Post-bond annealing acts on the copper and interfaces inherited from earlier processing. It can promote useful connection, yet an unfavorable incoming copper state can also contribute to later void formation. The later operation must be interpreted as part of a material history.
Two interfaces meet at bonding
The cited hybrid-bonding study distinguishes copper connections from dielectric bonding. Surface geometry affects whether opposing copper pads make the intended contact. The state of copper within those pads also matters; flatness alone is not the complete incoming condition.
The research reports a relation between earlier electrodeposited-copper treatment and voids observed after later bonding heat. Its interpretation involves condensation of inherited vacancies near the bonding interface. This is a reported mechanism in that study, not a claim that every post-bond void has the same cause.
Process checkpoint
Understand CIS/ISP Wafer TC Bond in context
Identify the meeting copper and dielectric interfaces and explain why their incoming states matter to bonding.
Process context for “Why Post-Bond Annealing Cannot Be Judged Without the Earlier Copper State”: 40nm BSI CMOS Image Sensor · BOND · Step 290
More heat is not a one-direction outcome
Atomic mobility can support connection and microstructural evolution. The same ability to move atoms also allows an unfavorable initial state to evolve into a different defect structure. A thermal step therefore needs an intended outcome and an evaluated risk, not the assumption that every structural change is healing.
This is different from treating a bond as merely two clean surfaces pressed together. Bulk and near-interface history can survive into the meeting surfaces' later evolution.
Conditions and counterexamples
A mechanically joined stack does not independently certify every electrical connection. A flat pad does not certify low vacancy content, and a resistance observation does not alone identify the microscopic origin of a void. The study evaluates structural and electrical evidence separately.
Do not transfer its anneal thresholds, connection pitch or measured reliability to this Flow. The supported lesson is the dependence on incoming state and the need for distinct endpoints.
Trace the handoff between stations
At bonding, identify the copper and dielectric interfaces. At the later anneal, explain the desired strengthening and one conditional inherited-state risk. State what evidence would distinguish a joined stack from reliable electrical connections.
The task adds a prehistory-dependent defect mechanism to the existing thermocompression overview, rather than rewriting its joining sequence.
References
Novel stacked CMOS image sensor with advanced Cu2Cu hybrid bonding
Y. Kagawa, N. Fujii, K. Aoyagi, Y. Kobayashi, S. Nishi, N. Todaka et al. · International Electron Devices Meeting