Encapsulation of Microelectronic Assemblies for use in Harsh Environments Open Access

The electronic assemblies studied in this paper must be rugged, robust, and reliable. The representative assembly is small in volume (2–5 cubic centimeters), contains a variety of surface mount components, and is built on multilayer flex and rigid-flex circuit boards. The flexible portion of the board is shown folded and stacked to fully utilize the available volume in the system. Given the environmental demands placed on the product, FEA of the encapsulated assembly is essential early in the design process to ensure that the system is adequately supported by its housing and mounting points in an airframe, and that the system integrity within the package will be maintained throughout deployment.

The electronics assemblies are encapsulated with Loctite Eccobond FP 4450 (formerly Hysol FP 4450), which is chosen for its relatively low coefficient of thermal expansion (CTE) of 22 ppm, large working temperature range, and high glass transition temperature (>150 C)[1]. Encapsulation is performed under vacuum and at elevated temperature. The chamber can be rapidly vented and evacuated to work encapsulant into small volumes between components and prevent entrapment of air bubbles. The temperature of the electronic assembly is controlled during filling to optimize the viscosity and wetting characteristics of the encapsulant.

CFD software is a critical tool during design[2] to minimize the possibility of void formation, as it enables the evaluation of alternate fill locations, rearrangement of components, changes to the fold configuration of the circuit boards, and surface treatments to modify the energy of component interfaces. Cross sections are used to compare the locations of actual filling problems with model predictions.

The encapsulant is characterized to validate input parameters to the CFD models and perform lot acceptance. A modified sessile drop method [3][4] is used to quantify the contact angle of cured material on various substrates. Residual stress of the encapsulant is measured by depositing a thin (~.005”) layer of the encapsulant on comparably thin pieces of metal. After curing at multiple temperatures, the material parameters are extracted from the curvature of the sample and the equation for a bimetallic strip.

The deleterious effects of a void in an encapsulated assembly can be demonstrated with a two-dimensional model of a chip capacitor in a filled epoxy potting material. The model geometry is shown in Figure 1. The chip capacitor is adhesively bonded to its circuit board and surrounded by potting material, except on its left end, which is exposed by a large void. The bottom surfaces of the potting material and adhesive are fixed, while an inertial body load equivalent to an acceleration of 20 kilo-g is applied to the assembly, as shown on the meshed model in Figure 2.

Figure 3 and Figure 4 show the calculated von Mises stress distributions in the assemblies with and without a void. The maximum stress in the void-free assembly is 17.5 MPa, which increases to 22.6 MPa when the void is present. Moreover, the presence of the void increases the overall level of stress in the chip capacitor.

Voids are most likely to form in corners defined by surfaces that meet at sharp angles, such as folds in flex cables or components on circuit boards. The following section develops an estimate of the wetting angle of the material as a function of the corner angle.

In the model, the epoxy is advancing into a corner space defined by two planar substrate surfaces that intersect at an angle of 2ξ as shown in Figure 5. The wetting angle of the epoxy on the substrate surface is θ. The epoxy-substrate interface is shown at a distance, l, from the corner point (a).

The height, h, of epoxy-substrate intersection (c) above the corner centerline is

The angle, ψ, between the substrate surface and the line normal to the corner centerline is obtained by setting the sum of the interior angles of triangle (a,b,c) equal to π.

In the figure, R designates the radius of the epoxy-air interface that passes between the points (o) and (c). Radius, R, is perpendicular to the line that is tangent to the epoxy surface at its intersection with the substrate. Since the angle between this tangent and the substrate is θ, the angle between its perpendicular, R, and the substrate is π/(2- θ). The angle, η, between R and the vertical line, h, is given by

The radius, R is obtained from (o,b,c)

Upon substituting for h and η, R becomes

When the epoxy interface advances towards point (a), from a distance l₁ to l_2, the change, ΔE in the interface energy of the system is

where the interfacial energies are given by γ_ES, γ_AS, and γ_EA for the epoxy-substrate, air-substrate, and epoxy-air interfaces, respectively. Substitution for R in this expression yields

The interface advances as long as ΔE is less than zero. The parameters that result in a stationary interface are obtained by setting ΔE = 0,

Figure 6 describes the relationship between surface energy vectors at the edge of a wetted drop. The energy balance is therefore given by

The ratio of the surface energies, is equal to -cos(θ) from the Laplace-Young equation[3] and substituting for η yields a two parameter transcendental equation:

Solving these equations results in the plots shown in Figure 7.

A series of CFD models were developed in COMSOL to simulate simplified two-dimensional flow of our potting compound in varied mold configurations[5][6]. The material data sheet density (1.77 × 10³ kg/m³) is used in all models[1]. The data sheet notes a dynamic viscosity of approximately 44 Pa·s at room temperature; the modeling below, however, is performed using lower viscosities (1–5 Pa·s). This decision was made in order to reach solutions with the computational engine. It is assumed to be valid given that the epoxy becomes less viscous at higher temperatures, though the magnitude of this change has not yet been quantified. All models use a two-fluid system with the starting fluid as air and the fill fluid modelled as epoxy. All models neglect gravitational forces.

First, a 10 mm × 20 mm cavity is modeled with a 1 mm × 2 mm obstruction near the inflow. The inlet pressure is 100 Pa and the epoxy viscosity is 5 Pa·s. Figure 8 shows the results at sequential time steps with the encapsulant movement initially perturbed by the obstruction, then proceeding to a more uniform filling.

The second set of simulations are designed to model flow around solder balls in a ball grid array. The geometry is simplified to a 2D representation of a “channel” obstructed by two circles, where the top and bottom plans represent the chip and substrate. The channel is modeled as 3 mm long, 0.4 mm tall. A dynamic viscosity of 2 Pa·s is used. For the first case, an inlet pressure of 100 Pa is applied on the left hand side of Figure 9. Persistent voids are observable near the circular obstructions.

The geometry of this setup is modified in Figure 10 to use an inlet velocity of 1 mm/s. This provides similar final results to those seen in the “Inlet Pressure” case, with the acute angle on the leading side of the second obstruction showing better filling and the top edge of the channel showing more variable coverage by the epoxy. The time scale of this flow is also about twice as long as in the “Inlet Pressure” case because of the initial conditions.

For the case depicted in Figure 11 and Figure 12, the dynamic viscosity is decreased to 1 Pa·s and the condition of constant inlet velocity and no inlet pressure is used, with U_i = 1 mm/s. Voids are again visible around the circular obstructions. These results are comparable to those of Lee et al.[7] predicting void formation within a ball grid array.

The model is expanded to study a cavity filled from a center column and vented through two outer openings as shown in the geometry of Figure 13. The rectangular obstructions mimic internal surface mount components within a cavity. The inlet width is modeled at 2 mm and the square obstructions are 6 mm square. The fill frontier and voiding are both observed to be significantly more complicated than seen in the “channel” type models and reinforce the need for critical inspection of the edges and corners of all physical samples.

A series of tests is performed to characterize encapsulant as received and over time. The first test is based on a composite beam bimetallic strip analysis. A composite beam is comprised of two beams with respective thicknesses, widths, coefficients of thermal expansion, and elastic moduli of t₁, w₁, γ₁, E₁ and t₂, w₂, γ₂, E₂. Figure 14 shows the separate, unbounded beams after a temperature change of ΔT.

When bound together, continuity requires the upper beam to be elongated by δL and the lower beam to be compressed by a different length equal to γ₂LΔT – δL. The force required to extend the upper beam is

and that to compress the lower beam is

In the absence of any external forces, F₁ = F₂, so equations [1] and [2] can be combined to obtain an expression for δL:

By letting

this force balance equation becomes

Substitution of this result into equation [1] yields the following expression for F₁:

This force is applied as a shear stress across the interface side of the beam and hence is equivalent to applying a bending moment, M₁, on the ends of the beam that is equal to

From linear beam theory, the second derivative of a beam's displacement w, with respect to position z, along its length is

in which I represents the beam's moment of inertia. The quantity ∂²w/∂z² is approximately equal to the reciprocal of the radius of curvature, R, of the beam, measured at its centerline. Thus, equation [5] is approximately equal to

For a beam of rectangular cross section,

Substitution of equations [4] and [7] into [6] yields the following expression for 1/R₁

The radius of curvature, R_B, of the composite beam is

This radius of curvature can be estimated by fixing one end of the composite beam in a cantilever configuration and measuring the deflection of the free end, as shown in the construction of Figure 15. A toolmaker's microscope is used to measure L₁, L₂ and h to obtain R_B.

Assuming that the δ² is small and can be ignored, so that this equation can be solved for R_B to yield

From the construction in Figure 15, L₁ is given by equation [11] and angle θ is given by

The values of h₁ and h₂ can then be obtained from

The composite beam bending radius is then written as

If L₂ – L₁ is sufficiently small, then δ can be set equal to zero in equation [10]. The validity of this approximation can be assessed by comparing the values of R_B obtained from equations [10] and [15].

To validate the model described in 4.1, composite beam samples are prepared by inserting a 4 in × 0.5 in steel shim, 0.004 in thick, into a mold and casting potting material on top of it. A straight edge is dragged across the open top of the mold to remove excess material and create a uniform thickness strip. The depth of the mold cavity sets the thickness of the epoxy (0.020 in). Test samples are prepared and cured, along with encapsulated hardware. Figure 16 shows the two-piece mold along with a sample epoxy strip.

A completed test sample is clamped to a glass slide at one end and a tool maker's microscope is used to measure the height of the other end of the beam above the slide. This measurement of h is recorded and processed using equations [9] and [15], and the requisite material properties and cure temperature. The Excel “solver” is used to adjust the γ and E values of the elastomer until the two calculated geometric and property based equations for bending radius (R_B) in equations [9] and [15] are equal. In Figure 17, the product of γ and E is plotted against time for four different samples. Initially, the values from the four samples are tightly grouped together, but they then disperse over time. This is presumably due to some combination of delamination, hydration and relaxation in the material, but we do not have a definitive explanation for these observations.

The contact angle of the encapsulant within the assembly was deemed to be a critical factor in both modeling and processing. This section builds on the work of Cheng and Lin described in their IEEE Transactions work[4]. Contact profilometry is performed on a cured drop of epoxy to quantify the contact angle on substrates present within the microelectronics. A substrate is prepared with a representative material and then a single drop of encapsulant is dispensed. The drop is cured and then measured with contact profilometry (Taylor Hobson Form Talysurf). Slope analysis is performed to determine the contact angle (see Figure 18).

Using this method, the contact angle between the FP4450 compound and Kapton is measured to be 17.0 ± 0.1°. The contact angle to aluminum is measured to be 54.4 ± 0.8°.

Further analysis is performed on cured epoxy using a modified sessile drop method. Using a microscope and beam splitter as shown in the cartoon below (Figure 19), the profile of the encapsulant is measured after cure. This method is more versatile than contact profilometry, as it allows examination of uncured drops as well. Figure 20 shows an image captured using this technique. The drop is measured using computer software at 165° outside angle, 15° internal angle.

This analysis is not performed regularly at this time given the preeminent importance of the encapsulant's cured characteristics on the movement profile through the microelectronics package and the stress it imparts on the final assembly. These methods of evaluation are relatively simple and quick to perform and can therefore be used as lot characterization methods throughout production. Future work includes expanding the range of substrate materials to include packaged integrated circuit components, stainless steel and soldermask.

The encapsulation station is designed to be versatile in accommodating various package sizes and encapsulation needs. The build platform is a custom-built adapter, designed to ensure a tight fit around the electronics package. The fixture is wrapped in magnet wire, which is then supplied a current in order to elevate the temperature of the fixture and electronics package. The heating process is controlled by a K-type thermocouple in the fixture and Omega temperature controller. A piezoelectric buzzer is also affixed to the fixture mount as a vibration source during encapsulation to further reduce voids.

Encapsulant dispense is performed using an extended Luer lock needle in a swage lock fitting. Two Swage locks are placed on either end of a flexible hose. One is threaded into a large Plexiglas top window, and the other is secured around the dispense needle. This configuration enables motion and control of the dispense inside the chamber while maintaining vacuum. Dispense rate is controlled manually with a foot pedal. The chamber is designed to accommodate a vacuum of 25 inHg. A ball valve fitted to the side of the chamber allows for rapid cycling of the vacuum.

Three packaging methods are employed to produce the data in this paper. The first is the encapsulation of the microelectronics package on a hotplate set between 80° and 100° C, per the epoxy data sheet recommendation for the material. Neither dispensing nor venting of the epoxy takes place under vacuum in this case. This method is used with the potting visualization samples discussed in Section 6.1 and allowed for full control of the encapsulant compound and access to all sides of the package for photographic documentation.

The second scenario is the encapsulation of the package again at temperature and ambient pressure with degassing steps added throughout the fill process. After partially filling the package, the assembly is moved into the vacuum station described in Section 5.1, the vacuum is cycled to 15 ± 5 inHg at least five times, until bubbles appear at the top surface of the package, indicating a release of entrapped air. After the air is satisfactorily removed with the unit partially filled, the process of adding epoxy and cycling vacuum is repeated. This method is used when fine location control of the encapsulant dispense is required.

The third method for encapsulation occurrs with the package under vacuum for the entire potting process. The package is at elevated temperature and 20 ± 5 inHg vacuum throughout dispense and can be vibrated in the fixture to increase settling. Production encapsulations are performed in this station given the superior void removal.

Visualizing the encapsulation process via mechanical mockups of the microelectronics package is integral to informing the geometry of the CFD models and designing the optimal flow path for filling and venting of the assembly. The mechanical mock-ups are volumetrically identical to the final package, the housing of the package is modified, however, so that a glass viewing “window” (taken from a standard microscope slide) can be affixed to the flat surfaces. Two windows are attached to each package, on opposite sides of the assembly. Using a high-fidelity mock-up provides a clear picture of the encapsulant flow within the cavity and removed from the results variables dependent on the geometry of the system.

Figure 23 shows the wetting front of the encapsulant (from upper left to lower right, in sequence). Encapsulant is added to the top left of the package via a fill hole and the main vent is mirrored on the right hand side of the top surface, per the description in Section 5.2. The first image of the set shows evidence of the epoxy encapsulant flowing more readily around the corner of the Integrated Circuit (IC) package into the larger channels, validating the flow profile seen in Figure 8. The fourth image shows that the encapsulant exhibits preferential flow over the IC chips instead of through the more exposed channel at the top of the package. This effect is likely due to a combination of gravity (not included in the modeling in Section 3.2), the directional nature of the dispense process, and to capillary effects across the large X-Y area of the glass and IC gap.

After encapsulating a sample with each of the experimental methods described in Section 5.2, cross sections are obtained to evaluate void formation and prevalence. Figure 24 depicts a representative cross-section of the encapsulation process when no vacuum was used to vent the epoxy. A three dimensional surface scan is used to aid the visualization of void locations. Voids in the encapsulation are common at the edges of the IC packages, particularly in the location of solder connections. This behavior is predicted by the CFD model (Figure 9 through Figure 12) where acute angles in the flow of the encapsulant tend to trap voids and resist filling.

An improvement in void reduction is seen in Figure 25, which was encapsulated under a constant vacuum of 20 ± 5 inHg. A single void is seen at the edge of an IC package indicated by the red circle. The green boxes indicate component pull out, not voids.

Given the continuing demand for encapsulation of microelectronics packages for harsh environments we anticipate the methods and frameworks described in this paper to be useful going forward. The simplified models discussed above can be expanded to describe additional geometries of interest that might arise in new packages and the baseline results can inform process and design decisions prior to prototyping and building future hardware. Periodic process monitoring will be continued for contact angle and cure stress. In addition, these methods will be used to expand the library of results going forward as they are adaptable to many materials and geometries.