Spatially-selective cooling by liquid jet impinging orthogonally on a wettability-patterned surface

International Journal of Heat and Mass Transfer 95 (2016) 142–152

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer journal homepage: www.elsevier.com/locate/ijhmt

Spatially-selective cooling by liquid jet impinging orthogonally on a wettability-patterned surface Theodore P. Koukoravas a, Aritra Ghosh a, Pallab Sinha Mahapatra a, Ranjan Ganguly b, Constantine M. Megaridis a,⇑ a b

Department of Mechanical and Industrial Engineering, University of Illinois at Chicago, Chicago, IL 60607, United States Department of Power Engineering, Jadavpur University, Kolkata 700098, India

a r t i c l e

i n f o

Article history: Received 13 September 2015 Received in revised form 10 November 2015 Accepted 11 November 2015

a b s t r a c t Jet impingement finds application in high-rate cooling because of its numerous merits, which, however, do not include selective directionality. The present study introduces a new configuration employing a wettability-patterning approach to divert an orthogonally-impinging laminar water jet onto a predetermined portion of the target surface. Diverging wettable tracks on a superhydrophobic background provide the means to re-direct the impinging jet and enable spatially-selective cooling on the heated surface. An open-surface heat exchanger is constructed using this approach, and its heat transfer performance is characterized. Sensible heat transfer is quantified in terms of the extracted cooling flux and the heat transfer coefficient. Since this approach facilitates prolonged liquid contact with the underlying heated surface through thin-film spreading, evaporative cooling is also promoted. Thus, phase-change heat transfer is also facilitated, and results in the extraction of 12.4 W/cm2 at water flow rate of
1.5 mL/min. By comparing with other jet-impingement cooling approaches, the present method provides roughly four times more efficient cooling by using less amount of coolant. The reduced coolant use, combined with the gravity-independent character of this technique, offer a new paradigm for compact heat transfer devices designed to operate in reduced- or zero-gravity environments. Multiple hot spot cooling is also demonstrated using a single jet to feed two different tracks by minimally displacing or splitting the impinging jet. Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction Dissipation of heat from small areas is a critical requirement in many engineering applications, such as heat treatment of metal parts [1], thermal management in electronic and photonic devices [2], power devices [3], electric vehicles [3,4], advanced military avionics [3], nuclear reactors [5], or even photovoltaic cells [6]. In many applications, air has been used as a coolant due to the associated low cost. However, air cooling is limited to low heat flux applications owing to the poor thermal transport properties of the gas. Better results are achieved by using high Prandtl number liquids and by harnessing mixed convection that augments the cooling efficiency of natural and forced convections. Even higher heat transfer performance requires phase change (boiling) of the liquid coolant. Cooling overheads in datacenters account for about 33% of their total electricity consumption [7]. Thus, proficient cooling can significantly cut the energy bill as well as extend hardware life [8,9]. ⇑ Corresponding author. E-mail address: [email protected] (C.M. Megaridis). http://dx.doi.org/10.1016/j.ijheatmasstransfer.2015.11.057 0017-9310/Ó 2015 Elsevier Ltd. All rights reserved.

The requirements of intense heat removal have motivated researchers to design and implement advanced thermal management techniques [10–16]. Jet impingement is one such technique and has been widely examined by researchers as a way to achieve high performance with high heat transfer coefficients (HTC) [17]. In this classical configuration, a high-speed jet issuing from a nozzle hits a target held at a certain distance from the nozzle. As the fluid moves radially outward after impinging on the plate, a very thin boundary layer develops along the surface, facilitating heat transfer [8]. Common impingement cooling strategies include free surface jets [17], submerged jets in a confined [18] or unconfined [19] configuration, and synthetic jets [13,14]. Li and Garimella [18] experimentally studied the heat removal capacity of confined and submerged impinging jets of different fluids and proposed a generalized correlation for heat transfer rates. From their experiments with a multi-jet array, Wang et al. [20] demonstrated twophase heat fluxes up to 90 W/cm2. This study offered a direction as a potential technology for VLSI chip cooling. Kanokjaruvijit and Martinez-Botas [21] used an eight-by-eight array of air jets impinging on a dimpled surface at Reynolds number
11,500

T.P. Koukoravas et al. / International Journal of Heat and Mass Transfer 95 (2016) 142–152

143

Nomenclature A An Cp D h H l q_ q00 Q Re TC T V

superhydrophilic track area on Al plate (wedge-shaped design) nozzle cross sectional area specific heat of liquid jet diameter heat transfer coefficient nozzle-to-plate distance length of wedge track extracted heat transfer rate extracted heat flux liquid jet volume flow rate Reynolds number thermocouple temperature velocity

and reported that the dimpled surface did not always enhance heat transfer. Conventional thermal management techniques, designed for providing uniform chip temperature, would result in either lower allowable chip power or unnecessary overcooling of the adjacent (unheated) areas at the expense of increased coolant consumption, pressure drop and pumping power [22]. These factors are largely inhibiting, because they are energetically demanding. This diminishes the returns of these thermal management techniques in electronics cooling, where the least possible percentage of data processing energy should be spent on cooling [23]. Therefore, new thermal management approaches are required for hotspots, where spatially non-uniform heat flux dissipation must be facilitated. Often the orientation and geometrical disposition of heat dissipating chips in an electronic package may limit easy access to the hot spots. In a traditional free-jet impingement scenario, the jet spreads radially in all directions. Therefore, any specific directionality cannot be imparted to the impinging jet. Joshi and Dede [24] proposed a cooling strategy using a wettabilitypatterned surface where the liquid droplets, produced by spray nozzles, were transported upon impact radially inward and towards the hot spots; this was achieved by applying a radiallyvarying wettability gradient (hydrophobic in the periphery, hydrophilic towards the center) on the substrate. While the efficacy of the device in transporting liquid was not portrayed in terms of the liquid velocity or heat transfer enhancement, this scheme could not warrant a unidirectional transport on the substrate. In the present work, an improvement over the classical form of orthogonal jet impingement is demonstrated and analyzed. Wettability patterns on the target hot surface cause the impinging jet to deflect and be transported by capillary forces in a predefined direction. This provides a directionally-controlled and continuous wall jet over the heated area. Specifically-designed superhydrophilic wedge-shaped tracks are constructed on a superhydrophobic background on the horizontally-placed impingement surface, confining the flow in the transverse direction and providing Laplace-pressure driven rapid transport of the thin water layer (the coolant) in the longitudinal direction. It has been shown that wedge-shaped wettability-confined tracks can rapidly transport high liquid volume flow rates without any external pumping or the assistance of gravity [25]. The ability to fully control the flow direction by modifying the location of the superhydrophilic tracks allows flexibility in designing an adjustable cooling system with spatiallyselective cooling capabilities. The present approach can also be

Greek letters angle of wedge track d(x) local width of wedge track at x m kinematic viscosity of liquid DT temperature difference

a

Sub/superscripts avg average in leaving the nozzle I impingement j jet max maximum n nozzle out leaving the plate p plate w water

implemented in directionally-selective thin-film spreading with phase-change heat transfer, which is advantageous because of the much higher value of latent heat, as compared to sensible heat. Thus, significantly lower flow rates can be utilized to achieve the same heat removal rates as compared to an identical singlephase configuration. An added advantage of the present approach is that jet impingement with low flow rate also eliminates the possibility of erosion on the cooled surface [26,27].

2. Materials and methods An aluminum plate, functionalized with a fluoro-alkyl silane (FAS) and then etched by CO2 laser irradiation to create the requisite wettability patterned substrate, was used as the jet impingement surface. The superhydrophobization of aluminum using FAS deposition was inspired by Zhijia et al. [28] and Yang et al. [29]. Multipurpose 6061 mirror-finish aluminum (McMaster-Carr), 2 mm-thick, 89 mm  25.4 mm rectangular plates were used. The samples were initially immersed in a 4 M HCl (Sigma–Aldrich, ACS reagent 37%) solution at room temperature for 8 min to impart surface micro-roughness, as required for durable FAS attachment. Next, the samples were rinsed with de-ionized (DI) water and placed in boiling DI water for 1 h, which facilitated further growth of hierarchical micro-nano structures. This step also passivated the textured surface by the formation of a thin layer of aluminum oxide hydroxide (AlO(OH)), more commonly known as boehmite, on the plate surface [30,31]. At this point of the fabrication process, the samples were superhydrophilic with a water contact angle of
0° (Fig. 1(a)). The passivated aluminum plates were subsequently dried with nitrogen gas and immersed in a 2 wt.% solution of 1H, 1H, 2H, 2H-Perfluorodecyltriethoxysilane (Sigma–Aldrich, 97%) in ethanol (Decon labs, 200 proof) for 1 h. The samples were left overnight to dry in order to complete the hydrophobizing process, resulting in a superhydrophobic Al surface (Fig. 1(b)). Sessile contact angles for water droplets on the FAS-coated aluminum were found to be 161.2 ± 2.0°, with a contact angle hysteresis of 12.2 ± 3.8° (measured with an in-house contact angle goniometer setup). Finally, the superhydrophobic Al surface was ablated with a CO2 laser (Universal laser systems V-460, 50 W at 80% power) beam to selectively remove the fluorinated coating and create the wedge-shaped track by re-exposing the underlying superhydrophilic structure (Fig. 1(c)). The SEM images in Fig. 1 (left column) show the existence of micro- and nano-structures, while

144

T.P. Koukoravas et al. / International Journal of Heat and Mass Transfer 95 (2016) 142–152

Fig. 1. Scanning electron micrographs (left column) and optical profiler (right column) images of the Al substrate: (a) after acid-etching and passivation in boiling water (superhydrophilic), (b) after FAS-coating (superhydrophobic), and (c) after laser ablation (superhydrophilic). The insets in the SEM images show the sessile state of 4.7 lL water droplets placed on the respective surface. All scale bars denote 50 lm. Images do not depict the same sample area. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 2. (a) Patterned side of Al sample plate with superhydrophilic wedge-shaped track (light grey) laid on a superhydrophobic background (dark grey). The positions of thermocouple mounting holes (embedded in the plate from the two sides) are also shown. (b) Experimental setup component schematic. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

T.P. Koukoravas et al. / International Journal of Heat and Mass Transfer 95 (2016) 142–152

the profilometer contour plots (right column) show a maximum roughness feature size of ±40 lm. The roughness contrast appears to decline slightly after FAS coating, as a comparison of Fig. 1 (a) and (b) indicates. The images for the laser-ablated surface (Fig. 1(c)) reveal the existence of small (
5–25 lm) islands that may have originated from remnant clusters of organo-silane molecules. Fig. 2(a) shows the final form of the patterned sample. The wedge-shaped superhydrophilic track used in this experiment had a width d(0) = 1 mm at its narrow end, wedge angle a = 5° and overall length l = 86 mm. 3. Experimental setup Fig. 2 depicts schematics of the sample (a), and the experimental setup (b). The wettability-patterned side of the plate is mounted facing down, on which the vertical jet impinges from below. The inverted orientation of the setup was intended to demonstrate the ability of the technique to work against gravity and to simplify collection of coolant liquid from the other end of the track by gravity. In addition, this configuration allowed the testing of cases where the effect of momentum of the impacting jet could be minimized by suitably lowering the nozzle below the plate, or adjusting the jet’s pre-impact length through regulating the flow rate. The Al plate is uniformly heated by mounting it on a strip heater (Omega, MSH00001) by way of double-sided thermally conductive adhesive tape (McMaster Carr, 6838A11). The heater was held with thermal insulation (McMaster Carr, extreme temperature silicone rubber) on the mounting side to prevent conductive losses through the holder, as shown in Fig. 2(b). Heat supplied to the system was adjusted using a variable transformer (Staco Energy, 3PN1010) through which the heater was connected to the power supply. Eight K-type thermocouples (Omega, bead diameter 0.13 mm) were used for measuring the plate temperature at several locations on the sample: six thermocouples (TC2–TC7) were positioned along the longitudinal direction of the liquid-carrying superhydrophilic track as shown in Fig. 2(b), while two others (TC1 and TC8) measured the outlet and inlet temperatures of the coolant liquid. Thermocouple holes were drilled into the plate such that the TCs could be embedded and reach the centrally-placed superhydrophilic track on the plate. Temperature data were recorded using a differential data acquisition system (Omega DAQ, USB 2400 series) at a sampling frequency of 1 Hz. Plate temperature was also recorded using a thermal imaging camera (Testo 885, Make: Testo, AG; Resolution: 320  240 pixels, thermal sensitivity <30 mK at 30 °C, spectral range 8–14 lm) and the temperature contours of the plate surface were obtained with an image analysis software (Testo IRSoft 0501 8809). Appropriate emissivity correction was performed by heating the plate uniformly in air within the range of 30–90 °C and matching the recorded temperature readings with the acquired thermocouple data (TC2–TC7). A magnetically coupled gear pump (Micropump, GA-V23) was used to drive the cooling liquid from a constant head tank (CHT in Fig. 2(b)) through a vertically mounted syringe forming a continuous jet. The distance (H) between the plate and the syringe tip (Fig. 2(b)) was adjusted to observe the effect of jet impact momentum on heat transfer. Different diameter nozzles were also

Table 1 Different configurations employed to study the effects of jet momentum transfer to the flow confined in the hydrophilic wedge area of the Al plate.

Configuration 1 Configuration 2 Configuration 3

Dn (mm)

H (mm)

0.51 0.41 0.41

15 15 20

145

employed to allow for a wide range of jet velocities and flow rates. Two nozzles with diameters (Dn) of 0.51 mm and 0.41 mm were alternately used with the syringe. The syringe was placed at H = 15 mm and H = 20 mm from the plate, in order to vary the momentum transferred from the impacting jet onto the plate. The three different configurations studied are presented in Table 1. The water dripping from the superhydrophilic well at the wider end of the track was collected inside a thermally insulated collection cup, in which TC1 was inserted. The collection cup was constructed so that it overflows, holding only a small volume of liquid, in order to avoid stagnating water which could potentially give a faulty reading of the temperature of the water leaving the plate. The distance of the collection cup from the plate was kept at 15 mm so as to minimize heat loss during the liquid’s free fall (see Appendix A). Two minor modifications were made in the setup for the experiments that involved phase change of the liquid coolant on the heated plate: the configuration was inverted (patterned side of plate facing up), and a syringe pump (Cole-Parmer, 74,900 series) was used instead of the aforementioned gear pump. The rationale for these modifications is explained in section 4.5. 4. Results and discussion 4.1. Fluid transport The jet emerging from the tip of the syringe was aimed at the center of the narrow end (local width d(0)) of the superhydrophilic track on the underside of the plate. We observed a range of flow rates (11.3–34.1 mL/min) for which the jet turned 90° as it hit the plate with all liquid transported smoothly, without any splashing, on the wedge-shaped track. Only this range of flow rates was examined in the following. As the water jet strikes the narrow end of the superhydrophilic wedge track, it is unable to spread radially outwards, as it normally would upon impact on an unpatterned (uniform-wettability) surface. Upon impact, the water comes in contact with the superhydrophilic part on the plate, where spreading is promoted by the wettable substrate and the Laplace pressure differential created by the diverging track [25]. However, at the boundary of the wettability-confined track, the water encounters a superhydrophobic barrier which prevents lateral spreading. Thus, the impinging water jet conforms and preferentially spreads along the wedgeshaped superhydrophilic track (see Fig. 3), provided that the liquid flow rate is not high enough to cause splashing. As more liquid impinges on the superhydrophilic track, the Laplace pressure and the capillary force take over and propel the water on the plate, in a fully-controlled manner. The confined liquid advances until it reaches the elliptical well at the end of the track, where water starts accumulating, as shown in Fig. 4. Eventually the accumulated water builds up and starts dripping into the aforementioned collection cup under the effect of gravity. Non-orthogonal jet impingement is also possible and is discussed in Appendix B. 4.2. Transient non-isothermal run As a first step, the overall response and cooling effectiveness of the system was examined. In this test, the heater was activated (
103 V), causing the plate temperature to rise quickly. To curtail further heating, the pump was turned on. As seen in Fig. 5, after the heater was turned on, the temperature started rising almost linearly at an average rate of 1.15 °C/s. After 48 s, when the average plate temperature had reached roughly 85 °C, the pump was turned on at a volume flow rate of 25.8 mL/min. This caused the average plate temperature to start dropping immediately at a rate

146

T.P. Koukoravas et al. / International Journal of Heat and Mass Transfer 95 (2016) 142–152

Fig. 3. Snapshots with time stamps of the vertical water jet impinging with
2.9 m/ s on the patterned plate (horizontal), getting diverted laterally and transported along the length of the superhydrophilic track (right to left in this image). The isothermal case of Configuration 1 (Table 1) is depicted here (heater turned off). The white scale bar denotes 5 mm.

of about 0.9 °C/s, thus exhibiting the rapid cooling response of the heat exchanger. Due to adhesion limitations (degradation beyond
100 °C) of the thermally conductive tape used in the setup, the pump was turned on before the heater assumed its peak temperature. This affirms that the heater, due to its thermal inertia, keeps heating up further even after the jet impinges on the plate. As a result, after the initial steep decline (see Fig. 5), the plate temperature starts rising again, but in a more moderate rate this time, until the heater temperature achieves steady state (not shown here). The different thermocouple readings (TC2–TC7) in Fig. 5 indicate that under steady state, the plate experiences a spatial temperature gradient (in the lengthwise direction). This will be discussed in more detail in the next section.

For lower flow rates or higher heater power input, the liquid film on the track could be made to experience phase change, as in Fig. 6, which shows thermal (IR) images that capture the plate temperature distribution for transient runs at a water flow rate of
2.7 mL/min dispensed dropwise. It is important to mention that each IR image shows the solid surface temperature within ±1 °C for a calibrated surface emissivity of 0.93. The temperature scale (90–200 °C) in Fig. 6 is chosen to resolve best the temperature profile on the plate. Therefore, the displayed temperature colors of the dispensing needle (which has a different emissivity from the plate, and is almost at the ambient temperature) and the pendant droplet (which is at liquid temperature T8) at the needle tip (on right side of the IR images) do not represent correctly scaled temperatures. Fig. 6(a) shows evidence that a propagating water film (right to left) brings down the plate temperature along the track. As the flow continues and the entire track is covered by the liquid film (evaporation also occurs from the film), the temperature field becomes more uniform (
113 °C with a slightly higher temperature at bottom left corner) on the plate, while the superhydrophilic track is maintained at the lowest temperature (see Fig. 6(b)). Also, commensurate to the trend of TC readings in Fig. 5, the upstream regions of the track and plate are found to be at lower temperatures than the downstream side. In Fig. 6(c), where the liquid dispension has been turned off, the plate temperature is seen to rise to
190 °C near the wider end of the wedge and the subsequent propagation of the thermal front is observed from left to right on the plate. The IR temperature fields clearly show that the superhydrophilic track region is cooler by at least
20 °C due to convection by the unidirectionally-transported liquid film, while the remaining part of the plate is comparatively hotter, being cooled by conduction. The temperature jump across the wettability contrast line is notable in all three frames of Fig. 6. 4.3. Steady state heat transfer run Fig. 7 shows how the plate temperature at different locations behaved for a typical case that was examined. It takes roughly 10 min for the system to reach steady state. During that time, both the heater and pump operate simultaneously, while the temperature of the plate is being monitored. When the system achieves steady state, the insulated collection cup is introduced. The cooling water collection time is varied from 5 to 10 min to minimize the error induced due to small random flow fluctuations (±3.8%) caused by the operation of the pump at low rpm. For every run, the volume flow rate (Q) is measured by gravimetric method over a period of 1 min.

Fig. 4. Side-view snapshots with time stamps of the liquid flow (right to left; see white arrow in top frame) during continuous isothermal operation of Configuration 1 (Table 1). Liquid is shed periodically from the outlet (left) end in the form of dripping droplets. The horizontal dotted line in each frame marks the surface of the patterned plate. The white scale bar denotes 5 mm.

T.P. Koukoravas et al. / International Journal of Heat and Mass Transfer 95 (2016) 142–152

Fig. 5. Transient response of thermocouples embedded in sample plate (see Fig. 2 (a)). Vertical dotted lines denote the instants when the heater and water jet (flow rate of 25.8 mL/min) are turned on. TC1 is not shown, as the collection cup was not used for this transient case. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

147

Fig. 7. Steady-state thermocouple measurements indicate less than 3% fluctuation from the mean value of each reading. The initial fluctuation in TC 1 is caused by the gradual accumulation of hot water and some splashing, until the thermocouple is fully submerged in the collection vessel (Fig. 2(b)). The volume flow rate for this case was fixed at 22 mL/min. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

4.4. Sensible heat transfer

Fig. 6. Typical IR images of the heated plate undergoing liquid impingement cooling in the presence of phase change. The dispensing tube can be seen at the top right corner of each image. The impinging flow rate in this case was
2.7 mL/min, which was attained by dropwise dispension. The scale bar denotes 5 mm and applies to all three frames. (a) Initial transience due to liquid cooling a few seconds after the plate heater has been turned on, (b) thermal field after a few minutes of operation, and (c) liquid dispension has been turned off, causing the plate temperature to rise due to heating from the bottom. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

From Fig. 7, it is clear that a temperature gradient gets established lengthwise along the track. This is expected for a forcedconvection experiment with uniform heat flux applied along the flow. Considering the location of each thermocouple in Fig. 2, Fig. 7 reveals that the impingement side of the plate is the coolest, with the plate temperature rising along the direction of the flow, reaching its highest value closer to the wider end of the track. Temperature differences between thermocouples placed at the same lengthwise location (2 and 3, 4 and 5, 6 and 7 in Fig. 2(a)) are due to slight asymmetries in the depths of the blind holes where the TC beads were placed, as well as asymmetries in the heating pattern (see thermal fields in Fig. 6).

The performance of the open-surface heat exchanger was evaluated under different operating conditions, viz., by varying the voltage supplied to the heater from 80 to 115 V in
10% increments, in order to modulate the heat flux delivered to the plate. At the same time, the effect of different volume flow rates in cooling efficiency was also studied. This produced a wide range of operating parameters for two different nozzles, two different water jet heights (H in Fig. 2(b)), four different heat fluxes and flow rates ranging from 11.3 mL/min to 34.1 mL/min. The heat taken away from the plate by the water jet is evaluated in out _ p ðT out as q_ out ¼ mC and T in w  T w Þ, where T w w denote the timeaveraged thermocouple readings of TC1 and TC8 respectively; Cp = 4.18 J/g K is considered to be a good approximation for water over the range of temperatures observed in this experiment. The average heat transfer coefficient (havg) is calculated from q_ out ¼ hav g AðT apv g  T awv g Þ, where A is the area of the wedge-shaped hydrophilic track confining the water, T apv g is the averaged value of the thermocouple readings from the plate (TC2 through TC7), and T awv g denotes the average water temperature, which is evaluated by averaging the values of the thermocouple readings TC1 and TC8. The variations of heat transfer rate and heat flux removed from the plate by the liquid with volume flow rate are presented in Fig. 8; the scatter plots are obtained for four heater-power settings and three different configurations of the impinging jet. The heat transfer rate rises with volume flow rate for the highest heater power setting (88 W), but the same trend is not as evident in the other three cases. In addition, for a specific flow rate, the more heat is provided to the system (increased voltage) the higher the heat transfer rate and heat flux become. The highest values of heat transfer rate and heat flux observed are 53.4 W and 10.5 W/cm2 respectively, with a volume flow rate of 31.6 mL/min and correspond to the heater working in full power (88 W). The lowest values are 21.5 W and 4.2 W/cm2 at 16.8 mL/min flow rate for the lowest power setting (43 W) examined. For the lowest heater power setting (43 W) the heat transfer rate ranges from 4.1 W/cm2 to 5.2 W/cm2. For the other power

148

T.P. Koukoravas et al. / International Journal of Heat and Mass Transfer 95 (2016) 142–152

Fig. 8. Heat removal rate and heat flux as functions of impinging jet volume flow rate. Different symbols correspond to different jet configurations, while color denotes the power setting of the heater; see legend. The flow rate uncertainty for each point attributed to the pump operation was ±3.8%. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

settings, viz., 57 W, 71 W and 88 W, heat flux ranges between 5.3–6.6 W/cm2, 6.9–8.4 W/cm2 and 8.4–10.5 W/cm2, respectively. For each of the four power settings of the heater, the impinging jets are found to extract a maximum of about 60% of the input heat. By comparing the results of the three different configurations (Table 1) among the four different power settings, it is clear that these results overlap. This indicates that the configuration does not have any effect on the heat transfer observed. This conclusion is further verified by analyzing the havg data inferred from Fig. 8 and the corresponding thermocouple readings. The average heat transfer coefficient versus volume flow rate plot (Fig. 9) contains three types of data representing the three different system operating configurations. Each data point, shown either in blue square (Configuration 1), red circle (Configuration 2) or black triangle (Configuration 3), represents a steady state run at a specific heater operating condition and a fixed jet volume flow rate. It is important to note that the volume flow rate, along with the nozzle-to-plate distance (H) and the nozzle cross sectional area are indicative of the impact velocity of the jet on the plate; thus,

they are directly correlated with the momentum transferred to the flow over the plate. The lowest volume flow rate values in each of the three configurations correspond to the situations where the upward jet barely reaches the plate, implying an impingement with almost zero momentum. On the other hand, the highest flow rates are limited to values up to which the orthogonal deflection of the jet still occurs without any splashing. For splash-free transport, it is also important that the diameter of the jet does not exceed the width of the wedge-shaped track where the jet impinges, i.e., Dj 6 d(x). Within these lower and upper bounds of allowable flow, the mass flow rate is found (Fig. 9) to influence the heat transfer coefficient. It is important to note in Fig. 9 that the maximum allowable (for splash-free operation) flow rates are lower for the smaller diameter nozzles than the wider-nozzle configurations. With the smaller-diameter nozzle, the impact speed is approximately 60% higher than the speed with the larger-diameter nozzle, hence splashing is observed at a lower volume flow rate. For the same reason, on the very low end of flow rates, the largerdiameter nozzle is incapable of producing a jet forceful enough to reach the plate. The operating jet Reynolds numbers h i n Ren ¼ QD for the present experiments were found to range mAn between 688–1415, 649–989 and 743–1024 for Configurations 1, 2 and 3, respectively. Fig. 9 shows that, regardless of the configuration, the havg value increased monotonically with Q. To investigate if this increase in heat transfer coefficient was merely due to the increasing film velocity or momentum of the impinging jet, the havg values of the three configurations were compared. It is apparent from Fig. 9 that the havg data for all three configurations nearly overlap. Comparing results from Configurations 1 and 2 shows the influence of jet impact momentum, since both jets operate at the same jet length but different nozzle diameters. At a given flow rate, Configuration 2 (the narrower nozzle configuration), has higher jet momentum than Configuration 1. A close look of the common operational flow regimes (16.6–19.4 mL/min) of the two configurations in Fig. 9 indicates that the havg values for both configurations lie within a narrow band of 11%. This clearly suggests that the influence of the jet impingement momentum on havg is insignificant. Similarly, havg values for Configurations 2 and 3 are comparable, which indicates that the influence of nozzle-to-plate distance on havg is also negligible. Therefore, the only factor found to influence havg is film velocity. Consequently, the decisive factor for heat removal is the pumpless capillarydriven flow [25] along the wedge-shaped track. This introduces an interesting opportunity for controlling heat transfer by altering the wedge-track design on the plate (e.g., track width, wedge angle or the track length), which, however, is left for a future study. 4.5. Phase change heat transfer

Fig. 9. Average heat transfer coefficient (havg) plotted as a function of jet volume flow rate (Q). Different symbols correspond to different experimental configurations, as indicated in the legend. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

As already mentioned in Section 4.2, phase change heat transfer is attained on the heated surface under very high heater power input or extremely low coolant flow rate and adequately high temperature. This regime was further explored to estimate the elevated havg values under two-phase heat transfer condition. For a maximum heater input power of 88 W, the volume flow rate of the jet had to be significantly reduced in order to induce phase change. For example, the jet volume flow rate range entailing phase change lies roughly in the range of 1 mL/min to 1.5 mL/min, as opposed to the single phase experiments where the rate was 11.3 mL/min to 34.1 mL/min. This extremely low volume flow rate meant that water was dispensed (via a syringe pump) in the form of isolated droplets on the plate, rather than as a continuous jet. Also, the liquid momentum could no longer be harnessed to carry the liquid up to the plate; the droplets issued from the nozzle with extremely low momentum, so the setup had to be inverted

149

T.P. Koukoravas et al. / International Journal of Heat and Mass Transfer 95 (2016) 142–152

_ volume flow rate (Q), average plate temperature Fig. 10. Variation of ‘‘incipient phase change” operating parameters, e.g., extracted heat flux (q00 ) and heat transfer rate (q), max (T pav g ), and maximum plate temperature difference (DT max þ 100) and T pav g are plotted on the p ), plotted as functions of input power to the plate heater. For convenience, (DT p av g same scale T p . Data points were obtained by running each experiment at least five times for each condition (error bars denote one standard deviation). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

(patterned side of the plate facing up) to allow gravity-aided deposition of droplets. The arrangement with the plate atop the heater also allowed the use of mounting thermal paste, which endured higher temperatures without compromising thermal contact between the plate and the underlying heater. It is intuitive that the highest heat abstraction with minimum water usage would correspond to the incipient two-phase heat transfer condition, where the full volume of water is vaporized precisely over the entire length of the superhydrophilic track, leaving no local ‘‘dryout” or water dripping from the wider end of the hydrophilic track. A high-speed video (slowed 20 times) of the incipient two-phase heat transfer condition is provided in the Supplementary Information (SI1). Each data point in Fig. 10 represents such a stable incipient two-phase operating point, achieved by striking a balance between the input power and the supplied liquid volume flow rate. For example, at 67 W input power it is possible to fully vaporize roughly 1 mL/min, while for an input power of 88 W, this flow rate is no longer adequate; the wide end of the track starts drying up, leading to a runaway situation where the average plate temperature shoots up rapidly in time. To avoid this thermal runaway, the flow rate had to be increased to 1.47 mL/min so that another steady state incipient two phase heat transfer condition was obtained at 88 W. The highest extracted heat flux observed here was 12.4 W/cm2 with 1.47 mL/min flow rate. In comparison, the single-phase mode of heat transfer discussed in the previous section was found to reach 10.5 W/cm2 at the much higher flow rate of 31.6 mL/min. This clearly shows the efficacy of using the incipient two-phase heat transfer mode in curbing the water consumption for a given heat load. However, the average plate temperature for the incipient phase change condition is slightly above 110 °C, as opposed to no more than 79 °C for the single-phase runs. As seen from Fig. 10, the average plate temperature for the incipient twophase cases investigated here ranged from 112 °C to 114 °C, rising as the input power increased, while for the single-phase runs the range was 42–79 °C. The maximum plate temperature difference remains fairly constant at roughly 5.5 °C for the input power range examined here; for the single-phase runs, the maximum plate temperature difference can go up to about 20 °C. Therefore, the incipient two-phase heat transfer cases offer better plate temperature uniformity, albeit at a higher average plate temperature than the single-phase run. It is important to note that the maximum cooling rate obtained herein (12.4 W/cm2 with 1.47 mL/min liquid flow

rate) is not indicative of the actual limit of this technique, but rather of the power limit for the plate heater utilized in the present experimental setup. A more powerful heater would cause a higher temperature rise, thus requiring a higher cooling liquid flow rate, and in turn, a higher cooling capacity of the heat exchanger. 4.6. Electronics cooling perspective Exponential growth in internet data traffic demands significant cooling fluxes for electronic and opto-electronic equipment. Since liquid is a better heat carrier compared to conventional air cooling, recent research interest [11,12,20,32–36] is heavily inclined towards improving heat transfer performance of those devices. A comparative review of the liquid cooling performance of several approaches is shown in Table 2 for both single- and two-phase flow scenarios. The various methods are compared in terms of power removed to liquid flow rate ratio, which quantifies the benefit-tocost ratio for each cooling approach. The higher this ratio, the more competitive is the corresponding approach. Evidently, the latent heat during two-phase heat transfer provides a significant advantage for the present method over other approaches. It is evident from Table 2 that the power-to-flowrate ratio of the present method is nearly four times higher than the closest competing method. This increased efficiency over other methods can result

Table 2 Performance of jet impingement cooling approaches. Authors

Heat transfer area (cm2)

Heat (W)/flow rate (mL/min), water Singlephase

Tuckerman and Pease [32] Zampino et al. [33] Silverman and Nagler [12] Fabbri and Dhir [34] Han et al. [35] Wang et al. [20] Wolf et al. [36] Bintoro et al. [11] Present work

1


1.53

1 10


1.33
4.5


2.92 0.25 1 92.82
1.23 5.1


1.24–3.49
1.06


1.6

Twophase


11.25
0.25
0.69
42.9

150

T.P. Koukoravas et al. / International Journal of Heat and Mass Transfer 95 (2016) 142–152

Fig. 11. (a) Design of wettability-patterned plate for jet splitting. The inset shows how the two opposing tracks are separated by a superhydrophilic spot at the point of impingement, intended to suppress splashing. (b) An upward-directed liquid jet impinging in between two oppositely-placed identical wedge tracks is split equally and transported to two different remote spots. No splashing is observed. The volume buildup at each end spot before the onset of pinch-off is shown in (c). The jet is operating in Configuration 2 (Table 1). Scale bars denote 5 mm. See Movie SI3 in Supplementary Information. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

to substantial reduction of coolant consumption, while keeping the temperature of the substrate fairly constant. Although the foregoing sections in this work have dealt with a single jet, several electronic cooling applications deploy multiple heat-generating components on a single motherboard, demanding more efficient multiple-hotspot cooling. Conventional liquid cooling methods impose challenges in integrating the cooling circuit with existing electronic system architectures [37]. The spatiallyselective cooling demonstrated in the present work can be extended to multiple-track designs that can offer a viable solution to this problem. For example, track designs may be laid on the heated surface in such a way that a single jet feeding the coolant to one track can be minimally maneuvered (e.g., by a very small linear or angular displacement) to supply another neighboring track, resulting in the cooling of a distinctly different area on the substrate. A high-speed video (slowed 10 times) of this configuration can be found in the Supplementary Information (SI2). Besides, simultaneous supply of more than one track is also possible. For example, Fig. 11 shows the splitting of a single jet in two oppositely-oriented tracks that transport liquid to two different locations of the heated surface (70 mm apart). The two linearly-opposing tracks in Fig. 11 were laid in such a way that the jet impinged on a superhydrophilic spot interspacing the narrow ends of the tracks. Upon impingement, the jet split in half and subsequently each jet moved towards the respective downstream direction of each track. The relevant video (slowed 10 times) can be found in the Supplementary Information (SI3).

5. Conclusions We demonstrated the capability to deflect a cooling water jet impinging orthogonally on a heated metal plate, and directionally guiding it along the surface by deploying contrast-wettability patterns. A wedge-shaped superhydrophilic track on a superhydrophobic background was laid on the substrate so that the impinging liquid got transported along the track (from the narrow end to the wider one) harnessing Laplace pressure gradient, while facilitating plate-to-liquid heat transfer. The liquid transport and heat removal do not rely on gravity assistance, and thus could be used even in reduced- or zero-gravity environments (e.g., space applications). The heat transfer performance of this device was quantified in terms of its heat transfer rate, extracted heat flux and average heat transfer coefficient. For a volume flow rate of 31.6 mL/min, a maximum extracted heat flux of 10.5 W/cm2 was

attained along with a heat transfer coefficient of 0.43 W/(cm2 K) under no noticeable phase change. A plate temperature gradient was observed in the direction of the flow, with a recorded plate temperature difference (DT max p ) up to
20 °C. The jet impingement momentum was found to have no measurable effect on the heat transfer coefficient, although the latter increased monotonically with the jet flowrate. Phase-change heat transfer was realized on the substrate when the liquid evaporated as it traversed along the wedge track. Although incipient two-phase heat transfer, resulting in complete evaporation of the liquid over the entire length of the wedge-track, was observed at much lower flow rates compared to the single-phase scenario, the corresponding extracted heat transfer rate and heat flux values were higher. For example, a low flowrate of 1.47 mL/min yielded a two-phase heat flux of 12.4 W/cm2; at this condition, all of the liquid was converted to vapor over the track. Our cooling method was compared with other jet impingement cooling approaches and, with phase change, was found to be roughly 4 times more efficient than its closest competitor. This increased efficiency can lead to designs of compact and lightweight cooling systems due to the overall reduced volume of coolant required, which in combination with the gravity-independent nature of this approach can be of utmost importance in earth-orbit or outer-space applications. In comparison to single-phase scenarios, DT max for phase-change heat transp fer was much lower and approximately constant for each case at
5.5 °C, suggesting a more uniform substrate temperature in the presence of phase change. The results indicate the feasibility of even higher cooling fluxes using the present wettabilitypatterning approach coupled with phase change at higher plate heating power combined with higher jet flow rates. In addition, cooling of multiple spots was demonstrated by minimal lateral displacement of the impinging jet between two closely-spaced tracks. Besides, simultaneous liquid transport on more than one tracks was shown by splitting the jet along two opposing tracks originating from the same impact spot. The study offers guidance for designing and implementing a jet-impingement-based, spatiallyselective, reconfigurable, multiple hot-spot cooling technology requiring very low coolant flow rates. Appendix A In order to estimate how closely TC1 in the thermally insulated collection cup (Fig. 2) represents the water temperature at the wider (outlet) end of the wedge track, we produce an estimate of

T.P. Koukoravas et al. / International Journal of Heat and Mass Transfer 95 (2016) 142–152

the extent of cooling of a water droplet as it drips from the plate into the cup. It is assumed that the shape of the drop remains spherical after pinch-off from the outlet end of the track and there is no evaporative mass loss during the brief free fall. Assuming an average convective heat transfer coefficient h over the falling droplet, a simple energy balance shows

  Tb þ Ta mC p ðT b  T a Þ ¼ hA  T 1 Dt 2

ðA1Þ

where m denotes the mass of the droplet, Tb and Ta its temperature at the plate (we assume that the post pinch-off free fall starts at the plate itself) and the collection cup, respectively, and Dt the free fall duration of the droplet over the plate-to-cup height. The mass of the droplet upon pinch-off in the present design is
0.17 g (measured gravimetrically in a precision balance, A&D EJ 610, Accuracy ±0.01 g). The droplet mass was found to remain practically unchanged over the entire range of flow rates examined (only the droplet generation frequency changed with flowrate). The convective heat transfer coefficient h around the falling spherical drop was calculated using the Ranz and Marshall correlation [38] for the Nusselt number, i.e.,

Nu ¼ 2 þ 0:6Re1=2 Pr 1=3

ðA2Þ

where Pr = 0.713 for air at T1 = 20 °C and the Reynolds number (Re = 114) was calculated from the average free-fall droplet velocity
0.27 m/s (assuming zero detachment velocity from the plate surface, and a free-fall height of 15 mm). The properties of water were calculated at 70 °C, which corresponds to the case of the strongest cooling of the falling drop. Solving Eq. (A1), the value of Τb corresponding to the largest possible cooling of the droplet was calculated, yielding a maximum (Tb  Ta) = 0.02 °C. This fell well within the limits of experimental error, thus justifying the assumption that the TC1 reading faithfully represented the water temperature at the track outlet. Appendix B Besides studying vertical jet impingement on the wettabilitypatterned surface, impingement at an angle was also investigated. Orthogonal impingement of the jet on the heated substrate is not always possible due to component packaging issues. The wide range of variation of the jet impingement angle, while maintaining fully controlled diversion of the jet, makes for a more flexible design and leads to a more compact device in a large variety of cooling applications. Two different cases were examined to demonstrate the flexibility of the approach. In one configuration, the jet impinged at an angle so that the horizontal component of jet-momentum was directed from the narrow to the wide end of the wedge-shaped track (flow-assisting oblique impact). The other case had the jet impinging at an angle such that the horizontal component was in the opposite direction (flow-opposing oblique impact). In both cases, the jet was on the orthogonal plane passing through the track axis. Nozzle-to-plate vertical distance in these experiments was adjusted for different inclination angles in order to maintain the same jet length H = 15 mm (corresponding settings for Configuration 1 of Table 1). With the flow-assisting alignment of the jet, inclination angles up to
40° from the vertical were investigated. The horizontal component of the impinging jet momentum facilitated the liquid transport on the track. The flow confined in the track appeared smooth, with a much thinner profile than the cases previously reported in Fig. 4. The high-speed video (slowed 60 times) with 40° inclination with respect to the vertical can be found in the Supplementary Information (SI4). With the same inclination angle, temperature measurements on a small subset of the original flow rates were carried out at four heater power settings

151

(43, 57, 71 and 88 W, i.e., as in Fig. 8) to determine how the inclined-jet data compare with the vertical impingement data. At flow rates of 27.8–29.2 mL/min, the average heat transfer coefficients (havg) were found to range between 0.35 W/(cm2 K) and 0.42 W/(cm2 K), which is close to the vertical impingement havg data reported in Fig. 9. When the flow-opposing configuration was examined, the liquid on the wedge showed a more unstable flow than the previous case, not only at the point of jet impingement, but also throughout the whole track. As the jet impinged at the narrow end of the track directed against the flow in the track, water started accumulating at the impinging point forming a distinct bulge, which eventually got pumped by the Laplacepressure driven flow before the next bulge started developing (see video in the Supplementary Information (SI5)). Throughout the entire length of the track, the flow profile was significantly thicker than in the previous case and showed distinct traveling bulges. The upper limit of inclination in the flow-opposing configuration was determined by gradually tilting the impinging jet at higher angles (with respect to the vertical) and observing the flow up to the point when the diverging track was unable to divert the jet (without splashing) and transport the liquid away fast enough from the impingement point. The highest angle achieved for Configuration 1 settings in the flow-opposing alignment was 18° (negative), which is the case shown in Video SI5 (slowed 60 times). The upper bound (in terms of the jet-inclination angle) was obviously not encountered in the flow-assisting impingement case, since the horizontal component of the jet momentum aided the flow away from the impact spot. The heat transfer data for a flowassisting jet alignment with 10° inclination (negative, with respect to the vertical) showed that the havg varied from 0.29 W/(cm2 K) to 0.33 W/(cm2 K) in the flow regime of 22.3–24.7 mL/min. This, once again, falls in the same range observed in the corresponding flow regimes of the vertical impingement (as reported in Fig. 9). In conclusion, the impact angle was not critical in the performance of the device, but imposed limits in the device operation range. Appendix C. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.ijheatmasstransfer.2015.11.057. References [1] D.A. Zumbrunnen, Convective heat and mass-transfer in the stagnation region of a laminar planar jet impinging on a moving surface, J. Heat Transfer – Trans. ASME 113 (1991) 563–570. [2] C.F. Ma, A.E. Bergles, Jet impingement nucleate boiling, Int. J. Heat Mass Transf. 29 (1986) 1095–1101. [3] I. Mudawar, Assessment of high-heat-flux thermal management schemes, IEEE Trans. Compon. Packag. Technol. 24 (2001) 122–141. [4] K. Gould, S.Q. Cai, C. Neft, A. Bhunia, Liquid jet impingement cooling of a silicon carbide power conversion module for vehicle applications, IEEE Trans. Power Electron. 30 (2015) 2975–2984. [5] M. Draksler, B. Koncˇar, Analysis of heat transfer and flow characteristics in turbulent impinging jet, Nucl. Eng. Des. 241 (2011) 1248–1254. [6] M. Ruiz, V.P. Carey, Experimental study of single phase heat transfer and pressure loss in a spiraling radial inflow microchannel heat sink, J. Heat Transfer 137 (2015) 071702. [7] S.V. Garimella, T. Persoons, J. Weibel, L.-T. Yeh, Technological drivers in data centers and telecom systems: multiscale thermal, electrical, and energy management, Appl. Energy 107 (2013) 66–80. [8] S.G. Kandlikar, A.V. Bapat, Evaluation of jet impingement, spray and microchannel chip cooling options for high heat flux removal, Heat Transfer Eng. 28 (2007) 911–923. [9] H.F. Sheikh, I. Ahmad, Z. Wang, S. Ranka, An overview and classification of thermal-aware scheduling techniques for multi-core processing systems, Sustainable Comput.: Inf. Syst. 2 (2012) 151–169. [10] L.N. Jiang, J. Mikkelsen, J.M. Koo, D. Huber, S.H. Yao, L. Zhang, P. Zhou, J.G. Maveety, R. Prasher, J.G. Santiago, T.W. Kenny, K.E. Goodson, Closed-loop electroosmotic microchannel cooling system for VLSI circuits, IEEE Trans. Compon. Packag. Technol. 25 (2002) 347–355.

152

T.P. Koukoravas et al. / International Journal of Heat and Mass Transfer 95 (2016) 142–152

[11] J.S. Bintoro, A. Akbarzadeh, M. Mochizuki, T. Nguyen, A closed loop impinging jet cooling for computer chip, Proc. ASME 2003 International Mechanical Engineering Congress and Exposition, American Society of Mechanical Engineers, pp. 349–355. [12] I. Silverman, A. Nagler, High heat flux cooling with water jet impingement, Proc. ASME 2004 Heat Transfer/Fluids Engineering Summer Conference, American Society of Mechanical Engineers, pp. 277–282. [13] A. Pavlova, M. Amitay, Electronic cooling using synthetic jet impingement, J. Heat Transfer – Trans. ASME 128 (2006) 897–907. [14] A. Lee, G.H. Yeoh, V. Timchenko, J.A. Reizes, Heat transfer enhancement in micro-channel with multiple synthetic jets, Appl. Therm. Eng. 48 (2012) 275– 288. [15] A.M. Haywood, J. Sherbeck, P. Phelan, G. Varsamopoulos, S.K.S. Gupta, The relationship among CPU utilization, temperature, and thermal power for waste heat utilization, Energy Convers. Manage. 95 (2015) 297–303. [16] C.S. Sharma, M.K. Tiwari, S. Zimmermann, T. Brunschwiler, G. Schlottig, B. Michel, D. Poulikakos, Energy efficient hotspot-targeted embedded liquid cooling of electronics, Appl. Energy 138 (2015) 414–422. [17] A.M.T. Omar, M.S. Hamed, M. Shoukri, Modeling of nucleate boiling heat transfer under an impinging free jet, Int. J. Heat Mass Transf. 52 (2009) 5557– 5566. [18] C.Y. Li, S.V. Garimella, Prandtl-number effects and generalized correlations for confined and submerged jet impingement, Int. J. Heat Mass Transf. 44 (2001) 3471–3480. [19] E. Baydar, Y. Ozmen, An experimental investigation on flow structures of confined and unconfined impinging air jets, Heat Mass Transf. 42 (2006) 338– 346. [20] E.N. Wang, L. Zhang, L.N. Jiang, J.M. Koo, J.G. Maveety, E.A. Sanchez, K.E. Goodson, T.W. Kenny, Micromachined jets for liquid impingement cooling of VLSI chips, J. Microelectromech. Syst. 13 (2004) 833–842. [21] K. Kanokjaruvijit, R.F. Martinez-Botas, Jet impingement on a dimpled surface with different crossflow schemes, Int. J. Heat Mass Transf. 48 (2005) 161–170. [22] Y.J. Lee, P.S. Lee, S.K. Chou, Hotspot mitigating with obliquely finned microchannel heat sink-an experimental study, IEEE Trans. Compon. Packag. Manuf. Technol. 3 (2013) 1332–1341. [23] R.R. Schmidt, E.E. Cruz, M.K. Iyengar, Challenges of data center thermal management, IBM J. Res. Dev. 49 (2005) 709–723.

[24] S.N. Joshi, E.M. Dede, 2014, Two-phase heat transfer assemblies and power electronics incorporating the same, Google Patents. [25] A. Ghosh, R. Ganguly, T.M. Schutzius, C.M. Megaridis, Wettability patterning for high-rate, pumpless fluid transport on open, non-planar microfluidic platforms, Lab Chip 14 (2014) 1538–1550. [26] P.V. Rao, D.H. Buckley, Unified empirical relations for cavitation and liquid impingement erosion processes, Wear 120 (1987) 253–288. [27] B.S. Mann, V. Arya, An experimental study to correlate water jet impingement erosion resistance and properties of metallic materials and coatings, Wear 253 (2002) 650–661. [28] Z. Yu, Y. Yu, Y. Li, S. Song, S. Huo, X. Han, Preparation and characterization of super-hydrophobic surfaces on aluminum and stainless steel substrates, Surf. Rev. Lett. 17 (2010) 375–381. [29] J. Yang, Z. Zhang, X. Xu, X. Men, X. Zhu, X. Zhou, Superoleophobic textured aluminum surfaces, New J. Chem. 35 (2011) 2422–2426. [30] W. Vedder, D. Vermilyea, Aluminum + water reaction, Trans. Faraday Soc. 65 (1969) 561–584. [31] M. He, X. Zhou, X. Zeng, D. Cui, Q. Zhang, J. Chen, H. Li, J. Wang, Z. Cao, Y. Song, Hierarchically structured porous aluminum surfaces for high-efficient removal of condensed water, Soft Matter 8 (2012) 6680–6683. [32] D.B. Tuckerman, R.F.W. Pease, High-performance heat sinking for VLSI, Electron Device Lett. 2 (1981) 126–129. [33] M. Zampino, W.K. Jones, S. Kappagantula, A closed loop submerged jet impingement cooling system in low temperature cofired ceramic substrates, Proc. Thermal and Thermomechanical Phenomena in Electronics Systems, 2006. ITHERM’06. The Tenth Intersociety Conference on, IEEE, pp. 161–166. [34] M. Fabbri, V.K. Dhir, Optimized heat transfer for high power electronic cooling using arrays of microjets, J. Heat Transfer – Trans. ASME 127 (2005) 760–769. [35] Y. Han, Thermal management of hotspots using upstream laminar micro-jet impinging array, 15th IEEE EPTC, Singapore, 2013, pp. 83–87. [36] D.H. Wolf, F.P. Incropera, R. Viskanta, Local jet impingement boiling heat transfer, Int. J. Heat Mass Transf. 39 (1996) 1395–1406. [37] S. Zimmermann, I. Meijer, M.K. Tiwari, S. Paredes, B. Michel, D. Poulikakos, Aquasar: a hot water cooled data center with direct energy reuse, Energy 43 (2012) 237–245. [38] W. Ranz, W. Marshall, Evaporation from drops, Chem. Eng. Prog. 48 (1952) 141–146.