Economic, versatile MEMS prototyping
New wafer-bonding equipment enables rapid development
BY SHARI FARRENS, PhD, SUSS MICROTEC
Researchers and engineers in university and start-up labs all around the world are creating innovative MEMS devices on shoestring budgets. Economic and versatile tools for rapid prototyping, including a new type of wafer bonder, are key to enabling their productivity. For instance, SUSS MicroTec’s ELAN CB6L has all the features of a production bonder-except for maximum force and manual adjustments-and can perform every type of bond imaginable, including metal diffusion, metal eutectic, anodic, adhesive, glass frit, and fusion. Detailed understanding of the needs of each type of bond will help to explain why less is more.
The four variables
Most wafer-bonding applications involve only four variables: temperature, applied force, time, and control of ambient atmosphere/vacuum. All factors must be just right to produce the desired outcome: Temperature and force must be extremely uniform, and timing and pump/purge cycles must be properly controlled.
Because up to 80% of the cost of a MEMS device can be consumed by packaging, the wafer-level bonding step-which typically occurs near the end of the fabrication process with full-value wafers on the line-is critical.
In thermal compression bonding, temperature is the driving force for chemical reactions and diffusion at the bond interface. Metal diffusion bonds are done at 390ºC to 450ºC; glass-frit bonds generally require slightly lower temperatures, although the new lead-free frit materials are processed at 550ºC. These represent the upper limits of the thermal processing range required of any commercial bonding machine. All other types of bonding will be done at temperatures below this and the industry trend is to continually lower processing temperatures for bonding to allow for integration with other materials and CMOS components.
Temperature uniformity need not be sacrificed to reduce the cost of bonding tools. Nearly all processes require at least a ±1% range, which is standard on bonding equipment today. The most sensitive processes are the eutectics with very narrow solid + liquid, two-phase, region. The most notable of these is the Au-Sn system.
The other critical performance parameter related to temperature is alignment accuracy. Absent contributions to all other effects, silicon wafers will expand ~100μm for every 100ºC of elevated temperature. If the temperature change across the diameter of the wafer is uncontrolled, alignment shifts will affect overlay accuracy.
Application of force during bonding is a rather misunderstood parameter. Force is used solely to maintain intimate contact between wafers. But most substrates are not flat; instead they have significant surface topography, wafer thickness variations, and bow or warp conditions. Hence the application of force is used to “enhance” the conformity of the materials to facilitate contact.
This should not be misinterpreted to mean that simply because a wafer is forced flat and into contact with an opposing substrate that the final bonded pair will remain flat. The material with the dominant personality-meaning the thicker wafer, the stiffer wafer, or the most bowed wafer-will generally determine the shape of the bonded pair.
The force required to maintain contact and flatten substrates together also depends upon force uniformity. If the force is only 50% of set point in some locations, a total force of twice the needed amount will be required to achieve desired results in all areas of the interface. This is particularly important for metal-to-metal diffusion bonds. This type of substrate bonding generally involves wafers with significant shape problems due to stress from the metal deposition process, and the surface roughness of the metals limits physical contact. Contact must be maintained during a diffusion bond because atoms cannot “jump” across the bond interface. Thus, uniformity of force and ultimate force are uniquely intertwined in wafer bonding.
A huge benefit to working with soft materials like eutectics, glass frit, and adhesives is the ability to apply the force slowly over time. Force ramping, as it is called, allows gentle application of the force so that the substrates can conform to one another without slipping. In eutectic bonding force ramping also prevents the metal from flowing into unwanted areas. The wetting of the metal to the substrate is naturally confined unless the force is applied so quickly and with such magnitude that the metal is simply squeezed everywhere. In the end, <10KN of very uniform force applied in a controlled manner is sufficient for most applications involving substrates that are 150mm or less. The cost of higher force tooling is not necessary and is a burden to development NRE budgets.
Environment and time
Many of the new advanced MEMS structures are requiring increased hermeticity levels. This requires a transition from anodic and glass-frit bonding to metal systems or internal getter systems. In addition this work is facilitated by polished stainless-steel chambers. Pump and purge cycles are also employed to remove reactive gas components, such as oxygen and hydrocarbons, in the chamber. With unpolished surfaces the chamber will become contaminated over time, especially in R&D running various different types of bonds. Performance and maintenance are significantly improved with polished sidewalls.
Processing time, including alarm conditions, are very important to bonding. The SUSS ELAN CB6L uses manual loading and unloading to keep tool costs low, but an optional cooling station can be integrated, and the open-fixture design helps reduce cooling cycles inside the bond chamber.
As you have seen, a new generation of bonding tools provides all the functionality of automated and semi-automated production equipment for R&D work. In most 150mm applications, these tools can not only effectively facilitate development, but also dramatically reduce non-recurring engineering (NRE) charges-and thus help researchers and design engineers achieve successful functionality.
Shari Farrens PhD is chief scientist at SUSS MicroTec (www.suss.com). Contact her at firstname.lastname@example.org.