The long-term reliability of a flip chip (FC) device mounted onto a substrate highly depends on the quality of the underfill material between the chip and the substrate. Adhesive and the underfill process need much attention, which can be attained by a unique image system.
Underfill material serves several purposes. First, it compensates for the differences in coefficient of thermal expansion between the silicon chip and the substrate. The difference in thermal expansion is usually so large that it is likely to cause damage if only the solder bump interconnects must tolerate the stress created as the device is cycled through its normal temperature range of about 85K. Adding underfill reduces the stress on the solder bumps by a factor of about 10.
Dispensing and long-term reliability
Underfill serves additional purposes as well. By completely surrounding each solder bump, underfill holds each bump in hydrostatic compression and effectively prevents the solder from creeping, as it would do if there were an adjacent open space. This is one of the reasons why voids in underfill material can be serious defects. Underfill also limits the initiation of cracks in the solder bumps. Finally, the addition of filler particles adds stiffness to the cured underfill, and limits the potential flexing of the device package. A stiffer package is less likely to incur such defects as broken interconnects or a cracked die. Since underfill performs so many vital functions, correct application of underfill and careful inspection of the results are necessary.
Underfill adhesives (Figure 1) are manufactured to exacting filler tolerances that are measured by the material's specific gravity (SG) or density variation and the particle size and shape distribution. A good adhesive is one that demonstrates no discernible density variation from batch to batch or from syringe to syringe. The density of an underfill plays a crucial role in the establishment of the thermo mechanical properties of the FC assembly. If it varies, so may the ultimate reliability of the device. Consider the material's density as a weighted average of all the SG values in the formula. Epoxy resins have a SG of 1.24, while fillers have a SG above 4.0; therefore, a resin/filler mixture obtains the density of the ratio of its individual components. If the resin/filler ratio varies, the thermal properties vary, which in turn causes the material supplier to reject the batch. A good supplier does not let a material with an improper resin/filler ratio get into the hands of the customer.
The particle size distribution also affects the mechanical properties, the flow-out ability, and the separation/settling characteristics of the adhesive. The pump used to dispense underfill adhesives must not affect the particle size distribution. Lab tests have shown that rotary pumps constructed of hard materials such as tungsten carbide can grind the filler particles to finer sizes, thereby affecting the viscosity of the encapsulation adhesive. Linear positive displacement pumps and time pressure dispensing valves do not affect the adhesive characteristics because there are no grinding wetted parts in the systems.
In planning the production of a FC device, the design of the underfill process is an important step. There are many ways in which the fluid underfill adhesive can be dispensed around the periphery of the chip. Successful dispensing designs will be both reasonably fast and unlikely to result in underfill-related defects.
The simplest method is to dispense the fluid underfill material along a single edge of the device. Since the material has a single wavefront as it flows by capillary action between the substrate and the chip face, the formation of bubbles or unfilled regions is minimized. Drawbacks are that flow along a single wavefront may be slower than other methods, and that the initial fillet where the fluid material is dispensed may become large and unwieldy.
More commonly, dispensing takes place along two adjacent edges, or even along two adjacent edges and a portion of a third edge. There are also patterns that employ dispensing single dots of material along the die corners. During process development it is very helpful to observe the course of underfill through a glass slide or clear die that takes the place of the silicon chip, which is opaque to light. In this way the interaction of the wave fronts, and the interaction of each wave front with solder bumps, can be observed and adjusted. Later, inspection of features in the cured underfill can be made with ultrasound, to which the silicon is very transparent.
The volume of material to be dispensed for proper underfilling may be predicted by performing a tolerance analysis on the die size, bump height, and fillet size requirements. If the tolerances on fillet size are too tight, the design may be difficult to manufacture. Application notes on tolerance analysis are available from Asymtek. In the production environment it is very important to assure proper volume dispensing. Linear positive displacement pumps provide accurate volume dispensing while rotary pumps and pressure-time valves dispense volumes that change with material viscosity. A mass calibration feature is preferred to assure the proper amount of material is dispensed, because mass calibration works in conjunction with standard production quality assurance methods. A pure volume measurement system is difficult to verify in production because after dispensing under the part, the amount of material dispensed can only be verified by measuring the mass of the pre-dispensed part compared to the mass of the post-dispensed part. Therefore, the best dispensing practice is to use a linear positive displacement pump to assure proper volume and mass calibration to verify dispensing under proper quality assurance methods.
Underfill defects can be generated by factors that have little to do with the actual dispensing design. Variations in temperature along the substrate can lead to variations in the speed of converging wavefronts; these variations in turn can lead to the formation of voids. Some underfill adhesives exhibit narrow temperature (±5K) bands where successful flow-out occurs. Contamination of the die, the substrate, or the solder bumps with residue (usually flux material) can result in incomplete wetting of these surfaces. Although the underfill material may nominally fill the complete under-chip volume, incomplete wetting will cause delaminations along these surfaces. Acoustic imaging is especially useful in identifying the proper flow-out temperature.
Acoustic micro imaging of cured underfill
Acoustic micro imaging permits non-destructive viewing of the underfill as well as solder bumps and their bonds, the substrate, the die, and any defects that are present. In a sense this method is similar to the viewing of underfill through a glass slide during process design; the difference is that acoustic micro imaging views internal features post-cure through the silicon die or, in some instances, through the substrate.
The small size of the internal features in FC devices has been a major factor in the recent development of very high resolution acoustic methods. Acoustic micro imaging systems in general employ ultrasonic frequencies ranging from 5MHz to above 200MHz. The higher frequencies give greater spatial resolution but have less ability to penetrate into relatively thick samples. A few years ago the highest practical frequency was 100MHz. The demands of flip chips have led Sonoscan to develop transducers putting out 180MHz and more recently 230MHz. The differences between these frequencies become very significant when it is necessary to image and analyze very small features such as a tiny void next to a solder bump or the distribution of filler particles within the underfill material.
Ultrasound, which is emitted by the transducer above a FC, travels downward through the device. When the ultrasound meets with an internal interface (between the die face and the underfill material, for example), part of the ultrasound is reflected back to the transducer, where it is received. The remaining ultrasound continues downward through the device. The amount of ultrasound reflected at each interface is a function of the acoustic impedance of the two materials involved, and can be calculated. A defect-free FC package will thus produce images of various internal interfaces. The most dramatic differences in acoustic impedance are those where one of the interfaces is a gap-type defect, which behaves like non-transmissive air. Delaminations, voids, disbonds, and cracks are all examples of gap-type defects. The relatively huge differences in acoustic impedance at these interfaces make gap-type internal defects very easy to image. But the sensitivity of very high frequency (VHF) ultrasound to differences in acoustic impedance also easily images non-gap features such as the interfaces between filler particles and the surrounding epoxy. Because the return echoes are separated in time, an image can be made using only the return echoes from a given level within the device. The term "electronic gating" is used to describe this method, and electronic gating is often used with FCs to restrict an image to the level at which the underfill material interfaces with the chip, to give one example.
Many underfill defects (missing, voids, delaminations, and cracks) represent gap-type defects. The distribution of filler particles, however, does not involve a gap. Filler particles, like larger internal features, are imaged because of the difference in acoustic impedance between the individual filler particle and the surrounding epoxy.
The acoustic image of filler particle segregation is seen in figure 2. The lighter areas in the underfill represent higher-than-normal concentrations of filler particles. The segregation into such concentrations poses a hazard to future reliability because of the altered thermal characteristics of the cured underfill. In addition, Sonoscan's applications laboratory has repeatedly observed that particle segregation is very often the site of voids; in the device shown here, two such voids are present.
The FC shown in the acoustic image in figure 3 has two essentially different types of defects. The black areas are very strong concentrations of filler particles; in the center of one of the areas of particle concentration, a large void has formed. Smaller voids have also formed around many of the peripheral solder bumps. Some of the voids near the solder bumps are halo defects, or thin, vertical voids adjacent to and partly or entirely surrounding a solder bump. Halo defects were first seen in Sonoscan's laboratory. Because of their thinness, their vertical orientation, and their location, they would be very difficult to find by non-acoustic methods such as grinding. Halo defects probably form when flux residue prevents the fluid underfill from thoroughly wetting a solder bump. Their location permits solder to creep into the void.
By giving a non-destructive and rapid means to visualize and characterize internal underfill features post-cure, acoustic micro imaging allows manufacturers to identify both individual defects and the probable causes of those defects. It is especially useful in identifying those hidden internal defects, which, although not immediately damaging, will eventually grow and cause the device to fail.
Fax Sonoscan +44-1747-855-938; Fax Asymtek +44-161-495-4272
Die Zuverlässigkeit montierter Flip-Chips hängt entscheidend von dem als Underfill zwischen IC und Board dosiertem Material ab. Bei dem Temperaturunterschied von 85K wirken sich die unterschiedlichen thermischen Ausdehnungskoeffizienten als mechanische Belastung auf die Kontaktbumps aus. Guter Underfill kann den Streß um den Faktor 10 reduzieren.
La fiabilité des flip-chips montés dépend en grande partie du dosage du matériau employé comme underfill entre le IC et le board. Avec des différences de température de 85 K, les différences de coefficients de dilatation thermique ont pour effet des contraintes mécaniques sur les bumps de contact. Un bon underfill peut réduire ce stress d’environ dix fois.
L’affidabilità dei Flip-Chips montati dipende dal materiale dotato come underfill tra il circuito integrato ed il board. In caso di escursione termica di 85K i vari coefficienti di dilatazione termica influiscono come carico meccanico sulla memoria ausiliaria di contatto. Un buon underfill può ridurre lo stress del 10%