Hygroscopic Swelling of Encapsulated Microcircuits
06/01/2003
Part I: New Measurement Technique Shows the deformation
BY BONGTAE HAN, ERIC STELLRECHT AND MICHAEL PECHT
The hygroscopic swelling of five commercially available mold compounds is analyzed and the coefficient of hygroscopic swelling is determined for each mold compound using moiré interferometry. The results indicate that the deformation caused by hygroscopic swelling can be as significant as the deformation caused by a thermal expansion of 90°C. The comparison indicates that hygroscopic strains must be considered for accurate reliability assessment when plastic encapsulated microcircuits (PEMs) are subjected to environments where the relative humidity fluctuates.
PEMs dominate the market share of microcircuit sales worldwide due to advantages over hermetic packages in terms of size, weight, availability and cost. Despite these many advantages, one important disadvantage is that polymeric mold compounds absorb moisture and, thus, swell when exposed to a humid environment. Hygroscopic stresses arise in a PEM when the mold compound swells upon absorbing moisture and the lead frame, die paddle and silicon die do not experience swelling. The hygroscopic stress increases as the hygroscopic swelling coefficient of the mold compound increases similar to the thermal stress produced by the mismatch in coefficient of thermal expansion (CTE) between adjacent materials. Accurate measurement of hygroscopic swelling is essential in assessing the effect of hygroscopic stresses on package reliability.
How to Measure Hygroscopic Swelling
The measurement techniques used in the existing literature include essentially point-measurement techniques, and the results showed large discrepancies in the magnitudes of swelling coefficients. Although some of the discrepancy should be attributed to the different properties of the tested polymeric materials, the significant variation warrants a more precise measurement technique to document the swelling coefficient of polymeric materials. Recently, the authors have developed a technique to measure hygroscopic swelling on a whole field basis using moiré interferometry.
Two samples of a particular mold compound were first subjected to a 125°C bake to remove any initial moisture that may have existed in the samples. When the bake was completed, the samples were temporarily removed from the baking oven and a cross-line diffraction grating of 1,200 lines/mm was replicated onto the samples at an elevated temperature.
 Figure 1. Experimental apparatus detailing the mold compound reference and test samples during measurement. |
null
One of the two samples was selected and left in the baking oven to prevent moisture gain after grating replication. This sample, referred to as the reference sample, was needed to compensate for the thermal expansion during the moirÈ measurement. The second sample, referred to as the test sample, was subjected to an 85°C/85 percent RH condition and its weight was periodically monitored until a virtual saturation state was reached. The virtual saturation state is defined as the occurrence of no additional weight gain within the resolution of the balance for two to three days. The true saturation state would have required many additional days of sorption.
Once the virtual saturation state was achieved, the hygroscopic swelling measurement procedure was initiated using moiré interferometry. The moiré setup used in the experiment is illustrated in Figure 1. The two major components in this setup were a portable moiré interferometer (PEMI II, Photomechanics Inc.) and a computer controlled environmental chamber (EC1A, Sun Systems).
It was desired that only hygroscopic swelling-induced deformations were documented, so it was vital to eliminate any thermal expansions during measurement. This was accomplished by the following procedure:
1. The reference sample and the test sample were removed from the baking chamber and the relative humidity chamber, respectively, and they were placed side by side in the convection oven of the real-time moiré system.
2. The samples were allowed to reach thermal equilibrium at the desorption temperature of 85°C.
3. The moiré setup was tuned to produce a null field (devoid of fringes) on the reference sample.
4. The test sample was viewed by the moiré interferometer and the corresponding fringe patterns were documented.
5. The test sample was removed immediately after Step (4) and was weighed. It was then placed back in the convection oven.
6. Steps (2) to (5) were repeated at preset time intervals until the desorption process was completed.
null
The fringe patterns obtained from the moiré measurement are shown in Figure 2. The null field patterns of the reference sample are shown in (a), and the fringe patterns of the test sample at time intervals of zero and four hundred hours are shown in (b) and (c), respectively. The fringe patterns at the zero hour (Figure 2b) represent the hygroscopic swelling at the virtual saturation point. The test specimen contracted as desorption progressed, as evidenced by a decrease in the number of fringes in the patterns.
The reference sample remained in the baking oven at 125°C after the grating was replicated. It was used to set the initial null field and to check the null field at each subsequent measurement. Step (3) in the above measurement procedure was equivalent to the optical subtraction of a uniform strain. This step canceled any thermally induced deformations in the test sample by using the deformed state of the reference sample as a reference. The re-tuning of the null field before each measurement was required to make the measurements completely free from any errors associated with temperature instability of the instruments.
The hygroscopic strains, εh, were determined from the fringe patterns by
 |
null
where fs is the frequency of the specimen grating (1,200 lines/mm), ΔN is the change of fringe orders in the moiré pattern and ΔL is any gauge length across which ΔN is determined.
Hygroscopic Swelling vs. Thermal Strain
Hygroscopic swelling in the V direction of the mold compound is plotted with respect to the moisture concentration (Figure 3). It is evident that a linear relationship exists between swelling and moisture concentration. The constant of linearity, called the coefficient of hygroscopic swelling (CHS), is defined as:
 |
null
where β is the coefficient of hygroscopic swelling, and C is the moisture content percentage. The CHS is a material property of the mold compound and, if known, the strain can be determined by measuring the moisture content in the mold compound. Only the V field is shown in Figure 3; the U field also showed a linear trend and had virtually the same CHS.
 Figure 3. Swelling vs. moisture content (percent) for mold compound sample EME 7720-TA obtained from the V field moiré fringes. |
null
The above procedure was used to analyze five commercially available mold compounds manufactured by Sumitomo Bakelite Co. Ltd. The material properties are shown in Table 1. Test results are summarized in Table 2, which includes CHS values, virtual equilibrium moisture content and the corresponding hygroscopic swelling obtained from Equation 2. The results show a significant variation in the magnitude of hygroscopic swelling.
EME-6300H and EME-7720TA exhibit hygroscopic swelling nearly twice as large as that of EME-6600CS, EME-7351LS and EME-G700. This was attributed to the combined effect of the amount of ash content and the resin/hardener system.
The temperature excursion required to produce thermal expansion equal to hygroscopic swelling can be calculated as:
 |
null
where α is the coefficient of thermal expansion. The results are also shown in Table 2. The deformation caused by hygroscopic swelling can be as significant as the thermal deformation caused by a thermal excursion of 92°C.
The hygroscopic mismatch strains at the mold compound/chip interface are identical to hygroscopic swelling as the chip does not absorb moisture and does not swell. The thermal mismatch strains at the interface, εσ, can be determined as
 |
null
 |
null
 |
null
The results are shown in the last row of Table 2, where the chip CTE of 3 ppm/°C and a temperature excursion of 100°C were used. The hygroscopic mismatch strains are compatible with the thermal strains induced by the considerable thermal excursion.
Conclusion
The above results show that hygroscopic swelling effects can have a significant impact on PEM reliability. In environments such as in automotive applications where packages are subjected to relative humidity excursions, the hygroscopic-induced strains must be considered for reliability assessment. In fact, these strains can be as significant as those found by thermal excursions.
Accelerated life testing conditions such as a highly accelerated stress test (HAST) chamber, where temperature, humidity and pressure are used, may also witness complications due to hygroscopic swelling issues. The temperature conditions in a HAST chamber are typically from 100°C to 150°C, the relative humidity is typically over 70 percent, and the pressure can be up to 50 psi. These conditions will drastically increase the amount of moisture absorbed by the polymeric materials in a package, and therefore greatly increase the hygroscopic swelling. The experimental results presented here imply that the hygroscopic swelling would play an important role in the failure of the package under the test. Just as thermal cycling is important for reliability analysis, we also recommend a relative humidity cycling test for encapsulated microcircuits.
At the time of writing this article, an actual plastic package was being tested to further investigate the combined effect of hygroscopic swelling and thermal deformation on PEMs reliability. The results will be reported in Part II in the July issue.
References
For a complete list of references, please contact the author.
Dr. Bongtae Han, associate professor, Eric Stellrecht, graduate research assistant, and Michael Pecht, professor, may be contacted at CALCE Electronics Products and Systems Center Department of Mechanical Engineering, University of Maryland, College Park, Md. 20742. E-mail: [email protected], [email protected], [email protected].