Temperature Control for Semiconductors
Some processes, like photolithography and track, require moderate amounts of cooling that utilize point-of-use temperature control such as thermoelectric devices. In some of the thin film processes, the requirements are for elevated-temperature heat removal using heat exchangers. Other thin film processes require extremely low temperature control utilizing cascade refrigeration systems to achieve temperatures down to -94oF (-70oC). Table 1 shows typical temperature ranges along with the temperature control method typically used for that manufacturing process. In addition to the wide range of temperatures, the requirements for temperature control often vary between heating and cooling with temperature control requirements of +/-0.18oF (+/-0.1oC).
As varied as these temperature requirements are, it is appropriate to call a chiller or heat exchanger a temperature control system. The temperature control system routinely is called on to provide heat to elevate the process temperature and then just as quickly to cool and maintain the process temperature as processing gets underway. The fundamental requirements for a temperature control system in the semiconductor industry are high reliability, compact size, cost-effective operation and reasonable first cost.
Reliability. With the cost of a state-of-the-art fabrication facility exceeding $1.3 billion, the requirement for highly reliable temperature control systems cannot be overemphasized. System uptime is so critical that mean time between failure (MTBF) and mean time to repair (MTTR) are key markers used for purchasing temperature control systems. For a single piece of fabrication equipment, which may cost $1.5 million, the amount of money spent on the temperature control system is relatively small. However, if the temperature control system is off-line, so is the $1.5 million piece of equipment; downtime can exceed $100,000/hr depending on the number of process layers and the size of the wafer. This drives the typical MTBF requirements for a temperature control system to 40,000 hr. To achieve this type of reliability, the temperature control system manufacturer must invest significantly in system development and component testing, and then meticulously document and monitor system reliability data.
Compact Size. With the cost of a fabrication facility being what it is, the need to effectively utilize production floor real estate is extremely high. The cost of floor space not only involves the current production layout but also future expansion requirements. This drives the price of floor space to between $3,000 and $5,000/ft2. To this end, a compact temperature control system solution is paramount. An effective approach to aid in the development of a compact design solution is concurrent engineering with key component suppliers. This helps to ensure that the most compact and cost-effective components are selected by enlisting the most knowledgeable people in their respective fields. Equally important is an aggressive research and development effort to develop innovative, commercially viable solutions that address the rapidly changing needs of the industry.
Cost Effective. Like all maturing industries, as the semiconductor market expands, so do the competitive pressures. Add to this the evolving low-cost manufacturing centers like Taiwan and China, and one can see that to stay competitive, component costs and manufacturing costs must be closely monitored and kept in check. New strategies in lean manufacturing and supply chain management are essential to remain viable in a competitive global environment.
To aggressively meet the goals of high reliability, compact size and cost effectiveness, one needs to look at the entire design of the temperature control system and determine where technology and innovation yield the most dramatic technical and financial impact. Discussions with industry experts lead to several key areas, but all of these areas focus on the core requirements of high reliability, small footprint and cost-effective solutions.
Pumping SystemA survey of numerous professionals in the semiconductor industry has made it apparent that the single largest issue facing temperature control system manufacturers is building a system that is leak-free during its service life. With the wide range of operating temperatures facing the temperature control system, creating a robust, leak-free system is a formidable challenge. Seals and joining methods that operate flawlessly at 68oF (20oC) do not fare so well at sub-zero or elevated temperatures.
Many of the temperature control applications in the fabrication facility rely on a pumping system to move heat transfer fluid from the temperature control system to the fabrication equipment. In production environments, temperature control systems often are remotely located due to the cost of production floor space, cleanroom requirements and for vibration isolation. Heat transfer fluid is pumped up to the fabrication equipment, which is located two stories above the sub-fab where the temperature control system is located. The heat transfer fluid lines may be as long as 150' with numerous fittings that result in a relatively high pressure drop. To maintain the processing equipment at the required temperature, flow rates of 15 to 30 gal/min at 100 psi often are required.
A single- or multistage centrifugal pump that can generate this pressure and flow is readily available, but the footprint of these pumps is quite large; with floor space at such a premium, they are not the preferred solution. A shaft-seal turbine pump is quite compact and can develop the pressure and flow required for the application, but the reliability of this type of pump is not acceptable as the shaft seal is prone to leaking after a relatively short service life, requiring frequent rebuilding.
A more robust alternative that solves the leaking shaft seal is a magnetically coupled pump. This type of pump effectively isolates the heat transfer fluid from dynamic seals and creates a leak-free solution, but a distinct drawback is evident as the operating envelope is pushed in lower temperature applications. As the process temperatures are pushed lower, the viscosity of the heat transfer fluid increases. Because the bearings of a magnetically coupled pump are at the heat transfer fluid temperature, this increase in viscosity causes a significant increase in the heat generated in the bearings. The increased heat load generated in the bearings effectively robs cooling power from the system, thereby reducing the amount of available cooling power for the process. Figure 2 shows the estimated heat generated in the bearings of a magnetically coupled pump compared to that of a static seal canned pump. To offset this decrease in available cooling power, the temperature control system must be made larger, sacrificing energy efficiency and floor space.
One solution to this design challenge is a thermally isolated pump (figure 3). This pump uses a thermally isolated canned motor/pump design. The bearings of the pump are not at the process temperature; therefore, the increase in viscosity of the heat transfer fluid has a significantly lower impact on bearing friction as the process temperature is lowered. The second and equally important benefit is that of the static seal. Because the entire rotor is surrounded by process fluid, there is no leak path. The static O-ring seal provides a robust sealing method that will not wear with time like a shaft-sealed pump. Moreover, the regenerative turbine design produces high pressure and flow in a small footprint, making it highly reliable and compact.
Heat Transfer FluidThe heat transfer fluid of choice for a number of years has been deionized (DI) water or DI water and glycol mixtures. After all, DI water is low cost, readily available, has good thermodynamic properties and is easily disposed of in a fabrication environment. However, with the broad range of temperature control applications, water has serious drawbacks. Pure water has a fairly narrow operating range of 32 to 212oF (0oC to 100oC). If it is not purified properly, it absorbs radio frequency and microwave energy and has low resistivity. Also, water is not particularly corrosion resistant. Adding glycol to the water improves the low temperature operating range to approximately -4oF (-20oC). However, in certain regions, glycol is not practical due to health and disposal issues, and glycol also is a potential contaminant to the process chamber. Deionizing filters are required to keep the resistivity of the water at acceptable levels. These filters have an upper temperature limit of 149oF (65oC), which limits the upper end processing temperature. If processing occurs above 149oF, without proper routine maintenance, the deionizing resin filters rapidly break down, causing a loss of resistivity in the water, which quickly leads to a decrease in production yield. More-over, it is difficult to trace the yield loss back to the process cooling water. To overcome these drawbacks, new heat transfer fluids have been developed and are becoming more commonplace in the semiconductor industry.
Fluorinated heat transfer fluids, for instance, can overcome these drawbacks by offering a range of fluids tailored for different temperature operating ranges. Because of the exceptional dielectric properties of these fluids, there is no electrolytic corrosion and there is no interaction with radio frequency and microwave energy. Although the initial system startup cost using fluorinated fluids is higher than that of using deionized water, the cost of ownership is greatly reduced with the elimination of the deionizing filters and the elimination of the resistivity monitoring hardware.
System Interface and ControlIn the highly integrated manufacturing systems used in the semiconductor industry, the temperature control system must be readily configurable to meet the needs of numerous unique OEM and end-user platforms. These varied and unique system interfaces can be a significant obstacle when responding to changes in the market. This requires a very flexible and precise control scheme that can quickly -- and with minimal effort and cost -- be configured to meet the demands of these different platforms. A temperature control system must have good diagnostic tools to maximize uptime, a user-friendly interface and superb temperature stability. Temp-erature control system manufacturers have been driven to develop various communication protocols such as Modbus, Devicenet, Ethernet, LON Works and a host of custom interfaces to communicate with unique host systems. Additional networking requirements for specialized monitoring and advanced diagnostics are becoming mainstream. Custom temperature control proportional-integral-derivative (PID) algorithms with adaptive technology are being used to optimize system performance Couple these advanced PID algorithms to variable-speed drive technology for pump and compressor control, and significant energy efficiency and temperature control gains are being realized.
As the geometry of integrated circuits becomes smaller and smaller, the requirements for accurate temperature control in the semiconductor industry will continue to increase. With the expanding market bringing ever increased competitive pressure to produce highly reliable, compact and cost-effective temperature control systems, utilizing the latest technology and developing new technology will be paramount to meeting the rigorous demands of this high-tech industry.