The last specimen was a needle, carved so sharp that its tip held only a few million atoms. Running through them, the suspect: one grain boundary from bar B, crossing the point like an old scar.
First the needle has to exist. An ion beam turns a splinter of bar B into a standing cylinder, then mills it in shrinking rings until the suspect boundary rides inside a tip sharper than a virus.
An atom probe does not look at the needle. It asks the needle to leave, one atom per pulse. Each atom streaks to the detector, and a clock names it on arrival: iron, iron, iron, carbon, iron. To read the needle is to destroy it.
Every arrival is weighed. Peak by peak the needle's census assembles itself, and the smallest peaks matter most: boron is barely there, and it is everything.
Inside the computer, the demolition runs backwards. Atom by atom, hit by hit, the needle reassembles: a building un-demolished, every brick back in its place, only now with every brick's name attached.
The rebuild is honest about its own eyesight. In depth it can count atomic planes one by one; sideways, trajectories blur. Knowing which direction to trust is half of reading a reconstruction.
Most of the needle was honest gray iron. Then the boundary surfaced: a sheet of boron, one or two atoms thick, riding a border no drawing ever showed. Rotate to face it, and the scar becomes a wall.
The proxigram is the verdict: composition against distance from the boundary plane. Boron spikes exactly at zero, about two atoms wide. Bar A's control needle drew a flat line.
Every instrument before this one saw the boundary. Only this one could read what was written on it. The recipe was right, the steel was honest, and the sentence that condemned bar B was one atom thick, written where nobody looks.
Some secrets are one atom thick.
The needle kept them until the end.
Case closed.