How to Mix Water and Air for Electronics Cooling (part 2 of 2)

In part 1 of our 2 part series on Hybrid Liquid/Air Cooling Systems and how you can use them to cool some of your toughest thermal challenges, we covered the air portion of our system, here in part 2 we’ll consider the liquid portion and how to integrate them.

Unless one is using a natural body of water for coolant, or operating in space, at the end of every cooling solution there is a liquid to air heat exchanger, when the generated heat is transferred.

Typical Liquid Cooling SystemFigure 2. Typical Liquid Cooling System

As shown in Figure 2, a conventional liquid cooling system consists of a cold plate, external plumbing, and a heat exchanger. The advantage of this type of system is to increase flexibility in packaging by allowing remote placement of the heat exchanger. Remote mounting does introduce disadvantages, as the external plumbing increases pressure drop throughout the system, which increases the required pumping power. The piping itself is a potential source of leakage at the plumbed junctions, as is the permeability of the piping system.

To appreciate the importance of spreading resistance, lets assume a high heat flux component generates 500 W/cm2 in a 10 x 10 mm package. The size of the copper heat sink used is 80 x 80 mm, with a base thickness of 5 mm. The spreading resistance alone for this case is 0.140C/W. Even if the thermal resistance of the heat sink is 0 (thermodynamically impossible), the temperature rise of the component above ambient is 700C. Considering an ambient of 500C, the above proposed heat sink will not cool the device adequately to prevent it from failure. The above example shows how important the spreading resistance is, especially in high heat flux applications.

There are many methods for reducing the thermal resistance. Among these methods are:

Use of a high conductivity material as the base plate of the heat sink to reduce thermal resistance. These materials include aluminum (k = 180 W/mK), copper (k = 380 W/mK), and CVD diamond (k = 2000 W/mK).
Using passive, high conductivity devices, like heat pipes, thermosyphons, or vapor chambers
Use of thermoelectric devices whose heat spreader structures consist simply of an electrically conductive heat sink with an applied external electric potential.

This induces a Thomson Effect, and provides heat transfer through the device. Of the above, the vapor chamber has been the most desirable method. Basically, a vapor chamber works like a heat pipe. The heat transfer to the base vaporizes the liquid and reaches the cold section of the chamber. The vapor condenses and returns back to the base with the help of the wick structure. But, even though the spreading resistance of a vapor chamber is theoretically appealing, it has been found that, under certain conditions, a solid copper spreader can have lower thermal spreading resistance [2,3,4].

comparison of solid copper heat sink and vapor chamberFigure 3. Comparison of a Solid Copper Heat Sink and a Vapor Chamber

In order to alleviate spreading resistance issues, Advanced Thermal Solutions, Inc. (ATS) has developed a new technology, the Forced Thermal Spreader, or FTS [5]. A schematic picture of the FTS design is shown in Figure 4.

Structure of Advanced Thermal Solutions Thermal SpreadersFigure 4. Structure of Advanced Thermal Solutions’ Forced Thermal Spreader

The FTS design is a combination of mini- and microchannels.

The heat transfer coefficient in the micro-channels is about 500,000 W/m20C. This high heat transfer coefficient creates a very small resistance between the heat source and the incoming liquid. The heat is then transferred to the bottom of the heat sink with the minifins attached to the top plate. Heat then transfers from the top plate to the ambient through the heat sink. The experimental test set up is shown in figure 5.

An experiment with an FTS was performed using an HFC-100 test equipment developed by ATS. The HFC-100 is a computerized data acquisition system capable of controlling up to 1KW of heat generated on a 1 cm2 simulated chip. This instrument is capable of ramping the heat with specified dwell times. The size of the FTS was 100 x 120 mm. Table 1 shows the experimental data from tests performed at power levels of 100, 200, and 300 W/cm2. The results show that the data is independent of the power. The experiment was conducted several times at each power level to ensure data repeatability.

Forced Thermal Spreader Test Set UpFigure 5. Forced Thermal Spreader Test Setup

Experimental Data  for the Thermal Resistance of the FTSTable 1b: Experimental Data for the Thermal Resistance of the Forced Thermal Spreader

The thermal resistance from the FTS is around 0.1370C/W on average. The water flow rate is set as 1.0~1.2 L/min, and it was observed that increasing the flow rate beyond 0.3 L/min had no noticeable change in temperature. One very noticeable phenomenon is the interfacial resistance. Because the heat source is small, 1 cm2, this resistance value is significant even under best contact conditions. It is expected that this number would be much higher in a real device application. For proof, a second experiment was carried out. In this experiment the heat source was made part of the FTS, which eliminated the spreading resistance.

Table 2 shows the data for this case. The thermal resistance of the FTS is about 0.14 to 0.15oC/W.


Experimental Data for the Thermal Resistance of the FTS, With No Interfacial ResistanceTable 2. Experimental Data for the Thermal Resistance of the FTS, With No Interfacial Resistance

The importance of the above numbers shows itself when calculating the spreading resistance. For a 100mm x 120mm base size copper heat sink, and a 10mm x 10mm heat source, the spreading resistance is about 0.12 degree C/W.  To achieve the total resistance of 0.14 degree C/W with a copper heat sink, we need a heat sink resistance of 0.02 degree C/W, which is not feasible using just air. To show this, we can look at the thermal resistance of a heat sink:

thermal resistance of a heat sink

Where Cp is the fluid heat capacitance, h is heat transfer coefficient, and is the mass flow rate. Assuming a heat transfer coefficient of infinity (thermodynamically impossible):

coefficient of infinity

To reach a resistance of 0.02oC/W, a velocity of 25 m/sec is required for a heat sink that is 100 mm wide and 20 mm high. In a typical systems environment, the h value is about 100-200 W/m2oC for very high speed flows. Assuming this heat sink has 65 fins at 1 mm spacing, the convective resistance will be around 0.02oC/W, but with an enormous amount of pressure drop of about 20 kpa (80H2O). This is clearly an impractical situation.

The data presented in this article shows that an effective hybrid system (liquid-assisted air cooling) has enormous capability for high heat flux applications. However the reader should not forget that the interfacial thermal resistance will always exist unless the interface is eliminated by integrating the cooling and the package systems. Equally important is the reliability of the cooling loops, as well as active control of the device functionality should the cooling system fail.

Got a question on part 1 or 2? Contact us and lets see how ATS’ thermal engineers can make your next project a success! Email us at , call us at 781-769-2800 or visit our Design Services

1. Soul, C., The Benefits of Liquid Cooling over Air Cooling for Power Electronics,
2. Sauciu, I., Chrysler, G., Mahajan, R., Spreading in the Heat Sink Base: Phase Change Systems or Solid Metals?, IEEE Transactionson Components and Packaging Technologies, Vol 23, No.4., 2002.
3. Jeung, S., Quantitative Thermal Performance Evaluation of a Cost-effective Vapor Chamber Heat Sink Containing a Metaletched Microwick Structure for Advanced Microprocessor Cooling, Sensors and Actuators, A: Physical Volume 121, Issue 2, 2005.
4. Wei, J., Cha, A., Copeland, D., Measurement of Vapor Chamber Performance, IEEE SEMI-THERM Symposium, 2003.
5. Xiong, D., Azar, K., Tavossoli, B., Experimental Study on a Hybrid Liquid/Air Cooling System, IEEE, Semiconductor Thermal Measurement
and Management Symposium 2006.


Leave a Reply

Your email address will not be published. Required fields are marked *