Highly time-resolved propagation dynamics of adhesion-front and defect in wafer bonding

The realization of imec’s CMOS 2.0 [1] ambitions critically depends on multiple wafer bonding steps. Low distortion bonding enables today high yield device fabrication with robust back-side connectivity, while hybrid bonding enables fine pitch interconnects. These bonding capabilities form the foundation of new device concepts and architectures both in logic as well as in memory applications. Bonded CFETs (complementary FETs) represent one of the promising device architectures for extending CMOS scaling. Instead of placing nFETs and pFETs laterally, they are stacked vertically to reduce footprint and improve performance.  A possible way to realize this is through wafer bonding, where separately optimized n-type and p-type transistor layers are fabricated on different wafers and then aligned and bonded together. Commercially available bonders already achieve astonishingly accurate alignments of sub 100 nm, while the requirements of bonded CFET are sub 10 nm. This cannot be achieved without a fundamental understanding of bonding mechanics and defect propagation. With this project we would like to investigate the dynamics of bond front using high speed imaging, and explain the mechanics of bond front propagation from a continuum mechanics perspective. The ultimate goal is to explain the propagation of in-plane strain which directly creates bonding overlay, and explain the mechanics of defect formation (such as voids, high distortion regions) in terms of scalable parameters. This will create robust and predictive models, so that the expected defects could be compensated at litho or mitigated during other process steps.

Scope:

Research applicability

1.     Fundamental understanding of mechanics at the bond front

One of the causes of bonding overlay errors (i.e., misalignments) between wafers are ascribed to in-plane strain in the vicinity of the bond front, and in the remainder of the unbonded region. With the current state-of-the-art equipment, the bonder overlay fails to meet the criteria for bonded CFET. Misalignments caused by in-plane strain are not fully understood. We aim to measure the time evolution of these strain fields using Synthetic Schlieren [2,3] and DIC [4] (digital image correlation) approaches. The experiments, once perfected on a desktop experiment, will be performed in a tool with infrared capabilities to see through silicon. 

2.     Mechanics and condensation voids at the edge

As the bonding front approaches the wafer edge, it accelerates together with the air in the unbonded gap that is expelled. This rapid outflow can lead to condensation voids forming along the wafer perimeter.  In this work, we aim to measure the acceleration in the ambient air, and directly correlate it to the void sizes and radial distributions detected with scanning acoustic microscopy of bonded wafers. 

3.     Particle induced voids and the tail of voids created

Particle-induced bond defects manifest as relatively large voids trapped between wafers and split the oncoming bond front. Such voids not only create locally unbonded gaps in the vicinity of the particle, but also create a tail of voids downstream, in the direction of the front’s propagation, as the split-bond-fronts collide with each other in closing the unbonded region [5]. We aim to visualize this parasitic effect of void-tail formation using high-speed imaging, and fully characterize their formation in terms of the mechanical parameters of the system.