Robust, Low-Cost Flip-Chip Encapsulation

Flip-chip packaging has the potential for use in automotive and robotic applications, where the need for low production costs conflicts with high reliability demands. To facilitate the migration of flip-chip packaging into these and similar applications, a study was conducted at the Fraunhofer Institute for Reliability and Microintegration. Building on earlier investigations, this study focused on materials and processes suitable for the simultaneous underfilling and overmolding of the flip-chip.
 
Currently, molded flip-chips involve a single chip in a chip-scale package (CSP) or BGA format, but there is potential to integrate multiple devices into a molded package of this type. Two long-standing problems in flip-chip generally are coefficient of thermal expansion (CTE) mismatches between the underfill material and adjacent materials, and the underfilling of very small gaps, particularly where a topographically irregular substrate can cause void formation. Simultaneous overmolding and underfilling with an appropriate material can alleviate these problems because encapsulation under pressure allows materials to be used that give a better CTE match and that fill small, irregular gaps more efficiently.

Five encapsulation materials were investigated, using both destructive and nondestructive evaluation techniques. The encapsulation processes can be performed with conventional transfer molding equipment, and fall into three categories: standard molding, vacuum molding and vent molding. Only vent molding was used in this study for the evaluation of the materials and processes.

Two general concerns apply to all three of these methods. First, excess fluid molding compound could flash around the perimeter if process parameters are not carefully defined. Second, pressure and flow speed during encapsulation must be somewhat restricted to avoid reaching temperatures at or above the 183°C melting point of eutectic SnPb solder (both SnPb and SnAgCu solders were used in this study).

Five commercially available encapsulation materials were selected for this study. In general, these materials are appropriate for all three methods described above. The materials selected met these requirements: They could be processed at temperatures below 175°C; they had relatively low viscosity to permit optimum flow during processing; and when cured they had a CTE of <20 ppm/K. All five materials are epoxy-based, and have filler content by weight of about 80%.

The test vehicle, developed at the Fraunhofer Institute, uses an FR4 substrate with contact pads for 10 × 10 mm flip-chips as well as other package types. The flip-chip die used in the study were peripherally bumped with pitches of 200 µm. A daisy chain test structure on each die was used for testing contact integrity.

The flip-chip molding tool (Fig. 1 ) was manufactured with electrical discharge machining (EDM) to provide the geometric precision needed to permit intimate contact of the tool with the substrate to minimize flashing of the fluid encapsulant. A small cavity was made in the top mold tool for encapsulant flow.

1. Molding tool for simultaneous underfill and encapsulation of flip-chips.

Three process parameters must be carefully controlled for successful encapsulation of a flip-chip. First, control of mold temperature permits optimum viscosity of the fluid molding compound and promotes fast curing. For the materials used here, mold temperature typical values were 165-175°C. Second, the clamp force of the molding tool against the substrate ensures a tight seal to prevent flashing (Fig. 2 ). Third, the injection pressure and transfer speed ensure that the mold will be filled homogeneously within the process window of the molding compound.

2. Diagram showing the molding tool in operation. Note the vent cavity in the substrate.

Test sequence

To optimize process parameters, the five encapsulant materials were evaluated by several means, with emphasis on the thermomechanical evaluation of both the encapsulant materials and the substrate. Differential scanning calorimetry was used to determine cure behavior and reaction onset/reaction enthalpy in uncured material, and to determine heat capacity and phase transitions in cured material. Dynamic mechanical analysis determined Young's modulus (E) and the glass transition temperature (Tg). Thermomechanical analysis was used to find the CTE and glass transition temperature of the materials. Thermogravimetrical analysis measured humidity content, filler content and decomposition temperature.

Flip-chips mounted on test vehicles were then encapsulated to perform empirical measurements of clamp force adjustment, transfer speed and material flow optimization. Cured packages were subjected to a popcorning test conforming to JEDEC 020, Level 1. Storage was for 168 hours at 85°C/85% relative humidity and three reflow cycles for eutectic SnPb. Thermal cycling was performed in accordance to MIL STD 883e from -55°C to +125°C.

At appropriate points during the test sequence, electrical measurements were made to determine daisy chain integrity. Acoustic micro imaging was used to nondestructively evaluate packages for internal delaminations and voids. X-ray microscopy was used to evaluate solder joints. Acoustic micro imaging was also used at regular intervals during reliability testing to reveal the progress of any internal anomalies. The acoustic micro imaging was achieved with a Sonoscan D-9000 C-SAM, operating with a specialized ultrasonic transducer for die-level imaging.

Differences in resin type and in the size and geometry of filler particles have considerable impact on flow behavior, filling rate, air entrapment, agglomeration of filler particles and the stresses placed on solder bumps, the die and the substrate.

There are significant differences among the material properties of the five molding compounds, designated A through E, and shown optically in Figure 3 . Material A has a resin formulation not found in the other four, and has a low Tg, good flow properties, a detectable high spiral flow length, and very low melt viscosity. Materials C and E are somewhat similar in that they have low viscosities and a high spiral flow length. Material B has a relatively high viscosity.

3. Particle size and structure in the five test materials. Size of the optical microscope images is 100 × 100 µm.

Materials A, B and C contain spherical filler particles of various diameters. Material D contains a mixture of spheres and flakes but, like materials A, B and C, has maximum particle diameters in the 25-35 µm range. Material E differs in having spheres and flakes with maximum diameters up to 75 µm. The CTE of the various materials is 14-19 ppm/K.

All five materials were capable of simultaneously overmolding and underfilling the flip-chips with both eutectic SnPb and SnAgCu solders, although only material C appears to be a candidate for further process optimization.

Optimizing the process parameters greatly reduced or eliminated internal defects in the packages. Reducing the transfer speed to values of 0.2-0.3 mm/sec avoided damage during the molding process. Higher speeds resulted in the deformation of SnPb solder bumps. When SnAgCu solder bumps are used, higher speeds tend to crack the die.

A mold temperature of 165°C prevented damage to the substrate and to solder joints of both alloys. A clamp force of 2 tons was adequate to prevent flashing of the molding compound from around the periphery, while avoiding damage to the substrate that occurred when higher forces were applied.

Even with optimized process parameters, materials A, B and D sometimes had voids in the mold cap. Material C sometimes trapped air that formed voids in the underfill. Only material E, which differs from the other materials in filler particle configuration and which has a high spiral flow length, formed voids in neither of these locations. Material B also caused slight deformation in the substrate layer.

Most of the materials survived the popcorning test (equivalent to three reflow cycles) with no damage. Only material C showed in acoustic images the formation of delaminations in the underfill layer. Humidity storage resulted in no delaminations in any samples, and in only slight increases in contact resistance that occurred only in material C.

Thermal cycling also resulted in very limited increases in contact resistance, and the occasional failures were noted only after 1000 or more cycles. Materials A and E survived with no electrical failures until more than 2750 cycles. In material C, cross sectioning after 2250 cycles showed cracks in solder bumps and in the substrate beneath the bump. Delaminations or cracks in the underfill layer were likewise unusual.

4. C-SAM acoustic imaging showing initiation and growth of cracks in the mold cap during thermal cycling of packages using material A. At about 3000 cycles, the cracks reach the die and begin to form delaminations (white) between the molding compound and the die.

5. C-SAM acoustic imaging showing mold cap cracks and delaminations (red) in material A after 7250 thermal cycles.