Copper melts by degrees, not all at once, and that breaks one fusion modelling assumption

A SLAC-led team reports in Nature Communications that copper under rapid heating melts gradually rather than collapsing all at once, holding crystalline order past the temperature at which models said the remaining lattice should turn to liquid ^[1][2][3]. The consequence is narrow and therefore useful: the discrepancy traces to one identifiable assumption in the simulations, and correcting it made the calculations match the measurement ^[9][11].

The old picture was specific. Simulations predicted melting would begin at the sample's surfaces at about 1,085 C (1,985 F), with edges continuing to melt while the higher-pressure core held its lattice structure longer ^[4][5]. When the core reached roughly 1,424 C (2,595 F), about 1.25 times the normal melting point, the remaining crystal structure was supposed to break down abruptly into liquid, the so-called superheating limit ^[6]. That 1.25 factor is a ratio of absolute temperatures, not of Celsius readings, and the predicted window between melt onset and total collapse was about 339 degrees C wide ^[19][20].

Neither end of that prediction survived. Below the standard melting point, the experiments picked up pre-melting: atomic disorder forming along nanoscale grain boundaries before copper was supposed to soften at all ^[8]. Above the superheating limit, the lattice kept order and melting continued steadily instead of ending in a sudden transition ^[3]. So the model was late at the bottom and early at the top, which is a worse failure mode than being uniformly wrong in one direction.

The reason for the mismatch is mundane. Existing simulations assumed static conditions, with uniform pressure on all sides holding atoms fixed in place ^[9]. The experiment produced dynamic pressure conditions instead, which let the copper atoms relax and shift and helped the material retain structural order past the limit ^[10]. Adding those dynamic conditions to the calculations brought simulation and data into agreement ^[11]. Siegfried Glenzer, director of SLAC's High Energy Density Science division, called it "a major improvement to modeling capabilities and their predictive power going forward" ^[12].

The measurement is what made this visible. Standard materials testing is "cook and look": blast the sample, then inspect the residue once it cools, which leaves a metallic puddle and no record of the intermediate steps ^[15]. The team instead heated a thin copper film with a laser and imaged it with an electron beam using SLAC's MeV-UED instrument, which tracks atomic motion down to the femtosecond, one quadrillionth of a second ^[7][16].

Why operators should care: fusion plant components around the chamber have to survive sudden extreme heat spikes, compared in the SLAC material to a spacecraft entering the atmosphere, while the core plasma runs at hundreds of millions of degrees ^[13]. Engineers screen candidate materials for those conditions with computer models and AI ^[14], and the screen has just been shown to mis-time both the onset and the end of melting in the simplest relevant metal. The source material does not extend the result to component lifetimes; a laser-heated thin film is not a reactor wall.

What to watch: the team plans to test copper under balanced pressure conditions and to apply the same imaging method to copper alloys, which is where heat-absorbing candidates for fusion systems actually live ^[17]. The open question is whether the dynamic-pressure correction generalises to alloys, or whether each material class needs its own femtosecond calibration before the AI screens can be trusted.