Like us, materials respond to stress in many ways. Like us, some materials simply break. Metals may relieve stress by spawning dislocations or other defects that permit plastic flow. Other solids, such as ice and silica, can undergo amorphization when compressed, or may be transformed to different crystal phases. It now seems that melting is another option for crystals seeking to relax under stress.

For example, simulations of copper subjected to shock waves — that is, to non-hydrostatic stress — suggest that, along some crystal axes, melting can occur a little below the equilibrium melting temperature Tm (ref. 1). It's not been clear how this happens. Is the liquid state produced by massive accumulation of stress-induced defects? Can we be sure that this apparent melting is not something that looks superficially like it, such as a special kind of plastic flow? Levitas and Ravelo have now explained how melting can indeed happen at very high strain rates in response to shockwave propagation in crystals, even at temperatures far (by as much as 80 per cent, or 4,000 K) below Tm (ref. 2).

The relationships between defect formation, disordering, amorphization and melting have long been a topic of inquiry and controversy, and it is fair to say that they are still not fully understood. True melting, as opposed to solid-state disordering, requires a thermodynamic driving force. Levitas and co-workers have previously argued that a solid–solid phase transition can induce transient local melting at the moving interface between the phases, more than 100 K below Tm, to relieve stresses there and allow the elastic energy to relax3; they have very recently reported experimental evidence of such effects in a crystalline nanofibre4. Such transient melting might also facilitate high-pressure amorphization as much as 1,000 K below Tm (ref. 5).

The curious thing about this kind of melting is that, once it has happened, it negates its own driving force, so that recrystallization happens almost at once. For this reason, the researchers called the phenomenon virtual melting.


Levitas and Ravelo now propose that virtual melting can also account for the response of metals like copper to high non-hydrostatic stresses like those induced by shockwaves. Their simulations of copper subjected to a shock pressure of around 160 GPa reveal a molten state just in front of the wave, with a corresponding large drop in non-hydrostatic stress there, even though the local temperature can be just a fifth of Tm. The metal recrystallizes in a matter of picoseconds, but a genuine (supercooled) liquid state can still be distinguished from a hot amorphous one by calculating the atomic self-diffusion rates.

Spotting this phenomenon experimentally will be challenging, given the short timescales. But ultrafast electron diffraction has already been used to study comparably ephemeral effects in surface phase transitions6, so it's not a vain hope. And the conditions considered by Levitas and Ravelo are by no means unrealistic, being relevant for example to meteorite impacts, nuclear explosions and the sort of laser-induced compression explored in nuclear fusion.