Die bonding with polyimide tape
01/01/2003
Reducing Air Bubbles During the Process
BY STEFAN BEHLER
The reliability of a plastic package is reduced by the presence of bubbles, since moisture can accumulate there. Solder reflow causes the moisture to vaporize and can lead to cracks in the mold compound. What generates bubbles in polyimide adhesive tapes during die attach? A mechanism for the formation of decompression bubbles is discussed, and it is suggested that bubble formation can be reduced by a slower pressure release and by lead frame geometries with less uncovered tape area between leads.
Polyimide Tape Die Attach
Polyimide adhesive tape is used in the lead-on-chip (LOC) package for DRAM die attach.1,2 Moreover, it has been applied to other packages such as ball grid array (BGA)3 and chip-on-lead (COL)4 packages. The tape consists of a base film coated with thermoplastic polyimide on both sides. Lead frame fingers are attached under application of pressure and heat directly to the die surface using this double-sided tape.
A key factor that differentiates epoxy die attach from polyimide tape die attach is that polyimide tape requires temperatures higher than 300°C for melting.5,6 In this temperature range, water vapor pressure rises to values up to 100 bar. During the bond process, the vapor pressure is balanced by the hydrostatic pressure within the adhesive, but this is not the case after pressure release. Water vapor pressure, thus, is the driving force for adhesive bubble formation.
Three types of bubbles can be distinguished: (a) trapped air bubbles, (b) moisture bubbles and (c) decompression bubbles (Figure 1). Note that decompression bubbles are located directly underneath the lead fingers, while moisture bubbles are not affected by the lead frame. Trapped air bubbles can be avoided by applying a high enough bond force, controlling the tilt of the bond tools and by placing the lead fingers close together. Moisture bubbles are avoided by drying the tape prior to bonding.
null
Experiment
The origin of decompression bubbles has not been well understood. They are difficult to see since they are hidden underneath the lead fingers. On the other hand, decompression bubble formation is not affected by the tape drying time. This puzzling phenomenon was investigated by an experimental set-up that allowed in-situ observation of the tape area underneath the lead fingers during the bond process through a glass plate. The experiments showed that bubbles begin forming when the pressure is released, which is why they can be described as "decompression" bubbles. The experimental results can be summarized as follows:
- No bubbles are formed as long as the pressure is not released, regardless of bonding time length.
- When the pressure is released, bubbles are formed starting from the center of lead fingers. Bubbles are formed only in the adhesive facing the glass plate.
- As the pressure decreases, the bubbles grow towards the edges of the lead fingers. If the pressure is increased at any point, the bubbles shrink inward.
- If pressure is applied again, the bubbles disappear completely. They appear at roughly the same position and shape as before if pressure is released finally.
- The faster the pressure is released, the bigger the bubbles grow.
- In most cases, the bubbles have a stripe-like or elongated shape as shown in Figure 1c.
 Figure 2. a) Numerical calculation of squeeze flow in an adhesive under pressure as created in (b) a cross-section of the stamping process. |
null
Numerical Calculation
Apparently, decompression bubbles start forming during pressure release, and simulation can be used to understand the mechanism. Using a one-dimensional model, the squeeze flow was calculated by the finite difference method.7 The arrangement in Figure 2a represents the die-side adhesive of a three-layer tape that is squeezed by lead fingers into areas between the lead fingers. In these areas, the pressure increases and reflects the elastic deformation of the base film. The base film bends upward by about 1 µm, as can be seen in cross-sectional cuts of bonded units.
The lead frame side adhesive is left out in the model because decompression bubbles were not observed on that side. The adhesive is squeezed between a bottom plate (die) and a stamp (lead finger). The stamp applies a bond force to the adhesive underneath it. The fixed top plate to the left and right of the stamp represents the base film carrying the adhesive. The tape edge is simulated by the outermost cells, which can move freely and are kept at ambient pressure. The pressure (p) is defined as the total hydrostatic pressure minus the ambient pressure (1 bar). Squeeze flow in the y direction is neglected because the lead finger length (L) is much greater than the width.
The adhesive itself is modeled by individual cells, which build up a hydrostatic pressure due to compression by the applied bond force. The outermost cells are kept at ambient pressure, causing a pressure gradient and finally a viscous flow of neighboring cells. The elastic deformation is neglected in the model.
 |
Figure 2b shows the result of the calculation. At t = 0, the applied force causes a homogenous pressure underneath the stamp. After t = 0.5 seconds (the bond time), the pressure has been distributed to areas beneath the stamp, giving a bell-like pressure distribution. The adhesive thickness underneath the stamp has been reduced. Note that the integral ∫pdx (area) underneath the stamp is the same at t = 0 and t = 0.5 seconds because L∫pdx = F. Now the force is ramped to zero in 10 steps of 1 msec. Thus, the value of Úpdx is decreasing. When the force is finally zero at t = 0.501 seconds, an area of negative pressure is formed starting from the center of the stamp.
An interpretation of this phenomenon is simple: the negative pressure is caused by the pressure reservoir that builds underneath the fixed top plate (= base film) during the bond. After the bond, this pressure reservoir pushes the stamp up and generates a pressure sink in the center.
Decompression Bubble Formation
The numerical analysis shows that during pressure release, negative pressure areas are generated. These negative pressure areas are localized at sites where bubbles in the adhesive are observed experimentally, i.e., at the center of the lead finger.
A possible mechanism for the formation of decompression bubbles, therefore, can be described as follows: The areas of negative pressure (relative to ambient) are origins of bubbles in the adhesive. The negative pressure causes a tensile stress that breaks the adhesive at a weak point (hyphenation point), and a cavity results. Why do the cavities not collapse after the adhesive was given enough time to reach pressure equilibrium? Once formed, the residual moisture in the adhesive fills the volume of the cavity. After bond, the tape temperature drops by 30° to 40°C/second.8 So even when the pressure in the adhesive has reached equilibrium after the bond, the water vapor pressure keeps the bubbles from collapsing until the adhesive hardens.
The pressure reservoir underneath the base film in between lead fingers causes the adhesive to form cavities underneath lead fingers once the pressure is released. However, the water vapor pressure prevents them from collapsing, resulting in decompression bubbles.
Reducing Decompression Bubbles
Why is it not possible to dry the tape down to a level where moisture no longer affects bubble formation? Generally, a substance containing water is dried in a completely dry environment under heat or vacuum application. Heat enforces diffusion and accelerates drying.9 Zero humidity conditions are not realistic for the lead frame/tape prebake (drying) process in the assembly line. Thus, the adhesive tape is exposed to an environment with low, but non-zero humidity.
The prebake process can accomplish a drying down to a level where the water vapor pressure in the adhesive is in equilibrium with the environment. On the other hand, when water molecules are dissolved within the polymer, their vapor pressure is reduced. For example, the equilibrium moisture content of a commercial LOC tape might be 0.5 percent in an environment with 0.02 bar partial pressure of water vapor (50% R.H., 25°C). When the water would not be dissolved in the polymer, the theoretical pressure of "free" water vapor of the same amount and volume would be 2 bar at ambient temperature. Thus, the polymer bulk can dissolve 100 times more water than the air environment. This factor decreases with increasing temperature, which is the origin of drying.
On the other hand, a moisture content of only 0.03 percent is sufficient to give a bubble concentration of 1 percent at 330°C. It is a difficult task (maybe even impossible) to achieve a moisture content lower than this by drying — and even hold it until the die is attached. However, there might be better ways to reduce formation of decompression bubbles. According to the suggested mechanism for bubble formation, these include:
Reducing the speed or relaxing the pressure after bond
- Avoiding lead frame/tape geometries where pressure reservoirs are possible (e.g., unsupported adhesive).
These approaches should lead to a higher reliability polyimide tape die attach process.
REFERENCES
- T.Lowrey et al., "The 64 Megabit DRAM Challenge," Semiconductor International, Vol. 16 (1993), p. 48.
- N.Taketani et al., "CSP with LOC Technology," 1996 Proc. International Symposium on Microelectronics, SPIE Vol. 2920 (1996), p. 594.
- R.D. Schueller and J. Geissinger, "New Chip Scale Package with CTE Matching to the Board," Proc. 1st Electronic Packaging Technology Conference EPTC (1997), p. 219 (IEEE Catalog No: 97TH8308).
- T.Suzumura et al., "Development of Lead Frame for COL and LOC Package," Proc. 41st Electronic Components and Technology Conference ECTC (1991), p. 210.
- Technical data sheet Ableloc 5500, Ablestik Electronic Materials & Adhesives, Rancho Dominguez, CA 90221.
- Hitachi Chemical data sheet HM-122U, Hitachi Chemical Co., Ltd. Tokyo.
- S. Behler, "The Formation of Decompression Bubbles in Polyimide Adhesive Tapes," Proc. 49th ECTC (1999), p. 727.
- S. Behler, "Understanding Heat Transfer In The Lead-on-chip Die Bonding Process," Proc. 1st Electronic Packaging Technology Conference EPTC (1997), p. 141 (IEEE Catalog No: 97TH8308).
- Arun S. Mujumdar (ed.), Handbook of Industrial Drying, Marcel Dekker, New York (1995).
Illustration by Gregor Bernard
Stefan Behler, senior process engineer, may be contacted at ESEC, Hinterbergstrasse 32, CH-6330 Cham, Switzerland; 41 41 749 5111; Fax: 41 41 741 64 84; E-mail [email protected]; Web site: www.esec.com.