Material Set Comparison in Moisture Sensitivity Classification of Nonhermetic Organic Packages William R. Schildgen 3M Company Microinterconnect Systems Division 2020 Prairie Lane, P.O Box 2200 Eau Claire, WI 54702 E-mail: wrschildgen@mmm.com Cameron T. Murray 3M Company Electronic Adhesives and Specialties Department 209-1C-30 3M Center St. Paul, MN 55144 E-mail: ctmurray@mmm.com Abstract Avoiding mechanical damage, such as die delamination, due to moisture-induced stress during the assembly solder reflow is a high priority for many package assemblers. Moisture/reflow sensitivity classification allows board assembly operations to exercise proper storage and handling precautions to avoid damage caused by moisture. The choice of materials and optimization of process parameters can significantly influence the moisture sensitivity classification of chip packages. This paper directly compares two types of heat spreaders and two types of adhesives used in a cavity-down nonhermetic organic wire bond package. Specifically, a nickel/gold plated copper heat spreader with a glass cloth reinforced, silver-loaded epoxy thermoset film is contrasted to a heat spreader treated (zincated) on the side facing the substrate with nickel plating on the opposite side, and bonded with an epoxy/acrylate hybrid thermoset adhesive film filled with silver coated glass particles. The responses analyzed include voiding and delamination as determined by reflective C-mode Scanning Acoustic Microscopy (C-SAM), and visual analysis for cracks. Introduction Materials selection in organic integrated chip (IC) packages is crucial to how well the package performs, especially in reliability testing. Moisture and reflow sensitivity testing allows nonhermetic solid state Surface Mount Devices (SMDs) to be classified so they can be packaged, stored, and handled properly. A joint IPC/JEDEC standard [1] details the process of moisture/reflow classification testing and outlines whether substrates (packages without die) must be dry packed from the manufacturer, what the floor life is at the package assembler, and if the packages must be baked prior to reflow at the package assembler. By carefully handling the SMDs according to the classification, mechanical damage from moisture can be prevented. A common failure mode is referred to as a popcorn failure, where the vapor pressure of moisture inside a package quickly increases when exposed to high temperatures, such as solder reflow. As the moisture quickly expands within the package, many types of mechanical damage can occur, including: delamination, internal and external cracks and bond damage. C-mode scanning acoustic microscopy (C-SAM) is one technique capable of detecting such mechanical damage [2]. Prior to describing the damage in the package, a brief description of scanning acoustic microscopy is necessary. C-SAM is a nondestructive test method in which an ultrasonic transducer is mechanically rastered over a sample. Typical transducers emit a focused spot of ultrasound at frequencies ranging from 10 MHz to 200 MHz. Usually water is used as a coupling medium to bring the ultrasonic energy in contact with the sample. In reflective C-SAM, the transducer acts as both a sender and receiver, switching electronically between the transmit and receive modes. As the acoustic waves enter the sample, specific interfaces produce return echoes. The return times of the echoes are a function of the distance from the transducer to the sample. An electronic gate can be defined to capture the echo from a period of time that corresponds to a specific interface, while excluding all other echoes. Voids, cracks, disbonds, density variations, and delaminations can be easily distinguished from the background, provided the instrument has adequate resolution and has been properly set up.
3M s high performance cavity-down wire bond organic chip packages were used in the testing described throughout this paper (Figure 1) [3]. The packages are being used in networking and computing applications that require high frequency/data rates (typically 2.5 Gbps and greater), superior I/O density, advanced power distribution impedance, full grounding/shielding, or optimized cost/performance. The simultaneous switching noise performance of these packages has been discussed previously [4]. Heat Spreader Die Attach Adhesive Heat Spreader Attach Adhesive Substrate Encapsulant Die Figure 2. Cross-section of cavity-down wire bond organic chip package. Figure 1. 3M 35mm cavity-down wire bond package after die placement and encapsulant cure. The package substrates are made with Microlam dielectric, a composite consisting of PTFE, a thermosetting resin, and inorganic filler. The thin (58 µm) dielectric, combined with its low dielectric constant of 3.4 and a tailored coefficient of thermal expansion make it an ideal material for high performance wire bond packages [5]. Cavity-down wire bond substrates are generally attached to some type of heat spreader with a heat spreader attach adhesive (Figure 2). The heat spreader serves as an area for die attach, increases the rigidity of the package, aids in preventing excessive warping during assembly, and creates a flat surface for mounting a heat sink. Many applications require a grounded heat spreader and/or heat sink to minimize electromagnetic interference that could be an effect from high frequency signals in the package [6]. Experimental Procedure Cavity-down organic wire bond substrates were assembled with a low stress, electrically conductive die attach paste, a 10 mm by 10 mm silicon die, and a low stress epoxy encapsulant. The 500 µm thick substrates studied were a 7 layer construction. Two non-consecutive assembly lots were used for the moisture sensitivity testing. Half of the samples in each experimental cell were from a given assembly lot. The sample size was 24 packages per material, per moisture soak level, for a total of 120 devices. Wires were not bonded to either the substrate or the silicon die, since the interfaces of interest for this study were the heat spreader-to-die and heat spreader-to-substrate interface. One heat spreader attach adhesive in this organic package has historically been a 125 µm thick, glass cloth reinforced, silver-loaded epoxy thermoset film. A cross-section image in which glass bundles are surrounded with the silver-loaded epoxy is shown in Figure 3. The electrically conductive film was furnished as a prepreg and cured within the manufacturer s recommendations. A challenge in using any adhesive in this application was optimizing the adhesive flow or squeeze-out while completely wetting out the surfaces to which the adhesive bonds. Adhesive lot-to-lot variation, processing variation, and adhesive structural variation, such as bundles of glass fibers, affected adhesive flow on different levels. Many experiments were performed in the past to increase the amount of wetting while maintaining specifications around adhesive flow into the cavity of the package. The silver-loaded adhesive described above is referred to throughout the work as a component in material set A as shown in Table 1. 2
Material Set A B Table 1: Material set comparison details of adhesive and surface finish. Heat Spreader Attach Adhesive Silver-loaded epoxy thermoset film 3M 7373 hybrid bonding film Die Side Heat Spreader Surface Finish Gold plated surface finish Zincated surface finish 3M 7373 Grounded Heat Sink Bonding Film is a 50 µm thick thermoset/elastomer hybrid bonding film loaded with silver coated glass spheres as the filler. Film 7373 is cured with a conventional PCB lamination profile to form an electrically conductive connection between substrates along with good mechanical adhesion and wet-out of the substrates. The flow is controlled by the size distribution of the conductive filler and the cure profile. This bonding film is a component of material set B. It is well known that elastomeric components add toughness to thermoset formulations [7, 8]. Thus the combination of an elastomer with a thermoset retains all the good properties of thermoset resins (good wetting, crosslinking into a strong mechanical network) while improving upon their deficits (poor handling, brittle bonds for fast curing systems). Film 7373 survives multiple solder reflow cycles, and has been shown to have good reliability under difficult environmental conditions [9]. Figure 3. Cross-section optical images comparing the silver-loaded epoxy thermoset film (left) to the 3M thermoset/elastomer hybrid film with silvercoated glass spheres (right). Heat spreader surface finish plays an important role as the die attach adhesive, encapsulant, and heat spreader attach adhesive must reliably bond to it. One of the heat spreader surface finishes investigated includes a copper base sequentially plated with nickel, then gold (Figure 4). The other heat spreader surface finish consists of a copper base material with a zincated finish on the side facing the die and substrate. The zincated surface contains a treatment with 1-3 µm nodules that provide a rough surface, in addition to an 800-1000 Å thick coating of brass. Figure 4 contrasts the rough surface topography of the zincated heat spreader to the smooth surface of the nickel/gold plated heat spreader. The heat spreader is the lighter component in the images. The backside, or heat sink side of the zincated heat spreader, has a nickel-plated surface. Figure 4. SEM images of nickel/gold plated copper heat spreader (left) and zincated heat spreader (right) taken at the same magnification. Initial inspection of the assembled packages was performed at 40X magnification with an optical microscope. Following optical inspection, the assembled packages were analyzed by reflective C- SAM (model D6000, Sonoscan). The samples were subjected to a 24-hour bake at 125 C prior to the moisture soak. Moisture soak conditions are outlined in Table 2. Material Set Table 2: Moisture soak conditions investigated for the two material sets. Moisture Moisture Soak Soak Level Conditions Time (hours) A 3 30 C/60 %RH 192 A 4 30 C/60 %RH 96 A 5a 30 C/60 %RH 48 B 1 85 C/85 %RH 168 B 2 85 C/60 %RH 168 After moisture soak, all samples were subjected to three 220 C convection reflows. The packages were then inspected, both visually and acoustically. A 100 MHz transducer was used to analyze the samples with reflective C-SAM. Results and Discussion Acoustic imagery was performed on all five experimental cells immediately after assembly and again after the packages had been subjected to the appropriate moisture/reflow environment. Figure 5 shows how material set A compared both before and after moisture soak level 5a and reflows. After only 48 hours at 30 C and 60 %RH, and 3 reflows, 3
delamination was observed around the periphery of the die. The delamination was present in the C-SAM image as lighter (whiter) areas within the outline of the package. More specifically, delamination occurred between the encapsulant/die attach adhesive and the heat spreader on most of the packages in the experimental cell after Moisture Soak at Level 5a (MSL-5a). It is also evident that some amount of voiding is present at the heat spreader attach adhesive interface immediately after assembly. heat spreader adhesive attach area, perhaps due to incomplete surface wetting. The extent of delamination does not appear to be any worse with MSL-4 than with MSL-5a. Figure 7. Material set A C-SAM images before any conditioning (left) and after MSL-4 soak and Figure 5. Material set A C-SAM images before any conditioning (left) and after MSL-5a and To verify the delamination shown in the C- SAM images above, an optical cross-section analysis was used. In Figure 6, the delamination around the periphery of the die is clearly shown as the dark region between the die attach adhesive and heat spreader. The delaminated area as determined by C- SAM corresponded well with the optical crosssection analysis. The propagation of the delamination at the encapsulant to heat spreader interface is evident in Figure 8 after MSL-3 and reflows. The ring of delamination around the border of the die has begun to spread through the die attach adhesive and the heat spreader attach adhesive. It also appears that small areas of voiding in the heat spreader attach adhesive have coalesced. Figure 8. Material set A C-SAM images before any conditioning (left) and after MSL-3 and Figure 6: Cross-section verification of delamination near the periphery of the die. After a more aggressive moisture soak in level 2 and reflows, material set B shows no signs of delamination in any of the areas that were susceptible to delamination with material set A. In Figure 9, no significant changes at the interfaces of interest are noticed in the package prior to and after moisture soak and reflows. A few small areas of disbonds or voids are present in the heat spreader attach adhesive, but much less in number and area that that of material set A. Figure 7 is very similar to Figure 5 in that delamination is present between the encapsulant/die attach adhesive and heat spreader after MSL-4. A small amount of voiding is also scattered across the 4
sample size of 11 devices for each moisture sensitivity level is also outlined in the standard. The die attach region of material set B passed both MSL- 1 and MSL-2, while that of material set A failed MSL-3, MSL-4, and MSL-5a. Figure 9. Material set B C-SAM images before any conditioning (left) and after MSL-2 and After level 1 soak and reflows with material set B (Figure 10), there was no significant change in any of the interfaces that contact the heat spreader. When compared to material set A, the number and area of the disparities in the heat spreader attach adhesive were very small, and did not expand appreciably after moisture soak and reflows. Material set B typically had much less of the lighter (whiter) areas that corresponded to voiding, delamination, or disbond when compared to material set A. Table 3: Die area voiding for the five experimental cells before and after moisture soak and reflows. Mean Die Mean Die Material MSL Area Voiding Area Voiding Set @ t=0 after MS A 3 2.9 39.3 A 4 3.5 28.2 A 5a 3.0 21.8 B 1 0.5 1.5 B 2 0.3 0.3 Figure 10. Material set B C-SAM images before any conditioning (left) and after MSL-1 and The area of the package was separated into two regions for analysis of the C-SAM images; the die area, which includes the die adhesive to heat spreader and encapsulant to heat spreader interface, and the heat spreader attach adhesive area. The separation into two areas enables further quantitative investigation into the delamination mechanism. From Table 3, it is clear that under more aggressive moisture soak conditions, material set B changes significantly less than material set A. An additional advantage of material set B is the lower initial die area void content prior to any environmental exposure. For the die attach region, a failure is defined in the joint IPC/JEDEC standard [1] as a 10% absolute change in area between pre- and post-reflow measured on one or more devices. A minimum Figure 11. Material set comparison of void content in heat spreader attach adhesive area. Voiding analysis in the heat spreader adhesive attach area as summarized in Figure 11 shows how immediately after assembly, material set B has a significantly lower initial void content. Although devices with material set A exhibited an increase in void content after moisture soak and reflows, the means did not change significantly. The initial mean void content of material set A was 7.1 %, and the initial mean void content of material set B was 0.5%, but the variation associated with material set B also appears to be less than that of material set A. After the moisture soak and reflows, none of the devices showed visible cracks using 40X optical magnification. 5
Conclusions Material set B, with the 3M thermoset/elastomer hybrid adhesive and the zincated heat spreader offers superior performance in moisture/reflow sensitivity testing. Both die area voiding and heat spreader adhesive attach area voiding were significantly improved with material set B as measured by C-SAM. The package with material set B would have been classified as nonmoisture sensitive, had it contained wire bonds and passed electrical test. Materials selection and material processing optimization can be important factors not only in how a package performs in moisture sensitivity testing, but also other reliability stress testing. Offering a high performance cavity-down wire bond product that can be classified as nonmoisture sensitive will not only afford board assembly operations more flexibility in regards to storage and handling, but will also lead to better performing product under reliability stress testing. Further work is underway to include electrical test as a response for material set B in additional moisture sensitivity classification testing. Acknowledgments The authors would like to thank Scott Bahe, Donald Banks, Kim Cramer, Robin Gorrell, Michael Holcomb, Lorie Knaack, and Sandy Walter. References [1] IPC/JEDEC Joint Industry Standard, Moisture/Reflow Sensitivity Classification for Nonhermetic Solid State Surface Mount Devices, IPC/JEDEC J-STD-020A, April 1999 [2] J. Lau (ed), Ball Grid Array Technology, McGraw-Hill, Chapter 10, pp. 327-328, 1995. [3] http://www.3m.com/us/electronics_mfg/microflex/p roducts/ic/multilayer/ [4] W. G. Petefish and E. C. Priest, High Performance Organic Single Chip package, International Symposium on Microelectronics, 1996. [5] J. Korleski, R. Gorrell, C. Bowen, D. Noddin, New Composite Organic Dielectric for High Performance Flip Chip Single Chip Packages, Proc. 47 th Electronic Components and Technology Conference (ECTC), San Jose, California, May 1997, pp.1015-1023. [6] J. Diepenbrock, B. Archambeault, L. Hobgood, Improved Grounding Method for Heat Sinks of High Speed Processors, 51 st Electronic Components Florida, May 2001, pp.993-996. [7] R. Y. Ting, Epoxy Resins, Chemistry and Technology, Ed. C. A. May (Marcel Dekker, Inc., 1988), Chapter 5. [8] M. Savla, Handbook of Adhesives, Ed. I. Skeist (Van Nostrand Reinhold, 1977), Chapter 26. [9] S. Tead, C. Murray and B. Rudman, Using Conductive Adhesives for Grounding Applications, Circuits Assembly, February 2000 Copyright, 2002 by IMAPS - International Microelectronics And Packaging Society. Reprinted with permission from the 35th International Symposium on Microelectronics Proceedings, pg. 679-684, Denver, CO, Sept. 4-6, 2002. 6