The importance of high accuracy adhesive bonding has grown with mobile phone and virtual reality headset applications and is still widely used in microfluidics, MEMS/MOEMS, capping of cavity structures such BAW and SAW devices and also in wafer level packaging. The well-known advantages of this bonding technique are: relatively low temperature process, no need for voltage or current, possibility of joining any types of material, possible low cost and even, to some extent, more tolerance for particles which presence doesn’t necessarily mean a failure as it would in the unforgiving direct bonding.

From our experience though we know that adhesive bonding can be quite problematic in many aspects, starting from application and problems with adhesion to the substrate, through establishing precise lithography parameters – the material needs to be cross linked to a certain level not to flow too much but can’t be overcured as this will prevent bond formation; some materials flow so much that the bond line broadening may not be acceptable and so the bonding pressure/force can be critical; and one of the biggest challenges is maintaining the alignment thorough the process while the adhesive goes into low viscosity state and acts like a lubricant.

In the first part of this study we would like to share our work on characterisation of PermiNex 1000 series that can be a simple solution to problems listed above. This material is manufactured and was recommended to us by MicroChem, and with MicroChem support we have developed a void free and high accuracy bonding process using this material. We have performed all work on AML Aligner Wafer Bonder, using in-situ optics and are sharing live view of the alignment and bond formation, highlighting how much easier and more effective qualification of any adhesive bonding process can be, when using this tool.

We present our data on <1µm alignment accuracy that we achieved and prove no alignment change while wafers are brought into contact. Figure below shows bond formation while the wafers are in contact and the adhesive is heated up.

In the second part of this work we are presenting some data from attempts of BCB bonding as an example for those applications that can’t use the PermiNex because of its optical properties. This shows how much we can learn about any other adhesive by observing its behaviour in-situ during bonding.

In our experience, the information obtained from the live view can speed up process development significantly from months to even days.

We can learn about the viscosity of the material at different temperatures and this knowledge is needed to be able to align accurately and maintain the alignment until the adhesive is fully cured and also for adjusting the heating rates and hence controlling the cross linking.

For example aligning and contacting hot BCB may introduce some misalignment at contact whereas contacting at room temperature can overcome this problem.

We can also watch the bond frame broadening and so adjust the bonding pressure and moreover we are able to tell at what point/temperature to apply the force.

Finally we are able to observe unwanted voids forming and react to those defects that may be created at any time during the process. This for example may tell us if the adhesives are outgassing and the voids may be trapped in the interface if wafers are contacted too early.