The Marker
In 1947, Ernest Kirkendall and Alice Smigelskas placed thin molybdenum wires at the interface between a block of copper and a block of brass. Molybdenum was chosen because it does not participate in copper-zinc diffusion — it sits inert at the boundary, marking the original position of the interface. The assembly was heated to 785°C and held for weeks.
When they sectioned the sample, the wires had moved. Not by expansion, not by creep — the markers had shifted measurably toward the zinc-rich side. The interface itself had migrated.
The prevailing theory of solid-state diffusion assumed atoms moved by direct exchange: two neighboring atoms swap positions simultaneously. If this were true, the fluxes would be equal in both directions and the interface would stay put. Kirkendall's moving markers proved otherwise. Copper atoms were jumping into the brass faster than zinc atoms were jumping into the copper. The fast-moving copper left behind empty lattice sites — vacancies — that coalesced into microscopic voids at the original boundary. Diffusion did not work by exchange. It worked by one atom jumping into the space left by another. The void was the mechanism.
Modern integrated circuits connect silicon to the circuit board through solder bumps — tiny spheres of tin-based alloy bonded to copper pads. During the chip's operating life, copper and tin interdiffuse at the joint, forming intermetallic compound layers. Copper diffuses into the tin-rich solder faster than tin diffuses into the copper pad. The differential rate creates Kirkendall voids at the copper-intermetallic interface — the same voids Kirkendall found with his molybdenum markers, now forming inside a structure designed to last a decade.
The voids accumulate over thousands of hours of thermal cycling. Each void is a break in the conduction path, a stress concentrator, a nucleation site for cracks. A sufficient density of voids can cause the joint to fail electrically. The mechanism that proved a theory in 1947 is a reliability problem in 2026. Billions of solder joints in billions of devices, each one a diffusion couple where copper and tin move at different speeds and the difference creates hollows.
In 2004, Yadong Yin and colleagues at Berkeley demonstrated something unexpected. They exposed solid cobalt nanoparticles to sulfur and oxygen at moderate temperatures. The chalcogens diffused inward through the cobalt lattice. Cobalt diffused outward. The rates were unequal. The interior emptied.
What remained was a hollow nanoshell — a cobalt sulfide or cobalt oxide sphere with a void in the center where solid cobalt had been. The Kirkendall effect, which creates defects in bulk materials, had been repurposed to manufacture hollow nanostructures. The shells had high surface area, chemical reactivity, and the ability to encapsulate other materials. Drug delivery, catalysis, lithium-ion battery electrodes — applications that require a container emerged from a mechanism that creates a flaw.
The void is the same in all three cases. Differential diffusion across an interface leaves empty space where the faster-moving species departed. In 1947, the void proved how atoms move. In microelectronics, the void is what breaks the joint. In nanoscience, the void is the product. The mechanism does not change. What changes is whether the void is a discovery, a failure, or a design.