Wafer Bond Aligner

A wafer bond aligner precisely registers two wafers (or a die and a wafer) to each other before bonding, and wafer-to-wafer and die-to-die bonding are two of the resulting bond configurations — distinguished mainly by whether both pieces are full wafers or individual dies, which changes how forgiving the alignment and throughput requirements are.

Die-to-die bonding exists specifically to avoid wasting a known-good die on a defective partner, a risk wafer-to-wafer bonding can't avoid. Bonding two complete wafers together commits every single die location on both wafers to its exact counterpart at the same coordinates, so a defective die on either wafer locks in a bad pairing; die-to-die (and die-to-wafer) bonding lets a fab test each die first and bond only known-good dies together, trading wafer-to-wafer's higher throughput for the ability to avoid pairing a good die with a bad one.

The bond aligner's optical registration has to hold sub-micron accuracy because 3D-stacked interconnects depend entirely on that alignment staying correct through the actual bonding step. Modern hybrid-bonded or TSV-connected stacks rely on microscopic contact pads lining up precisely between the two bonded surfaces; the aligner uses optical alignment marks on both pieces to register them to each other before any physical bonding force or heat is applied, and once that alignment is locked in and bonding proceeds, there's no correcting a misalignment afterward — making the aligner's precision as critical to final yield as the bonding process itself.

Wafer bond aligners, wafer-to-wafer bonding, and die-to-die bonding together cover the full spectrum of 3D integration choices — from highest-throughput full-wafer stacking to the slower but defect-tolerant selective die-level approach — all depending on the same underlying requirement for precise, permanent sub-micron alignment before the bond is made.