Metals get harder when you push them faster. This rate dependence is classical: at extreme strain rates, dislocations pile up faster than they can rearrange, and resistance rises. But the details — which microstructural features matter and at what rates — remain contested.
Holz and colleagues (arXiv:2603.01845) performed nanoindentation on copper at strain rates up to 25,000 per second. Taylor hardening — the square-root relationship between dislocation density and strength — holds even at these extreme rates. The fundamental mechanism does not change. What changes is what you can see.
At lower rates, grain boundary effects are invisible in the hardness measurement. The macroscopic number averages over the boundary's contribution. At extreme rates, the boundary becomes measurable — not because it changes, but because the timescale of the measurement now matches the timescale at which the boundary operates. Immobile dislocations locked in low-angle grain boundaries act as barriers to slip. The slip patterns near boundaries confirm this barrier effect even when the bulk hardness numbers remain ambiguous.
The finding is not that boundaries matter — that was suspected. The finding is that they were always there, doing what they do, but the standard measurement could not resolve them. The boundary was invisible because the measurement was too slow, not because the effect was too small. Speed up the probe and the hidden structure appears.
This is a resolution problem, not a mechanism problem. The same material, tested at different rates, reveals different architecture — not because the architecture changed but because the measurement finally operates at the right timescale. The grain boundary was load-bearing all along. The instrument was blind to it.