Endlos · circular polymer design
A plastic that survives two years on a shelf and comes apart in five minutes.
Those two requirements pull against each other, and only a narrow range of catalytic barriers satisfies both. Below the band the article degrades in storage. Above it, the recycler cannot get the material back. Drag to place a design.
The window is not a fitted result. Both edges invert in closed form from the storage and recovery constraints — which is why it can vanish entirely. Set the wash to plain water below and watch it close.
Selectivity budget · orders of magnitude
Where the suppression comes from
Storage stability is the ratio between two rates. Each gate below contributes independently, and they multiply — so the bar is additive in decades. The line is what this design needs to clear.
clears the requirement
8.14 orders supplied against 7.12 required — 1.02 orders of headroom. Alkalinity carries 74% of it.
Design inputs
Specify the material and the process
Material
Ester backbone throughout. Glassy at ambient, rubbery in a 60 °C wash.
Fragment floor: DP 14.0 (crystallinity-limited)
Also set by dragging the band above
Process
0 = bone dry, 1 = immersed
Caustic — 1–2 % NaOH, the standard bottle wash
Does the window contain any chemistry that exists?
Eleven measured bond-exchange barriers from the vitrimer literature, drawn on the same window the engine inverts from the storage and recovery constraints. The band and the marks are computed from different things and never see each other.
Marks are hollow because they measure a different reaction. A vitrimer’s dynamic bond exchanges partners — the network flows and can be reprocessed, but the backbone survives and no monomer is released. Read this as “is the required barrier in the range chemistry reaches?”, not as “these materials would depolymerise”. The clustering below the floor is the real signal: published networks are optimised to rearrange under moderate heat, which is the same property as failing a shelf-life requirement.
Property surrogate · active learning with an uncertainty gate
Learning the glass transition on a budget
A Gaussian process predicts Tg across comonomer loading. Where its own uncertainty exceeds the gate it spends an oracle call; where it is confident it costs nothing. The truth is drawn alongside the posterior, because the honest claim is that the method recovered a known function — not that it predicted an unknown one.
Uncertainty per query · gate fires above the line
Tg from the V–T curve
39.2±0.4°C
Segmented regression with the smeared transition window withheld from both branches. True value 39.0 °C.
A simulated Tg is not a measured one: molecular dynamics cools ~12 decades faster than a lab scan, worth about 35 K of systematic offset. Quoting a simulated Tg to one decimal against literature would misrepresent the method.
Stochastic reactor · Gillespie SSA
Put it in each environment
Same material, same catalyst. In caustic at 60 °C the gated rate is 2.10e-3 s⁻¹ per bond and the population fragments in real time.
Mechanism · scission vs depolymerization
Two ways to take a polymer apart
Random scission cuts the backbone at internal esters. Depolymerization releases monomer from an activated chain end. Both make product — but at the same yield they leave completely different material behind, which is how the two are told apart by GPC.
Ceiling temperature · PLA
647°C
The wash runs at 60 °C, below the ceiling. Propagation wins: any monomer that leaves a chain end re-adds, so there is no net depolymerization at all. This host has to be taken apart by scission.
What each leaves behind
- Scission
- Molar mass collapses long before much product appears. Good for making waxes, useless if you wanted the monomer back.
- Depolymerization
- Chains are consumed whole. The residue stays high in molar mass because only a rare activated subset is unzipping at any moment — and conversion stops at 52 % because a finite zip length kills each chain before it is fully consumed. Recovering the rest needs re-initiation, not more time.
What this engine can and cannot answer
The scoping document’s 25 claims, scored for coverage and for whether the result agrees. Most of what is missing is the packaging system rather than the polymer, which is the document’s own first criticism of itself.
The reading corpus
23 sources, grouped by the question each answers. Every literature-anchored number in this engine resolves to one of them, and a datum with no entry here cannot carry the demonstrated grade.
- Design principles for intrinsically circular polymers with tunable propertiesin specShi · 2021 · Chem 7, 2896–2912 · 10.1016/j.chempr.2021.10.004 · sha256 a4e702076b3d9b45Primary source for the iCP ΔH°p/ΔS°p/Tc table, the kinetic-trapping argument, the DFT→kMC modelling chain, and the TEA/LCA framing.