New materials sets and design innovations are needed to handle the higher heat transfer demands of today's denser circuitry and increased clock speeds.

Thermal management is an increasingly important enabling technology in the development of advanced microelectronic packages and systems. As circuits become more dense and clock rates increase, so do heat fluxes and other thermal demands. These thermal issues are propagated through the electronics consumption chain (chip, module, system, groups of systems, etc.). Historically, thermal solutions have addressed the immediate level of concern (i.e. a manufacturer of a desktop PC worries only about cooling the chips in its enclosure), and there has been little attention given to the impact of local decisions on the next level of assembly. Today, however, evidence is growing that a global thermal management solution strategy can be more effective than a series of localized solutions. This is just one of the changes in thermal management for electronics that is noted in the Thermal Management chapter of the 2007 iNEMI Roadmap. The chapter addresses the need to develop improved cooling technology in terms of heat transfer processes, materials and innovative designs. In particular, further advances and developments are needed in:

• Thermal materials and thermal spreaders
  
• Thermal interfaces
  
• Heat pipes
  
• Air cooling
  
• Liquid cooling
  
• Direct immersion cooling
  
• Refrigeration cooling
  
• Thermoelectric cooling
  
• Advanced thermal design tools
  

This article highlights the technical challenges in some of these areas.

Situation Analysis

Air cooling with aluminum heatsinks continues to dominate the application space, but OEMs are increasingly turning to liquid cooling to meet the demand for higher performance electronics in smaller and smaller packages. However, there is a cost penalty associated with this and other higher performance thermal management solutions, which is driving manufacturers to take a closer look at the sources of heat in electronic devices.

One of the newest areas to become a significant factor in the design of electronic devices is the non-uniform generation of heat on a die, which causes localized “hot spots.”

To combat this problem, some manufacturers are turning to changes in CPU design to lower the thermal impact of higher performance devices. Multi-core CPUs are an example of this strategy and are credited for maintaining a low-cost infrastructure for thermal management of electronics. Packaging materials are also being recognized as a limiting factor in the performance of high-performance devices.

There are thermal solutions to handle some of the most demanding needs, but the technology to do so does not always fit within the expected level of cost. Unfortunately, there is a non-linear relationship between the cooling capabilities and the cost of the solution. As heat dissipation increases, so does the cost of cooling in dollars per watt but, at a certain point, the thermal management solution cost increases by 200%. At 65 watts, for example, the cost is only $0.18 per watt, but at 75 watts, it is more than $0.45.

The continued use of air cooling to maintain die temperatures below 100?C with higher die power will necessitate advanced fan/blower technology that is engineered for low acoustic noise and/or high-head operation, as well as further advances in the optimization and manufacture of heatsinks. Even with such advances, it is anticipated that more effective cooling technologies, such as water cooling and, possibly, direct immersion, will be necessary to meet growing power dissipation requirements.

FIGURE 1. Relative cooling potential of various modes of cooling.

Figure 1 illustrates the cooling potential of various modes of cooling relative to natural convection as a base. For example, natural convection with a heatsink confined to the chip footprint might typically be in the range of 0.1 to 1.5 W/cm². The chart dramatically illustrates the wide range of cooling leverage (up to 4000X) offered by higher performance cooling schemes, albeit at increased levels of complexity and cost.

In addition to the need to accommodate higher heat flux, assembly and packaging technologies are being driven to simultaneously meet very demanding requirements in the areas of performance, power, junction temperature, package geometry and cost. These demands, along with increased reliability expectations, the emergence of on-chip hot spots, and the growing interest in stacked-chip packages can be expected to push thermal and packaging needs well past the limits of today’s electronic products. Technology generations beyond 100 nm will either require materials beyond conventional metals and dielectrics, or new approaches to interconnect.

All of these areas will require increased focus in terms of thermal and packaging evaluation. Increases in the number of processors packaged within a system frame, coupled with higher chip power, will drive the total frame-level power dissipation to higher and higher levels, and will necessitate the development of more efficient means of rejecting the heat load from the system to avoid overheating the room. Today’s packaging trends are driving the need for multiple system-level thermal management solutions. Continued improvements in the thermal performance of electronic packages and systems to accommodate increased heat flux and to maintain or even reduce, junction temperatures, will require improvement of existing cooling technology, as well as the development and implementation of new techniques.

The Thermal Challenge

There are three primary approaches currently being pursued to provide improved thermal management solutions for tomorrow’s semiconductor chips and the electronic devices that use them:

1. Development of low-cost mechanisms to spread heat from hot spots that develop on the chip and PCB (chip/package level).
2. Development of quiet, more effective cooling fans that consume less power and more effective heatsinks (system level).
3. Development of compact, reliable and easy-to-use liquid cooling systems to reject heat directly to the outside environment without significantly impacting the building cooling load (facility level).

Chip- and package-level thermal management is primarily concerned with conduction heat transfer mechanisms. As a result, the metric commonly used to evaluate thermal solutions is the thermal conductivity of the packaging materials. In the past, the low power dissipation of semiconductor chips allowed the use of materials with thermal conductivities (k) on the order of 1 W/m-K. Today, these materials need to have values for k exceeding 10 W/m-K. In the near future, packaging and PCB materials with a CTE near that of silicon and k > 100 W/m-K are needed.

Table 1 [PDF format] shows some of the advanced thermal materials that are currently available or under development. One key factor in the use of these materials will be a better understanding of how to incorporate these costly materials into a cost-constrained process. Thermal simulation software that is able to handle the intricate details of package and PCB construction will also be needed to accomplish this goal of improved thermal materials.

Another potential solution for chip-level cooling will be the development of internal liquid-cooling passages to spread and transport heat from on-chip hot spots and chips with high average heat flux to the external package/PCB surfaces through the introduction of liquid flow-through chip packages.

System-level thermal management continues to be dominated by convection heat transfer; that is, air cooling. It is the preferred solution when cost is a concern; however, significant engineering development will be needed to extend air cooling limits to accommodate the power increases projected for the near future. At a chip-level heat flux of about 50 W/cm2, conventional heatsink designs become too large.

High-performance heatsink designs with high fin density and high aspect ratio fins are continuing to evolve and may provide some relief. Greater attention must be devoted to manufacturability considerations and to reducing manufacturing costs for high-performance designs. System fans and active heatsink fans with enhanced airflow and pressure drop characteristics are needed. In addition, fans need to be designed to reduce acoustic emissions or provide acoustic suppression to absorb the fan and flow noise. Research is also needed to develop models and correlations to predict heat transfer in transition and perturbed low Reynolds number flow over packages and in heatsink passages.

Facility-level thermal management is practiced today by routing cold building air to racks of servers or rows of cubicles. This is a challenging situation, and in some large server farms, the cost of cooling the building is higher than the cost of all of the electronic equipment. It is easy to see that this is an unpopular position and is driving many facility engineers to consider liquid cooling systems that bypass the facility and reject heat into a separate system.

In this area, the development of new thermal components, including low-cost, compact cold plates, pumps, and heat exchangers is the primary goal. In addition, a framework is needed within which building designers, systems integrators and OEMs can work together to optimize the thermal solution for an entire installation. Other liquid cooling technologies that can have an impact on facility-level thermal management are direct immersion cooling of CPUs and refrigeration cooling. For both of these, the major requirement is to develop a technology that is low-cost, reliable and occupies a minimum volume within the system.

Research Needs

The continued development of new and improved thermal management technology will require the combined efforts of industry-based development and university-based research with a focus on practical application. Extensive heat transfer, thermo-fluid and thermo-mechanical research is needed to define new opportunities (i.e. path breaking) and to improve predictability and reliability (i.e. gap filling). Recommendations for research required to satisfy the thermal technology needs identified in the 2007 iNEMI Thermal Management Roadmap are summarized below.

• Low-cost, higher-thermal-conductivity packaging materials, such as adhesives, thermal pastes and thermal spreaders, need to be developed for use in products ranging from high-performance computers to automotive applications.
  
• Advanced cooling technology in the form of high-performance heat pipe/vapor chamber cooling technology, thermoelectric cooling technology, direct liquid cooling technology, as well as high-performance air cooling and air-moving technologies, need to be investigated.
  
• Closed loop, liquid-cooling solutions, which are compact, cost-effective, and reliable should be developed.
  
• High-performance cooling systems, which will minimize the impact to the environment within the customer’s room and beyond, are needed.
  
• Advanced modeling tools that integrate the electrical, thermal and mechanical aspects of package/product function, while providing enhanced usability and minimizing interface incompatibilities, need to be developed.
  
• Advanced experimental tools for flow, temperature and thermomechanical measurements are needed for obtaining local and in-situ measurements in micro-cooling systems.
  
• It is further recommended that action should be taken to pool resources to fund cooling technology development, promote the involvement of university/research labs and establish a closer working relationship with vendors.
  

The iNEMI Roadmap also looks at seven product categories and identifies the thermal technology improvements needed for each product sector to fill gaps and avoid “showstoppers.” These needs are summarized in Table 2 [PDF format].

Conclusion

As electronic circuitry becomes smaller and denser, and clock rates continue to rise, thermal issues will be increasingly important and potentially limiting to performance. The iNEMI Roadmap addresses the need to develop improved cooling technology in terms of heat transfer processes, materials and innovative designs. If successfully implemented, enhanced thermal management will contribute to the increased competitiveness of packaged electronic products. PCD&M

Chuck Richardson is director, iNEMI Roadmapping. He can be reached at This email address is being protected from spambots. You need JavaScript enabled to view it.. For more information about the 2007 iNEMI Roadmap, go to inemi.org/cms/roadmapping/2007_iNEMI_Roadmap.html.