Evaluation of a ThermaIly-Enhanced Molding Compound Containing Silica-Coated Aluminum Nitride Filler

The use of this new molding compound improves the thermal and electrical performance of packaged power devices.

By Keith M. Edwards, Motorola SPS, Phoenix, Arizona, and Kevin E. Howard, The Dow Chemical Company, Midland, Michigan

This paper presents the results of our investigation into the performance of a new, thermally-conductive molding compound containing silica-coated aluminum nitride (SCAN®) filler for use with power MOSFET (metal-oxide silicon field effect transistor) devices in an S0-8 package.

The devices were evaluated for shifts in electrical characteristics and we performed extensive reliability testing. We found that the SCAN filler material reduced the R0ja by 8 to 13% (compared to the standard angular silica filler material) while causing no shifts in electrical characteristics and no degradation in device reliability. Thermal performance was determined through the measurement of junction-to-ambient thermal resistance (Rja), both in still air and in a wind tunnel environment.

We concluded that while our version of the S0-8 package (with fused leadframe) would not realize the maximum potential benefit from a thermally-enhanced molding compound, the product containing SCAN filler provides a measurable increase in thermal performance which allows for an increased current rating of the device and lower typical junction temperatures. Although we have not done extensive testing using the compound with the CSP format, it appears that SCAN may be very beneficial in the production of chipscale packages.

Perhaps the most viable candidates for molding compounds with SCAN are thin packages (TSSOP, TQFP) where package and leadframe size make the incorporation of heat sinks impractical, and the influence of thermal dissipation by the leadframe minimized. It is expected that these applications will see significant improvements in thermal performance, with little to no effect on device performance or reliability.

Introduction

The performance and reliability of power devices are strongly influenced by the thermal characteristics of the package. As heat is generated at a device junction, the path for dissipation is through a combination of the device material itself (Si), the die attach (Ag filled epoxy), the leadframe (Cu alloy) and the molding compound.

Most of this heat is then transferred to the mounting media (circuit board, heat sink, etc.) and eventually to the surrounding ambient. The junction-to-ambient thermal resistance (R0ja) quantifies the combined thermal performance of package's physical attributes under a given set of environmental conditions. For a particular ambient temperature (T_a), the power (P) applied to the device must be regulated to keep the device junction temperature (T_J) lower than its maximum specification, where T_J = T_a + P (R0_Ja).

Although mounting and environmental conditions ultimately determine thermal performance capability, the thermal resistance of the package is a primary component of R0ja. It is therefore a critical factor in determining how well the device will perform in a particular application.

For a given device and input current, more thermally resistive packages (higher R0ja) cause higher device junction temperatures with a corresponding degradation of critical electrical parameters, such as leakage current and on-resistance. These changes in electrical parameters, in turn, cause more power to be dissipated as well as further increases in junction temperatures. Assuming that the circuit was designed to avoid thermal runaway of the MOSFET, the more thermally-resistive package simply causes a reduction in the efficiency of the circuit. If not, the junction temperature can increase beyond its capability, causing catastrophic overstress failure.

Device Reliability

In addition to reduced electrical performance and the potential for overstress failure, higher junction temperatures also cause a marked decrease in device reliability, another important factor in circuit design.

Devices for a particular application must be selected based on electrical performance, thermal performance and reliability, all of which are influenced directly or indirectly by the thermal resistance of the package. In practice, therefore, the increased thermal performance of a package may be used to increase the device reliability or allow more efficient use of space (increased current or power density).

From an engineering standpoint, success of a new thermally-conductive mold compound is measured by its increase in thermal performance. However, this improved performance must be achieved without degradation of the electrical characteristics or device reliability, which could be caused by chemical interaction or thermal mismatch.

Silica-Coated Aluminum-Nitride Filler

A new, thermally-conductive filler material has been developed by The Dow Chemical Co. for use in thermally enhanced molding compounds and glob-tops. Silica-coated aluminum nitride (SCAN) is a hydrolytically-stable aluminum nitride powder with a particle size distribution and ionic purity appropriate for use in the encapsulation of microelectronic devices.

Aluminum nitride is a thermally conductive, electrically insulating material which, to-date, has been unacceptable for use as a filler due to its reactivity with water. SCAN incorporates a patented coating process, which produces an amorphous silica surface that provides excellent moisture stability and resin compatibility, similar to commonly used angular silica particulates.

SCAN is a thermally conductive filler which can significantly enhance the thermal dissipation characteristics of an epoxy molding compound without compromising other performance parameters such as reliability. The molding compound used in this study was a formulation (75 wt.% SCAN) based on a standard ortho-cresol novolac epoxy formulation and is available commercially from a major Japanese molding compound supplier.

Assembly of MOSFET Devices

All MOSFET devices used for this experimentation were fabricated at Motorola facilities in Phoenix, Arizona. After fabrication, the silicon wafers were shipped to a subcontractor for assembly in the S0-8 package. Standard assembly procedures were used, beginning with sawing the wafers into individual die.

Each die was then mounted to a copper leadframe using silver filled epoxy. Gold ball bonds are used to connect the source and gate leads to the aluminum topmetal of the die. Each assembly lot was split so that approximately one-half of the devices received the standard molding compound, while the other half received the molding compound containing SCAN filler. After molding, devices were externally marked so the molding compounds could be distinguished easily.

We evaluated both a single die and a dual-die configuration of the S0-8 package (Figure 1) using both standard molding compound and the thermally-enhanced molding compound containing SCAN filler. The dual-die split leadframe device shown in Figure 1 (b) was selected for reliability evaluation.

Figure 1-Schematic of the S0-8
package (a) exterior, (b) dual leadframe
configuration, and (c) single leadframe configuration.

As an established production device, it provided a long history of electrical performance and reliability in the S0-8 package using standard molding compound. The single die configuration, Figure 1 (c) was selected for thermal performance evaluation due to its higher thermal demands. Additionally, although it had no history of electrical performance or reliability, the single die device was a new product which would push the thermal limits of the S0-8 package.

Figure 2 shows a typical S0-8 device cross section. Note that both the single- and the dual leadframes have fused leads on one side to reduce the electrical and thermal resistance of the package. The source and gate connections to the MOSFET device are made by wirebonds to the floating leads. The drain connection is made directly by attaching the backside of the die to the fused leadframe.

Electrical Performance

After assembly, all devices were evaluated to verify that electrical characteristics were within specification. All parameters were measured at room temperature, and current was applied in short pulses, so that thermal performance was not a critical factor. However, if the thermally-enhanced molding compound had changed the contaminant level, die stress or other physical attributes, the electrical parameters would likely have indicated the change. The data in Table 1 shows that the SCAN molding compound filler caused no significant shifts in the critical electrical parameters. Also, assembly yields for all lots were greater than 98%.

Figure 2-Cross-section of the S0-8 package with the primary components Indicated.

Reliability

After final electrical testing at the assembly site, all units were shipped to Phoenix for additional testing and reliability evaluation. Prior to this testing and evaluation, devices were solder mounted to individual FR-4 "snap-out" boards, using the minimum recommended footprint, to make electrical testing easier. Each lot was then randomly divided into groups and processed through the following tests.

• High Humidity, High Temperature, Reverse Bias (H3TRB), 85% RH, 85°C, VDS=20V, 1000 hours.
• Autoclave, 121°C, 100% RH, 14.7psi, 96 hours.
• Temperature Cycle (TC), -65°C to 150°C, 15 minute dwell at each extreme, 1000 cycles.
• Intermittent Operating Life (IOL), 10 kcycles, 3 minutes on/3 minutes off per cycle, current adjusted for Dtj = 100°C.

During testing, all devices were monitored at regular intervals to identify any individual device failures and to determine if any parametric shifts had occurred. As shown in Table 1, no failures were seen in two separate lots of experimental devices (molding compound w/SCAN) or in one lot of control devices (standard molding compound).

Thermal Performance

We determined thermal performance of devices by measuring their transient thermal response in a still air chamber and the steady-state thermal response in a wind tunnel environment. As noted, all devices were mounted on FR-4 boards using the minimum recommended footprint, similar to most engineering applications. Power (P) was applied by running current (I) through the intrinsic diode of the MOSFET device. Forward voltage (VDS) was measured and power dissipation was calculated by P = IV_DS.

Forward voltage (V_DS) behavior is well characterized over temperature, which allows accurate monitoring of actual device junction temperatures. Figure 3 shows the transient thermal response (in still air) of devices assembled with the mold compound containing SCAN, compared to those assembled with the standard mold compound. It is clear that the molding compound containing SCAN has lower thermal resistance-about 8% lower under steady-state DC conditions.

Figure 3 -Transient thermal response curves
for devices assembled with mold compound
containing SCAN versus those assembled with
standard mold compound.

While this is a positive result, it also verifies that while the SCAN filler material improves the thermal conductivity of the molding compound by over 30% compared to the standard molding compound, the majority of the heat generated in the S0-8 packaged device is not dissipated through the mold compound. The fused leads provide a lower resistance heat path. However, it is also important to note that having a more thermally-conductive material close to the device junction has a larger impact on transient thermal response. The reduction in R0ja for shorter pulse durations can be up to approximately 25%. This is significant for most applications as the device will be switched on and off for short pulse durations. This implies that the true thermal "improvement," using the thermally enhanced molding compound containing SCAN, is likely greater than the 8% indicated by the steady-state thermal resistance measurement.

Figure 4-Percent improvement in thermal resistance
(R0ja) vs. air velocity (wind tunnel environment)

However, because of the dynamics of heat flow through the package components, we might expect increased air velocity to improve the effectiveness of the molding compound containing SCAN by taking heat out the top side of the package.

Figure 4 shows the magnitude of this effect. At zero air velocity, similar to the result obtained above, the devices assembled with the molding compound containing SCAN have about 8% lower thermal resistance than those assembled with the standard molding compound. However, as air velocity increases, the effect of the molding compound containing SCAN indeed becomes greater.

At 500 FPM air velocity, the thermal resistance is improved by about 13%. It is obvious, however, that most of the heat is dissipated through fused leads on the bottom side of the package. While the molding compound containing SCAN improves overall thermal conductivity, heat transfer from the molding compound to the ambient limits effectiveness of this thermal path.

Summary and Conclusions

The thermal enhancement for the S0-8 packages (with fused leadframes) incorporating thermal molding compound with SCAN, was an approximately 8% reduction in R0ja in a still air ambient and 13% under forced convection (500 lfpm). These improvements in thermal performance directly effect an improvement in electrical performance characteristics in the form of higher power/current ratings as well as improved efficiency at a given current/power level. Additionally, no changes in DC electrical parameters or decreases in package reliability were observed for devices assembled using the molding compound containing SCAN.

In the area of low-cost power devices, cost increases for incremental thermal performance improvements are difficult to tolerate. Based upon the economics currently available for thermal molding compounds, specialized high-end devices, although still utilizing SOIC packages, would be more likely to achieve favorable price/performance tradeoffs. Technologies where the packaging of the IC device is valueadded, or where space and thermal budget limitations are more critical, are the most probable candidates for the use of thermal molding compounds containing SCAN, especially in small, dual-sided packages.

Acknowledgments

The authors wish to thank Asif Chowdhury, Tanya Fowler, David Shumate, Larry Walker and Bret Zahn for their contributions to this work.

Keith Edwardsis (left> a new product development engineer at Motorola's Semiconductor Products Sector, Phoenix, Arizona. He received a bachelor's degree in metallurgical engineering from the University of Illinois, Urbana-Champaign, and a master's degree in materials science and engineering from the University of Michigan. He can be reached at 602.244.3572, fax 602.244.5967 or r19812@email.sps. mot. com.

Dr. Kevin Howard is a senior technical specialist in electronic materials at the Dow Chemical Co. He earned his Ph.D. in inorganic chemistry from the University of Illinois, Urbana Champaign, and was a post-doctoral associate at the Beckman Institute. Dr. Howard joined Dow Chemical's Central Research & Development Laboratory in 1989. He is the inventor of SCAN, has authored many technical papers and holds 14 U.S. patents. He can be contacted at 517.636.1170, fax 517.638.7092 or kehoward@dow. com.