Thermal Conductivity In Advanced Chips
EMERGING GENERATION OF THERMAL GREASES OFFERS ADVANTAGES
BY GREG BECKER, CHRIS LEE, AND ZUCHEN LIN
As electronic devices continue to pack more power into smaller components, temperature control has become one of the critical design challenges: how to efficiently remove the increasing amount of heat generated by greater power densities, despite shrinking architectures and smaller operating spaces. Microprocessors face an ongoing need for improved thermal dissipation, as designers continue to seek higher speeds and more power for CPUs in desktop machines and servers. But there also is a growing demand for greater computing power in other applications, such as graphics devices for video game consoles and digital applications in which more powerful devices are required to support high-quality images.
As a result, chip manufacturers are placing an unprecedented focus on thermal interface materials (TIMs) and other techniques capable of removing excess heat that has an adverse effect on component reliability and longevity. It’s well recognized that circuit (transistor) reliability is exponentially dependent on the operating temperature of the junction, and relatively small reductions (on the order of 10° to 15°C) can result in a ~2X difference in device life.1 Lower operating temperatures also mean reduced gate delay, contributing to higher processing speeds. Lower temperatures have been related to a reduction in idle power dissipation (leakage power) of the device, reducing overall power dissipation.
Figure 1. Typical CPU architecture for desktops, servers, and workstations.
Flip chip devices offer the greatest packaging efficiency and power density, and there are two applications requiring a TIM to improve the thermal path from the device to the surrounding environment. One application is the interface between the flip chip backside and the lid (or cap). Another is to improve thermal flow from the lid to the heatsink. Figure 1 shows a typical CPU architecture for desktops, servers and workstations. A key assumption is that the bulk of the heat transfer occurs through the inactive side of the silicon die, and that thermal management addresses this primary heat transfer path with energy ultimately removed by a heatsink to the ambient air.
Several types of thermal interface materials are currently in use, including epoxies, phase change materials (PCMs), greases, and gels. Among these various technologies, thermal greases offer several advantages for CPU applications: excellent wetting, high bulk thermal conductivity, a thin bond line with minimal attach pressure, and reworkability.
TIMs serve to connect the different parts in a package design, and once they are inserted between solid surfaces, effective thermal resistance is based on three factors: 1) the bulk resistance of the TIM material as determined by its finite thermal conductivity, 2) the contact resistance between the TIM and the adjoining substrates, and 3) the bond line thickness (BLT).
Thermal greases are comprised of two primary components: the matrix polymer and filler. Polymer selection typically is driven by filler compatibility, ability to wet the mating surfaces, viscosity, and other performance considerations. Silicone is a common choice of base polymer because of its temperature stability, low surface energy for wettability at the mating surfaces, and low modulus (stress-absorbing properties).
The key ingredient in TIMs is the filler material, dispersed in the polymeric matrix for handling and processability, as it is responsible for conducting heat. Important filler properties include bulk thermal conductivity, particle morphology (size and shape) and particle distribution. Ceramic powders such as alumina and zinc oxide are frequently used, due to their moderate cost and good dielectric properties. Metal particles such as silver and aluminum are also common.
In general, the higher the filler loading in a thermal grease, the greater the conductivity. Maximum filler loading is dictated by the polymer viscosity and thermodynamic wettability of the filler by the matrix. Filler loading and application pressure are also factors in determining bond line thickness, which is one of the determinants of overall thermal performance.
Table 1. Thermal conductivity of selected fillers.
Particle size plays a key role in determining the BLT, as large filler particles can act as spacers that prevent an extremely thin bond line. Thermal greases made with smaller particles can achieve thinner bonds and lower thermal resistance, as shown in Figure 2. The total volume fraction of fillers in a thermal grease has a significant effect on its thermal properties. Researchers have found that filler loading can be enhanced by as much as 50% by using a variety of filler particle sizes, with significant conductivity gains as a result.2
Figure 2. Controlled filler size enables low BLT and thermal resistance.
Current research indicates that optimized particle size is a key to the processing benefits of a lower viscosity material. When particles of different sizes are mixed, a much lower viscosity can be achieved than with suspensions containing the same volume fraction of mono-sized particles. To improve grease flowability, filler surface treatments (coatings) are also used. These are typically silanes, siloxanes, or fatty acids that will bond to the fillers to improve their flowability.
Since thermal greases can be viewed as suspensions of solid particles in a liquid medium, early attempts to model the contact resistance of thermal greases assumed them to behave like pure liquids.3,4 The approach has proven insufficient, however, as thermal greases with high filler loading (above 30% volume) exhibit viscoelastic behavior, so the viscosity of the grease depends on the shear rate applied (Figure 3).
Figure 3. Typical shear thinning behavior of a thermal grease.
Because of this viscoelastic nature, capillary action and enhanced surface wetting from applied pressure also reduce contact resistance and improve thermal performance. Simple viscosity measurement is not sufficient to describe either thermal performance or rheological behavior, which is critical to screen printing and dispensability.
Figure 4. Illustration of thermal test instrument.
While there is no standardized test method for evaluating TIM performance in CPU applications, a guarded hot plate method has been developed to measure thermal resistance across two mated surfaces under constant pressure or thickness (Figure 4).
Pressure is applied by the lower axis and measured with the load cell on the upper axis. Sample thickness is obtained by measuring the distance between heating and cooling axes with a digital micrometer. Thermal resistance is calculated from temperature values taken at five thermocouples placed on each axis. The load, film thickness, and temperature values can all be logged into a computer, which allows them to be displayed and/or saved. With this instrument, thermal resistance can be measured at various temperatures and pressures, with a controlled bond line thickness. The device accurately determines thermal resistance within ±0.03°C/W and measures the bond line within ±5 µm, making it a useful tool in screening and ranking multiple materials.
Figure 5. Power cycling of a thermally conductive grease.
Equal to thermal resistance in importance is reliability over time, as grease pump-out during device operation is a recognized failure mechanism. Power cycling is a direct method of examining grease reliability, and Figure 5 shows the results from testing one thermal grease formulation. A power cycle of 8 min. on and 2 min. off was used, producing temperature cycles between 20° and 92°C. The grease has been shown to be stable through approximately 16,000 test cycles so far, which has been estimated at being equivalent to several years of actual service for a CPU. Microprocessors are typically designed to survive normal operating conditions for 7 to 10 years.
To facilitate efficient processing, thermal greases are required to have viscosities in a range that allows application by screen or stencil printing, which forces a tradeoff in product development. Higher filler loading and smaller particle sizes that are used to reduce thermal resistance also increase viscosity, which has a negative impact on printability. However, no quantitative correlation has been established between screen printability and rheological profiles of thermal greases.
A material’s ability to eliminate voiding at the interface is also critical in CPU applications, since these air gaps act as heat transfer barriers. Solvents can be used to lower grease viscosity, although greases made with solvents often require special storage (such as refrigeration) and additional viscosity adjustment during printing as solvents evaporate. Shelf life (i.e. stable viscosity) is another consideration for ease of manufacturing. Thermal greases should demonstrate little or no viscosity creep for a significant period of time.
Formulating and optimizing effective thermal interface materials involve an in-depth understanding of chemical constituents, physical characteristics, and interactions within a polymer matrix to achieve the necessary thermal conductivity, processability, and reliability. In the absence of real-time test capabilities on actual CPUs, the use of test methods that closely simulate the actual service environment is essential to product selection and comparison.
Performance considerations and processing requirements demand tradeoffs during product development, but the emerging generation of thermal greases offers superior thermal conductivity, reliability, and manufacturability compared to existing materials. These products are being formulated specifically to overcome the limitations of traditional thermal greases, which enables the development of faster, more powerful CPUs that take advantage of the materials’ enhanced heat dissipation and process efficiency.
- Lin, Zuchen; Becker, Gregory S.; and Zhang, S. Mark, “Thermally Conductive Grease for CPU Applications,” Compotech, June 4, 2004, 103-111.
- Elliott, J. A., Kelly, A., and Windle, A. H., “Recursive Packing of Dense Particle Mixtures,” Journal of Material Science Letters 21, 1249-1251.
- Prasher, R., “Surface Chemistry and Characteristics Based Model for Thermal Contact Resistance of Fluidic Interstitial Thermal Interface Materials,” Journal of Heat Transfer, 123 (2001), 969-975.
- Das, A.K., Sadhal, S.S., “Analytical Solution for Constriction Resistance with Interstitial Fluid,” Heat and Mass Transfer 34 (1998), 111-119.
GREG BECKER, research chemist, CHRIS LEE, market leader, and ZUCHEN LIN, program leader, may be contacted at Dow Corning Corp., Electronics & Advanced Technologies, P.O. Box 995, Mail #AUB100, Midland, MI 48686-0995; 989/496-7181; e-mail: firstname.lastname@example.org.