- Imec researcher derives self-consistent formulation for bond front velocity in flexible substrate bonding
- Model addresses lubrication flow and elasticity mechanics without explicit adhesion mechanisms
- Framework scales from continuum to molecular limits for wafer and die bonding
- Work published in Proceedings of the Royal Society A, May 2026
A new mathematical framework tracks how bonded regions propagate across flexible substrates during semiconductor manufacturing processes. Utkarsh Jain, a researcher at imec, published the work in May 2026 in Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, developing a lubrication-elasticity model that predicts bond front velocity without requiring detailed adhesion chemistry.
Bonding and assembly processes now occupy center stage in semiconductor manufacturing because they enable 3D stacking and heterogeneous integration for novel interconnect architectures. The progression of a bond front hinges on the balance of adhesion and dissipation forces. Imec and partner EV Group recently demonstrated wafer-to-wafer hybrid bonding at 200nm copper interconnect pad pitch, advancing the overlay performance required for logic-to-logic and memory-to-logic tier stacking.
Power-law substrate deformation governs gap closure
The lubrication-elasticity model elucidates near-field dynamics while staying within continuum limits, excluding adhesion mechanisms. When governed by lubrication flow in the gap and linear elasticity in the substrates, the unbonded part assumes a power-law shape as a function of the distance from the bond front. The approach isolates viscous dissipation from elastic effects, providing a tractable path to predict how quickly two wafers join.
The formulation crucially relies on a re-parameterization in terms of a ‘high viscous dissipation region’ ahead of the bond front, which travels along with the adhesion front. This region drives the fluid mechanics that squeeze out trapped gas between approaching surfaces. The mathematical treatment avoids case-by-case simulations and instead provides general scaling laws.
Velocity accelerates as unbonded region shrinks
The approach for expressing bond front velocity allows for its acceleration as the unbonded region shrinks, for example with the bond front approaching the edge of a finite substrate. Real wafers and dies have boundaries, and process engineers see faster bonding at edges—a behavior this model captures from first principles rather than empirical curve-fitting.
The model’s predictive power matters for process control. Bonding speeds affect throughput, and uncontrolled variations can trap voids or create stress concentrations that degrade electrical connections. A self-consistent velocity formulation lets engineers set temperature, pressure, and substrate stiffness to hit target bond front speeds, especially relevant as imec achieved copper pad-to-pad post-bond overlay vector below 40 nanometers for all dies across a full 300-millimeter wafer.
Continuum framework bridges to molecular scales
Unifying velocity and length scales are proposed, incorporating a lateral adhesion front length scale that effectively captures velocity variations in the limit of vanishing viscous dissipation, bridging continuum and molecular scales. This bridge addresses a persistent challenge: continuum mechanics fails when gaps approach molecular dimensions, yet molecular dynamics simulations can’t handle wafer-scale geometries.
The framework offers a handoff point. Process engineers can apply the lubrication-elasticity model for most of the bonding process, then switch to molecular-scale treatments only in the final nanometers of contact. The framework provides scalable insights for better understanding and process control in applications with wafer and die bonding. It gives fabs a physics-based tool rather than purely empirical recipes, which becomes essential as imec’s CMOS 2.0 scaling paradigm drives the wafer-to-wafer hybrid bonding roadmap toward 200nm interconnect pitch, partitioning a system-on-chip into heterogeneous functional tiers reconnected using 3D interconnect technologies.
Process engineers scaling down to sub-200nm interconnect pitches should revisit bonding time budgets and void-detection protocols. The lubrication-elasticity model shows that bond front velocity is not constant but accelerates as the unbonded region shrinks—meaning edge effects intensify at finer pitches. That acceleration changes the stress distribution during bonding and can shift the thermal profile, both of which affect yield. The self-consistent velocity formulation gives you a basis to tune pressure and temperature ramps rather than rely on legacy recipes developed for coarser pitches.
Why does bond front velocity matter for 3D chip stacking?
Bond front velocity determines how quickly trapped gas escapes between substrates. Too slow and throughput drops; too fast and voids form or stress concentrations crack interconnects. At 200nm pitch, alignment tolerances below 40nm leave no margin for uncontrolled velocity variations that distort substrates during bonding.
How does the lubrication-elasticity model differ from adhesion-based bonding models?
Traditional models explicitly calculate adhesion forces at the molecular level, which is computationally expensive and material-specific. The lubrication-elasticity approach isolates viscous dissipation and elastic bending, deriving velocity formulas that apply across different substrate materials. It bridges continuum mechanics to molecular scales without requiring full atomistic simulations for the entire wafer.
Article Source: Model Tracks Bond Front Velocity in Lubrication-Mediated Bonding of Flexible Substrates (imec)
