Thermal performance of flat vapor chamber heat spreader

(1)

Thermal performance of ﬂat vapor chamber heat spreader

Shou-Shing Hsieh

a,*

, Ron-Yu Lee

a

, Jin-Cherng Shyu

b

, Shao-Wen Chen

b

a_{Department of Mechanical and Electro-Mechanical Engineering, National Sun Yat-Sen University, Kaohsiung 80424, Taiwan, ROC} b_{Micro-System Technology Center, Industrial Technology Research Institute, Tainan 70955, Taiwan, ROC}

Received 11 August 2006; received in revised form 21 March 2007; accepted 30 October 2007 Available online 20 February 2008

Abstract

Experiments were performed to examine the spreading thermal resistance of centrally positioned heat sources and the thermal per-formance of a water charged, gravity assisted flat vapor chamber to be used for electronic cooling. Parametric studies including different heat fluxes and operating temperatures were conducted, and the effect of the relevant parameters on the cooling performance in terms of the spreading resistance was presented and discussed. The present vapor chamber heat spreader showed a heat removal capacity of 220 W/cm2_{with a thermal spreading resistance of 0.2}_C/W.

Keywords: IC cooling technology; Vapor chamber heat spreader; Evaporation and condensation

1. Introduction

A problem commonly encountered in thermal analysis of electronic packages is that of the thermal spreading resistance. Thermal spreading resistance occurs as heat ﬂows by conduction between a source and a sink with dif-ferent cross-sectional areas. Typical applications include cooling of electronic devices, both at the package and sys-tem level, and cooling of power semi-conductors using heat sinks. Fig. 1a and b, two frequently used (solid metal vs vapor chamber) spreaders, shows that a chip mounted on the bottom surface of a substrate in which heat ﬂows from the chip across on area A1into the substrate and spreads

out to leave across on area A2.

As computer systems continue to become more compact and oﬀer more functionality, the components such as pro-cessors and chipsets have experienced an increase in power dissipation. The result of this is that the ambient tempera-tures next to the microprocessor heat sinks have increased and, in some cases, temperatures in excess of 45C have

been reported [1]. In order to ensure device performance and reliability as well as durability, the junction (die) tem-peratures of the processors should not be over 90–110C. These two thermal boundary constraints on heat sink design, associated with the increasing power demand of processors have led to the understanding that improve-ments are essentially needed in all aspects of the design of cooling considerations, which mostly include the increase of the heat transfer coeﬃcient of the working med-ium and reduction of the thermal spreading resistance of the heat sink base. In order to achieve this goal, one of the approaches frequently encountered recently has been the use of a ﬂat vapor chamber instead of a traditional solid metal heat sink.

Generally, a vapor chamber is a vacuum container with/without a wick structure lining the inside walls that is saturated with a working fluid (typically, water). As heat is supplied, the liquid at that location immediately vapor-izes, and the vapor rushes to fill the vacuum. Wherever the vapor comes into contact with a cooler wall surface, it will condense, releasing its latent heat of evaporation. The condensed fluid returns to the heat source via capillary force/or gravity force, ready to be vaporized again and repeat the cycle[2].

*

Corresponding author. Tel.: +886 7 5252000x4215; fax: +886 7 5254215.

E-mail address:[email protected](S.-S. Hsieh).

www.elsevier.com/locate/enconman Energy Conversion and Management 49 (2008) 1774–1784

(2)

Exact analytical solutions of the partial differential equations governing thermal spreading for situations simi-lar to that mentioned above have been reported in the lit-erature [3]. These solutions address the general problems of determining the temperatures throughout multi-layer plates with multiple heat sources located on one side of a multi-layer structure with a uniform heat transfer coeffi-cient applied to the opposite side. Actually, some analytical and numerical solutions have been performed for symmet-ric and non-symmetsymmet-ric, singly connected, planar geometries subjected to a uniform heat flux based on a semi-infinite body model [4]. A three dimensional analytical solution using the method of Fourier expansion was developed for determination of the thermal spreading resistance of a cubic heat spreader[5].

There is a very limited number of experimental studies available regarding the ﬂat vapor chamber heat sinks [6]. Moreover, little has been done to quantify the spreading resistance of a vapor chamber heat spreader due to phase change and its comparison with that using solid metals as base materials.

In this work, an experimental study is conducted to characterize the thermal performance of a vapor chamber heat spreader and its associated thermal spreading resis-tance with centrally located heat sources of diﬀerent power inputs. The temperature distribution and heat transfer

coefficients are obtained. The thermal spreading resistance is defined as the temperature difference between the centroi-dal temperature at the heating area and the reference tem-perature, divided by the total heat from the heat source. With the definition above, the thermal operating resistance is, thus, calculated. Finally, the steady state experimental results are compared with the analytical results.

1.1. Experimental

Fig. 2is the schematic of the vapor chamber heat sprea-der to simulate the real applications of Fig. 1b. As stated previously, a vapor chamber consists of copper and glass shells (if any, for visualization purposes) and working flu-ids. One plate shell is the evaporator section that may be mounted on electronic elements to absorb heat, and the other is the condenser section from which heat is trans-ferred to air. The working fluid is evaporated on the heated side and condensed on the cooling side and then returns to the evaporator section under gravity. It is known that phase change heat transfer can be enhanced in a narrow space under certain combinations of the related parameters of heat flux, chamber size/geometry, working fluid and operating temperature. The chamber is 300 mm in length, 300 mm in width and 100 mm in thickness. The chamber top and bottom walls were made of 5 mm thick copper

CPU cross sectional area A1

Solid metal / vapor chamber cross sectional area A2

CPU CPU Motherboard Motherboard

a

_b

Fig. 1. A chip mounted on the bottom surface of a substrate.

Nomenclature

A surface area, m2

h phase change heat transfer coeﬃcient, W/m2 C

hfg latent heat, J/kg

Q total heat input, W q heat ﬂux, W/m2

Rc condenser thermal resistance,C/W

Re evaporator thermal resistance,C/W

Rsp spreading thermal resistance, C/W

Rt total thermal resistance, C/W

Tavg heater well temperature,C

Texp junction temperature measured by experiments

(inTable 2),C

Tj junction temperature,C

Tnum junction temperature obtained by analytical

solutions (inTable 2),C Tsat saturation temperature,C

x x direction, m y y direction, m z z direction, m

(3)

plate. The remainder of the walls was made of quartz glass. The evaporation section was located at the center of the outside surfaces (at the bottom) of the vapor chamber. Therefore, the chamber can be divided into two sections, i.e. one is for the evaporator section and the other is for the condenser part.

Experiments have been conducted to examine the per-formance of ﬂat vapor chambers. The experimental rig consists of an electric heater, a vapor chamber, thermocou-ples and data acquisition system. The heater, which is mounted on the center of the lower side of the vapor cham-ber, is used to simulate the electronic elements that produce heat during operation. The upper surface was cooled by natural convection. A thin layer of thermal silicone grease is inserted between the heater and the vapor chamber to reduce the contact resistance.

Fig. 3indicates the temperature measuring positions and the respective thermocouple placements. A flexible film heater with different sizes of 80, 100 and 80 mm in length and 80, 200 and 300 mm in width, respectively, was attached on the center of the bottom vapor chamber sur-face. The other side of the heater was insulated. Twenty K type thermocouples with 0.1 mm ID were installed to measure the upper/lower surface temperatures of the vapor chamber with ten on each surface of the chamber. On the lower surface, ten thermocouples were instrumented on the upper surface of the film heater to measure the junction temperature of the heat source. A 10· 0.5 mm groove was machined in the chamber walls and a high conductivity cement was utilized to embed the thermocouples within

the chamber wall. The spacing between adjacent thermo-couples was 12.5 mm except for thermothermo-couples at the end, which were separated 5 mm from each other. Four thermocouples were used on the right/left vertical walls of the chamber, as shown inFig. 3.

Two thermocouples are used to measure the vapor tem-perature inside the chamber. One more thermocouple was used to observe the history of the ambient temperature. The detailed installation can also be seen in Fig. 3. The liquid temperature in the chamber was maintained at its saturation temperature at the corresponding pressure. The temperature difference between these two temperature readings was within ±0.1C at the maximum power input. The experimental rig consists of a film heater, a flat vapor chamber, thermocouples and data acquisition system with a PC as illustrated inFig. 4.

The optimum charging amount of distilled water (work-ing fluid) was examined and obtained by vary(work-ing its charg-ing ratio in the vapor chamber for different heat inputs as shown in Fig. 5. The total volume of the vapor chamber was measured to be about 8000 ml. The minimum junction temperature can be secured following this examination, which would indicate the minimum charging volume of about 2200 ml was required to avoid system dry out. The prepared test sections of the chamber were cleaned with chlorinol and water and finally with acetone. The chamber was cleaned with acetone before each run. Once the heater was installed, the system was evacuated. In order to mon-itor the heat loss through the insulation surfaces, thermo-couples were also installed on both the inner and outer

(4)

surfaces of the polystyrofoam thermal insulation. The room temperature was maintained at 25C so that the losses from the test section were almost constant during all experiments.

A variac controlled AC power supply, a current shunt (15 X with 1% accuracy) and two precision multimeters, one for current and one for voltage measurement, provided the measurement and control pf the input electric power with an accuracy of ±1%. The other two thermocouples, as stated previously, were used to measure the vapor tem-perature and were positioned at the centroid of the cham-ber about 5 and 20 mm, respectively above the liquid free surface.

During all the tests, the saturation temperature in the vapor chamber, at least for 5 min, was kept within

±0.2C for the diﬀerent working conditions with water. Test liquids were carefully prepared and kept clean to avoid contamination. All the data were obtained and reduced with a computer controlled data acquisition sys-tem. Room temperature was maintained at 25C, so that heat losses from the chamber were almost constant during all experiments.

2. Data reduction and uncertainty analysis

For each power input, the heat transfer coefficient was calculated on the basis of bulk fluid saturation temperature (Tsat), heat flux, (q) and the average value (Tavg) of the

hea-ter and condenser wall temperatures. The heat transfer coef-ﬁcient at each power input was then calculated, following

(5)

vapor chamber without wick with a single heat source was assembled and tested for a heat sink footprint size of 300· 300 mm, and the thermal performance was evaluated in terms of evaporator and condenser temperature and thermal resistances at two diﬀerent working temperatures of 50C and 70 C. Distilled water was the working ﬂuid. The major results can be listed below:

1. The evaporation/condensation heat transfer coefficient increases as the applied heat power increases. It is found that the evaporation heat transfer coefficient is about 6000 W/m2C, and the condensation heat transfer coef-ficient is 10,000 W/m2C at 140 W applied power. 2. The maximum heat input of 140 W with a heater size of

80· 80 mm was found with a total thermal resistance (Rt) of 0.5C/W at an operating temperature of 70 C;

while the thermal spreading resistance (Rsp) is about

0.20C/W and is independent of the operating temperature.

3. The junction temperature of the die can reach 97C (74C) for 140 W heat input at the operating tempera-ture of 70C (50 C), and the heat removal rate can be up to 220 W/cm2.

4. Based on the present experimental study, it is found that the vapor chamber heat spreader is a good replacement for the traditional solid metal heat sink under the cases studied herein.

References

[1] Sauciua I, Chrysler G, Mahajon R. Spreading in the heat sink base: phase change systems or solid metals? IEEE Trans Compon Pack Technol 2002;25:621–8.

[2] D. Meh1, Vapor chamber heat sinks eliminate heat spots,<www. thermancore.com/pdfs/vapor.pdf>[05/2004].

[3] Simons RE. Simple formulas for estimating thermal spreading resistance. Electron Cool 2004;10(2).

[4] Muzychka YS, Culham JR, Yovanovich MM. Thermal spreading resistance of eccentric heat sources on rectangular ﬂux channels. J Electron Pack 2003;125:178–85.

[5] Feng TQ, Xu JL. An analytical solution of thermal resistance of cubic heat spreaders of electronic cooling. Appl Therm Eng 2004;24:323–37. [6] Go JS. Quantitative thermal performance evaluation of a cost-eﬀective vapor chamber heat sink containing a metal-etched microwick structure for advanced microprocessor cooling. Sensors Actuators A 2005;121:549–56.

[7] Hsieh S-S, Lee RY, Hsu CT, Chen CW. An analytical solution of thermal resistance of vapor chamber heat sink with/without a pillar. Energy Convers Manage 2007;48(10):2708–17.