Enhancing Flip Chip Reliability
THE FLUX UNDERFILL INTERFACE
BY FRITZ BYLE AND DAVID EICHSTADT
The electronics industry is not known for being reluctant when it comes to change. Far from it, the industry embrace changes, even seems to need it. The industry constantly is driven by demands for smaller, better, faster, cheaper, and more ecologically sound product designs. The packaging world is currently experiencing its most rapid rate of change in many years due to the confluence of several factors - some evolutionary and some disruptive. The Reduction of Hazardous Substances (RoHS) initiative is demanding changes in materials that are not based directly on performance, but on environmental needs, while electrical performance requirements are driving a much-delayed transition to flip-chip-in-package (FCiP) for new designs.
The increasing complexity of many chips is driving higher I/O density, and better dielectrics are required to meet on-chip performance goals. The implication of all this is that the currently specified material sets for FCiP are inadequate to meet near-term demands. The higher solidus temperature and stiffer nature of the lead-free solders result in higher residual stresses post-soldering. The mechanically weak nature of the new low-k dielectric materials incorporated in high-performance die are demanding lower stresses be placed on the chip at the same time that the lead-free initiative implies higher stresses. Lower standoff heights for mounted die, a result of higher I/O counts and lower I/O pitches, also raise the stress levels in both the solder and the underfill.
Implications of Disruptive Influences
There are specific challenges created by the disruptive influences for underfills, fluxes, flux residues, and the less than perfectly understood interactions between underfill and flux residue. The transition to high tin content, lead-free solders offers problems for no-clean flux residues in FCiP applications before considering the interaction with the underfill. The increased amounts of tin oxides associated with lead-free solders require increased flux activity to reduce. Conversely, decreasing gap heights and bump pitches imply that less flux is available to do the work, and increases the difficulty of complete cleaning of the residues. No-clean fluxes intrinsically remove the cleaning problems; however, the residues must not impede capillary underfill flow in packages with 75 µm and less gap height and 150-µm pitch. Further, residues should not flow and maintain mechanical stability at temperatures as high as 260°C. This high-temperature stability combined with increased activation requirements constrain the parameter space of a successful flux candidate before interaction with the underfill is even considered.
Underfill interaction with a rosin-based no-clean flux residue has the potential to aggravate the initiation of underfill delamination. Initiation, not propagation, is the critical step in delamination failures, in that it takes many more thermal cycles to create than propagate. The low contact angle associated with flux residues creates a natural interface for delamination initiation if the underfill is not compatible with the residue (Figure 1). In addition to poor adhesion of the underfill to the residue, mechanically weak residues at the high temperatures of thermal cycling are defect singularities that are already initiated. This underscores the need for flux residue that is both mechanically robust and compatible with the underfill.
Changes in Underfill Technology
There is no material in the package that has had more severe demands placed on it by recent events than the capillary underfill. It is being forced to flow faster, through smaller gaps, and to reduce interfacial stresses at the same time the underfill must prevent excessive strain energy from accumulating within the solder joints. More difficult lead-free versions of the JEDEC MSL (Moisture Sensitivity Level) tests demand additional resistance to the diffusion of moisture through both the bulk of the underfill and at the interfaces.
Straight epoxy resins are giving way to blends of epoxies and other polymers in the newest generation of underfills. The resin blends help optimize the mechanical properties of the underfill and mitigate diffusion of water through the underfill material. Filler systems are becoming highly engineered, with material morphology and size distributions under tight control. This is required to maintain the high filler loadings necessary to optimize mechanical properties while simultaneously avoiding unwanted increases in viscosity that could negatively impact capillary flow.
Figure 1. No-clean flux cross-sectional view.
The most recent generation of capillary underfill materials are highly sophisticated materials. They rely, however, on a solid bond at the die and substrate interfaces to successfully accomplish the job. The weakest link in this bond has always been the flux-underfill interface. This interface lies in a crucial area to reliability: the area around the base of the bump. Delamination in this area initiates mechanical decoupling between the die and the substrate. Figure 1 shows how stress is concentrated at the edges of the flux residue, and how this can greatly increase the tendency for crack initiation. Delamination around the bumps rapidly leads to accumulation of large stresses within the solder, ultimately leading to bump failure through fatigue cracking. The traditional approach of qualifying fluxes and underfills separately implies less than optimal behavior of the resultant ad-hoc system. Eliminating this weak interface demands an engineered flux-underfill system.
Impacts on Flux Technology
The transition to lead-free packaging has significant implications on flux technology. Traditional rosin-based tacky soldering fluxes (TSF) have been used with mixed success in lead-free packages. The higher reflow temperatures and greater oxide levels of the high-tin alloys, as well as slower wetting of the lead-free alloys, has made it important for the activity level of the flux to be higher than that required for SnPb eutectic packages. Smaller bumps also demand that the flux amount used for each bump is reduced, further reducing the available fluxing action and raising susceptibility to flux depletion prior to completion of the reflow process. Although increased activity is a requirement, ion mobility in the residue cannot be increased or there will be a risk of degraded electrical performance. These two flux requirements are directly at odds.
The rosins traditionally used in flip chip fluxes also can decompose at the higher lead-free processing temperatures. The resulting residues will be glassy and brittle, with a surface poorly suited for underfill adhesion. This is especially troublesome because the flux will be exposed to the higher reflow temperatures on two and possibly more occasions: first, when the chip is mounted, and second, when the package is mounted in the final assembly. A third reflow pass is possible if the package is mounted on the first side of a two-sided assembly, and additional reflow cycles are possible if rework is performed. This is why JEDEC specifies that a package must pass 3 × 260°C reflow cycles without delamination. Some package manufacturers test to 5 × 260°C reflow cycles.
The mechanical and physical properties of the flux during and after reflow are critical to the performance of the subsequent underfill process. If the liquid underfill does not readily wet the flux residue, it will slow the flow of the underfill and affect the shape of the flow front. If there is too much flux residue and this residue is concentrated around bumps, it may mechanically impede flow, as well as displace the more mechanically robust underfill in the critical volume around the bump. The result may be a package with flow voids or separation of filler and resin in the underfill. Both phenomena are detrimental to package reliability. After capillary flow and curing of the underfill is complete, the nature of the flux residue and its interaction with the underfill can still impact package reliability to a great degree because the flux lies in the critically-important area around the bump.
For high-performance packages (large die, high power, high I/O, or high-reliability applications), rosin-based fluxes are not compatible with lead-free assembly. The demands on the flux-underfill interface are too great, and the packages may not survive multiple reflow cycles, much less have acceptable lifetimes in the field. For these aggressive package designs, a new paradigm is required. A concept recently introduced is a no-clean polymer flip-chip flux (NPF). This type of flux differs dramatically from traditional rosin-based fluxes in several ways. The NPF flux materials are more fluid than rosin-based systems, and have very low wetting angles on virtually all substrate materials. During reflow, the NPF flux flows out into a thin layer. Although the amount of residue as a percentage of flux applied is as great as 80%, (compared to 40 to 60% for rosin fluxes), there is less flux initially applied (lower viscosity flux), and the thin, uniform layer of the epoxy-based flux residue does not interfere with underfill flow. Figure 2 compares traditional no-clean and NPF impact on underfill flow.
Figure 2. Comparison of traditional no-clean and NPF impact on underfill flow.
Figure 3. NPF flux before underfill.
As the flow front progresses, the NPF ensures that the wetting angles on the substrate and die are similar, reducing chances of flow voids caused by the splitting and re-joining of the underfill flow front. Figure 3 shows a cross-section of a package prior to underfill with the NPF residues in place. Note the difference in the area of the flow channel between bumps, compared to the no-clean scenario shown in Figure 1. After reflow, the NPF residues are cross-linked, essentially forming a mostly-cured epoxy layer on the substrate. During underfill curing, these residues crosslink with the underfill, forming a robust mechanical bond. Figure 4 shows a cross section of underfilled and cured chip. The NPF residue has interdiffused with the underfill, and is not visible in cross sections performed after underfill cure. The absence of a concentration of residue in the immediate vicinity of the solder bump ensures that the bump is constrained as effectively as possible by the underfill, limiting the strain energy imparted during thermal cycling.
Figure 4. NPF flux after underfill.
While the NPF flux system seems to be an ideal solution, there are some tradeoffs. Although the NPFs are effectively a drop-in replacement in the flip chip attach process, their lower apparent viscosities can lead to different behavior on thin film fluxers, and to less adhesive force after chip placement. For higher I/O die, the reduced tack is not a problem, and the thin-film fluxers can be tuned for the different viscosity. More important, NPFs have moderate activity levels; therefore, the substrates and bumps must be maintained in a relatively clean state prior to joining. Proper control and storage of incoming materials and attention to detail in process development can easily surmount these issues.
When successfully processed, packages built with an NPF show significantly increased mechanical robustness and resistance to delamination. For some package designs, it may not be possible to pass JEDEC MSL tests without use of NPF. In other cases, the NPF may help the package pass at a higher level. The end result is more flexibility for the end user of the package and increased confidence in reliability in use.
For packages using smaller die and/or die with larger bumps/pitches, it is still possible to obtain acceptable reliability with rosin-based flux materials. The latest generation of rosin-based fluxes have been formulated to be heat-stable to 275°C and for greater activity, while leaving reduced amounts of residue. This combination of lower amounts of residue and increased thermal stability enhances package performance compared to previous generations of no-clean fluxes. The rosin-based flux system maintains full drop-in compatibility with existing processes, while providing a path to reliable Pb-free packages. Higher activity ensures great process robustness and compatibility with the widest array of substrate metallizations and bump alloys short of a water-soluble TSF.
Based on the severe demands placed on the flux-underfill material system, it may be assumed that rosin-based no-clean fluxes have no place in flip chip packages. For high-end packages, this is the case; however, for many packages, the latest generation of rosin-based fluxes can provide more than adequate package reliability while retaining all of the positive processing attributes that have made rosin-based fluxes the industry standard for decades.
For packages with more difficult geometries, material sets, and reliability targets, a different solution is necessary. Unfortunately, it is not possible to clean flux from the small gaps under large die, so water-soluble fluxes are not an option. No-clean polymer fluxes are one solution to this conundrum. Although these products are in their first generation, there are strong indications that they will be able to provide significant gains in package robustness while maintaining high levels of process compatibility.
It is clear that as flip chip packages continue evolving to support upcoming generations of semiconductor technology, the flux-underfill system also will need to continue evolving. It will be imperative that flux and underfill are designed to work together as a system. To date, they have simply inhabited the same ecological niche with a sometimes-strained relationship. What is envisioned for future solutions is more of a symbiosis, where one material’s function is enhanced by the presence of the other.
FRITZ BYLE, corporate technologist, and DAVID EICHSTADT, advanced products engineer, may be contacted at Kester, 515 East Touhy Ave., Des Plaines, IL 60018; (847) 297-1600; e-mail: firstname.lastname@example.org, email@example.com.