Comparative Study of Spray and Multiple Micro Jets Cooling for High Power Density Electronic Applications

Direct cooling by means of jets and sprays has been considered a solution to the problem of cooling high power density electronic devices. Although both methods are capable of very high heat removal rates it is necessary to be able to decide which one is more convenient than the other when designing a cooling system for electronic applications. In this work the results of an investigation of the performances of sprays and arrays micro jets are reported. Experiments have been conducted using HAGO nozzles and orifice plates to create droplet sprays and arrays of micro jets, respectively. The liquid jets had diameters ranging from 50 to 150 μm and the pitches between the jets were 1, 2, and 3 mm. The test fluid was deionized water and the jet Reynolds number ranged between 90 and 2600. A comparison of the results obtained employing both sprays and jets has been carried out. The micro jets have dimensions of the same magnitude or smaller than those of the electronic components. The micro jet arrays give better heat transfer rates than the large diameter ones studied in the past and they use liquid mass flow rates similar to those used by the sprays. INTRODUCTION As the chip fabrication technology keeps improving, smaller and more powerful components are introduced in the market. The traditional air cooling techniques struggle to remove the high heat fluxes generated by these new microchips and new ways are sought to cool the components. Active cooling methods are taken into considerations, and in particular those that can provide high heat transfer rates. Liquid droplet spray and jet impingement cooling have been widely used in the metal manufacturing industry and have been shown capable of high heat removal rates. Researchers have investigated the possibility of applying such techniques to the cooling of electronic components. In this study the goal is to develop a close loop system where the liquid is sprayed directly on the back surface of the microchip thereby removing the heat and is then collected and cooled down in a small heat exchanger, and finally is recirculated by a small pump. The droplet sprays can have the form of a mist, and impinge on the surface with a random pattern or they can be formed by one or more streams of droplets which hit the surface on a fixed pattern. If the frequency of the streams is high enough, the droplets merge forming continuous liquid jets. After hitting the surface, the liquid droplets spread and if they are close enough they merge covering the surface with a thin liquid film. If the wall superheat is high, a thin vapor layer is present underneath the droplets or the thin liquid film. The heat transfer process is transient and it involves liquid and vapor convection, thin film evaporation, and air convection. The areas not covered by the droplets dry out. When continuous liquid jets are employed, the liquid film covering the surface is continuous and the heat is removed mainly by convection. Evaporation from the thin film may occur at high heat fluxes or low flow rates. The physics governing the heat removal process by sprays or jets is very complex and still not completely understood, and few theoretical models are available in the literature. Hence, it has turned out to be easier to investigate the various aspects of the problem by performing experimental work. Several studies have been conducted in the past on sprays, but most of them deal with the boiling regime, which is not of interest in the present work. Air driven sprays obtained using atomizer 1 Copyright © 2003 by ASME All properties except for hfg were evaluated at the mean film temperature. The range of parameters for their experiments is reported in table 1. nozzles are not considered in this study because they would be impractical to use in a closed system for electronic cooling. Bonacina, et al. [1] presented a study in which they developed a one dimensional conduction model for multi drop evaporation. They compared the predictions from the model to the experimental results they obtained using water droplets impinging on an aluminum surface at low wall superheats. The average droplet size was approximately 400 μm and the impinging velocity was between 1 and 2 m/s. They used camera and video camera to obtain information on the fraction of the heater’s area covered by the droplets. The highest heat transfer coefficient achieved in the experiments was 150 kW/mK, and the maximum heat flux was equal to 220 W/cm. The experimental data matched well with their prediction. Cho and Ponzel [5] experimentally investigated spray cooling of a heated solid surface using subcooled and saturated water. They tested three full cone nozzles having orifice diameter equal to 0.51, 0.61, and 0.76 mm, respectively. The distance between the nozzle and the 50 mm diameter copper heated surface was kept at 30 mm. Three liquid flowrates were tested (8.7, 5.4, and 3.7 ml/s). From the analysis of their data, Cho and Ponzel concluded that the droplet size was important only when evaporation occurred on the liquid film deposited on the impinged surface. Even though most of the data showed that the liquid flow rate had negligible effect on the heat transfer in single phase, for the case of the 0.51 mm diameter nozzle, where a flowrate of 1.8 ml/s was also tested, the heat transfer improved as the flowrate increased. They correlated their single phase heat transfer data as a function of Reynolds and Prandtl number only, as shown below, Ghodbane and Holman [2] first, and Holman and Kendall [3] later studied spray cooling on constant heat flux vertical surfaces. They tested full cone circular and square hydraulic nozzles using Freon-113. Two square heat transfer surfaces were studied (7.62 x 7.62 cm and 15.24 x 15.24 cm) and from that they concluded that the heat transfer is independent of the impinged area as long as the spray is uniform. They also found that the heat transfer increases almost linearly with the droplet mass flux, and that a high degree of subcooling causes higher onset of nucleate boiling and CHF. They correlated all the data obtained in [2] and [3] as, 309 . 0 667 . 0 Pr Re 531 . 2 = Nu (5)