Warpage Behaviors of System-in-Packages on a Substrate Strip

Electronics packaging technologies are facing more and more complex integration challenges as consumers want their electronics to be smaller and faster with better performance and functionality packed into a single device. With these requirements, microchip encapsulation technologies are moving toward system-in-package (SiP). In SiP solutions, a number of integrated circuits are enclosed in a single module. The SiP may perform all or most of the functions of an electronic system and is typically used inside a mobile phone and digital music player. Silicon dies containing integrated circuits may be stacked vertically or horizontally on a substrate. They may be connected with each other by fine wires that are bonded to the package or, alternatively, with a flip chip technology which uses solder bumps for electrical connections. Although SiP technology has been used for a range of electronic devices, the warpage behavior of the package can be difficult to control and predict due to complex manufacturing parameters, design rules, and material properties. Previous research on this topic primarily focused only on either strip warpage or on the unit package while the combination of these two different warpages has not been studied and seems almost to be completely ignored in the industry [1, 2].

In this paper, the substrate strips, with two-layer stack-ups, are prepared to study their impact on the warpage. Table 1 shows the thicknesses and materials of the substrate layers, where SR stands for solder resist, M1 and M2 are metal layers, PP stands for prepreg, and core is the central layer in the substrate.

The copper trace patterns on each layer may impact the warpage of the strip warpage. For this study, they are fixed, not changed, to simplify the analysis. The patterns on each layer are shown in Fig. 1.

The substrate strip size is 240mm × 95mm × 0.26mm to accommodate 126 SiP modules as shown in Fig. 2.

The size of the SiP module size is 9mm × 15mm × 1.320mm, as shown in Fig. 3. Three different densities of passive components were prepared to test their impact on the strip warpage.

Four different epoxy molding compound (EMC) materials, with different filler sizes and filler contents, are used to study the impact of strip warpage. Their material properties are shown in Table II.

Twelve legs of designs of experiment, as shown at Table III, were studied to check the impact of molding compound materials and component density.

The manufacturing of the IC packages involve complex procedures, and Fig. 4 illustrates the detailed process flow. With careful observations and measurements of warpage, we have identified that the molding and post mold cure are the critical steps dominating the warpage behaviors, while the impact from other process steps are small. In this regard, other process steps are ignored and not considered. From a practical viewpoint of process improvement, we focused only on the critical steps which predominately impact the warpage.

Fig. 5 illustrates the configuration of the mold chase which has 24 gates. The mold compound material was injected into the cavity from the bottom of the strip, and the venting holes were at the other side of the strip. The temperatures of mold compound and steel mold were carefully controlled according to the recommended specifications for high volume production of SiP modules.

After the mold compound was applied on the substrate, the deformation of strip was visually checked first. To measure the warpage, the strip was flipped to measure the profile of the bottom substrate area because the top mold compound surface area is not smooth due to the shrinkage. The warpage of unit SiP modules was measured the same way as strip, but only after each unit SiP module was diced from strip.

3D mold flow modeling tool, Moldex3D R15, was used to study the molding process. Theoretically, the microchip encapsulation process is a three-dimensional, transient, reactive problem with moving resin front. The non-isothermal resin flow in mold cavity can be mathematically described by the following equations:

where u is the velocity vector, T is the temperature, t is the time, p is the pressure, σ is the total stress tensor, ρ is the fluid density, k is the thermal conductivity, Cp is the specific heat, and Φ is the energy source. In this work, the energy source contains two contributions:

Where η is the viscosity, γ˙ is the magnitude of the rate of deformation tensor, α˙ is the conversion rate and ΔH is the exothermic heat of polymerization.

The curing reaction of epoxy resins has received much attention using different analyses. In this work, the combined model was applied to investigate the curing kinetics of the given EMC because of its ability to accurately predict the experimental data. The combined model can be expressed as follows:

where α is the conversion rate of the reaction, A₁, A₂, E₁, E₂, m, n are model parameters. During the curing process, the viscosity of epoxy resins changes with temperature and conversion rate. The Castro-Macosko model was adopted to describe the rheological properties of epoxy resins:

where A, E_a, C₁, C₂, are model parameters, α_g denotes gelation conversion at which viscosity curve grows up because of the formation of three-dimensional network structure of the epoxy resins.

After the product was ejected from the mold, a free thermal and cure shrinkage happened due to the temperature and conversion rate difference. The mechanical properties are described in Table II. The equilibrium equation with representing the stress is expressed as follows:

And the relation between stress and strain,

where σ is the stress, C is a 4^th tensor and function of relaxation modulus E(t, T, α), ε is the strain tensor and U is the displacement vector, representatively. Thermal and cure induced strains can be expressed in the following:

α_CLTE is CTE tensor, VS(P,V,T,C) can be calculated from the chemical volume shrinkage by measurement.

3D mold flow modeling tool, Moldex3D R15, was used to study the molding process with the focus on strip warpage and SiP module warpage. Fig. 6 shows the FEM model of the lower half strip, with 8 culls for molding compound to be pressed into the gate and into the cavity. The upper half strip was symmetrical to the bottom strip, thus only half of the strip was modeled to shorten the simulation time and to save computational resources. The designed gates and cavity were modeled in detail to reflect the real world structures.

Fig. 7 illustrates different passive component densities on the SiP modules. The passive components affect not only the warpage but also the voiding behaviors of the mold compound during the manufacturing process [3].

In theory, the warpage of SiP modules comes from the strip warpage which occurs during the manufacturing processes. In this regard, a numerical scheme was developed to extract the warpage of SiP modules from the strip warpage. This numerical scheme was different from a traditional mechanical warpage simulation approach. The traditional approaches, in general, consider only the material properties of the SiP components while the manufacturing processes and kinetic behaviors of molding compound are ignored. The numerical scheme contains two steps. The first step is to map the displacement values from the strip to the unit SiP modules, as shown in Fig. 8. The second step is to use the “best fitting” theory to combine the rotation and translation to minimize the rigid body motion. Fig.9 illustrates the procedures to record the warpage of SiP modules from the warpage of strip.

The numerical scheme does not consider the impact of the dicing. The dicing and cutting of strip to SiP modules may induce mechanical stresses on the modules.

To better demonstrate and compare the simulation vs. experimental data, for the warpage of both strip and SiP modules, the legs of A, D, G, and J were grouped together, as shown in Fig. 10 and Fig. 11 to check the impact of material properties. These legs have the same 42% component density, but with different molding compound material properties. The result shows that the trends of simulation and experiment warpage match with each other, although the absolute warpage values are not perfectly matched. The match verifies that the simulation approach is reasonable, but more study is needed to improve the correlation. At the time of the publication of this paper, the root causes of mismatch in between the simulation and experiment data were not yet clearly identified, and continuing research is still on going. For the other two component densities, 49% and 65%, the results were similar, thus the data is not repeated in this paper.

Fig. 12 places the simulation results of strip warpage and the averaged SiP module warpage together. The corresponding leg number, EMC material, and component density are noted at X axis at the bottom. The warpage values of SiP modules were in the range of 10 to 40um, while the warpage values of the entire strip were roughly in between 2mm to 9mm. The trend of strip warpage and the corresponding SiP modules warpage was consistent with each other.

In this paper, the warpage behaviors of the strip and SiP modules were studied numerically and experimentally. The warpage trends of numerical simulation data match with the experimental data. However, the absolute warpage values were not perfectly matched. To improve the numerical accuracy, work will be continued in the following areas: (a) A better viscoelastic material model is needed in order to consider the nonlinear and kinetic behaviors, especially when mold compound is cured from a high temperature to room temperature; (b) a better control and monitoring of the manufacturing conditions to ensure the stability and repeatability; (c) development of a more advanced material characterization hardware and software.