What the merging melt retains is an orientation. The two surfaces that were pressed together leave a faint directional bias behind, and that bias acts as a seed for crystal growth as the temperature falls [8]. The familiar account of weld strength stops earlier, at intermingling: chains from each side have to cross the join and retangle [4], and in glass-like plastics that really is the whole test, because a joint whose chains fail to mix stays weak [5]. Polyethylene gets a second act on the way down, when sections of its chains lock into ordered, crystal-like structures that carry much of the material's toughness [6]. The simulations put the joint's strength in that second act.
This is a question that nearly demands a model. Nobody had watched crystallization behave at the weld line itself [7], which is how a property engineers have leaned on for decades, a good joint being as strong as the pipe or stronger, went so long without a molecular explanation [2]. The paper, "Molecular Insights on the Thermal Welding of Semicrystalline Polymers," carries a 2026 Physical Review Letters DOI [14], and the stakes are ordinary rather than exotic: buried water and gas mains that must not leak [1].
Rate matters here for a reason particular to crystallization, which is that nucleation and growth are separate steps. The authors' own list of levers contains one temperature and one rate, heating temperature and cooling speed [12]. Seeds decide where crystals start; the cooling curve decides how long they have to grow. If the ramp is what governs the outcome, a procedure has to say something about the conditions a joint cools in, which is a harder sentence to write than a heater setting.
Size is what the account leaves out. It gives no figure for how much crystallinity the seam gained, none for the stiffness difference against the pipe wall, none for the simulated cooling rates, and nothing on fatigue or slow crack growth across a service life [2]. Interfacial stiffness is a gradient with a length scale, and whether that stiff band is microns or millimetres is what decides how much of a real joint's geometry an engineer should think about differently.
The recycled-feedstock line is the one I would hold loosest. The mechanism needs no virgin resin, so on physics alone a remelted recyclate should crystallize across a join the same way [13]. That is an argument about polymer chains, not about a waste stream [3], and clean simulated chains are precisely where contamination and grade mixing do not appear. My reading: this should change what you measure on a test weld. Rewriting a weld procedure will take more evidence than this.
