Stress Migration in Copper
Stress migration is a reliability failure mechanism in copper interconnect where atoms slowly diffuse under purely mechanical stress — no current flow required — eventually opening voids near via connections that can interrupt the circuit entirely. Unlike electromigration, which needs current flowing through a wire to push atoms around, stress migration can happen even in a wire that's never carrying any current at all, driven purely by mismatched thermal contraction between copper and the dielectric surrounding it.
Stress migration's root cause is a simple mismatch in how much copper and the surrounding dielectric want to shrink as the chip cools from the elevated temperatures used during processing. Copper has a significantly higher coefficient of thermal expansion than the oxide or low-k dielectric packed around it — meaning copper naturally wants to contract much more than the surrounding dielectric does as temperature drops. But the copper is embedded in that dielectric and mechanically constrained by it, so it physically can't contract as much as it "wants" to. The result is tensile stress locked into the copper itself, left over from the simple fact that it's trapped inside a material that won't shrink along with it.
Atoms under sustained mechanical stress will slowly migrate toward any available free surface that lets them relieve that stress, and that migration is the direct mechanism behind stress-migration voiding. Via bottoms, grain boundaries, and other interfaces inside the copper structure act as relief points — a copper atom sitting near one of these surfaces can move toward it, gradually relieving a small amount of the local stress each time it does. Multiply this atom-by-atom migration over the chip's operating lifetime, and material accumulates at certain points while emptying out at others, eventually opening a genuine void where enough atoms have migrated away. If that void grows large enough at a critical point — commonly right at a via bottom, where geometry concentrates stress most severely — it can interrupt the electrical connection entirely.
Stress migration is worst at intermediate temperatures, not at the highest temperatures a chip ever experiences, which is a genuinely counterintuitive aspect of the failure mechanism. At very high temperatures, mechanical stress in the copper actually relaxes somewhat, because the metal becomes more pliable and better able to accommodate the mismatch without staying rigidly stressed. At very low temperatures, atomic diffusion itself becomes too slow for migration to proceed at any meaningful rate, even though stress may be high. Stress migration is most dangerous specifically in the intermediate range — roughly 150 to 250 degrees Celsius — where stress remains substantial and diffusion is still fast enough to actually move atoms meaningfully over realistic timescales.
Wider metal lines are counterintuitively more vulnerable to stress migration than narrower ones, which runs against the usual assumption that bigger wires are always more robust. A wider copper line contains a larger total volume of copper under stress, and therefore has a larger total reservoir of atoms available to migrate when a relief path opens up. This larger available supply of mobile atoms makes voiding more severe once it starts, even though the same wire's cross-sectional area might otherwise suggest it should be the more reliable, less failure-prone geometry — a reminder that stress-migration risk tracks total stressed volume rather than simple wire width alone.
Design-level mitigation for stress migration looks almost identical to the fix for a completely different failure mode, electromigration, which is a useful coincidence for reliability engineers. Redundant vias and via-array design rules — placing multiple vias at a single connection rather than relying on one — reduce stress-migration risk the same way they reduce via-failure risk generally: a void opening at any single via doesn't interrupt the connection if parallel vias remain intact. Burn-in testing, which deliberately runs chips at elevated stress conditions to accelerate and screen out weak devices before shipment, can also trigger or accelerate stress migration specifically, which is why reliability qualification programs have to account for stress migration as a distinct failure mode alongside electromigration, rather than assuming that passing electromigration qualification alone is sufficient proof against long-term copper interconnect failure.