be milled — the two are exclusive by construction, and a mill choosing one is choosing which fabrics it can make.

Yarn-to-yarn friction falls, which matters for the fabric rather than the fibre. The capstan argument says a pick is held by friction at its crossings, so a treated wool cloth grips its weft less well and is more prone to seam slippage and to fraying at a cut edge.

Fibre cohesion in the yarn falls, so a treated wool needs more twist to spin to the same strength, and more twist is a harsher, less lofty yarn.

And the surface takes up dye differently, which is why the treatment is placed at a defined point in the sequence rather than wherever is convenient.

So the treatment is not free anywhere. What it buys is a fabric that survives a washing machine, and what it costs is a share of every property wool’s surface was providing.

The specification the mechanism writes #

Because the requirement is exact — close the gap — it can be stated as a target for a treatment that does not exist yet, and this is the most useful thing the mechanism produces.

Any process that equalises the two coefficients will work, whatever it is made of. It does not need to coat the fibre, it does not need to remove the scales, and it does not need to make the fibre smooth. Enzymatic treatments, plasma treatments and various polymer routes have all been pursued on exactly that basis, and each is an attempt to hit the same target without chlorine — and each has to reach it in water, where the gap is at its widest.

And any process that lowers both coefficients without changing their ratio will make things worse, which rules out a whole family of otherwise attractive approaches. That is a negative result and it is the kind a mechanism is most useful for: it says where not to look, which is information a purely empirical search does not have.

The forty-year search for a chlorine-free route is that specification being worked, and none of the alternatives has yet matched Hercosett on cost and reliability together. The mechanism says the target is reachable; it does not say cheaply.

There is one more reason the exactness matters, and it is about what the claim licenses. A treatment that reduced felting by ninety per cent would still felt a garment eventually, over enough launderings, and a care label would have to bound the number of them. A treatment that removes the mechanism does not, and the label can say machine washable without a count. The difference between a large reduction and a removal is the difference between a qualified claim and an unqualified one, and it is worth the assertion being written to tell them apart.

Who found it, and when #

Chlorination as a shrink-resist treatment predates the understanding of why it works by decades; it was found empirically, like most textile chemistry. Martin’s measurement of the directional friction effect in 1944 and the work that followed gave the explanation, and the Hercosett process of the 1960s was the first treatment designed with the mechanism in view rather than discovered against it.

That order — treatment first, mechanism second, better treatment third — is the ordinary order in this industry and is worth not being condescending about. An empirical process that works is a real achievement, and the mechanism’s contribution was not to validate it but to say what a replacement would have to do: close the gap, by any means, without oxidising the fibre. The forty-year search for a chlorine-free route is that specification being worked on, and it is still open.

Where the ladder goes next #

The felting ladder ends here. The field’s remaining ladder is the surface: the yarn flattened, which turns out to move a cloth from one of this site’s thread models to the other, and the fibre swollen, which is a single parameter with a long list of consequences.