Thermal management becomes more critical as higher electronics performance is designed into smaller packages. As a result, heat removal mechanisms need to become more efficient to prevent overheating and failure. Recently, there has been increased attention to the details of heat sinks, heat spreaders, and heat removal methods in military, medical, and other sectors of electronics manufacturing.
Heat moves through various materials by radiation, convection, or conduction. Numerous devices have been developed to remove heat from electronic components and systems. And while they may to some extent use all three of these routes, most heat-removal devices rely primarily on conduction, the transfer of heat across the interface between two materials that are in intimate contact. Heat sinks and heat spreaders are examples of heat removal by conduction.
Heat conduction requires that the heat sink or heat spreader be well bonded to the heat source. The heat sinks adhesively bonded to the back side of flip chips typically are copper or aluminum. It is imperative that there be no voids or even significant variations in thickness in the adhesive layer, which typically is a thermal interface material or a thermal grease.
Each material has its own coefficient of thermal conductivity, which expresses the amount of thermal energy transferred through a given area and the thickness of the material in a given time under a specified temperature difference. Silica-filled epoxy, or mold compound, has a rather low value of 0.30 W/(mK), which suggests that parts of a chip in contact with mold compound dissipate relatively little heat by that route.
Copper, on the other hand, has a value of 380 W/(mK) while aluminum has a value of 200 W/(mK). Heat that is to be dissipated by a metal heat spreader embedded in the mold compound underneath the die paddle in a PQFP or similar package must first cross the layer of minimally conductive mold compound between the die paddle and heat spreader (Figure 1). Sonoscan application laboratories have noted that the mold compound frequently fails to flow between the die paddle and the heat spreader, resulting in a barrier of insulating air between the two metal elements.
The acoustic image of such a PQFP is shown in Figure 2. This image was made by scanning the component from the back side. The star-shaped feature is the heat spreader, which is larger than the square die that is below the depth shown here. Gray regions of the heat spreader are in good contact with the molding compound that lies between the heat spreader and the die, but white regions are delaminated. The large area of the heat spreader that is delaminated would have a negative effect on heat transfer.
The path of heat transfer runs from the silicon to the adhesive layer to the heat sink and from there to the surrounding atmosphere. The need to dissipate increased power per unit volume means that heat sinks and heat spreaders need to be very efficient.
Efficiency can be greatly diminished by the presence of air-filled gaps along the transfer path because air has an extremely low thermal conductivity value of 0.025 W/(mK). The thermal grease in a flip chip assembly may contain voids. Voids of a sufficiently large area can reduce thermal conductivity to the point where the chip, or a portion of the chip, overheats and fails.
Because voids and similar anomalies are gaps, they can be detected by acoustic micro-imaging systems, which are sensitive to the differences in material properties at the interfaces. When the ultrasonic transducer of an acoustic micro-imaging system scans the heat sink attach, it receives return echoes of modest amplitude from the copper-to-grease interface, but far higher amplitude echoes—close to 100% of their transmitted strength—from the copper-to-air interface.
As a result, acoustic imaging can rapidly and nondestructively evaluate the probable performance of a heat sink. In assembling high-end, high-power microprocessors, automated acoustic micro-imaging systems often are used in production lines to detect bondline anomalies such as voids and simultaneously measure the thickness of the adhesive grease.
Figure 3 is the acoustic image made through the heat sink to the depth between the heat sink and the flip-chip microprocessor. The X pattern at the center of the acoustic image (left) is an artifact of the deposition of the adhesive. The square feature at the center is the chip. During application of the heat sink, the adhesive has been extruded laterally into a roughly circular pattern that extends beyond the edges of the square die. Just beyond the edge of the die, there are rounded rectangular windows in the heat sink.
The most critical information in the acoustic image is the absence of voids in the adhesive layer over the area of the die where they would interfere with thermal transfer. Voids or other gaps in the adhesive layer would appear bright white if they were present.
There are, however, voids in the extruded adhesive in at least three of the four windows. These voids originally may have been located over the die and moved laterally during the application of the heat sink. In their present location, they pose no thermal management threat.
Figure 4 is the acoustic image through a heat sink that is bonded to the bottom side of a surface-mount printed wiring board. The board was turned upside down for acoustic imaging. The purpose of acoustic imaging was to determine the extent of bonding between the heat sink and the board.
To make the acoustic image, return echo signals from the interface between the heat sink and the board surface were used with some penetration into the body of the board. The weave pattern is visible, and the vertical arrangements of white circles are vias within the board.
The most significant features are the red and yellow areas in the adhesive, which will act as insulators and block heat transfer. Specifically, the 0.025 W/(mK) thermal conductivity of the air in the voids will limit heat transfer although presumably there also will be minor heat transfer across the voids through radiation and convection.
Figure 5 shows a side view of the board and the attached heat sink.There are 201 voids between the heat sink and the board. Some of these are tiny and probably have essentially no impact on thermal transfer. Automated analysis of the acoustic image shows that, collectively, the 201 voids cover 12.7% of the area of contact with the heat sink, a figure that could be high enough to permit overheating. The largest single void covers 0.91% of the contact area.
This data, along with the acoustic image, reveals several things. First, the overall bonding process designed to hold the PCB and the heat sink together needs improvement. Second, it indicates where the largest voids are. Third, the locations of voids of various sizes can be compared with the positions of heat-producing components on the board. Although the overall distribution of voids is somewhat random, there also are voids that clearly follow the outline of a component. Near the right end of the board, some voids are arranged in a vertical pattern that probably reflects features such as vias.
About the Author
Tom Adams is a freelance writer and photographer who has written more than 500 articles for semiconductor and microelectronics trade magazines. Sonoscan, 2149 E. Pratt Blvd., Elk Grove Village, IL 60007, 847-437-6400, e-mail: [email protected]
September 2009