Rapid Cooling of Aluminum Nitride Heaters
by Mark Everly and Dan Block, Watlow Electric Manufacturing Company
The continuing increase in the consumption of semiconductors has created a need for faster throughput in many semiconductor production processes. Some of these processes, such as die bonding and testing of integrated circuits, require cyclic heating and cooling. Increasing the throughput for these cyclic thermal processes requires decreasing heating and cooling times. Recent advancements in heater technology using a aluminum nitride (AlN) matrix material allows for substantial reductions in heating time and creates a need for cooling methods that contribute to an overall cycle time reduction.
Single Iteration, a division of Watlow, has conducted preliminary concept generation and evaluation for several alternative solutions for providing rapid cooling of AlN based heaters. Cost/benefit analysis of the various alternatives indicated that forced air cooling would be preferred for many applications of the latest AlN heating technology. While it is not the best solution available in terms of cooling performance, compressed air is widely available, does not typically create concerns related to leaks, accommodates the range of common heater operational temperatures and can be easily integrated into production equipment as it requires relatively simple, small, lightweight components and does not complicate electrical isolation requirements. Based on this, several configurations using compressed air to accelerate the cooling of AlN heaters were evaluated through testing for feasibility and relative performance. Representative heat transfer coefficients and a suggested design procedure are presented.
For testing purposes, a simple support fixture was machined to include a duct immediately below the heater location. A simplified illustration of the basic test configuration is shown in Figure 1. In the basic test configuration, the heater forms the fourth side of a rectangular duct and is supported by a groove machined into the duct walls. Compressed air is routed to this duct through a filter/regulator and a flow meter for controlling and measuring flow and is injected into the duct near the center of the heater. The air exits the duct at both ends of the heater. A thermocouple in the air stream was used to measure the temperature at the air inlet, and a thermocouple brazed into the heater was used to measure the heater temperature. These temperatures were recorded in approximately 0.2 second intervals using an automated data acquisition system and were subsequently used to evaluate the performance of the system under various conditions. Various air flows and air temperatures at the inlet to the duct were tested to determine the dependence of the cooling rate on these factors. Heaters with integral fins were tested in addition to the standard “flat-sided” heaters. The heaters used in these tests measured 50.0 mm x 10.0 mm x 2.5 mm (1.9 in. x 0.39 in. x 0.10 in.). The fins added an additional 1.0 mm to the thickness of the heater but increased the surface area exposed to the air flow by a factor of 3. Figure 2 is a photograph of the tested heaters.
Figure 1 Basic Test Configuration
Figure 2 HEATER TEST SAMPLES
Results and Observations
As expected, initial testing showed a strong relationship between the air flow rate and the cooling rate of the heater. The presence of fins also increased the cooling rate. Figure 3 includes time-temperature curves for the standard heater with 5 and 10 CFM air flow rates and for the finned heater with 5, 10 and 20 CFM air flow rates.
Figure 3 Cooling Rates of AlN Heaters under various conditions
For use in designing future cooling systems, it is useful to identify average coefficients of heat transfer corresponding to the data. Table I summarizes the information gained based on an analysis of the test data and on the following simple relationships.
Q = Qc Qo = [(ρ V Cp) dT/dt]
Qc = h A ΔT
qc = h ΔT = Qc / A
- Q is the total heat loss per second from the heater during cooling,
- Qo is the heat loss per second from the heater excluding the air in the duct,
- Qc is the heat loss per second to the air in the duct,
- (ρ V Cp) is an expression for the heat capacity of the heater (density x volume x specific heat capacity of the material),
- dT/dt is the cooling rate, or time rate of change of the heater temperature,
- h is the average coefficient of heat transfer over the heater surface area,
- qc is the average heat flux, or heat flow per unit area, to the air in the duct,
- A is the area of the heater exposed to the air flow in the duct, and
- ΔT is the temperature difference between the heater temperature and the bulk air temperature in the duct.
Table I TEST DATA SUMMARY
Average Air Velocity
Peak Cooling Rate
Average Coefficient of Heat Transfer
The calculated cooling rate and heat transfer coefficient data in Table I uses an approximation of the heater as a single lumped mass, implying an assumption of uniform temperature within the heater. This assumption is logical since the AlN matrix of the heater is very conductive, having a thermal conductivity of approximately 140 W/m oC. However, inspection of the temperature/time curves in Figure 3 indicates that the peak measured cooling rate for each heater tested is not achieved for over 0.5 seconds after power is removed from the heater. This is true even for cases where the air flow is on during the heating process, indicating that the sensor in the heater measures a temperature lower than the peak heater temperature at the start of cooling due to its position within the heater matrix. Additionally, it is likely that the local coefficient of heat transfer is greater near the center of the heater where the incoming air impinges on the underside of the heater, and lower near the ends. The heat transfer from the heater to the air is also retarded near the ends by an increase in the air temperature as it flows along the heater. The variation in local heat flux from the heater drives a temperature gradient in the heater during cooling that in turn causes an underestimate of the heat loss during the initial 0.5 to 1.0 seconds of cooling and an overestimate later in the cooling process. The data for average heat transfer coefficient presented in Table I is averaged over the time of the cooling process as well as over the area of the heater to compensate for the error in cooling rate measurements.
Inspection of the cooling rate of the heater without flow indicates that approximately 4 to 5 W of heat were lost to the mechanical support structure and at the working surface of the heater when the heater was near 200oC (°F) i.e. Qo is between 4 and 5 W. This heat loss is not significant relative to the heat flow of the forced air inside the duct and therefore moderate variations in the heat loss other than to the air flow in the duct are negligible.
The average coefficients of heat transfer calculated for the finned heaters are lower than for the standard heaters despite the fact that the reduction in cross sectional area for flow actually increases the flow velocity in the duct. The difference is likely due to the temperature variation along the length of the fins (the fins are undoubtedly cooler at their tips). However, even with the reduced heat flux per unit surface area of the heater, the amount of heat transferred from the heater to the air is significantly higher for the finned configurations. The simplified equation, Qc = h A ΔT, offers a basis for understanding this. The three-fold increase in surface area has a greater impact on the heat transferred than a 40 percent lower heat flux. It should also be noted that the fin profile selected for the tests was not optimized for performance. Rather, it was selected for manufacturing convenience. So, some improvement in cooling rate may be possible if increased cost is acceptable for an optimized fin profile. Longer fins with a tapered cross section may show improved results.
As expected from the equations presented above, a reduction of inlet air temperature improves the cooling rate proportionately to the increase in the difference between the heater temperature and the air temperature. A vortex tube was tested as a means of reducing the air temperature without the expense of adding a refrigeration system. The test set-up allowed the generation of air temperatures down to 1°C (34°F). Figure 4 shows that the effect of reduced air temperature to1°C (34°F) is less than the effect of a small change in the air velocity. Larger decreases in air temperature would have proportionately larger effects on cooling rate; however, achieving a very cold air stream may require significantly more capital equipment than using a larger flow of compressed air from an existing system. In plants with existing cryogenic nitrogen or argon delivery systems, air (or more precisely “coolant”) temperature reduction can be considered as a cost effective means of accelerating the cooling rate. Additional testing to determine the acceptable limit for cooling rate in such designs should be done prior to finalizing any equipment design.
A greater advantage for decreased air (or coolant) temperature occurs as the temperature of the heater is decreased toward the temperature of the air stream. Systems requiring cooling of the heater or associated equipment to temperatures near or below ambient would benefit from a reduced air inlet temperature to a greater degree than systems operating at temperatures significantly above the ambient temperature.
Figure 4 COOLING RATE FOR THREE AIR INLET TEMPERATURES
Accurate estimation of the cooling rate of a heater with a load (or heated part) attached can present a problem. The average coefficient of heat transfer data presented above can facilitate the estimation of heat flow from a surface of an AlN heater exposed to a specific air flow. However, the heated part may have significant temperature gradients associated with it, its thermal coupling to the heater may be less than perfect, and time dependent sources of heat gain or loss may be present. For these reasons, most designs must be verified with a more detailed approach.
- Finite element models using time stepping solvers is one common approach.
- Computational fluid dynamics is gaining popularity as a means of analytically examining convective flows.
- Physical testing is also an option if partial prototypes can be constructed at a reasonable cost.
Regardless of the verification approach, it is usually required to design a rough first approximation of the system for analysis or testing. More accurate first approximations allow the analyses and tests to gather more useful information about the desired system. In systems using forced air cooling the following approach can be used to identify a reasonable first approximation for a design that achieves a desired cooling time.
- Identify the work piece to be heated and cooled, including its dimensions and its material.
- Select a heater of appropriate heating capacity and dimensions to achieve the heating rate and uniformity objectives (see www.watlow.com/reference ). From these first two steps the components to be heated and cooled are identified. These components, including the work piece, the heater and any other component not thermally isolated from the heater constitute the control volume for cooling (and heating).
- Identify and quantify any significant sources of heat gain or loss from (or to) the control volume in addition to the forced air flow. Conduction through structural supports, radiation to the environment and convective heat transfer from exterior surfaces are examples of sources of heat gain or loss.
- Estimate the cooling power required by calculating the amount of heat to be removed by the forced air and identifying the time that can be allowed for heat removal. Cooling power can be estimated by multiplying the heat capacity of each component by the cooling rate required, summing the results and adding any heat gains to the control volume during the cooling time period (Q = Qc Qo = [(ρ V Cp) dT/dt].
- Select the volumetric flow rate of air to achieve a desirable air outlet temperature. Vf (air) = Q / (ρ(air) x Cp(air) x ΔT (inlet to outlet) )
- Calculate the average temperature difference between the heater and the air: ΔT = [(Heater high temperature heater low temperature)/2] – [(air inlet temperature air outlet temperature)/2].
- Calculate the coefficient of heat transfer needed to support the cooling power required: h = Q / (A(heater) x ΔT ) (where A(heater ) is determined in step 2 and ΔT in step 6).
- Select a flow passage size to achieve the flow velocity required to give the heat transfer coefficient needed. Refer to Table I above. (Note that higher velocity usually means greater pressure drop, and that the impingement on the heater surface is important if the values in the table are used.)
- If the heat transfer coefficient needed cannot be achieved with reasonable air velocity, fins or some other form of extended surface can be considered to increase the area (A(heater) in step 7) exposed to the flow. Keep in mind that the apparent coefficient of heat transfer will be reduced for an extended surface. Other adjustments such as increased volumetric air flow or increased heater length or width can also contribute to improved cooling and can be considered if the initial values selected do not offer the performance desired.
The development of AlN heaters has made rapid heating of small parts and assemblies possible. For applications requiring cyclic heating and cooling, compressed air is one method for cooling that should be considered. When a system operating on a thermal cycle is appropriately designed to use forced air as a coolant, good performance can be achieved without the disadvantages of high initial cost and complexity associated with many other cooling technologies.
Mark Everly is the principal systems engineer for Watlow. Everly holds a master’s of science in mechanical engineering from the University of Missouri – Rolla. Dan Block is a laboratory technician at Watlow and holds an associate of science degree in applied electrical technology from Vatterott College.