Design and process issues in optoelectronics
BY JOE BELL
In the last decade, microelectromechanical systems (MEMS) have moved from the university laboratory into mainstream use in a multitude of commercial products. New uses for these chip-level devices are being discovered daily in the information technology, military, environmental, space and biomedical industries. These new discoveries have fueled the rapid growth in existing product and emerging product markets. It is estimated that in 2000, MEMS represented a $2 to $5 billion industry and is expected to reach $8 to $15 billion by 2004. Table 1 shows MEMS proliferation in today's markets (Table 1).
Although these are promising numbers, growth rates far exceeding these estimates are expected when makers of consumer products find ways of reducing MEMS production costs. Costly manual manufacturing processes are the primary impediment to a potentially phenomenal growth rate in MEMS markets. For many in the MEMS packaging industry, then, the answer has become automation.
Today's manual packaging techniques allow costs to be driven down only so far. Automating these packaging processes will allow component providers to further reduce costs by taking advantage of the economies of scale, reduce yield losses by eliminating the need for human touch, and increase package reliability and performance through repetitive and accurate component placement and attachment.
MEMS Packaging Challenges
The semiconductor packaging industry has forged the path for automated component assembly, providing today's MEMS manufacturers with lessons that are directly transferable to MEMS packaging. Even with this foundation, automating a MEMS packaging process presents some unique challenges. By understanding these challenges and addressing their solutions, component manufacturers will have the knowledge they need to automate their MEMS packaging process successfully.
For today's MEMS manufacturer, low volume and high complexity have made process automation a low priority. Consequently, many MEMS device providers never consider automation when designing their components. Although the need for full automation may not be a requirement today, up-front design for automation will reduce future redesign costs when demand for these products grows.
Designing MEMS devices for automation means ensuring they possess the precise physical features and tight manufacturing tolerances needed for an automated assembly process. Generally, the machine can place parts only to the accuracy allowed by the quality of the components it is placing. For example, in the telecommunications industry where optical MEMS devices are used in switching arrays, the MEMS must be placed with a high degree of accuracy - on the order of 5 µm in the X, Y and Z directions, and 0.2 milliradians in theta. This and other similar applications require the components to be of the highest quality.
Figure 1. MEMS being placed into leadless chip carrier with customized vacuum pickup tool and bonded using thermocompression.
Physical features, such as metalization patterns or fiducial marks, are used by the machine's automated vision system to recognize the component's location and orientation. Often the required placement accuracy of a MEMS device is with respect to its microstructure. Therefore, MEMS manufacturers should ensure that the fiducials or metalization patterns are benchmarked to that structure. If the fiducial and structure variance is greater than the acceptable placement accuracy, the required accuracy will never be met.
When placing MEMS devices relative to their edges, size variance and chip-outs cause vision-guided placements to vary outside the acceptable limit. Chip-outs from poor dicing can cause the vision system to misidentify die corners, causing inaccurate theta placements. Variance in die size will cause the machine to skip parts or cause the critical microstructure to be placed inaccurately due to the edge-to-structure dimension variance. Vision systems with improved software capabilities can overcome some of these chip-outs and structural variances. However, correcting the problem before it arises requires significant attention be paid to the wafer dicing process to ensure that the highest quality methods and equipment are used.
Automation requirements must be considered when designing MEMS devices, keeping in mind that the highest degree of accuracy is achievable with the highest quality components. By designing components for automation from the beginning, MEMS manufacturers will save critical time and money when scaling their manufacturing operation to meet increased demand.
Presentation and Handling
Because MEMS devices typically are fragile, they must be handled with care. The slightest shock to these devices can damage the microstructures, causing altered performance. This is the reason that human intervention or manual packaging processes of such delicate components contribute to lower yield. In an automated packaging process, the need for human intervention to touch or handle these components is minimized.
Presenting components for assembly often is done in waffle or gel paks. These carriers then can be placed either on special tooling for the machine to access or presented automatically. Packages can be presented similarly or in standardized process carriers on an in-line conveyor (Figure 1). In any case, designing the proper presentation method can minimize the need for human intervention.
Figure 2. Vacuum pick-up tools can be customized to pick-and-place MEMS without damaging the fragile microstructures.
Once the parts have been presented for assembly, they must be picked and placed into the package for bonding. To ensure no damage occurs during this operation some automated assembly machines are equipped with as many as eight vacuum pick-up tools. Each tool can be custom designed to pick particular components in the process (Figure 2). Such tools are ideal for handling MEMS devices because they are designed to avoid the device's "no touch" areas yet provide ample vacuum to hold the part during the pick-and-place operation. Highly controllable force and movement trajectories are needed during the pick-and-place operation to move the component into the package and hold it for the bonding process. In some cases, the component must be turned over prior to being placed, requiring the assembly machine to have flip chip capability. Still others require the use of a "look-up" camera to reference fiducials or features on the bottom of the MEMS device.
An automated MEMS packaging process provides an immediate yield increase over a manual process. This is because an automated assembly cell is capable of interacting with various presentation methods, as well as moving and assembling the most delicate of components.
There are three main attachment methods for MEMS packaging. While each method carries its own benefits and drawbacks, all three typically are used in MEMS packaging.
In the first method, epoxy, either thermally curable or UV-light curable, can be daubed or dispensed to attach the device. Traditionally, daubing is used for dot sizes less than 15 to 20 mils, while dispensing is used for larger dot sizes and for lines - although automated nanoliter dispensing equipment now is available for smaller devices as well. Epoxy has long been used for die attachment purposes and comes in various types to suit almost any application. Outgassing resulting from the epoxy curing process has been a concern when applied to MEMS packaging. Outgassing can cause deposits to accumulate on the MEMS microstructures, inhibiting their operation. However, improvements have been made resulting in epoxies that minimize outgassing in almost any environment.
A key advantage to using epoxy is that it is flexible in its ability to adhere components in numerous complex applications, such as stacking optical MEMS devices. By using an in-situ UV cure, epoxy can provide the support mechanism for stacking MEMS devices. In-situ epoxy curing, either thermally or by UV light, also can ensure components do not move during the sequential build until final cure is complete.
The second method of attachment is by eutectic soldering. MEMS components can be attached using either a eutectic solder preform or by having the submount pad pre-metalized with the solder alloy while the MEMS itself is metalized with gold on its attaching surface. In deciding which method to use, cost analyses should be done on both options. Submounts with pre-metalized solder tend to result in higher throughput but are expensive. Preforms, on the other hand, result in the same quality bond but are cheaper. Preforms also can be more tedious to work with and decrease throughput due to an extra pick-and-place operation per die.
To perform the eutectic die attach correctly, the automated assembly cell must have the ability to pick-and-place the preforms if necessary, but it also must be able to tightly control the heat pulse while the bond head holds the part in place.
Some MEMS manufacturers feel that eutectic attach is too rigid or induces stresses at the bondline that will inhibit the microstructure's operation. However, it has been shown that using the correct eutectic solder will provide adequate flexibility at the bondline. Also, tightly controlling the eutectic solder reflow through accurate and repeatable temperature control minimizes bondline stresses during the attachment process. Through this control, voiding is minimized, further reducing the bondline stresses and increasing the quality of the mechanical, electrical and thermal characteristics of the attach.
Finally, in the third method of attach, ball bumping can be used to put small deposits of pure gold bumps on the back of the die (via a flip chip process) followed by a thermal-compression bond attachment procedure. Heat, force or ultrasonics cause the gold bumps on the die to diffuse with the gold on the submount pad, creating the bond. Another ball bumping method is to deposit solder bumps on the submount pad, place the MEMS device onto the pads and then reflow the solder.
MEMS devices are chip-level systems that use an electromechanical input to affect the physical environment. In some cases, such as in an optical switch, the MEMS component is used to affect the light path of an optical signal. Any variation in the MEMS placement causes the device to improperly alter the path of the laser beam, seriously affecting the process yield. To achieve the proper functionality, the MEMS must be placed to a high degree of accuracy in all six degrees of freedom: X, Y, Z, theta, pitch and roll.
Achieving accuracy to this degree requires not only quality parts, but also a machine with a high degree of control that can place and bond the MEMS components into the package repeatedly. X, Y, Z and theta accuracies depend on the machine's precise mechanical movement and its ability to accurately provide positional feedback for movement correction. Pitch and roll, or flatness, depends on the ability to achieve coplanarity between the working surface and the die itself, and to some degree on the mechanical ability of the machine to maintain planarity throughout the vertical motion of the placement. In all cases, a machine that is highly repeatable in all its motions can be programmed to place MEMS devices with the highest degree of accuracy required in all six axes.
The machine also ensures high placement accuracy by its ability to hold the part in place throughout the bonding process. The purpose of high placement accuracy is defeated if the machine moves the part during bonding. Whether it is in an epoxy, eutectic or thermal-compression bond, the highest degree of accuracy occurs when the part is held while the bond is made (in-situ cure/reflow).
Cleanliness is key for the packaging of MEMS devices because of the small scale of the MEMS structures. Dust particles and moisture can affect microstructure operability seriously or even cause inoperability. Whether the packaging process is manual or automated, cleanliness standards should be attained and followed to ensure the highest yield.
In anticipation of the market's growing demand for MEMS products, a high-volume, high-yield manufacturing solution is the answer - design for automation. MEMS component designers must do the up-front work in designing their MEMS devices for volume production. The process, as discussed, requires care and due diligence. The value brought to the customer in lowered costs, and to the manufacturer in increased throughput and yields, translates to success for all. AP
Joe Bell, applications engineer, may be contacted at Palomar Technologies, 2230 Oak Ridge Way, Vista, CA 92083; (760) 931-3763; Fax: (760) 931-5191; Email: email@example.com.