Key process controls for underfilling flip chips
04/01/1997
Key process controls for underfilling flip chips
Alec J. Babiarz, Asymtek, Carlsbad, California
Standard flip-chip packages require an underfill adhesive to reduce the stress on connections. The underfilling process is conceptually simple and fast, provided the process variables are understood and automatically controlled. The most important variables are dispense volume and adhesive viscosity. Viscosity changes rapidly with time and temperature, making temperature control critical throughout the system.
Advanced IC packaging is being driven toward smaller sizes with higher I/O requirements, especially for microprocessors, telecommunications, consumer devices, and automotive electronics. The most compact and efficient package for an IC, a flip chip, eliminates the need to wire bond connections from a die to a leadframe. In most flip-chip applications, the wire bonds are replaced by small metallic bumps on the circuit side of the die (Fig. 1). The connection from the die to a substrate is made by placing the chip circuit side down onto pads and reflowing the bumps to make the connection. This circuit-side-down placement of the die gave rise to the name "flip chip." The process was patented by IBM and called "C4," which stands for Controlled Collapsed Chip Connection.
The major advantages of flip-chip packages are higher I/O, higher speeds, lower emissions, smaller sizes, and better heat-transfer characteristics. The package allows an area array interconnect layout that provides more I/O than a perimeter interconnect in the same die size. Flip chips provide higher-speed performance because the interconnects are made on the inside of the die, allowing shorter transmission paths. The design has lower emissions because the wire bond loops are eliminated. Since the package only requires an underfill adhesive, the size of the package is the size of the die. The package also allows efficient heat transfer because a heat sink can be directly attached to the die, rather than working through an encapsulant cover or leadframe.
The actual assembly process is quite simple. A pick-and-place machine handles and places bare die and should be able to flux the pads or bumps prior to placing the die. After the die are placed, the bumps are reflowed onto the I/O pads. Next, the parts may go into a preheat station where they are heated to 70-100?C, then moved to an underfill dispensing station. The underfill dispenser delivers a volume of adhesive in a straight line along the die`s perimeter. Proper underfilling requires a heated substrate to heat the dispensed epoxy and lower its viscosity so that capillary force can "flow out" the epoxy under the die. After the first pass of dispensing, the die is moved to a post-heat station. In some cases, the post-heat station is used to flow out the epoxy. After flow out, another dispenser dispenses a bead of encapsulant around the entire die. This "seal" eliminates any induced stress caused by unsymmetrical fillets on the die. After the last seal dispense cycle, the parts are sent to a vertical curing oven.
Underfilling is necessary on most flip-chip assembly lines to reduce the stress on the solder bumps during temperature cycling. The stress results from the mismatch in thermal expansion between the substrate and the silicon die. As temperature increases, the substrate and die start to expand at different rates, but the solder bumps attempt to keep the die and substrate together. Therefore, large shear and tensile stresses are developed in the bumps. As a simple model, the stress is equal to the force divided by the summation of the cross-sectional areas of each bump. Assuming the force is constant, when an underfill adhesive is applied, the stress becomes the force divided by the surface area of the die. The difference is more than an order of magnitude of reduction in stress on the joints. Solder joints typically fail through fatigue. The adhesive also helps slow crack initiation in the solder bump/substrate interface, thereby increasing the fatigue life of the joint.
Figure 1. The traditional wire bonds are replaced by solder bumps on the circuit side of the die.
Underfill requirements
The optimum volume of underfill is the summation of the volume of adhesive in the fillet around the die and the volume of adhesive that is under the die. The volume of adhesive under the die is the area of the die times the bump height minus the volume of the bumps. The volume in the fillet is the sum of a triangular prism and a cone. The prism`s length is the perimeter of the die, and the cross section is a triangle with a height equal to the die thickness, and a base equal to the width of the fillet allowed around the die. The fillet at the corners of the die is modeled as a cone with a height equal to the die thickness and a base radius equal to the fillet width.
With the volume model, a tolerance analysis can be done on how much adhesive is required. If the spacing between die fillets is too small, and the bump height increases, the minimum required adhesive can actually exceed the maximum that can be dispensed without connecting dice. This is particularly evident in larger dice, and also shows that flip chips cannot be positioned too closely together or the adhesive in the fillets will overlap. The minimum spacing on a 15-mm square die is about 3 mm with the safety margin (Fig. 2).
Figure 2. The dispensing accuracy required depends on the geometry of the die and underfill fillet size; changing the fillet size from 0.75-1.5 mm in width allows wider dispense volume variation; a) 0.75-mm fillet width; and b) 1.5-mm fillet width. Note that in a) there is no dispense accuracy possible for die greater than 20-mm square.
The flow equations for underfilling have been worked out by Schwiebert and Leong of Hewlett-Packard [1]. The time to flow is given by:
t = (3 ?L2)/(hgcosq)
where
t = time in seconds
? = fluid viscosity
L = flow distance
h = gap or bump height
q = contact or wetting angle
g = surface tension of liquid vapor interface
The time of underfilling is proportional to the square of the length or width of the die. Also, by substituting typical underfill adhesive properties, the following equations show that gravity would only make a 20% difference in the time to flow under the die.
The capillary pressure drop is:
Dp = 2gcosq/h
The hydrostatic pressure drop is:
Dp = rgL/2
where
r = fluid density
g = acceleration of gravity
However, by substituting the same epoxy property numbers, vacuum could be used to increase the flow rate.
Dispensing process
Three key process control areas assure proper underfill dispensing. The volume control on the dispensing pump, the heat management of the dispenser environment, substrate, and valve, and efficient part management working in conjunction with the software control of the three process segments are all necessary.
Dispensing valve volume control
A rotary positive displacement valve (RPDV) may be used effectively in underfilling when volume control is added to the system. RPDVs have been called "positive displacement" in the industry for years; however, the valve is not true positive displacement because the pump`s flow rate is a function of the fluid`s viscosity, rotor speed, rotor geometry, and needle diameter. Additional nonlinear effects on flow rate appear if the fluid viscosity changes with shear rate or temperature. The auger pump flow is equal to the drag flow plus the pressure flow, which has to equal the flow out of the needle.
Underfill adhesives usually have viscosities in the 3-20 thousand centipoise (kcps) range after thawing. The epoxy`s pot life is typically 5 hr. (Pot life is usually the time in which the viscosity doubles. However, some underfill epoxy suppliers define the pot life as a 50% increase in viscosity.) These adhesive properties cause several issues with RPDVs. Since the rate of dispensing is dependent upon the auger speed, an increase in viscosity will cause an increase in drag on the auger. A closed-loop motor servo such as back EMF speed control is required to keep the auger speed constant. In addition, since there is a flow path from the syringe through the auger and needle, material can drip from the needle. The drip is prevented by placing a check valve on the output of the auger and using an auger valve that can reverse the auger direction to allow back flow. This type of arrangement will allow approximately 15% dispensing accuracy over the pot life of the epoxy.
To obtain greater dispensing accuracy, a mass flow calibration device is required to adjust the dispensing process as the viscosity of the epoxy changes and thereby obtain higher control. Using mass calibration, the dispensing accuracy of 5 kcps epoxies on multiple-pass dispensing applications of 50 mg can be better than 6% at 3 s with a Cpk of 1 (Fig. 3). The underfill
tolerance analysis should be used to determine what dispensing accuracy is required.
Mass calibration brings other benefits to the dispensing process. The history of the flow characteristics of the epoxy can be saved in statistical process control (SPC) data files for lot tracking. Also, the dispense programming process becomes easier because each dispensed line can have a mass associated with it. In this manner, trial and error are eliminated.
Figure 3. The graph shows that the flow rate of the valve changes due to the viscosity change of the adhesive with time; by using mass calibration, the amount dispensed remains constant.
Heat management
Since substrate heating influences the epoxy`s viscosity, the temperature of the substrate affects the flow-out time and each epoxy supplier has slightly different requirements. Heating mechanisms typically used are IR and contact heat.
Closed-loop IR can heat substrates that have pins or underside components without tooling. The closed-loop heating allows programming of the temperature ramp rate as well as the end temperature. In some FR4 FCOB (Flip Chip on Board) applications, the ramp rate of the substrate can affect the thermal cycling life of the component. By using an IR sensor, the actual temperature of the substrate is measured and the IR heat is regulated to keep the substrate temperature constant. Contact heating methods are inexpensive but the ramp rate cannot be well controlled. The contact method does produce very stable substrate temperatures.
The dispensing equipment must manage the heat in the system. Since the substrate is being heated, the entire environment inside the machine is subject to change. The viscosity of the epoxy changes greatly with temperature. An active heating and cooling mechanism for the valve must therefore be provided to maintain greater dispensing control. Further heat management requires multitasking software to control the dispenser and the proportional-integral-differential (PID) temperature controller for both the substrate and the dispensing head.
Part management
The last important step in the dispensing process is a good part-management system. Part management includes such issues as dispensing patterns, throughput optimization, flow-out timers, rework issues, and part queuing.
The dispensing pattern can be chosen to optimize flow-out time, minimize air entrapment probabilities, and equalize fillet size. The actual pattern to dispense and the amount of epoxy to dispense/pass is a process-development area that requires specific application tests. Some typical patterns are shown in Fig. 4. The best way to develop an optimum pattern is to use clear glass die. By using clear die, the underfill operation can be observed and the best pattern determined from the dispensing data. Larger die require multiple
Figure 4. When underfilling, multiple passes are used to control the fillet width, and software timers are used to assure that dispensing is not done too soon or too late; a) a multiple-straight-line pass and seal pass; b) a single L dispense and seal pass; and c) a modified U pass and straight-line seal that optimizes speed of flow out.
dispense passes to optimize the amount of material dispensed along the die edge. Software dispense timers are required for multiple dispense passes to ensure that the fluid has flowed under the die before another dispense pass is applied.
In FCOB applications, flip-chip attach and underfilling are the last assembly operations. Therefore, there is usually a high-cost investment in the PCB assembly before underfilling. Since, if a bad die is underfilled, it is nearly impossible to rework, part skipping is a required feature. The die can be skipped if they are misplaced, or a marked fiducial is used to indicate a bad part.
In most encapsulation applications, substrate heating is required for good flow. Depending upon the application, the dispensing equipment must be able to incorporate a preheat, dispense-heat, and post-heat station. In some cases, a preheat oven and a flow-out oven can be eliminated by managing the heating in the dispenser. The new underfilling adhesives can underfill a 10-mm square die with a 0.125-mm gap in 30 sec. Quick underfilling adhesives allow the flow out to occur so the seal pass can be made in the dispense station.
Conclusion
Flip-chip assembly is becoming more visible in mainstream surface mount technology (SMT) lines. The process fits well in standard SMT production lines, since no wire bonding is required. Advances in underfilling adhesives are making the process faster, thereby alleviating dispensing as the bottleneck. Production-ready equipment for underfill must control many variables to achieve reproducible results in high-volume production.n
Reference
1. M.K. Schwiebert, W.H. Leong, "Underfill Flow as Viscous Flow between Parallel Plates Driven by Capillary Action," ISHM Proceedings, October 1995, e-mail: schwiebe.corp.hp.com and [email protected].
ALEC J. BABIARZ received his BSME degree from Arizona State University and his MSME and MSEE degrees from Stanford University. He is a cofounder of Asymtek, and is currently senior VP overseeing marketing and sales. Asymtek, 2762 Loker Avenue West, Carlsbad, CA 92008; ph 619/431-1919, fax 619/431-2678, e-mail [email protected].