Nucleate boiling on ultra-smooth surfaces: Explosive incipience and homogeneous density of nucleation sites

Accepted Manuscript Nucleate boiling on ultra-smooth surfaces: explosive incipience and homogeneous density of nucleation sites M. Al Masri, S. Cioulachtjian, C. Veillas, I. Verrier, Y. Jourlin, J. Ibrahim, M. Martin, C. Pupier, F. Lefèvre PII: DOI: Reference:

S0894-1777(17)30148-6 http://dx.doi.org/10.1016/j.expthermflusci.2017.05.008 ETF 9101

To appear in:

Experimental Thermal and Fluid Science

Received Date: Revised Date: Accepted Date:

2 October 2016 20 February 2017 13 May 2017

Please cite this article as: M. Al Masri, S. Cioulachtjian, C. Veillas, I. Verrier, Y. Jourlin, J. Ibrahim, M. Martin, C. Pupier, F. Lefèvre, Nucleate boiling on ultra-smooth surfaces: explosive incipience and homogeneous density of nucleation sites, Experimental Thermal and Fluid Science (2017), doi: http://dx.doi.org/10.1016/j.expthermflusci. 2017.05.008

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Nucleate boiling on ultra-smooth surfaces: explosive incipience and homogeneous density of nucleation sites M. Al Masri1, S. Cioulachtjian1, C. Veillas2, I. Verrier2, Y. Jourlin2, J. Ibrahim2, M. Martin3, C. Pupier3, F. Lefèvre1* 1

Centre de Thermique de Lyon, UMR 5008 CNRS INSA Univ. Lyon, INSA-Lyon, F-69621 Villeurbanne, France 2 Laboratoire Hubert Curien UMR CNRS 5516, University of Lyon, 42000 Saint-Étienne, France 3 HEF R&D, ZI Sud, 42166 Andrezieux-Boutheon Cedex, France

ABSTRACT Pool boiling heat transfer characteristics in saturation conditions are investigated on rough and ultra-smooth surfaces of several aluminum samples, acetone being the working fluid. The topography of the surfaces is analyzed with a confocal microscope: the samples are methodically scanned in order to record the height or the depth of each point of the surface with respect to its mean altitude. This analysis enables to characterize both the mean roughness of the sample but also the presence of significant imperfections due to the polishing process. Depending on the characteristics of the surfaces, different types of boiling incipience - from gradual to explosive - are observed using a high speed camera. When the heat flux is gradually decreased following boiling incipience on an ultra-smooth surface, nucleate boiling subsists for very small heat fluxes, leading to a homogeneous distribution of very small bubbles on the entire surface. On the contrary, in the same conditions, rough surfaces or ultrasmooth surfaces with imperfections are not able to sustain nucleation: for decreasing heat fluxes, a decreasing number of dispersed nucleation sites producing large bubbles is observed, before the full extinction of the boiling process for a heat flux much higher than with the ultra-smooth surface. For increasing heat fluxes, the results show that ultra-smooth surfaces with a small number of imperfections deliver the best thermal performance. Keywords Pool boiling, Nucleate Boiling, Boiling incipience, ONB, Ultra-smooth surfaces 1. INTRODUCTION Many engineering applications requiring very high heat transfer rates are based on nucleate boiling heat transfer, which holds the potential of facilitating the transfer of a large amount of energy over a relatively narrow temperature range, with a small weight to power ratio. These characteristics are important in terms of practical industrial or domestic applications, such as heat exchangers in air conditioning systems or refrigerant plants, boilers, etc. For decades, significant advances have been made to optimize the overall thermal performance of these systems, by modifying their geometry, increasing their surface exchange area, etc. No matter what the scale of the technology, industries are continually demanding smarter and smaller *

Corresponding author: [email protected]

heat exchangers. Nowadays, this approach has reached its limits and amelioration of the efficiency requires the optimization of the boiling process itself. Enhancement of boiling surfaces has been the goal of numerous experimental and theoretical works over the past century [1, 2]. The practical interest of these investigations is to design heat exchangers or boilers operating with efficient heat transfer characteristics. Two main objectives are considered: increasing the heat flux at constant driving temperature difference between the wall and the liquid (also called wall superheat) and increasing the critical heat flux (i.e. the maximum heat flux that can be reached without causing a damage to the wall so that higher heat flux densities can be reached for a given size of heater). Basically, the first goal can be reached if the surface enables high nucleation site density for low wall superheat and the second if the properties of the bubbles (size, shape, frequency, velocity, etc.) enable a sufficient supply in fresh liquid to the wall. Thermal properties of the boiling surface [3] as well as surface morphology and especially its roughness [4-8] have an important effect on boiling heat transfer coefficients. As an example, Jones and Garimella [6] studied the effect of surface roughness on the pool boiling heat transfer of water under atmospheric pressure and saturation conditions. The root-mean-square roughness Rq was given to characterize the surfaces. Five aluminum surfaces were analyzed: a polished surface (Rq = 62 nm) and 4 rough surfaces (Rq = 1.37, 2.81, 7.37, and 12.53 µm). They concluded that heat transfer is improved when the roughness increases. Nonetheless, Jones et al. [7] suggested an upper limit, above which there is no noticeable heat transfer enhancement. Using ultra-smooth surfaces covered with different nano-coatings, Phan et al. [9] and Jo et al. [10] highlighted also the effect of surface wettability: compared to hydrophilic surfaces, enhancement of boiling heat transfer was obtained with hydrophobic surfaces, but the critical heat flux was lower. A comprehensive review on the parameters affecting boiling heat transfer can be found in the review paper by Pioro et al [11] with new insights in the recent paper by Paz et al. [12]. The latter concluded that the effect of material and surface morphology on boiling behavior is still an open issue that needs further research. Most of the previous described parameters were taken into account in the original attempt to model heat transfer from boiling data by Rohsenow in 1952 [13]. This model was the foundation of the modelling of boiling heat transfer coefficient in the next decades [14]. Recently, Paz et al [12] study the effect of surface roughness and material on the subcooled flow boiling of water. The authors studied surfaces made of three different metals with arithmetic mean roughness heights between 0.2 µm and 7 µm. A comprehensive analysis of the morphology of the surface is presented using several parameters such as arithmetic mean roughness height and arithmetic mean surface height Sa. Moreover, the authors introduces a modified correlation including a roughness correction factor in terms of the parameter Sa instead of Ra and a material factor in terms of the wall material effusivity. This correlation shows better prediction of the heat flux versus wall temperature in the range of pressure. In Further works [15-16], the authors increased their analyses by introducing other morphological parameters such as peak height, valley depth or peak-to-valley height. In these works, the effect of the surface morphology on the heat transfer, the size and the density of the bubbles are analyzed. The results enable to improve the accuracy of the previous developed correlations. Understand the parameters influencing the onset of nucleate boiling is also an important issue. To achieve this goal, it is necessary to understand and control the surface functional parameters on which the nucleation process takes place. Since the first theoretical analysis of Bankoff [17], cavities or defects on the surface are reported as locations where nucleation

occurs preferentially even with small wall superheat. These observations led to the hypothesis that vapor trapped in surface cavities is responsible for the boiling nucleation process. Griffith and Wallis [18] identified the minimum superheat required to start boiling as a function of the cavity dimension. This theory requires the preexistence of a liquid–vapor interface of macroscopic size and does not explain the formation of premature bubble embryos and the influence of parameters such as the wettability and the surface topography at submicronic scale on the bubble incipience. Qi and Klausner [19] manufactured artificial cylindrical cavities of varying sizes - ranging from 8 to 60 µm – on a silicon wafer (Rq ~ 0.5 nm) and measured the superheat required to initiate bubble incipience. They found that over the size range of cavities, the theory developed by Griffith and Wallis [18] was qualitatively satisfactory for water, but they were not able to predict the incipience conditions for ethanol. A similar analysis was presented by Whitharana et al. [20] but for smaller cavity sizes (from 90 nm to 4.5 µm in diameter) manufactured on a silicon wafer. They found that their data are in agreement with the predictions of the Young-Laplace equation and reaffirmed the correctness of the classic view of heterogeneous bubble nucleation for water. Bourdon et al. [21] present experimental data showing that nucleation usually occurs with lower superheat on hydrophobic surfaces than on hydrophilic surfaces. A surprising result was published recently by Bon et al. [22] who observed heterogeneous nucleate boiling on ultra-smooth metallic surfaces (30 < Rq < 365 nm) at very low wall superheat despite the Griffith and Wallis [18] theory for bubble incipience. The authors concluded that heterogeneous nucleation does not exclusively originate from vapor-trapping cavities and that theoretical considerations required for bubble incipience is more complex than previously hypothesized. These results are in agreement with those obtained by Theofanous et al. [23]. Bourdon et al. [21] also analyzed boiling characteristics of highly polished bronze surface (Ra < 10 nm). The incipience was obtained for low wall superheat and the authors observed that few nanometer differences in surface roughness, even at this scale, modify the superheat needed to initiate nucleate boiling. On the other side, Jones et al. [7] observed that a superheat of 40 K is required to initiate nucleate boiling with FC-77 on an aluminum surface (Rq = 27 nm). In this paper, pool boiling heat transfer characteristics in saturation conditions are investigated on ultra-smooth metallic surfaces (Rq < 60 nm) and ultra-smooth aluminum surfaces with some defects, acetone being the working fluid. A rough surface (Rq ~ 5 µm) is also studied as a purpose of comparison. The onset of nucleate boiling and the hysteresis of the boiling curve are investigated, but the experimental bench is not designed to study the critical heat flux. Indeed, the long term goal of this work is to improve reboiler heat exchangers - made of aluminum - for cryogenic air separation applications, in which the boiling process is far from the CHF. The paper is divided in three parts. The first part presents the aluminum sample and the surface characterization. In the second part, the experimental bench is presented and a thermal model of the sample is described in order to normalize the experimental data. Indeed, the study of the ONB requires an experiment in which no parasitic bubbles appear. Therefore, the diameter of the sample is higher than the diameter of the studied area, which causes heat losses that need to be evaluated. Finally, the normalized data are presented in the third part with the conclusion of this study.

2. TEST SAMPLES AND CARACTERISATION OF THE SURFACES 2.1. Test sample The test sample, shown in figure 1 is an aluminum (AU4G, λ = 160 Wm-1K-1) circular disk of diameter 80 mm and thickness 5 mm. The studied area is a circle of diameter 25.4 mm located in the center of the upper side of the sample, and polished with different surface finishing quality. The bottom of this central surface is heated with a flat cylindrical electrical heater of diameter 25.4 mm, while the periphery is insulated. A cylindrical groove of width 3 mm and depth 4.8 mm is etched around the heated area at the bottom side of the sample in order to reduce heat losses towards its periphery. This particular geometry of the test sample is designed to prevent parasite bubble activation at its periphery. Indeed, in most studies of pool boiling, the first bubbles generally grow at the periphery of the studied surface, at the junction with an insulated environment (Teflon or Nylon or glue filling the interstice between the sample and the environment). This junction usually constitutes a very favorable place for nucleate boiling. This is the reason why in this study, the studied surface is increased with a long fin which is also polished. The thickness (0.2 mm) of the sheet of metal separating the surface and the bottom of the groove is chosen so as to lessen heat losses toward the peripheral fin and maintain the temperature of the fin close to the saturation temperature. This is sufficient to avoid nucleation sites on the fins and enables to observe the birth of the first bubble on the area of interest. Nevertheless, as for mechanical reasons the sheet of metal cannot be lessened under 0.2 mm, radial heat losses through this separating region are not negligible and need to be evaluated using a numerical model, presented in section 3. 2.2. Surface treatments All the samples were mechanically polished to obtain the requested roughness using the lapping machine M.M. 8400 from LAMPLAN®. Rotating plates are in contact with the samples with a controlled load applied by the machine’s jacks. A diamond grains liquid solution produces an abrasive action on the side of the sample to be lapped. The sample is first grossly polished with a 15 µm NEOLAP® diamond grains abrasive. The duration of this first stage depends on the initial roughness of the surface. Polishing cloths (satin woven natural fibers) are used in a second and a third polishing stages with respectively 6 and 3 µm NEOLAP® diamond grains abrasive in order to remove the presence of micro-grooves at the surface. The duration of each stage ranges from 5 to 10 minutes. A final stage uses a liquid made of aqueous suspension of non-agglomerated nanometric silica stabilized in an alkaline medium. This liquid is used to obtain ultra-smooth surfaces. After polishing, the surfaces are cleaned using ultra-pure acetone inside an ultrasonic bath. Some samples were coated with ultra-pure aluminum to observe the effect of the material quality on the boiling characteristics. The Physical Vapour Deposition (PVD) technique was used to coat a thin layer (300 nm) of ultra-pure aluminum. Symbol + is used in the next sections to refer to surfaces being coated with pure aluminum. 2.3. Characterization of the surfaces The surface of each sample was analyzed with a confocal microscope STIL Micromesure 2 system. The optical sensor has a nominal measuring range of 350 µm. The microscope resolution in the vertical direction z is 60 nm. The entire studied area (diameter 25.4 mm) is

scanned by moving the optical sensor with a step of 5 to 10 µm in the x direction and 20 to 100 µm in the y direction. Therefore a map of the surface can be obtained using each measurement point. These measurements were realized just before the boiling tests. Figure 2 (a) presents an example of roughness profile measured along one line of a sample with a confocal microscope. The mean altitude of the sample is not perfectly horizontal, because the surface is not perfectly plane. Therefore, each profile made of several points in the x direction is fitted with a 6 order polynomial fitting curve in order to flatten the surface. The mean altitude of the surface being determined, the distribution of the holes and the peaks around this mean altitude can be plotted (figure 2 b). Several parameters are generally used to characterize the roughness of a surface (Rq, Ra, etc.) as presented in the introduction. These parameters are presented in table 1 one for each surface. For a flat surface, these parameters are calculated by considering a level of reference, corresponding to the mean elevation of each points of the surface. This level of reference represents the location of an idealized surface, assumed perfectly smooth. The difference between this mean level and the real elevation of each point of the surface represents the roughness of the surface. About 20 samples, having different characteristics (ultra-smooth surfaces, ultra-smooth surfaces with defects and rough surfaces with a Rq higher than 5 µm) were tested, but only some of them are going to be presented (Table 1). Indeed, the boiling characteristics are similar in many cases and the study can be restricted to a limited number of surfaces. In table 1, SS designates smooth surfaces, SSd designates smooth surfaces with defects, SR designates rough surfaces and the symbol + designates surfaces with ultra-pure aluminum coating. S S1, SS1+ and SS2+ are ultra-smooth without any defects; SSd3+ is ultra-smooth with ten defects. Surfaces SSd4 and SSd5 have increasing Rq and some defects. Surface SR6 is rough with many defects. Figure 3 presents four examples of profiles measured by confocal microcopy with a scale adapted to the roughness. Figure 4 presents the distribution of altitudes, on a logarithmic scale, of five surfaces having different Rq level. We observe a symmetrical distribution of the altitudes with an anomaly for the positive altitudes which is due to the presence of very small dusts. Indeed, the measurements were not realized in a perfectly clean room. Nonetheless, these dusts are removed by an acetone washing procedure before the boiling experiments. Several imperfections were measured on some samples. These imperfections are probably the result of some holes of microscopic size that were present within the material before the polishing process and that could not be removed during the successive polishing steps. Their diameter is higher than 5 µm. Surface roughness parameters such as Ra or Rq are not sufficient to characterize the role of surface topography on the boiling process. Indeed, these parameters do not able to take into account the presence of one or several imperfections on the surface, which can create artificial nucleate sites. Therefore, in this paper, the Rq and Ra parameters are calculated without taking into account the points considered as defects. Nonetheless, these imperfections being fundamentals for the study of boiling mechanisms, their number as well as their characteristic length is given separately for each sample. As an example, the Rq of surface SSd3+ is equal to 1100 nm if the 10 defects are not removed from the data, while this value decreases to 95 nm when defects are not taken into account, which shows that it is necessary to separate both surface finishing.

3. POOL BOILING EXPERIMENTAL SET-UP AND DATA PROCESSING In this section, the pool boiling experimental set up and the instrumentation developed to characterize the boiling characteristics of the different samples are presented. As the aluminum discs are not rigorously similar in terms of thermal characteristics, a 2D numerical model is developed to process the experimental data. 3.1. Experimental set-up and experimental procedure The experimental setup for pool boiling study is shown in figure 5. The pool boiling chamber is a transparent reservoir with a double wall connected to a thermostatic bath used to control the temperature (± 0.2 °C), and therefore the saturation pressure inside the reservoir. The reservoir is tightly sealed to an AU4G support using a PTFE O-ring and a clamp specifically designed to close this type of glass reactors. The aluminum sample is tightly plated against the support made of Au4G, using a PTFE sealing ring. Therefore, the pool boiling experiment is fully hermetically sealed with two O-rings. The flat heater (Captec, diameter 25.4 mm, 0.25 mm thick, resistance 7 Ω) located under the sample is connected to a controlled power supply. A disk shape fluxmeter (Captec) is located under the heater in order to evaluate the heat losses. It has the following characteristics: sensitivity 2.62 µV.W-1.m-2, response time = 0.3 s, diameter = 25.4 mm and thickness = 0.4 mm. Both the heater and the fluxmeter are plated against the sample using a piston made of PEEK (λ = 0.25 W/m.K) in order to reduce the heat losses. The temperature of the sample Tm is measured with a K type thermocouple (diameter 80 µm) which is inserted inside a micro groove of 300 µm width and 450 µm depth machined at the bottom of the sample. Once the thermocouple is inserted inside the microgroove a copper wire of diameter 315 µm is forced into the groove in order to seal it. Therefore, the contact thermal resistance between the thermocouple and the bottom of the device is small. Three tubes made of stainless steel (6 mm diameter) cross the AU4G support. The first tube is used to empty and fill the reservoir. The second tube is connected to a manometer and to a primary vacuum pump. The manometer measures the pressure inside the reservoir. The third tube contains a K-type thermocouple in order to measure and control the temperature of the reservoir Tsat. The heater is connected in series with a shunt of resistance 0.001 Ω in order to measure with precision the power. The temperature measurements of the fluid and of the test sample, as well as the measurements of the voltage across the shunt, the heater and the fluxmeter are recorded using a switch Keithley 7700 connected to a multimeter Keithley 2700 data acquisition system. A custom designed Labview Virtual Instrument is used to monitor and record the experimental data. A high speed camera (Photron Fastcam SA3) is used to observe the boiling phenomena illuminated by a square LED. The acquisition frequency of the camera is set to 1000 frames per second and the image size is 384×240 pixels, which enables to observe the ONB phenomenon that can appear anywhere on the studied surface. Therefore it is not possible to reduce the size of the image and thus increase the acquisition frequency. The working fluid is ultra-pure acetone, which is chemically compatible with aluminum and has good wetting properties on this material: measured contact angles of acetone droplets on the different samples are in the range 12° to 20°. The working fluid is introduced in the preliminary emptied reservoir, to avoid the presence of non-condensable gas. The reservoir is then heated to a temperature corresponding to a pressure higher than the atmospheric

pressure, and the fluid is degassed several times to ensure the absence of dissolved non condensable gas. The different samples are tested following the same experimental conditions. The reservoir is maintained at a temperature close to 60 °C, corresponding to a pressure of acetone equal to 1.15 Bar. The electrical power of the heater is increased step by step. Each step is followed by a period of transient conditions preceding a steady state of the system during which the temperatures are measured. When nucleate boiling occurs on the surface, a high speed camera movie of the nucleation process is also recorded with a frame every 1 ms. When a heat flux step is followed by the onset of nucleate boiling (ONB), a movie is also recorded to observe the boiling incipience phenomenon. After ONB, the heat flux of the heater is increased until a maximum of 15 W/cm², which is lower than the critical heat flux. The critical heat flux was found to be 38 W/cm2 for a saturation pressure of 1 bar [24]. As explained in the introduction, the experimental bench was not designed to observe the critical heat flux, but the boiling incipience. When 15 W/cm² is reached, the electrical power is decreased step by step following the same procedure in order to observe the boiling hysteresis and the repartition of the nucleation sites at low heat fluxes. Each experiment is realized four times, with a rest period of several hours in between, in order to observe the system behavior repeatability. In boiling conditions, we observe that bubbles are present only in the middle part of the samples, the peripheral area beyond the groove being always subjected to natural convection. In order to estimate the uncertainty of the temperature measurement inside the microgroove at the bottom of the sample, one test sample was instrumented with three calibrated thermocouples (uncertainty lower than 0.5 K) located in three parallel microgrooves at the bottom of the device. The test sample was heated following the previously described procedure several times. Between two experiments, the thermocouple located in the middle of the sample was removed and introduced again inside the microgroove using a 315 µm copper wire to seal the groove as described above, while the two other thermocouples were let in place. The difference of temperature measured between this thermocouple and the average value of the two other thermocouples was recorded for increasing and decreasing heat fluxes. The maximum range of temperature differences observed between these experiments is lower than 3 K. In the following, the measured superheat (Tm-Tsat) is presented versus the measured heat flux. Therefore, the maximum uncertainty on the superheat measurement, obtained by adding the localization uncertainty to the uncertainty of both thermocouples leads to a maximum uncertainty lower than 4K. The measured heat flux is lower than 1% and is therefore negligible compared to the temperature measurements. 3.2. Data processing using a 2D axisymmetric thermal model of the sample Figure 6 shows the boiling curves plotted for surfaces SS2+, SSd3+ and SR6 from raw data. As any classical boiling curve, it is composed of two different sections: a first section where the measured superheat (Tm-Tsat) increases rapidly with the increasing heat flux ϕ m, which is characteristic of natural convection phenomenon, followed by a second section where nucleate boiling enables a smaller superheat for the same increasing or decreasing heat fluxes. On figure 6, a dotted line roughly separates nucleate boiling data from natural convection data. Given the small roughness of the samples, the natural convection section of these curves should not depend on the surface characteristics and therefore should be superposed whatever the tested sample. However, some differences are observed experimentally, which are mainly due to the mean thickness of the sheet of metal separating the center of the sample and its

periphery. Indeed, this thickness is different from one sample to the other. Moreover, the thickness of this sheet of metal is not constant: at the bottom of the groove, a curvature is observed at the junction with the thick parts of the sample. Indeed, it was not possible to machine accurately the same groove inside each sample with a remaining fin of thickness exactly equal to 200 µm from one edge to the other. The accuracy of the machining process is of the order of 100 µm. As the remaining thickness of the thin sheet of metal above the groove cannot be lessened under 200 µm for mechanical reasons, the heat losses towards the periphery are not negligible, especially in natural convection conditions. Therefore, the heat losses towards the periphery are different between two samples, leading to different natural convection curves. In order to facilitate the comparison between the different samples, we have chosen to develop a numerical model enabling to calculate the heat flux really dissipated in the studied area - compared to the total energy of the heater - and thus to normalize the data. This model is presented in the next sections. 3.2.1

Model description

An axisymmetric 2D finite difference steady state numerical thermal model is developed to calculate the temperature and heat flux fields within the sample. An alternating-direction explicit finite difference method is used. This model enables to estimate the fraction of the heater power dissipated in the studied area and the heat losses towards the periphery of the sample. Since the model is used in an inverse approach to estimate the unknown parameters of the experiment, a program was developed using VBA and compared to the software COMSOL for validation. The SOLVER tool of VBA is used for the estimation of the unknown parameters. The boundary conditions and the nodes of the model are presented in Figure 7. The device is discretized into n lines i in the horizontal direction and m columns j in the vertical direction with constant space step. Each node (i, j) has a constant temperature Tij. At the bottom of the sample, a heat flux, ϕm is imposed at the center of the surface while the rest of the bottom surface is insulated. At the top of the sample, a heat transfer coefficient hi with a fluid at temperature Tsat is defined. A constant heat transfer coefficient in natural convection is considered at the periphery of the sample while the area of study located just above the heater is cooled down either by natural convection or by nucleate boiling. Therefore, the heat transfer coefficient hi is equal to hnc in natural convection conditions and hb in boiling conditions. As the thickness in each point of the fin cannot be known experimentally, the thickness of the fin is supposed to be constant in the model and equal to the mean thickness of the fin. Figure 8 presents an example of the calculated temperature difference Tij - Tsat within the sample in natural convection (hnc = 500 W/m²K). The applied heat flux is equal to 15 W/cm². The temperature fields at the bottom (Tn1≤Tnj≤ Tnm) and at the top (T11≤T1j≤T1m) of the device are plotted for three different thicknesses (Hr = 150 µm, Hr = 200 µm and Hr = 250 µm) of the sheet of metal separating the area of study and the peripheral surface. This figure shows two stages of temperature: a high temperature stage in the studied area and a low temperature stage in the peripheral zone separated by a high temperature gradient zone corresponding to the location of the groove. In the studied area, the temperature at the bottom of the device is slightly higher than at the top due to the thermal resistance in the vertical direction. In the fin and in the peripheral areal, these temperatures are equal since the bottom is considered as

adiabatic. The numerical results show that the difference of temperature between the center and the periphery of the device is large enough to prevent the formation of parasitic bubbles on the periphery as it was already shown in [25]. Therefore the boiling phenomenon occurs only in the central part of the sample. One can see that the fin width is not small enough to fully prevent the heat flux from the studied area to the peripheral surface, especially in natural convection conditions. Moreover, the results show that the thickness of the sheet of metal is a very sensitive parameter of the model. For a heat flux of 15 W/cm², increasing the thickness of the sheet of metal by 1 µm increases the temperature at the bottom of the device by 0.2 K. As explained above, the thickness of the sheet of metal is different from one sample to the other for machining reasons but also because of the polishing process. Furthermore, this thickness is not constant along the fin as explained above. As a result, the geometry of the fin is a major and critical unknown of the thermal problem in natural convection and must be estimated to analyze properly the results and compare the data one to the others. 3.2.2

Fin thickness and natural convection heat transfer coefficient estimation

The heat transfer coefficient hnc is not known during the experiment as well as the mean thickness of the sheet of metal between the center of the sample and its periphery. Nonetheless, at the bottom of the device, both the heat flux and the temperature are measured. Knowing these two boundary conditions enables to identify the unknowns of the problem using an inverse approach and all the data recorded during the different experiments. The estimation of these parameters requires solving the numerical model a multitude of times, which is time consuming. Therefore, it is necessary to reduce the model. The number of nodes in the vertical direction is therefore reduce to one node, while the number of node in the radial direction is equal to n = 400, minimum number for which the result of the model do not change when the number of nodes is increased. Compared to the complete numerical model, this introduces a bias, which is equivalent to an increase of 20 µm of the thickness of the fin. This bias being similar for all the samples, it does not modify the conclusions concerning the comparison between the different samples. To solve the inverse problem, we assume that hnc is similar for each experimental data in natural convection. Obviously, the thickness of the sheet of metal is supposed to be constant for each test concerning the same sample. The comparison between the temperature measured in natural convection conditions and the numerical model enables to estimate the unknown parameter for each sample using the VBA solver. A mean heat transfer coefficient equal to 500 W/m²K is estimated and the estimated metal thicknesses vary between 150 µm and 350 µm, depending on the samples, which is coherent with the machining process. Figure 9 shows an example of the data obtained in natural convection, with no correction of the heat losses (ϕm versus Tm-Tsat in figure 9a) and with correction of the heat losses (ϕ sc versus Tsc-Tsat in figure 9b). ϕsc and Tsc are respectively the heat flux and the temperature in the central part of the sample. Figure 9b shows that the three curves become mixed up after correction. One can note that a large part of the heat flux is dissipated on the periphery rather than in the central part of the sample in natural convection. Therefore the heat flux dissipated in the studied area ϕsc is much lower than the heat flux measured on the heater ϕm.

3.2.3

Boiling convection heat transfer coefficient estimation

In section 3.2.2, the boiling curves in natural convection were normalized by identifying the natural heat transfer coefficient and the thickness of the metal between the center of the sample and its periphery. Knowing the natural heat transfer coefficient in the peripheral part of the sample and the thickness of the fin, it is possible to identify the boiling heat transfer coefficient in the studied area by comparison of the numerical model with the experimental data. Therefore, the complete boiling curve with correction of the heat losses can be plotted as it will be shown in the next sections. Figure 10 presents an example of the calculated temperature difference Tij - Tsat within the sample in boiling conditions (hb = 2500 W/m²K and hnc = 500 W/m²K). The applied heat flux is equal to 15 W/cm². The temperature fields at the bottom (Tn1≤Tnj≤ Tnm) and at the top (T11≤T1j≤T1m) of the device are plotted for three different thicknesses (Hr = 150 µm, Hr = 200 µm and Hr = 250 µm) of the sheet of metal separating the area of study and the peripheral surface. In this case, the model is less sensitive to the thickness of the fin and therefore ϕsc is much similar to ϕm than in natural convection conditions. Therefore, the correction of the heat losses towards the periphery of the sample is less visible. 4. ANALYSES OF THE EXPERIMENTAL RESULTS This section presents boiling curves with correction of the heat losses for the different kind of samples: smooth surfaces, rough surfaces and smooth surfaces with defects. 4.1. Boiling curves for smooth surfaces Figure 11 presents four boiling curves obtained for increasing and decreasing heat fluxes for the smooth surface SS2+ without defects. The ageing of the surface does not explain the differences between the curves since there is a random distribution of the four curves. The wall superheat reached just before the onset of nucleate boiling (ONB) is very high and is within the range 70 K < Tsc – Tsat < 90 K for the four different tests. The corresponding heat flux in the studied area is within the range 3.5 Wcm-2 to 4.5 Wcm-2. Figure 12 presents the images recorded during the ONB for the test 1. For this test, the temperature superheat reaches Tsc -Tsat = 73.5 K and the heat flux in the studied area is 3.5 W/cm². Nucleate boiling begins in the middle of the surface, where the temperature is maximum and spreads in a few milliseconds on the entire area of study. ONB can be qualified of explosive. After the ONB, the wall temperature decreases brutally (figure 11), because liquid evaporation and local convection movements due to the growth of bubbles lead to a huge intensification of the heat transfer between the wall and the fluid. The measured superheat after ONB decreases brutally to the range 30 K < Tsc – Tsat < 40 K while the heat flux dissipated in the studied area reaches between 9 W/cm² and 12 W/cm². This result shows that in boiling conditions, the heat dissipated by the fin is much lower than in natural convection. Therefore, the heat flux dissipated in the studied area is much closer to the heat flux dissipated by the heater (12 W/cm² < ϕm <15 W/cm²). After ONB, increasing the heat flux leads to a small increase of the wall superheat because the number of nucleation sites increases. The data obtained for decreasing heat flux are similar to those obtained for increasing heat fluxes until the heat flux corresponding to the ONB. Then,

the decreasing curve follows a different trend until the heat flux reaches zero, showing a huge hysteresis phenomenon. For very low heat fluxes, the curve overlaps the natural convection curve obtained for increasing heat fluxes. Figure 11 also shows the boiling phenomena at different locations a, b and c of the decreasing boiling curves. At low heat fluxes, a huge number of very small bubbles (diameter close to 0.7 mm) are observed, which cover very uniformly the totality of the surface. This phenomenon can be observed even at very small heat fluxes, because of this particular geometry of the sample, which prevents from the parasitic bubbles. To our knowledge, such a behavior of the boiling phenomena for very smooth surfaces was never observed before. For higher heat fluxes, one can observe larger bubbles whose origin is undoubtedly due to the coalescence of the huge number of bubbles appearing on the surface. The same boiling behavior and boiling curves were observed for surfaces SS1 and SS1+, which are the same sample with two different coating (Au4G or pure aluminum). Similarly, we observe no differences between samples SS2+ and SS1+, which are two different samples with the same surface characteristics. For the three samples, the ONB occurs always for high superheat and the ONB is in all the cases explosive: the time required to cover the studied area with bubbles is within the range 5 ms to 30 ms. Similarly, a huge hysteresis is observed for all the surfaces: the boiling phenomenon is able to survive at very low heat flux in decreasing heat flux with a huge number of very small bubbles covering the entire surface. One can conclude from these experiments that the observations made for surface SS2+ are repeatable and that the difference of nature of the two types of aluminum does not affect the boiling characteristics of the surfaces. 4.2. Boiling curve for rough surfaces For the sake of comparison, a rough surface with a Rq close to 5 µm, SR6 was also tested in this study. Figure 13 presents the boiling curves in increasing and decreasing heat fluxes in the same experimental conditions as the ultra-smooth surfaces. For the rough surface, the behavior is completely different since bubbles appear progressively at localized places on the surface for a relatively low superheat (Tsc – Tsat = 16.7 K). Figure 14 presents the images recorded during the first seconds of the ONB. Time origin corresponds to the observation of the first bubble. At time t = 0 s, a single bubble appears on the surface. This bubble grows and pulls away from the surface at time t = 85 ms, with a diameter close to 0.8 mm. Then, nucleation on this site goes on regularly. At time t = 5.4 s, a second nucleation site appears. Initially, one single bubble forms and leaves from this site with the same diameter of 0.8 mm. However, as soon as the bubble leaves the surface, the nucleation surface increases around this second nucleation site, leading to coalescence phenomena and thus to vapor bubbles of larger diameter. This phenomenon is due to an increase of the local heat flux toward this zone. Compared to the ultra-smooth surface, the ONB phenomena is both very slow and localized. The brutal cooling, observed at the ONB for the ultra-smooth surfaces is not observed for the rough surface and the boiling curve is rather similar in increasing and decreasing heat fluxes (figure 13). Particularly, compared to the ultra-smooth surface, the onset of nucleate boiling at decreasing heat flux and the offset of nucleate boiling at decreasing heat flux occur at about the same superheat and with similar bubble shape, size and location. The small difference between the two curves is due to a number of active nucleation sites slightly higher in decreasing heat fluxes. This is in line with Griffith and Wallis theory, which expresses that the wall superheat increases with the decrease of the nucleation site diameter. Nevertheless, it has to be noted that for the ultra-smooth surfaces, this theory overestimates the wall superheat compared to the wall superheat measured experimentally.

4.3. Boiling curves for smooth surfaces with imperfections Figure 15 presents the boiling curve for surface SSd3+ in increasing and decreasing heat fluxes in the same experimental conditions as the ultra-smooth surfaces and the rough surface. Surface SSd3+, being ultra-smooth with 10 imperfections, has the same ONB characteristics as SS1, SS1+ and SS2+, but behaves as sample SR6 for decreasing heat fluxes. Surprisingly, the wall superheat reached just before the onset of nucleate boiling (ONB) is high and is within the range 50 K < Tsc – Tsat < 70 K for the four different tests. The corresponding heat flux in the studied area is within the range 2.8 Wcm-2 to 3.6 Wcm-2. The presence of the defects does not modify the ONB characteristics, which is maybe due to the shape of the defects that could prevent to trap gases [18]. Figure 16 presents the images recorded during the first milliseconds of the ONB for two different tests. As for ultra-smooth surfaces, the onset of nucleate boiling is very fast and covers the entire surface in a few milliseconds. ONB occurs in the very center of the studied area. At time t=1ms, a small change of contrast can be observed just above the surface. One can ask if ONB occurs within the superheated liquid above the surface or at the solid-liquid interface. In other words, one can ask if this is a homogeneous or a heterogeneous nucleation process. The same behavior was also observed for the ONB of ultra-smooth surfaces without defects. As for the ultra-smooth surfaces, the wall temperature decreases brutally (figure 15) after the ONB. The measured superheat after ONB decreases brutally to a superheat Tsc – Tsat close to 20 K for the 4 different tests. After ONB, increasing or decreasing the heat flux leads to a very small variation of the superheat, which varies between 21 K and 17 K for heat fluxes decreasing from 14 Wcm-2 to 0.8 Wcm-2. As for the ultra-smooth surface, the decreasing curve follows a different trend from the increasing curve, showing a huge hysteresis phenomenon. Nonetheless, the decreasing curve meets the natural convection curve for a superheat much higher than for the ultra-smooth surfaces. The energy required to maintain nucleate boiling within the defects is not sufficient and the decreasing curve overlaps the natural convection curve for a superheat equal to 18 K. Figure 15 also shows the boiling phenomena at different locations a, b and c of the decreasing boiling curves. The behavior of SSd3+ is in between ultra-smooth surfaces and rough surfaces. At the ONB, mechanisms are similar to an ultra-smooth surface: superheat is huge and nucleation is explosive with formation of a big vapor core in a few milliseconds. For decreasing heat fluxes, bubbles look more like those of a rough surface with bigger diameters and lower density of bubbles. When the heat flux is low, a small number of nucleation sites remain, corresponding most certainly to the location of the defects. With ultra-smooth surfaces, a huge number of small bubbles were observed in the same conditions. For surface Sd3+, defects prevent the formation of these small bubbles, by concentrating the heat flux lines as it was analyzed by Bonjour et al [26]. For ultra-smooth surfaces, the energy is uniformly spread on the surfaces leading to the formation of many small bubbles requiring less energy to survive. For the two other smooth surfaces with defects, presented in table 1, Sd4 and Sd5, ONB is in between ultra-smooth surfaces and rough surfaces. For Sd 4, ONB is similar to SR6. Defects create nucleation sites where bubbles appear progressively for a relatively low superheat and it is necessary to increase the heat flux to observe bubbles on the entire surface. For Sd5, ONB is more explosive and it is possible to observe bubble expansion mechanisms similar to those observe by Stutz and Moreira [27].

4.4. How to increase the boiling efficiency? Figure 17 gathers the boiling curves of the three types of samples: S S2+, SSd3+ and SR6. At low heat fluxes, ultra-smooth surface SS2+ is more efficient than SSd3+ and SR6, because bubbles survive even with a low superheat. When the superheat reaches 18 K, surface SSd3+ becomes more efficient: increasing the heat flux beyond this threshold has a very small influence on the superheat, which increases from 18 K to 21 K when the heat flux increases from 3.5 W/cm² to 13.5 W/cm². On the contrary, the superheat of the ultra-smooth surface increases from 18 K to 40 K when the heat flux increases from 3.5 W/cm² to 12.5 W/cm². This can probably be explained by the huge density of small bubbles which coalesce when the heat flux increases. The same behavior is observed for the rough surface. When defects are present on an ultrasmooth surface, they act as point heat sink, concentrating the heat flux lines coming from the bottom of the surface. As these defects are separated by ultra-smooth area, there are no parasitic bubbles around these spots and therefore, the coalescence process is postponed, leading to a small increase of the superheat when the heat flux increases. As a conclusion of this study, it seems that increasing the boiling efficiency of a surface requires an ultra-smooth surface with some nucleation points wisely located on the surface. 5. CONCLUSIONS An experimental study of pool boiling of acetone on an aluminum sample shows a strong dependence of the boiling mechanisms on the topography of the surface. In the case of an ultra-smooth surface of Au4G, being coated or not with pure aluminum, huge superheats (80 K to 90 K) are required to start nucleate boiling. At the ONB, a core of vapour increases very quickly and covers the entire surface in a few milliseconds. The birth of this bubble is followed by a huge decrease of the wall temperature (a decrease of 50 K is measured). As soon as the vapor core leaves the surface, a huge number of small bubbles appear on the surface and survive even at very low heat flux. The boiling curve presents an important hysteresis due to the cooling of the surface at the ONB. As a multitude of small bubbles are still present at low heat flux this hysteresis is observed nearly on the entire boiling curve for decreasing heat fluxes. In the case of a rough surface, the above described phenomena are not observed. For increasing heat fluxes, bubbles appear progressively on the surface due to the presence of numerous nucleation sites. The superheat at the ONB is relatively low and the cooling effect is negligible when the first bubbles appear. The boiling curve in increasing or decreasing heat fluxes are very close one to the other. The bubble diameters are also bigger than on the ultrasmooth surface. The behavior of an ultra-smooth surface with a small number of defects is in between ultrasmooth surfaces and rough surfaces. The superheat required to start nucleate boiling is important and a huge vapour core spreads on the surface at the ONB with characteristics similar to the ultra-smooth surface. Nonetheless, the diameter of the bubbles observed on the surface at decreasing heat fluxes is similar to the diameter observed with rough surface. For decreasing heat fluxes, the number of bubbles decreases and nucleate boiling stops for a relatively high superheat. The ultra-smooth surface with some defects has excellent thermal characteristics at high heat flux, which is due to the presence of some nucleation site, well distributed on the surface and separated by ultra-smooth zones, preventing the formation of bubbles. This postpones the coalescence phenomenon which occurs at relatively low heat flux

for ultra-smooth surfaces. This result shows that the boiling efficiency of a surface can be improved by creating nucleation points wisely located on the ultra-smooth surface. ACKNOWLEDGEMENTS The support from the French National Research Agency (ANR) within the ‘‘NUCLEI’’ project (ANR-12-SEED-0003) is gratefully acknowledged. REFERENCES [1] R.L. Webb, D. Q. Kern, Lecture Award Paper: Odyssey of the Enhanced Boiling Surface, J. Heat Transfer 126 (2004) 1051-1059. [2] J. Mitrovic, How to create an efficient surface for nucleate boiling?, Int. J. of Thermal Sciences 45 (2006), 1–15. [3] L. Zou, B. G. Jones, Heating surface material’s effect on subcooled flow boiling heat transfer of R134a, Int. J. Heat Mass Transfer (2013) 168-174. [4] A. Luke, Pool boiling heat transfer from horizontal tubes with different surface roughness, Int. J. Refrigeration 20 (1997) 561-574. [5] R.J. Benjamin, A.R. Balakrishnan, Nucleation site density in pool boiling of saturated pure liquids: Effect of surface microroughness and surface and liquid physical properties, Exp. Therm. Fluid Sci. 15 (1997) 32-42. [6] B.J. Jones, S.V. Garimella, Effects of surface roughness on the pool boiling of water. In : ASME/JSME 2007 Thermal Engineering Heat Transfer Summer Conference collocated with the ASME 2007 InterPACK Conference. American Society of Mechanical Engineers, 2007. p. 219-227. [7] B.J. Jones, J.P. McHale, S.V. Garimella, The Influence of Surface Roughness on Nucleate Pool Boiling Heat Transfer, J. Heat Transfer 131 (2009) p. 121009. [8] J. Jabardo, G. Ribatski, E. Stelute, Roughness and surface material effects on nucleate boiling heat transfer from cylindrical surfaces to refrigerants R-134a and R-123, Exp. Therm. Fluid Sci. 33 (2009) 579-590. [9] H.T. Phan, N. Caney, P. Marty, S. Colasson, J. Gavillet, Surface wettability control by nanocoating: the effects on pool boiling heat transfer and nucleation mechanism, Int. J. Heat Mass Transfer 52 (2009) 5459–5471. [10] H. Jo, H.S. Ahn, S. Kang, M.H. Kim, A study of nucleate boiling heat transfer on hydrophilic, hydrophobic and heterogeneous wetting surfaces, Int. J. Heat Mass Transfer 54 (2011) 5643– 5652. [11] I.L. Pioro, W. Rohsenow, S.S. Doerffer, Nucleate pool-boiling heat transfer. I: review of parametric effects of boiling surface, Int. J. Heat Mass Transfer 47 (2004) 5033-5044. [12] M.C. Paz, M. Conde, E. Suárez, M. Concheiro, On the effect of surface roughness and material on the subcooled flowboiling of water: experimental study and global correlation, Exp. Therm. Fluid Sci. 64 (2015) 114-124. [13] W. Rohsenow, A method of correlating heat transfer data for surface boiling of liquids, Trans. ASME 74 (1952) 969-976. [14] I.L. Pioro, W. Rohsenow, S.S. Doerffer, Nucleate pool-boiling heat transfer. II: assessment of prediction methods, Int. J. Heat Mass Transfer 47 (2004) 5045-5057. [15] M.C. Paz, M. Conde, J. Porteiro, M. Concheiro, Effect of heating surface morphology on the size of bubbles during the subcooled flow boiling of water at low pressure, Int. J. Heat Mass Transfer 89 (2015) 770-782.

[16] M.C. Paz, M. Conde, J. Porteiro, M. Concheiro, Effect of heating surface morphology on active site density in subcooled flow nucleated boiling, Exp. Therm. Fluid Sci. 82 (2017) 147159. [17] S.G. Bankoff, Entrapment of gas in the spreading of a liquid over a rough surface. AIChE Journal 4 (1958) 24–26. [18] P. Griffith, J.D. Wallis, The role of surface conditions in nucleate boiling, Cambridge, Mass. : Massachusetts Institute of Technology, Division of Industrial Cooperation (1958). [19] Y.Qi, J. Klausner, heterogeneous nucleation with artificial cavities, J. Heat Transfer 127 (2005) 1189-1196 [20] S. Witharana, B. Phillips, S. Strobel, H.D Kim, T. McKrell, J.B. Chang, J. Buongiorno, K.K. Berggren, L. Chen, Y. Ding, Bubble nucleation on nano- to micro-size cavities and posts: an experimental validation of classical theory, J. Applied Physics 112 (2012) p.064904. [21] B. Bourdon, R. Rioboo, M. Marengo, E. Gosselin, J. De Coninck, Influence of the wettability on the boiling onset. Langmuir 28 (2012) 1618–1624. [22] B. Bon, C.K. Guan, J.F. Klausner, Heterogeneous nucleation on ultra-smooth surfaces, Exp. Therm. Fluid Sci. 35 (2011) 746–752. [23] T. Theofanous, J. Tu, A. Dinh, T. Dinh, The boiling crisis phenomenon: Part I: nucleation and nucleate boiling heat transfer. Exp. Therm. Fluid Sci. 26 (2002) 775–792. [24] S.Cioulachtjian, M. Lallemand, Nucleate pool boiling of binary zeotropic mixtures, Heat Transfer Engineering 25 (2004) 32–44. [25] M.Al Masri, S. Cioulachtjian, C. Veillas, F. Celle, I. Verrier, Y. Jourlin, F. Lefèvre, Nucleate boiling on ultra-smooth surfaces: study of Bubble incipience at submicronic scale. In 9th International Conference on Boiling and Condensation Heat Transfer, April 26-30, 2015 – Boulder, Colorado. [26] J. Bonjour, M. Clausse, M. Lallemand, Experimental study of the coalescence phenomenon during nucleate pool boiling. Exp. Therm. Fluid Sci. 20 (2000) 180–187. [27] B.Stutz, J.R.Simões-Moreira, Onset of boiling and propagating mechanisms in a highly superheated liquid - the role of evaporation waves. Int. J. Heat Mass Transfer 56 (2013) 683 – 693.

Heated zone Annular groove to reduce heat losses

Micro groove for temperature measurement Figure 1. a) Bottom view of the aluminum sample b) Face view

180.8

6th degree

Z (µm)

180.6

180.4 180.2 0

2

4

6

8 X (mm)

10

12

0.3

14

16

Reference line

Z (µm)

0.2 0.1 0 -0.1 -0.2 0

2

4

6

8 X (mm)

10

12

14

Figure 2. Example of roughness measurement a) Raw profile with the mean altitude obtained using a six order polynomial fitting curve b) Distribution of the holes and the peaks around the mean altitude

16

200 Z (nm)

S

1

0 -200 0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

5

200 Z (nm)

S

Sd3+

0 -200 0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

5 S

200

Sd4

0 -200 -400 0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

5

5 R Z (µm)

Z (nm)

400

6

0 -5 0

0.5

1

1.5

2

2.5 X (mm)

3

3.5

4

4.5

5

Figure 3. Distribution of holes and peaks around the mean altitude for different samples

N°

Rq (nm) 58

Ra (nm) 34

Rv (µm)

Rp (µm)

Number of defects

Characteristic length of defects (µm)

Ss1 0.27 2 Ss1+ 0 60 0.7 2.8 Ss2+ 40 35.5 2.0 16.9 SSd3+ 95 10 5 to 25 SSd4 145 63.8 3.6 15.6 21 5 to 25 SSd5 196 91.8 2.8 9.3 23 5 to 25 SR6 5166 882.5 8.9 14.2 360 5 to 40 Table 1. Surface characteristics of the samples; Ra: Arithmetic mean roughness height; Rq: Root mean square roughness; Rv: Max valley depth; Rv: Max peak height

6

10

S

S1

SSd3+

5

10

SSd4 SSd5 SR6

4

Number

10

3

10

2

10

1

10

0

10 -5000

-4000

-3000

-2000

-1000

0 Z (nm)

1000

2000

3000

Figure 4. Distribution of corrected altitude for some samples

4000

5000

Reservoir Thermostatic bath for temperature control

Tubes

Transparent reservoir

O-ring

Engineered surface Thermocouple

Two-phase working fluid Tight clamp between the reservoir and the chassis

Aluminums upport

Heater Fluxmeter Insulation piston

Spring Nylon case

Figure 5. (a) Pool boiling setup (b) Schematic diagram of pool boiling setup

20

S_S2+

Max. T uncertainty

S_Sd3+

16

ϕ m (W/cm²)

S_R6 12 Nucleate boiling 8

4 Natural convection 0 0

10

20 30 Tm - Tsat (K)

40

50

Figure 6. Raw data plotted using temperature and heat flux measurements

Figure 7. Boundary conditions and finite difference nodes for the thermal model

120 Hr = 150 µm 100

Hr = 200 µm Hr = 250 µm

Tij - Tsat (K)

80

Tnj - Tsat : bottom

60

T1j - Tsat : top 40 20 Area of study

Groove

Peripheral surface

0 0

5

10

15

20 25 30 35 40 Radius (mm) Figure 8. Superheat at the bottom and at the top of the sample in natural convection for three different thicknesses (Hr = 150 µm, Hr = 200 µm and Hr = 250 µm) of the fin separating the area of study and the peripheral surface; ϕm = 15 W/cm²; hnc = 500 W/m²K

(a)

(b)

16

5 S_S2+ S_Sd3+ S_R6

4 ϕ sc (W/cm²)

ϕ m (W/cm²)

12

8

4

S_S2+ S_Sd3+ S_R6

3 2 1 Max. T uncertainty

Max. T uncertainty

0

0

0 25 50 75 100 50 75 100 T T (K) sc sat Tm - Tsat (K) Figure 9. Natural convection curves; a) raw data plotted using temperature sensor measurements; b) processed data with correction of the heat losses 0

25

120 Tnj - Tsat: bottom

100

T1j - Tsat: top

Tij - Tsat (K)

80 60

Hr = 150 µm Hr = 200 µm

40

Hr = 250 µm 20

Peripheral surface Area of study

Groove

0 0

5

10

15

20 25 30 35 40 Radius (mm) Figure 10. Superheat at the bottom and at the top of the sample in boiling conditions for three different thicknesses (Hr = 150 µm, Hr = 200 µm and Hr = 250 µm) of the fin; ϕ m = 15 W/cm²; hnc = 500 W/m²K; hb = 2500 W/m²K

20 Test1 Test2 Test3 Test4

ϕ sc (W/cm²)

16

increasing heat flux decreasing heat flux hysteresis surface max. T uncertainty

12

a

8

4 b c

0 0

20

40 60 80 100 Tsc - Tsat (K) Figure 11. Typical boiling curve and bubbles for smooth surfaces (surface SS2+)

2 ms

5 ms

10 ms

15 ms

20 ms

30 ms

40 ms

50 ms

60 ms

Figure 12. ONB for S2+ (Test 1; Tsc – Tsat = 73.5 K; ϕsc = 3.5 W/cm²)

20 increasing heat flux

16

a

decreasing heat flux

ϕ sc (W/cm²)

max. T uncertainty

12

a

b

8 c

4 b

0

c

0

20

40 60 Tsc - Tsat (K)

80

100

Figure 13. Typical boiling curve and bubbles for rough surfaces (surface SR6)

0 ms

1 ms

20 ms

85 ms

5.4 s

6s

T=6s

6.5 s

7.5 s

8.8 s

Figure 14. ONB for SR6 (Tsc – Tsat = 16.7 K; ϕsc = 0.8 W/cm²)

20 increasing heat flux decreasing heat flux

ϕ sc (W/cm²)

16

hysteresis surface

12 a

max. T uncertainty

8

ONB

Test1 Test2

4

Test3

b

Test4

c

0 0

20

40 60 Tsc - Tsat (K)

80

100

Figure 15. Typical boiling curve and bubbles for smooth surface with defects (surface SSd3+)

(a) Test2

1 ms

2 ms

3 ms

4 ms

5 ms

6 ms

(b) Test3

1 ms

2 ms

3 ms

5 ms

6 ms

7 ms

Figure 16. ONB for SSd3+; (a) Test 2: Tsc – Tsat = 70.6 K; ϕsc = 3.3 W/cm²; (b) Test 3: Tsc – Tsat = 61.3 K; ϕsc = 3.6 W/cm²

20 S_S2+ 16

Max. T uncertainty

S_Sd3+

ϕ sc (W/cm²)

S_R6 12

8

4

0 0

10

20 30 Tsc - Tsat (K)

40

50

Figure 17. Comparison of the boiling curves for a smooth surface SS2+, a smooth surface with defect SSd3+, and a rough surface SR6

Highlights Pool boiling heat transfer characteristics in saturation conditions are investigated Influence of surface topography from rough to ultra-smooth is investigated The topography of the surfaces is analyzed with a confocal microscope The results show that ultra-smooth surfaces with a small number of imperfections deliver the best thermal characteristics.