Pad Cratering

Pad cratering is a wire-bonding failure where the bond itself looks perfect but the fragile dielectric stack underneath the bond pad has fractured, so the chip can fail electrically with no visible defect anywhere on the surface. A wire bonder presses a ball or wedge bond onto the pad with enough force and ultrasonic energy to metallurgically weld the wire to the metal, and that same force pushes straight down through the pad into whatever low-k interlayer dielectric sits beneath it — if the dielectric is too brittle or the bond force too high, it cracks in a cone-shaped fracture that can propagate down toward the transistor layer, invisible from above because the bond pad itself still looks intact.

Low-k dielectrics made this problem much worse than it used to be. Older chips used denser, mechanically tougher dielectric materials under the bond pad, but the carbon-doped low-k and ultra-low-k films adopted to cut interconnect capacitance at advanced nodes are deliberately porous and mechanically weaker — exactly the property that helps electrical performance also makes the stack more prone to cracking under bonding force, so pad cratering became a much bigger reliability concern as fabs moved to low-k BEOL stacks at the same time packaging houses kept pushing bond force up for other reasons.

The industry's shift from gold to copper wire made the problem harder to avoid, not easier. Copper wire bonding largely replaced gold for cost reasons starting in the 2000s, but copper is harder than gold and needs more bonding force and ultrasonic energy to form an equally reliable weld, which pushes exactly the parameter that drives cratering risk in the opposite direction from where packaging engineers would want it; qualifying a copper-wire bonding recipe on a low-k part means tuning force, ultrasonic power, and bond time specifically to stay under the dielectric's fracture threshold while still forming a sound metallurgical bond.

Detection depends on destructive test, not inspection. Because the pad's top surface gives no visual clue, packaging houses qualify a bonding recipe using ball shear and die shear testing — pulling or shearing a sample of bonds to failure and cross-sectioning them to check whether the failure mode is a clean bond pull or a crater fracture into the dielectric — and wire bonder makers like Kulicke & Soffa, Shinkawa, and ASMPT build real-time force and ultrasonic-power monitoring into their bonders specifically so a drifting process parameter gets caught before it ships a wafer full of cratered pads.

Read pad cratering through a *force-versus-fragility* lens rather than a *bad-bond* lens — the bond on top can be perfect and the part can still be dead, because the failure that matters happened one layer down, in a dielectric that was made weaker on purpose for reasons that had nothing to do with how it would survive getting bonded to.