Microplastics are ice-nucleating particles · weathering makes them better at it · and they are already in the clouds, the Arctic snow and the Antarctic snow · the distributed aerosol programme WP41 proposed has been running for decades, without consent, dose control, or measurement
WP66 §3 proposed four tests a distributed intervention must pass — local, reversible, attributable, self-benefiting-or-oracle-scored — and concluded that atmospheric ice-nucleation deployment fails all four and must never be handed to a crowd. This paper reports that it already has been.
Microplastics nucleate ice. Atmospheric weathering makes them better at it: ten days of sunlight raises the freezing onset by about 5 °C, and after 90–180 days their ice-nucleation activity is comparable to mineral dust per unit surface area [documented]. They are present in high-altitude cloud water, in Arctic snow at up to 14 400 particles L⁻¹, and in Antarctic snow 6 000 km from any plausible source. Over 1 000 tonnes fall on the protected lands of the western United States each year.
Three consequences. First, the four tests are not a forecast — they are a post-mortem, and §8 scores the deployment against them. Second, by WP66 §2's own commuting argument this loading is not a gate, which is worse rather than better: it is an uncontrolled perturbation to a pointwise term of unmeasured magnitude. Third, and operationally: the ice-nucleation baseline WP66 §2.2 identified as the field's central unknown is not merely unmeasured but anthropogenically drifting. The observation network WP66 §10 listed as optional is now the only instrument that can say what the baseline is.
WP41 proposed training technicians to disperse ice-nucleating particles into the lower troposphere. A recurring amendment proposed scaling that to a billion deployers using cheap, easily manufactured materials. WP66 §3 argued that the second proposal is structurally unsafe: the atmosphere is a non-excludable commons, deployment is unattributable at DIY scale, irreversible, and returns nothing to the deployer.
That argument was made as a prediction about a hypothetical. It is not hypothetical. Roughly eight billion people are already dispersing ice-nucleating particles into the lower troposphere, continuously, in every populated region, with no dose control, no monitoring, no attribution and no off switch. The material is plastic, the mechanism is ordinary use and abrasion, and the programme has been running since roughly the middle of the twentieth century.
This paper does not argue that microplastics are cooling or warming the planet. The honest status of that question is [open], and §10 says why it is likely to stay open for some time. It argues something narrower and harder to escape: that the arc's own framework, applied to a perturbation already in place, returns an answer nobody has looked for.
The Antarctic result is the one that settles the transport question. Back-trajectory modelling indicates long-range transport of up to 6 000 km on a 6.5-day residence time, and remote field sites up to 20 km from any station still carried 22.5 ± 4.0 particles L⁻¹. Bergmann's comparison is the useful intuition: these particles travel like pollen grains of similar size, which routinely reach the Arctic from mid-latitudes.
For scale at the source end, the same study found 154 000 particles L⁻¹ in snow beside a rural Bavarian road. The gradient from roadside to pole is four orders of magnitude, and the pole end is not zero.
Presence aloft is not the claim. The claim is nucleation, and it has been measured directly in droplet-freezing assays on the four commodity polymers found in atmospheric samples — polyethylene, polypropylene, polystyrene and polyethylene terephthalate [documented].
Note the temperature band. Freezing onsets of −20 to −25 °C sit squarely in the mixed-phase cloud regime — which is to say, at the altitudes WP41 specified for deployment, 2–4 km. This is not the cirrus regime below −35 °C that WP66 §2.1 identified as the only radiatively relevant one. It is precisely, and unhelpfully, the regime WP41 was aiming at.
The finding that makes this more than a curiosity is that the atmosphere improves the particle. Sunlight exposure roughens the surface and promotes hygroscopic salt uptake, and the nucleation activity rises accordingly:
| Exposure | Effect on nucleation | Notes |
|---|---|---|
| 0 d | Freezing near −25 °C; water uptake at −11 °C | Already well above the hydrophobic control |
| 10 d | ~40% lower relative humidity required for water uptake; freezing ~5 °C warmer | PP and PS photooxidise rapidly; PET shows little change after one day |
| 90–180 d | Below −24 °C, activity comparable to mineral dust normalised by surface area | Mineral dust is the dominant natural INP — this is the reference class, reached |
A designed aerosol has a specification and degrades away from it. This one self-activates: residence time in the atmosphere is the activation step, and the particle's ice-nucleation efficacy is an increasing function of how long it has been aloft and how much sun it has seen. Any control scheme premised on a dose therefore has the sign wrong — the effective dose is not what was released, it is what was released convolved with its exposure history. Nobody has that history for any particle.
Two limits must be stated plainly, because the temptation to overread here is strong and this series has a paper (WP29) about what happens when it is indulged.
First, concentration. The authors of the weathering study state directly that current atmospheric microplastic concentrations remain lower than mineral dust. Per-particle comparability is not burden comparability. What the result establishes is that the loading is in the same efficacy class as the dominant natural INP, not that it currently rivals it in mass.
Second, mechanism attribution. Denise Mitrano's objection to the cloud-water work is correct as stated: finding microplastics in cloud water does not show they nucleated it — they may simply be transported with the air mass, and she argues only nanoplastics below 100 nm could plausibly matter for cloud processes. The cloud-water studies did not test the mechanism. The laboratory freezing assays did, on isolated particles. Neither closes the loop between them, and that unclosed loop is the paper's principal gap [open].
WP66 §2 argued that an ice-nucleating aerosol released inside the mixed layer acts pointwise on phase, and that by WP39's Theorem T2 a 0/1 gate commutes with pointwise loss — so such an aerosol does not inherit the order-dependence that made \(K\) worth deploying. Intellectual honesty requires applying that argument here, where it cuts against an alarming conclusion rather than a reassuring one.
That argument is formalised rather than asserted. ZeoliteCommutation.lean
(github.com/TOTOGT/io) closes
gate_commutes — the 0/1 aperture gate commutes with the pointwise fold, for every
state — alongside gate_fold_not_commute, which shows the gate fails to commute
once the fold carries the coupling term. Sorry-free, 0 admits, axioms
[propext, Classical.choice, Quot.sound]. The dichotomy holds at the level of the
algebra, and it does not care whether the release was deliberate.
It applies unchanged. Microplastic INP loading is an in-layer pointwise perturbation. It is not a gate, does not produce order-dependence, and does not constitute accidental geoengineering in the sense WP41 meant. Anyone wishing to claim that plastic pollution is inadvertently operating \(K\) must first defeat §2 of WP66 — and §2 does not distinguish between intentional and accidental release, because the box model does not know the difference.
"Not a gate" would be reassuring if the alternative were "no effect." It is not. The alternative is an uncontrolled, self-activating, globally distributed perturbation to a pointwise term whose magnitude has never been measured. A gate at least has a sign and a deployment window. This has neither. It cannot be scheduled relative to the fold, which means it cannot be reasoned about with the arc's central tool — and it does not stop while we work out how.
This conclusion has since acquired independent support from a discipline that has never heard of the operator chain. WP68 §2 observes that the 2024 stratigraphic ruling — the Anthropocene rejected as an epoch and characterised instead as a diachronous event admitting no Global boundary Stratotype Section and Point — is the same structural claim in another vocabulary. A GSSP is what a fold looks like in rock: one location, one horizon, one globally simultaneous crossing. To say the anthropogenic signal admits no GSSP is to say it is not a fold. Stratigraphy reached that from varves and cores; this arc reached it from a three-layer box model. Two formalisms with unrelated failure modes, one conclusion.
The nucleation pathway is contested. The deposition pathway is not, and it connects directly to WP66 §4, where the argument was that albedo, not enthalpy, is the planetary lever.
Microplastics reaching snow and ice darken it. Most plastic products are dyed or pigmented, and degradation drives them from transparent toward translucent and light-absorbing. Deposited on a high-albedo surface, they do what every light-absorbing particle does: reduce reflectance, accelerate melt, and expose lower-albedo substrate — the same feedback WP66 §4 identified as the reason ice is worth defending in the first place.
The mass of microplastic particles in the environment is very likely greater than that of black carbon, and one water-sample study found five times more microplastic particles than visible black-carbon particles. Reviewers of the cryosphere literature raise the consequent possibility that microplastics are responsible for part of the reduced snow albedo currently ascribed entirely to black carbon.
If that holds, existing snow-albedo forcing estimates are not necessarily wrong in magnitude but are misattributed in source — which matters enormously for policy, because the two have entirely different mitigation levers. Black carbon responds to combustion controls. Microplastics do not. Status: [open], and flagged by the reviewers themselves as unquantified.
This is the one place in the paper where WP66's §4 arithmetic bites directly. The albedo term was worth ~1 040 MJ m⁻² per melt season against 306 MJ m⁻² of fusion enthalpy — a factor of three, and recurring rather than one-off. Anything systematically degrading surface albedo across the cryosphere is operating on the larger term, and it is doing so in the same places WP66 §4 argued were worth the pumps.
WP66 §2.2 identified the field's central unmeasured quantity: what fraction of cirrus currently forms with versus without seed particles — the number required to size any deliberate dose, and the reason a distributed deployment has no dose control by construction.
This paper makes that gap qualitatively worse in a specific way.
The baseline against which any intervention would be measured is not a natural constant. It contains an anthropogenic component of unknown size that is growing with plastic production and improving with atmospheric residence. A dose computed against an assumed-natural baseline is computed against the wrong number, and the error moves in one direction over time.
WP66 §10 listed a distributed ice-nucleating-particle observation network as an option-preserving line item at 10⁷ $/yr. It is no longer optional. It is the only instrument that can establish what the baseline is, and the baseline is now known to be moving. WP66 §3.1 showed that large-scale unpaid technical work runs on an oracle — and this work has one: the WT-CRAFT freezing assay taught in WP40 returns a nucleation temperature, per sample, unarguably. Separable tasks, external verifier, visible ledger. It is the Galois campaign's structure applied to atmospheric chemistry, and it is the one job in this entire arc that a crowd can safely be handed and is needed for.
WP66 §3 offered four tests as a design filter for prospective interventions. Scored against the deployment that actually occurred, they read as a post-mortem.
| Test | Microplastic INP loading | Result |
|---|---|---|
| 1 · Local | Transported up to 6 000 km in 6.5 days; deposited in Antarctica from sources on other continents | FAIL |
| 2 · Reversible | No retrieval mechanism at any scale; polymer residence in the environment measured in centuries; weathering increases rather than decreases the relevant activity | FAIL |
| 3 · Attributable | No per-source accounting; polymer identity does not resolve emitter; possibly confounded with black carbon in the existing albedo record (§6) | FAIL |
| 4 · Self-benefiting / oracle-scored | The deployer receives nothing and is not informed they deployed; no oracle exists because no one is asking the question at the point of release | FAIL |
Four failures out of four, on a live global deployment. WP66 presented these tests as a filter for choosing among candidate interventions. They turn out to describe, exactly, the failure mode of an intervention that was never chosen at all — which is the strongest available evidence that the filter is measuring something real, and the least comfortable way to have obtained it.
| Action | Rationale | Status |
|---|---|---|
| Distributed INP observation network — promoted from optional to required | §7. The baseline is drifting and unmeasured; WP40's assay is the instrument; WP66 §3.1's oracle condition is satisfied | URGENT |
| Separate microplastic from black-carbon contributions in snow-albedo forcing | §6. If the attribution is wrong the mitigation lever is wrong; co-located sampling is the minimum experiment | TO DESIGN |
| Close the laboratory-to-cloud loop | §4.1. Freezing assays measured isolated particles; cloud-water studies measured presence. Neither shows in-situ nucleation. This is the field's actual gap | OPEN |
| The core-profile test: compare microplastic against plutonium and CFC vertical profiles in the same sediment core | WP68 §6. Perturbations that are attributable acquire an error signal and get corrected, so they deposit as a spike-and-decline; perturbations failing attribution and self-benefit deposit as an unbounded step. Plutonium and CFCs are the bounded controls, microplastics the predicted step. Same core removes site confounds | SHARPEST TEST |
| Treat plastic emission reduction as an atmospheric measure, not only a marine one | §2–§4. The framing is almost entirely oceanic; the ice-nucleation and cryosphere pathways are airborne and are not represented in that framing | TO ARGUE |
| Do not infer that deliberate deployment is therefore acceptable | §5. "It's already happening" is not a licence — it is the demonstration of why the licence was withheld. The accidental case fails the same four tests | STATED |
Stated in the series' four marks, so that the load-bearing claims are separable from the rest.
| Claim | Mark |
|---|---|
| Microplastics nucleate ice in droplet-freezing assays; weathering raises the onset ~5 °C at 10 days; 90–180 d activity comparable to mineral dust per unit area | [documented] |
| Microplastics are present in cloud water, Arctic snow, Antarctic snow, and deposit at >1 000 t/yr on western US protected lands | [documented] |
| In-layer microplastic INP loading is a pointwise term and therefore not a gate under T2 | [modeled] |
| Microplastics contribute measurably to in-situ cloud ice formation in the atmosphere | [open] |
| Part of the snow-albedo reduction attributed to black carbon is microplastic | [open] |
| The anthropogenic component of the INP baseline is large enough to invalidate dose calculations that assume a natural baseline | [prospective] |
| Net radiative sign of atmospheric microplastic loading | [open] — and likely to remain so; the cloud-formation and UV-degradation pathways have opposite signs |
Note on what would falsify the paper's core move: if in-situ nucleation by microplastics is shown to be negligible at atmospheric concentrations, §3–§5 reduce to a laboratory curiosity and only §6 and §8 survive. §8 survives regardless, because it scores transport and reversibility rather than nucleation efficacy.
| Destination | Content | Status |
|---|---|---|
book6/wp67-… (this paper) | Microplastics as unintended distributed INP deployment | WRITTEN |
wp66-… §10 | Promote the observation network from option-preserving to required; cite §7 | DONE |
wp66-… §3 | Add pointer to §8 — the four tests have a scored real-world instance | DONE |
wp66-… strip | Change WP67 → to the live link | DONE |
wp40-aerosol-engineering-pedagogy.html | WT-CRAFT assay module gains a second justification: baseline drift, not only deployment science | DONE |
wp39-smoke-transamerican.html | No change. The pointwise/transport distinction is used as proved | NO CHANGE |
index.html | WP67 entry; climate arc now WP39–45, WP66–67 | DONE |
Microplastics nucleate ice at temperatures that place them in the mixed-phase regime — the same 2–4 km band WP41 proposed to seed. Atmospheric weathering improves them, so the effective dose is a function of exposure history rather than release, and no one holds that history. They are in the cloud water, in Arctic snow at up to 14 400 particles L⁻¹, and in Antarctic snow 6 000 km from a source.
By WP66 §2's own argument this is not a gate, and that is the bad news rather than the good: it is an uncontrolled perturbation to a pointwise term of unmeasured size, which cannot be scheduled against the fold and therefore cannot be reasoned about with the arc's central tool. On the deposition side it degrades exactly the albedo term WP66 §4 identified as the larger one, and may be sitting inside the black-carbon attribution while it does so.
The practical residue is small and unambiguous. The ice-nucleation baseline is drifting anthropogenically, nobody is measuring it, the instrument exists and is already taught in this series, and the work is separable and oracle-scored — the one shape of task that WP66 §3.1 showed a crowd will genuinely do unpaid. The billion technicians were never going to cool the planet by dispersing anything. They might, plausibly, be the only way to find out what is already up there.