Glass frit wafer bonding remains a very attractive process for industrial applications. The main benefit is that the glass frit is an active bonding layer, which planarizes surface roughness and topography up to the direct sealing of metal lines at the bonding interface. This allows very simple process integration. The bonding yield and bonding strength are high, while the bonding interface is reliable regarding mechanical degradation and hermeticity. The processing costs for screen printing, thermal conditioning (firing) of the printed glass paste to a solid glass are moderate, but no extra costs arise from any specially-required preparations of the electrical connecting metallisation (no passivation or planarization is needed). The main drawback has been the relatively-broad sealing and bonding frame width of 300 µm or even more, but now this has been reduced, so that bonding frame widths of around 135 µm are possible in industrial glass frit bonding processes. The biggest challenge of glass frit bonding is the control of the pressure of the sealed MEMS devices. Resonant MEMS devices such as gyroscopes require a controlled cavity pressure which must not be higher than a certain value in consideration of the sensor design, since at higher pressures, the sensor loses functionality and precision. To investigate the inner cavity pressure, a good measurement device is needed, which we introduced at the ECS Wafer Bonding Symposium 2012 in the form of a heat transfer-based pressure gauge integrated in an Open Platform MEMS Technology utilising glass frit bonding. Using this measurement device, we were able to investigate and optimize the glass frit thermal conditioning process and the storage conditions before bonding, as published in 2014 at the ECS Wafer Bonding symposium. Based on these advances, a further improvement of the inner cavity pressure of glass frit-sealed MEMS devices, regarding pressure reduction and increased process stability (the target was a stable cavity pressure below 5mbar in mass production) was realized.The sealed cavity pressure in glass frit bonding is defined by the chamber pressure and the bonding temperature, both of which are controlled by the bonding equipment. During bonding at about 400...450°C, the chamber pressure is sealed, and in cooling down this value decreases by about 50%, related to the ideal gas equation. This is true, as shown in previous publications, for final pressures of around 15mbar. Below that, outgassing from the glass at the bonding temperature, during the time that the glass is still at the bonding temperature but has already flowed, then dominates – since the cavity is already sealed by the glass flowing, the gases emerging from the glass (mainly organics remaining after glass firing) are trapped in the cavity, and will result a higher pressure than expected. As mentioned earlier, glass firing and glass storage (glass frit is hygroscopic) were already considered to be optimal. Therefore, we continued investigating the bonding process to achieve an optimal overall process. The outgassing of the glass frit is a temperature- and time-driven process – the gas generation out from the glass needs a temperature which is the bonding temperature, while a time is required to get the gas out from the glass into the cavity. The influence of the bonding temperature was first investigated: the bonding temperature was reduced in steps of 10 °C, wafers were bonded, and the cavity pressure measured. The results showed very clearly that by reducing the temperature, the outgassing was reduced, to give a lower inner cavity pressure. The standard recipe showed maximum pressure values of about 1 bar, which could be reduced to values lower than 1 mbar. However, as expected, the reduction of the bonding temperature was accompanied by reduced bonding and sealing yield and lower bonding strength. From this, an optimal bonding temperature 10°C less than the so-far standard temperature was found as optimal – this gives a reduction in pressure values from 5 to 3.5 mbar, which is a considerable improvement. Next, the time duration at the temperature of the glass frit during bonding was investigated. We found this to be a rather minor influencing factor, as it appears that by the end of the heating-up process, glass is already flowing and the cavities can be sealed. Nevertheless, by reducing the time at a nominal bonding temperature, the typical cavity pressure was lowered by a further 1 mbar. With both measures, we could seal a cavity pressure of typically less than 3 mbar, which gives a good margin for the critically-defined value of 5 mbar that the cavity pressure must not exceed (see Figure). Figure 1
Search from 250 M+ research papersSearch from 250 M+ research papers
Open Access
Sep 1, 2016
Sihai Zhao + 4
