The back-end process: Step 5 - Encapsulation step by step
05/01/2001
BY CHRISTINE NAITO AND MICHAEL TODD
The demand for high-performance, high-density semiconductor assemblies continually pushes materials suppliers and packaging engineers to develop enhanced integrated circuit (IC) package designs. Economic constraints and manufacturing feasibility, however, further challenge the packaging engineer to optimize processing in an effort to improve manufacturing efficiencies and thereby reduce the total system cost.
 |
Many factors should be considered in a manufacturing process optimization, including its impact on process cycle time, product performance and manufacturing yields. Changes to various process parameters can often affect a number of key performance characteristics of the package. For example, changes to the processing parameters of typical epoxy encapsulant materials used in IC packaging may affect the package's thermal, mechanical and electrical properties.1 It is important to understand the effects of these process changes on the total system performance of the IC package.
Encapsulant materials protect bare die in a variety of packages and applications, each of which may impose different requirements on the properties of the encapsulant. Key material characteristics, including glass transition temperature (Tg), coefficient of thermal expansion (CTE), flexural strength, flexural modulus, ultimate strain and package warpage, often vary as a function of processing conditions. Important processing variables include cure temperature, cure time and type of cure schedule (single step, two step or discontinuous). Results of testing demonstrate that alternative cure schedules to those reported on manufacturers' data sheets are often available that provide an acceptable balance of material properties for specific applications.
Curing Options
During processing of liquid encapsulant, several types of cure schedules can be followed. The simplest is a single stage of curing at constant temperature. An alternative is step curing, where the curing step is preceded by a gel step.2 The gel step is of short duration and occurs at temperatures slightly above the initiation temperature of the catalyst. It is designed to dimensionally "set" the polymer into position before completion of the cure.3 Using a gelling step can increase yield by reducing the possibility of damage during processing.4 The second step, or cure step, lasts much longer and occurs at higher temperatures. Full development of the physical properties occurs during the cure step.
 Figure 1. Contour plot of Tg (°C) development. |
It has been demonstrated in previous work that if materials of the type used in this study are exposed to ambient conditions longer than one hour while in the uncured state, enough moisture will be absorbed to interfere with proper cure. Significant degradation of the performance of devices encapsulated with the material will result.5 If curing begins within one hour of dispense, moisture contamination is minimal and the level of protection is sufficient to allow devices to survive extensive environmental testing.6
 Figure 2. Contour plot of flexural strength (MPa) development when tested at 80°C. |
It may be difficult to complete the entire cure schedule continuously on some manufacturing lines if box ovens are used. Some engineers may want to index devices onto a heated platen immediately after encapsulant dispense, where an abbreviated gel would be given. After this, the devices would be loaded onto a magazine and stored under ambient conditions until the magazine is filled and moved to the box oven for final cure. Before proceeding with a discontinuous process, it is important to determine if gelling the material is sufficient to prevent moisture contamination from occurring. Only by preventing moisture contamination can an engineer be sure that the final properties of the encapsulant will not be compromised.
 Figure 3. Contour plot of flexural modulus (GPa) development when tested at 80°C. |
A set of experiments was performed on a liquid epoxy/anhydride encapsulant material filled 73 percent by weight with fused silica filler. This material is typical of those used to protect bare silicon ICs for high-performance, high-density semiconductor assemblies. The experiments investigated a range of processing conditions for the material based on recommendations made on the product data sheet. These experiments were designed to evaluate the effects of deviating from the recommended cure schedule on key physical characteristics often used to predict the material performance. This information can help a packaging engineer design a manufacturing process that will ensure development of material properties while meeting productivity goals.
 Figure 4. Contour plot of flexural strain to break (percent) development when tested at 80°C. |
Three sets of experiments were performed. The first explored the effects of cure time and cure temperature on the development of the physical properties of the encapsulant. The second determined the effect of step curing on the warpage of a semiconductor package as well as on the development of the previously studied physical properties. The final set of experiments examined the level of moisture protection an encapsulant would provide if a step cure process were discontinuous.
Time/Temperature Study
A design of experiments (DOE) study was created using an experimental design software. The design was based on a quadratic model, with cure time and cure temperature as independent variables. The limits for cure time were 15 to 120 minutes, and the limits for cure temperature were 130 to 180°C. Eleven trials with four replicates were evaluated. The responses under investigation included Tg, flexural strength, flexural modulus and ultimate flexural strain to break. The Tg for each specimen was determined by differential scanning calorimetry (DSC) on a single specimen. Flexural properties were measured at 80°C on six specimens per ASTM D 790. All data was analyzed using the design software to construct contour plots of predicted values over the entire range of values in the experiment.
A contour plot of Tg as a function of curing time and temperature was constructed (Figure 1). The plot shows that the Tg of this encapsulant is very dependent upon the temperature at which the material is cured, as well as upon the amount of time it is exposed to that temperature. Overall, Tg decreases substantially with cure times less than 90 minutes and temperatures lower than 150°C. Above these, Tg change is gradual and finally plateaus with a maximum value of approximately 170°C.
Figure 2 shows the contour plot of flexural strength at 80°C. This property is also time and temperature dependent. Curing 90 minutes at 165°C produces very high strength. Curing at even higher temperatures can increase the strength slightly. However, curing for longer than 90 minutes at temperatures greater than 150°C can cause the material to become somewhat brittle, reducing the flexural strength.
Plotting the flexural modulus requires a partial cubic model, which is more complex than the quadratic model used for the other properties. This plot is presented in Figure 3 and illustrates that modulus initially increases significantly with time and temperature, and then reaches a plateau. Curing for at least 60 minutes at any temperature will result in fairly constant modulus values.
Figure 4 shows the contour plot of flexural strain to break at 80°C. A partial cubic model was also used to construct this plot. The figure suggests that when this material is cured for short times at low temperatures, it is not fully cured and the strain to break is high, like that of a "flexible" material. As curing continues, the strain to break reaches a minimum as the material transitions from being flexible to being truly thermoset. Once this occurs, further curing causes the strain to increase as the material becomes stronger.
The material in this study will never achieve fully developed Tg or flexural properties if cured at low temperatures, even for extended periods of time. This is evident in Figures 1, 2 and 4, where maximum Tg strength and strain values are lower with low-temperature cures than with high-temperature cures. Curing at low temperatures creates a lower molecular weight polymer; only by curing at high temperatures can enough cross-linking occur to form a high molecular weight, fully developed polymer. Cross-linking will continue with extended cure times at very high temperatures. Tg does not degrade with increased cross-linking, but flexural strength and strain do as a result of brittleness from overcuring.
 Figure 5. Effect of step curing on device warpage. |
Although the optimum values for the individual properties may not occur using the same cure schedule, a good balance of properties can be achieved. The best cure schedule is one that offers a good development of properties with minimum risk of overcuring. In this case, 60 to 120 minutes at 160 to 170°C would be a good target processing range.
Step Cure Studies
The curing of liquid encapsulants is often performed in two steps, as mentioned previously, and the temperature of the gel step can affect device warpage. Warpage was measured by using a BT laminate board (15 x 40 x 0.28 mm) as the test vehicle. One gram of encapsulant was spread evenly over the surface of the board. No die or damming material was used on the test vehicle to eliminate die attach or damming materials as sources of shrinkage. Warpage of the test vehicle would therefore be caused only by the shrinkage of the encapsulant.
 Table 1. TMA results for step cure study. |
Samples were gelled for one hour at five different temperatures, ranging from 100 to 165°C, with three specimens per gel condition. The specimens were cooled to room temperature (RT), and warpage was measured. Warpage measurements were made by placing the specimens encapsulant-side up on a flat surface, clamping one of the short ends of the specimen to the surface, and measuring the distance between the bottom of the specimen at the opposite short end and the top of the flat surface with a feeler gauge. All of the specimens were then cured an additional three hours at 165°C, then cooled to RT, and warpage was measured again. An unencapsulated control board was exposed to the 165°C gel and cure conditions.
 Table 2. Flexural results for step cure study. |
The results of this gel temperature study are presented in Figure 5. The data indicate that all test specimens warped upon gel, but the lower the gel temperature, the lower the warpage. Once the specimens were exposed to the high-temperature cure step, the warpage was reduced in almost all cases. Even the specimen that was gelled and cured at 165°C saw a reduction in warpage over time. The warpage of the unencapsulated board was negligible and was not included in the figure.
 Table 3. TMA ratings of samples given discontinuous cures. |
The specimens that exhibited less warpage after the gel step tended also to have the lowest warpage at the completion of the cure schedule. The exception to this occurred on the specimens gelled at 100°C. In this case the warpage increased when the cure step was given. This is most likely explained by incomplete encapsulant gel before exposure to the high cure temperature. The portion of the encapsulant that was not already gelled reacted in a fast, uncontrolled manner typical of a straight high-temperature cure. Therefore, it is important to completely gel a material of this type before proceeding with the cure step of a two-step cure profile.
 Figure 6. Typical TMA for a material exhibiting an area of high expansion just above the Tg. |
To supplement this data, the CTE below the Tg was measured by thermo-mechanical analysis (TMA) on encapsulant cured according to low and high warpage cure conditions. CTE values were 21.1 and 22.8 ppm/°C, respectively. This demonstrates that the expansion characteristics of the material are not changed significantly by cure schedule, and any differences in warpage are due solely to shrinkage.7 These results raised the question of whether step curing would have any effect on the development of other physical properties.
 Figure 7. Discontinuous cure study ("good" TMA scan). |
To answer that question, Tg values and flexural properties were measured at three cure temperatures, with and without a gel step. 125°C was chosen as the gel temperature because it is well above the initiation temperature of the system. Those specimens receiving the straight high-temperature cure were cured 90 minutes at 150°C, 165°C or 175°C. Those receiving the step cure were gelled 30 minutes at 125°C, followed by curing 90 minutes at 150°C, 165°C or 175°C. One specimen per condition was tested. Tg values were determined by TMA, single scan. CTE1 values were measured between 40 and 120°C, and CTE2 values were measured between 190 and 220°C. Results are presented in Table 1.
All values for Tg are within experimental error and indicate that step curing does not affect the development of this property. CTE values above and below the Tg are reported and again demonstrate that this property is independent of cure schedule.
 Figure 8. Discontinuous cure study ("questionable" scan). |
One unexpected phenomenon occurred on specimens that experienced a straight high-temperature cure. TMAs of these specimens showed an area of high expansion just above the Tg. An example of this is presented in Figure 6. If material with this anomaly is scanned a second time by TMA, this area of high expansion disappears. This indicates that this area represents stress built into the polymer during cure. The high-temperature excursion of the first scan annealed this stress, as shrinkage was annealed in the warpage study.
Next, flexural properties were measured at RT. Eight specimens per condition were tested. Results are presented in Table 2. Test values for step cured specimens compared to those that were not step cured were within experimental error.
Discontinuous Step Cure Study
In the final study, uncured material was dispensed into metal dishes measuring 2.5 mm in diameter and 1.0 mm thick to simulate a cavity application. The dishes were transferred to a hot-plate preheated to 125°C, and the material was allowed to gel 5, 10, 15, 20, 25 or 30 minutes, providing a range from partial to complete gel. Three dishes were prepared per gel condition. One dish from each condition was placed into a 30°C/80% relative humidity (RH) environmental chamber for either one, three or five hours. These conditions were chosen because they represent a worst-case scenario for manufacturing in a tropical environment.
After exposure to humidity, all dishes were step cured 30 minutes at 125°C plus 90 minutes at 165°C in a box oven. A control specimen was included, which did not experience any gel or exposure to humidity, but which did undergo the same cure schedule as the other samples. The level of moisture contamination that occurred was analyzed via TMA, single scan.
 Figure 9. Discontinuous cure study ("bad" scan). |
The results of this study are summarized in Table 3. The control scan is characterized by a smooth Tg and linear expansion rates before and after the Tg (Figure 7). A "good" rating was given to samples with TMA scans that look like the control sample scan. A "questionable" or "?" rating was given to samples with scans that are good, but have lower expansion areas near the Tg. A "bad" rating was given to any sample with a scan that does not look like the control scan (Figures 8 and 9).
As expected, those samples that spent more time on the platen and were exposed to humidity the least amount of time experienced less moisture contamination than those that were minimally gelled and exposed to humidity for long periods of time. Within each humidity condition, the transition from good to bad was gradual, with some samples being questionable.
These results confirm that the more complete the gel of an encapsulant, the better the material's ability to resist moisture contamination. If devices are to be exposed to an environment of 30°C/80% RH, they should be gelled for at least 25 minutes at 125°C if full cure cannot be given immediately. Devices must not be exposed to this environment for longer than three hours prior to full cure, regardless of the gel conditions.
Conclusions
Three sets of experiments were performed on a model epoxy/anhydride IC encapsulant material. These experiments were designed to identify processing windows in which package engineers could work when designing their manufacturing lines.
It was determined that Tg and flexural properties are dependent on cure time and cure temperature. The development of these properties occurs quickly at first, but slows and eventually plateaus.
Step curing significantly reduces device warpage by reducing linear shrinkage. Extended time at high temperature anneals stresses built into the polymer and allows warpage to be reduced further. Step curing does not affect the development of Tg CTE or flexural properties. Finally, the sensitivity of the encapsulant to moisture contamination can be reduced if a complete gel is given to the devices immediately after dispense. This can occur on a heated platen or in an in-line oven.
As can be seen by the data, many processing conditions are available that result in acceptable physical characteristics for a given encapsulant. A materials supplier often recommends one or two cure schedules on a data sheet based on what will meet reliability requirements for most applications. Packaging engineers must decide if those cure schedules will meet their reliability requirements and productivity goals. If changes to the recommended cure schedule are needed, it is best to make those changes with a full understanding of the possible effects on material properties.
AP
References
- Jeffery Suhling et al, "Measurement of Backside Flip Chip Die Stresses Using Piezoresistive Test Die," Proceedings of the 32nd International Symposium on Microelectronics, October 26-28, pp. 298-303, 1999.
- Henry Lee and Kris Neville, Handbook of Epoxy Resins, McGraw-Hill, second edition, New York, Chapter 6, pp. 6-9, 1982.
- Wei H. Koh, Beverly H. Tai and Elizabeth A. Kolawa, "Low Stress Encapsulants for Reduced Failures in Plastic Packages," Proceedings of the International Mechanical Engineering Congress and Exposition, San Francisco, California, November 12-17, 1995.
- B. Anderson and B. Bacher, "Design and Process Guidelines for Plastic Ball Grid Array Dam and Fill Encapsulation," Dexter Electronic Materials, 1997.
- Bradley Goodrich and Bruce Toyama, "Liquid Encapsulation Application Considerations," Dexter Electronic Materials, December, 1994.
- Dale L. Robinson, Marc Papageorge and Christine Naito, "A New Epoxy Based Liquid Encapsulant with Perfomance Comparable to Mold Compounds," International Journal of Microcircuits & Electronic Packaging, Vol. 17, No. 2, pp. 176-183, 1994.
- Jim C.L. Wu, Han-hien Shiue, Sting Wu, Mike Hung and J.J. Lee, "Study of Rapid Cure BGA Mold Compound on Warpage with Shadow Moiré," Proceedings of the 49th Electronic Component and Technology Conference, San Diego, California, June 1-4, pp. 708-713, 1999.
CHRISTINE NAITO is a project leader and MICHAEL TODD is a research associate at Dexter Corporation, Electronic Materials, 15051 East Don Julian Road, Industry, CA 91746; 626-968-6511; Fax: 626-336-0526; E-mail: [email protected] and [email protected].
null