Quantitative Evaluation of the

Dependence of Pool Boiling Heat

Transfer Enhancement on
Suchismita Sarangi
School of Mechanical Engineering;
Birck Nanotechnology Center,
Sintered Particle Coating
Purdue University,
585 Purdue Mall,
Characteristics
West Lafayette, IN 47907
Immersion cooling strategies often employ surface enhancements to improve the pool
Justin A. Weibel boiling heat transfer performance. Sintered particle/powder coatings have been com-
School of Mechanical Engineering; monly used on smooth surfaces to reduce the wall superheat and increase the critical
Birck Nanotechnology Center, heat flux (CHF). However, there is no unified understanding of the role of coating char-
Purdue University, acteristics on pool boiling heat transfer enhancement. The morphology and size of the
585 Purdue Mall, particles affect the pore geometry, permeability, thermal conductivity, and other charac-
West Lafayette, IN 47907 teristics of the sintered coating. In turn, these characteristics impact the heat transfer
coefficient and CHF during boiling. In this study, pool boiling of FC-72 is experimentally
Suresh V. Garimella1 investigated using copper surfaces coated with a layer of sintered copper particles of
irregular and spherical morphologies for a range of porosities (40–80%). Particles of
School of Mechanical Engineering;
the same effective diameter (90–106 lm) are sintered to yield identical coating thick-
Birck Nanotechnology Center,
nesses (4 particle diameters). The porous structure formed by sintering is characterized
Purdue University,
using microcomputed tomography (l-CT) scanning to study the geometric and effective
585 Purdue Mall,
thermophysical properties of the coatings. The boiling performance of the porous coat-
West Lafayette, IN 47907
ings is analyzed. Coating characteristics that influence the boiling heat transfer coeffi-
cient and CHF are identified and their relative strength of dependence analyzed using
regression analysis. Irregular particles yield higher heat transfer coefficients compared
to spherical particles at similar porosity. The coating porosity, pore diameter, unit neck-
ing area, unit interfacial area, effective thermal conductivity, and effective permeability
are observed to be the most critical coating properties affecting the boiling heat transfer
coefficient and CHF. [DOI: 10.1115/1.4034901]

1 Introduction parameter at a time while keeping all others fixed. Nishikawa

et al. [7] studied the enhancement of refrigerant boiling from dif-
The semiconductor industry has witnessed a continuous minia-
ferent sizes and thicknesses of copper and bronze sintered pow-
turization of modern electronic circuits with increasing power
ders on copper tubes. The maximum heat transfer coefficient was
consumption per unit footprint area. To keep up with these power
obtained when the coating thickness was about four times the
density trends, advances in electronics thermal management are
mean particle diameter (d/d ¼ 4), with copper powder being more
essential. Immersion cooling by pool boiling from heated surfaces
effective than bronze. A model formulated by Ranjan et al. [8],
offers one of the more efficient methods for high-heat-flux dissi-
which considered the porous structure as being analogous to
pation, due to its passive nature and the high heat transfer coeffi-
extended fins, showed that the optimum wick thickness occurred
cients obtained. Many variants of surface modifications have been
when the effective area enhancement was maximized. Sarangi
developed and studied to further enhance pool boiling
et al. [9] studied the effect of particle size variation on copper
performance.
surfaces coated with sintered spherical copper particles, for parti-
A common modification for enhancing pool boiling heat trans-
cle sizes varying from 45 to 1000 lm. The sintered coatings were
fer is to sinter thermally conductive particles onto a surface to
manufactured with a constant thickness-to-diameter ratio (d/d ¼ 4)
form a porous structure, as investigated early on by Milton [1]. A
and porosity of 30–40%. It was observed that a particle size range
number of studies have since explored various processing meth-
of 90–106 lm provided the highest heat transfer coefficient, while
ods and geometric modifications for this structure, such as modu-
45–53 lm-diameter particles provided the maximum CHF for
lated porous copper coatings with periodic variations in layer
pool boiling of FC-72. Furberg and Palm [10] studied the boiling
thickness [2], a silver flake-based boiling enhancement paint [3],
heat transfer characteristics from a dendritic microporous copper
spray-coated alumina particulate layers [4], and porous graphite
surface in R134a and FC-72. The bubble departure size and fre-
layers with randomly interconnected microscale pores [5,6], and
quency were visualized, and the heat transfer coefficient (along
observed significant enhancement of the heat transfer coefficient
with delineation of the latent and sensible heat contributions) was
and critical heat flux (CHF).
measured over a range of heat fluxes. The latent heat contribution,
While different geometric surface modifications have been
estimated from visualization of the bubble departure frequency
postulated for boiling enhancement, few have expressly analyzed
and diameter, was enhanced by up to ten times for the micropo-
the parametric effects of individual sintered coating characteristics
rous surface as compared to a smooth surface. The dendritic
(e.g., particle size, porosity, and thickness), by varying one
porous structure was also said to increase agitation of the vapor
1
bubbles inside the porous structure, thus enhancing single-phase
Corresponding author.
Contributed by the Heat Transfer Division of ASME for publication in the JOURNAL OF
convection heat transfer by up to six times. O’Hanley et al. [11]
HEAT TRANSFER. Manuscript received April 12, 2016; final manuscript received September performed experiments to study the effect of porosity on CHF dur-
16, 2016; published online November 8, 2016. Assoc. Editor: Debjyoti Banerjee. ing boiling of water from surfaces with different geometric

Journal of Heat Transfer Copyright V

C 2017 by ASME FEBRUARY 2017, Vol. 139 / 021502-1

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jhtrao/935799/ on 04/24/2017 Terms of Use: http://www.asme.org/about

 properties (e.g., surface wettability and roughness). Silica nano- applications is provided in Ref. [20]. The accuracy of such model-
particles were deposited on sapphire substrates to achieve the ing is limited for porous media with a high degree of randomness
desired coating properties. Porous surfaces showed up to a 60% and irregularities, such as sintered particle coatings. An alternative
higher CHF than nonporous surfaces for hydrophilic surfaces. It modeling approach involves direct simulation of transport within
was concluded that for surfaces having similar wettability charac- the actual porous microstructure geometry obtained using imaging
teristics, porosity plays a major role in determining the CHF techniques such as microcomputed tomography (l-CT) [21,22].
behavior during boiling. Li and Peterson [12] performed experi- Tabor et al. [23] demonstrated a voxel-based three-dimensional
ments to study the effects of coating thickness, porosity, and mesh (3D) reconstruction method for generating finite-volume meshes
size on the CHF during pool boiling of water from surfaces coated from l-CT data to represent the complex structure of porous
with sintered copper screen mesh. For coatings with the same media. Bodla et al. used l-CT scanning to simulate the thermal
thickness and mesh sizes, there was an optimal porosity at which conductivity, permeability, and convective heat transport charac-
the CHF reached a maximum value due to a trade-off between teristics of open-celled aluminum foams [24] and sintered copper
capillary pressure within the pores and hydraulic flow resistance. particle coatings [25]. A strong dependence of the coating charac-
Since the morphology of the individual particles affects the teristics on the sintering conditions was demonstrated.
postsintering pore shape, permeability, surface roughness, and The present study aims to investigate the effects of geometric
effective conductivity and diffusivity of the coating, it can be properties (viz., particle morphology, porosity, pore size, postsin-
expected to significantly influence the heat transfer coefficient and tering particle size, particle sphericity, and interfacial and necking
CHF during boiling. Prior studies on pool boiling have not explic- areas per unit volume) and effective thermophysical properties
itly investigated the effect of particle morphology; however, sev- including permeability and thermal conductivity on the pool boil-
eral studies have explored the effects of particle morphology on ing performance of sintered particle coatings over a wide range of
effective geometric and thermophysical properties for other heat porosities. The surfaces are designed and fabricated to keep pre-
transfer applications. Chi et al. [13] studied the effects of shape sintering particle size, particle material, and coating thickness
variations on the thermal properties of coatings. Various particle constant while varying particle morphology and coating porosity
shapes such as angular/polyhedral, globular/rough-textured, and to evaluate these effects. A detailed l-CT characterization of the
hollow spheres were studied for powder-based yttria-stabilized coating is performed to obtain the geometric properties; numerical
zirconia (YSZ), molybdenum, and silicon thermal-spray coatings. simulations are performed on the reconstructed porous domains to
Different initial particle shapes and sintering conditions were used obtain the effective thermophysical properties. A regression anal-
for fabrication of the coating, resulting in different pore structure ysis is used to quantitatively evaluate sensitivities of the pool boil-
and porosities. The resulting thermal conductivity and diffusivity ing heat transfer coefficient and CHF to these coating
were shown to be more dependent on pore architecture than on characteristics.
the porosity of the coatings. The thermal conductivity of globular/
irregular particle coating was found to be almost twice that of the
hollow spherical powder coating. Dixon [14] studied the effect of 2 Experimental Methods
particle morphology on the convective heat transfer at varying
flow rates through packed beds with spherical, cylindrical, and 2.1 Pool Boiling Test Facility. Figure 1 shows a schematic
hollow cylindrical particles. The different particle shapes had a diagram of the pool boiling test facility. An abbreviated descrip-
significant effect on the effective conductivity and wall heat trans- tion is provided here for this modified version of the facility used
fer coefficient of the packed beds. Cylindrical particle beds by Sarangi et al. [9]. The base heater assembly consists of a cop-
showed higher thermal conductivity and heat transfer coefficient per heater block embedded in an insulation casing. The copper
compared to spherical particle beds, while minimal difference was block has a 25.4 mm  25.4 mm top surface coated with sintered
observed between the full and hollow cylinders. The effect of par- particles. The heater width is sufficiently large such that it should
ticle shape and size on flow boiling performance of porous copper not influence CHF [26]. The heat input is provided by 12 cartridge
microchannel heat sinks constructed from spherical and irregular- heaters (100 W each) inserted into holes in the bottom surface of
shaped particles (50–150 lm effective diameter) was investigated the block. Block temperatures are measured using T-type thermo-
by Deng et al. [15]. Irregular-shaped particles of the smallest size couples inserted into tapped holes along three rakes, as shown in
(50–75 lm) showed the highest heat transfer coefficient, while Fig. 1. The copper block is inserted into a block of PEEK, a mach-
spherical particles of the smallest particle size showed the lowest inable, low thermal conductivity (0.28 W/m K) thermoplastic that
heat transfer coefficient. Lin and Hwang [16] conducted experi-
ments to study the effect of particle morphology and sintering
conditions on the permeability of sintered powders for heat pipe
applications. Three different particle shapes (irregular, dendritic,
and spherical) were evaluated with varying compacting pressure
and temperature during sintering. It was observed that gas-
atomized spherical particles had the lowest porosity but the high-
est permeability. Permeability may be estimated by the
Kozeny–Carman equation [17]
d 3 e3
K¼ 2
(1)
Cð1  eÞ

where C is a geometric factor depending on the particle morphol-

ogy and manufacturing process. While Eq. (1) dictates that a
higher coating porosity would result in a higher permeability, the
morphology-dependent factor C can significantly alter the
permeability.
To determine the properties of porous media, several studies [18,19]
have used unit-cell models or simplified analytical/semi-empirical
models to evaluate the pore-scale transport characteristics, which Fig. 1 Schematic diagram of test facility with inset showing a
can then be used to solve macroscale governing equations using photograph of a copper block coated with spherical sintered
effective properties. A review of these methods and their particles

021502-2 / Vol. 139, FEBRUARY 2017 Transactions of the ASME

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jhtrao/935799/ on 04/24/2017 Terms of Use: http://www.asme.org/about

 Fig. 2 SEM images of (a) spherical and (b) irregular coatings at a low, intermediate, and high porosity, showing the pore
structure of the coatings

can withstand temperatures up to 250  C. A 0.8 mm-thick polycar- particles). The sintered-coating fabrication method is adjusted to
bonate plate is attached to the top of the PEEK block; the copper achieve a range of porosities for each particle shape. This requires
block is aligned such that it protrudes just above the polycarbon- either reducing or increasing the native packing density of the
ate plate as shown. The vertical location of the thermocouple rake powder. The porosity of the coating can be reduced by
and the thickness of polycarbonate plate have been slightly altered controlling the sintering time and temperature to achieve the
from the original design [9]. The lower part of the copper block is desired porosity [9]. In order to increase the porosity beyond the
insulated as described in Ref. [9]. native packing density, the most commonly used method is lost-
The pool boiling chamber comprises four polycarbonate walls carbonate sintering (LCS) [29]. A mixture of spherical metal par-
sealed together with RTV silicone that is cured under compression ticles, carbonate, and a binder is sintered at high temperature
with screws. The chamber is sealed to the base heater assembly under pressure until the carbonate decomposes or melts, leaving
with a compressed PTFE gasket. An aluminum cover plate seals behind a porous metal coating. The carbonate particle size and
the top of the chamber walls by compressing another PTFE gas- percentage, as well as the sintering temperature and time, can be
ket, forming a 30.48 mm  30.48 mm  93.9 mm sealed chamber controlled to produce porous coatings with the desired porosity
enclosing the working fluid on the heated surface. A T-type ther- and pore diameter.
mocouple routed through the cover plate is used to measure the The top surface of the copper heater block is prepared for sin-
pool temperature during testing. A chilled 50–50% water–glycol tering by sanding with progressively finer diamond sanding sheets
mixture is cycled through the condenser at a flow rate of 0.5 l/min (9 lm, 3 lm, and 0.1 lm) followed by metal polishing (Simi-
at 15  C using a chiller (ULT-80, Neslab) to condense vapor dur- chrome, Happich, Wuppertal, Germany). The irregular particles
ing testing. A 6.35 mm-diameter polyurethane tube connects the (ACuPowder International) and the spherical particles (Alfa
liquid pool to a vapor trap that is open to the ambient during test- Aesar) are first sieved into the desired size range of 90–106 lm.
ing and maintains the pool at atmospheric pressure. The chiller The range is selected based on the particle size that yielded the
flow rate and set-point temperature were selected to ensure that no best pool boiling heat transfer performance for spherical sintered
working fluid collects in the vapor trap during the experiment (to coatings at 40% porosity in a prior study [9]. The particles are
prevent fluid loss from the chamber over time). cleaned by soaking in acetone and rinsing with methanol. To sin-
ter the porous layers with a reduced porosity, loose copper powder
2.2 Fabrication of Test Surfaces. Copper particles having is first poured into a graphite mold and leveled off to the desired
two different morphologies—irregular and spherical—were used height. The initial layer thickness of particles is varied, as
to fabricate the test surfaces. The differences in morphology are required, to obtain a final layer thickness-to-particle diameter
brought about by the particle fabrication method. As described in ratio, d/d, of 4, which has been shown to have the maximum
Refs. [27,28], irregular particles are produced by water atomiza- heat transfer coefficient in previous studies [7,30]. The copper
tion and spherical particles by air atomization. Irregular particles block is then gently placed face-down on top of the powder and
appear as roughened spheroidal shapes, while spherical particles the assembly is placed in an evacuated furnace. The temperature
are smooth. Figure 2 shows representative images of the particle is steadily ramped up to 950  C and held at this temperature under
shapes observed using a scanning electron microscope (SEM). vacuum for the desired time as shown in Table 1. The assembly is
The SEM images are taken after testing. Due to the different mor- then allowed to gradually cool to room temperature under vacuum
phology of the particles, the native packing porosity before sinter- and inspected under an optical microscope at multiple locations to
ing is different (48% for spherical particles and 60% for irregular ensure a uniform coating thickness.

Journal of Heat Transfer FEBRUARY 2017, Vol. 139 / 021502-3

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jhtrao/935799/ on 04/24/2017 Terms of Use: http://www.asme.org/about

 Table 1 Sintering conditions and resulting porosity of the reached when the change in calculated surface temperature versus
spherical and irregular sintered particle coatings fabricated time at each heat flux setting is less than 0.02  C/s over 30 min.
Once this steady-state criterion is satisfied, the temperature data
Particle shape Carbonate weight fraction (%) d/d e (%) collected over 120 s are time-averaged for subsequent postpro-
cessing of the surface heat flux and temperature. The procedure is
Spherical 0 3.9 39
0 3.9 45
continued until CHF is reached, which is marked by a sudden
10 4.4 55 rapid rise in temperature, following which, the power to the heat-
45 4.1 68 ers is shut off and the system is allowed to cool down. The porous
60 3.7 72 coating was confirmed to maintain a highly wetting contact to the
80 4.2 79 fluid after fabrication and testing; multiple boiling curves were
Irregular 0 3.8 51 obtained to confirm the negligible effect of surface degradation
10 4.2 58 between consecutive tests. Tests were also conducted with the
45 3.5 66 heat flux being decreased in steps from the maximum value to
60 3.5 71 evaluate if significant hysteresis is observed in the system. As
80 3.9 74 described in Sarangi et al. [9], area-averaged wall temperature and
heat flux are calculated from the temperature gradients along the
three thermocouple rakes. The extrapolated wall temperature is
For the cases where the porosity needs to be increased, an LCS first calculated from each of these rakes by linear regression. The
method is used. The copper particles are mixed with spherical area-averaged wall temperature is then calculated using a weight-
potassium carbonate particles (90–150 lm), with a weight fraction ing of each local temperature based on the surface areas corre-
as shown in Table 1 to achieve the desired porosity. A small sponding to each rake. Note that this heat flux is based on the
amount of ethanol is added to act as a binder in the mold, and the sample footprint area, not the surface area of the porous coating.
particles are mechanically stirred to ensure uniform mixing. This Following the uncertainty analysis procedure described in Ref.
mixture is then poured into the mold to the desired height and [9], for a calibrated absolute thermocouple temperature measure-
placed in the evacuated furnace. The temperature of the furnace is ment uncertainty of 60.23  C, the area-averaged heat flux uncer-
steadily ramped up to 850  C and then held for 2 h to initiate the tainty is 64–5.5 kW/m2 over the range of heat fluxes investigated
sintering process. The temperature is then ramped up further to (4–250 kW/m2).
950  C for 30 min to decompose the carbonate particles (melting
point of 891  C), leaving behind open pores. The assembly is then
allowed to cool gradually. This sintering recipe was selected 3 Coating Characterization
because it has been shown to produce a clean and well-bonded 3.1 Sample Preparation and Imaging. For sintered particle
porous structure for the desired porosity range, and does not coatings such as those used in the present work, the high degree of
require any postprocessing to remove the potassium carbonate randomness and irregularity makes it difficult to use simplified
from the sintered matrix [29]. analytical or representative unit-cell models to accurately predict
Table 1 shows the coating porosity and thickness after the sin- the coating properties. In the present work, microcomputed
tering procedure for all the samples fabricated. The weight of the tomography (l-CT) is used to resolve the exact 3D structure of
copper block is measured before and after the sintering process. the porous coatings for subsequent characterization.
The height of the coating is measured from images observed After the pool boiling test is completed, samples are carefully
under SEM. From these measures, the porosity of the coating can excised from the copper block surface over a 5 mm  5 mm foot-
be obtained by comparing the density to that of solid copper. print area, preserving the original thickness of the coating. The
Based on propagation of the initial measurement uncertainties, the samples are lightly pressed between two smooth surfaces to flatten
uncertainty in measured porosity values is calculated to be 2–4% out any warping of the samples. The prepared samples are then
for the different cases. Figure 2 shows representative images of imaged (Skyscan 1172) at a resolution of 4 lm using a 100 kV X-
the pore structures observed using a scanning electron microscope ray source voltage. The scans were performed in the Department
(SEM). of Anatomy and Cell Biology, Indiana University School of Medi-
cine, Indianapolis, IN. The scanner software generates a two-
dimensional (2D) stack of images corresponding to the 3D object.
2.3 Experimental Procedure. To conduct an experiment, the
sintered copper block is first assembled into the test facility. The
0.8 mm-thick polycarbonate plate is mounted on the PEEK block 3.2 Image Processing. The stack of 2D images generated
to surround the test surface. A chamfer in the PEEK block creates from the l-CT scans is imported into the commercial software
a groove around the outer edges of the test block, which assists in package SIMPLEWARE [32] for image processing (ScanIP module)
sealing at this location. A layer of RTV silicone sealant is first and mesh generation (þFE module). The details of the image
filled in the groove and cured in an oven for 90 min at 75  C, fol- processing steps are briefly described here and a more detailed
lowed by the application of a topping layer of epoxy (E 5302 description may be found in Ref. [25]. Example images at each
High Temp Epoxylite). The silicone prevents the wicking of processing step are shown in Fig. 3 for a reduced portion of the
epoxy into the gap between the test piece and insulation, thereby domain. The image processing steps in Fig. 3 are shown for a sin-
facilitating easier removal of test pieces between tests; the high- gle 2D slice for clarity, but all the steps are performed on a 3D
viscosity paste epoxy, on the other hand, seals out the working volume. Each 2D image of unit-pixel thickness from the stack is
fluid, without itself seeping into the pores of the sintered surfaces. “stitched” with adjacent image slices to generate a 3D volume as
The test chamber walls are then secured onto the base and the shown in Fig. 3(b). Copper has much higher X-ray absorptivity
assembly is cured in the oven for 20 h at 80  C. than air that fills the surrounding pore region during scanning;
The assembled test chamber is mounted onto a support stand hence, the brighter regions correspond to copper and darker
and filled with 40 ml of FC-72 (Tsat ¼ 56  C, q ¼ 1594 kg/m3, regions correspond to the pore space in the image. The transition
l ¼ 4.3  104 kg/m s, k ¼ 0.054 W/m K, hfg ¼ 88 kJ/kg, between these regions is not discrete in the images. In order to
cp ¼ 1101 J/kg K, and c ¼ 0.010 N/m [31]). The cover plate is then accurately differentiate the two regions, a grayscale threshold
secured. The fluid is first degassed by boiling aggressively for 2 h. value is selected such that the porosity of the reconstructed 3D
After degassing, the heat flux is increased in predetermined steps volume matches the measured porosity of the sample. Figure 3(c)
using the cartridge heaters embedded in the copper block to obtain shows an image slice after thresholding. The domains are then
a boiling curve. Steady-state conditions are assumed to have been inspected for unphysical unconnected regions or islands, which

021502-4 / Vol. 139, FEBRUARY 2017 Transactions of the ASME

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jhtrao/935799/ on 04/24/2017 Terms of Use: http://www.asme.org/about

 sample) indicates the extent to which the constituent particles are
spherical (after sintering). The total interfacial area (Sint,tot) is
obtained from the total surface area of the unsegmented copper
domain. The segmentation process identifies the necking areas
where particles are adjoined, shown as red lines in the 2D images
in Fig. 3(e). The total necking area (Sneck,tot) is obtained by sub-
tracting the total surface area of unsegmented copper domain
from that of the segmented copper domain (equivalent to the area
of the segmented necks shown in red in Fig. 3(e)). The unit inter-
facial area (Sint) and unit necking area (Sneck) are obtained by nor-
malizing the total interfacial area and total necking area by the
total volume of the sample, respectively.

3.4 Mesh Generation. The SIMPLEWARE þFE module is used

for generating meshes of discretized volumes. The number of par-
ticles/pores in the scanned 3D volume is computationally intracta-
ble for mesh generation and subsequent numerical simulation
Fig. 3 Image processing steps showing (a) raw 2D slice from
l-CT scan, (b) 3D reconstructed volume, (c) 2D slice showing
processes. Hence, a reduced domain size is carefully chosen such
pore domain (lighter region) and copper regions after thresh- that the geometric properties are maintained and the volumetric
olding, (d) segmented pore domain, (e) segmented copper porosity matches that of the scanned sample; the length and width
domain, and (f) 3D reconstructed volume segmentation in both of each sample are cropped to seven pore lengths. A mesh size of
pore and copper domains. An overall domain size of 0.1  106 cells per pore is selected based on a mesh-independence
800 lm 3 800 lm 3 207 lm for the spherical particle coating at study performed on a reduced domain size, which shows a 1.6%
39% porosity is shown. deviation in effective permeability for 0.1  106 cells per pore
compared to a refined mesh size of 0.3  106 cells per pore. The
are removed (typically less than four particles or pores over the resulting domain sizes and number of mesh elements for each
entire volume scanned). Figure 4 shows the 3D solid volumes sample are shown in Table 2. The processed 3D volume used for
generated for several of the spherical and irregular particle coat- generating meshes consists of 3D voxels, which would generate a
ings over the range of porosities investigated. stair-stepped mesh if used directly. Hence, the data are smoothed
(as recommended by Ref. [32]) to generate high-quality confor-
3.3 Evaluation of Geometric Coating Properties. The mal meshes with tetrahedral elements, ensuring the volume skew
microscale particles fuse together during the process of sintering, limit to be lower than 0.01. Figure 5 shows the mesh generated for
forming one connected 3D volume. It is necessary to identify indi- the spherical particle coating at 39% porosity.
vidual particles and pores to characterize the geometric properties
of the coating. This is achieved by using the watershed algorithm 3.5 Numerical Modeling. In order to obtain the effective
available in the SIMPLEWARE ScanIP module [32]. After segmenta- thermophysical properties of the sintered coatings, viz., perme-
tion of the individual particles and pores, as shown in Figs ability and thermal conductivity, the physics modules available in
3(d)–3(f), the statistical distribution of particle diameter (dpart) and SIMPLEWARE are employed. The primary direction of heat transfer
pore diameter (dpore) is obtained for each coating, based on the through the sintered coating is in the z-direction. Similarly, during
equivalent spherical volume of each segmented particle/pore. The boiling, the departure of vapor bubbles and subsequent replenish-
sphericity (/) is defined as ment of liquid also occurs primarily in this direction. Hence, the
effective permeability and conductivity are calculated along the
2=3
p1=3 ð6V Þ z-direction.
/¼ (2) The SIMPLEWARE þFLOW module is used to calculate the effec-
A
tive permeability of the pore domain. The governing mass and
where V is the volume of the particle and A is the surface area of a momentum conservation equations are solved using a minimum
particle. The sphericity (and statistical distribution throughout the residual-based homogenization solver and linear pressure

Fig. 4 Three-dimensional reconstructed solid domain (after thresholding) for (a) spherical
and (b) irregular coatings at a low, intermediate, and high porosity

Journal of Heat Transfer FEBRUARY 2017, Vol. 139 / 021502-5

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jhtrao/935799/ on 04/24/2017 Terms of Use: http://www.asme.org/about

 Table 2 Numerical mesh and domain size

Particle shape e (%) Domain size x  y  z (lm  lm  lm) Number of cells (106)

Spherical 39 340  340  207 15.7

45 480  480  163 10.3
55 520  520  219 15.4
68 700  700  219 14.2
72 720  720  219 14.5
79 760  760  203 14.0
Irregular 51 340  340  195 20.2
58 420  420  199 18.6
66 540  540  199 16.4
71 620  620  203 15.5
74 680  680  219 16.0

Kz
U¼ rp (3)
l

where rp is the applied pressure gradient and Kz is the effective

permeability of the medium in the z-direction.
The SIMPLEWARE þLAPLACE module is used to calculate the
effective thermal conductivity of the liquid-saturated porous
medium. The governing Laplace equations for diffusion are
solved using the conjugate gradient method for the pore and cop-
per domains individually. For the pore domain, fluid properties of
FC-72 at room temperature are used whereas for the solid domain,
properties of copper at room temperature are used. A mixed
boundary condition is employed that assumes symmetry boundary
conditions along the lateral faces and a fixed-temperature bound-
ary at the top and bottom faces such that the temperature differ-
ence per unit length between the top and bottom faces is one unit.
Continuity of temperature and heat flux is imposed at the
pore–solid interface. The boundary conditions are shown in
Fig. 6(b). The effective thermal conductivity is computed from

Fig. 5 Meshed domain (340 lm 3 340 lm 3 207 lm) for the @T

q00 ¼ kz (4)
spherical particle coating at 39% porosity is shown @Z

variation boundary condition. This boundary condition fixes the where q00 is the heat flux on the bottom wall, @T=@z is the temper-
pressure on the inlet and outlet faces of the domain such that the ature gradient along the z-direction, and kz is the effective thermal
pressure drop per unit length between the inlet and outlet faces is conductivity of the medium along the z-direction.
one unit. Linear pressure variation is imposed across the lateral The physics modules in the software use a finite element-based
faces, which mimics the more conventionally used symmetry homogenization solver. The pressure and velocity components are
boundary condition [32]. These boundary conditions are shown in discretized with continuous linear nodal elements. Further details
Fig. 6(a). The effective permeability in the flow direction (z-direc- about the solution methods and boundary conditions may be found
tion) is obtained using Darcy’s law in Ref. [32].

Fig. 6 Boundary conditions for (a) conduction through copper and pore domains and (b)
flow through the pore domain, used to obtain the effective conductivity and effective perme-
ability, respectively

021502-6 / Vol. 139, FEBRUARY 2017 Transactions of the ASME

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jhtrao/935799/ on 04/24/2017 Terms of Use: http://www.asme.org/about

 3.6 Quantitative Regression Analysis. The statistical rela- consistently increases. The heat transfer coefficient, as shown in
tionship between the coating characteristics and the boiling per- Fig. 8, is a function of the wall superheat at any given heat flux as
formance is analyzed in order to identify critical sensitivities. This is typically observed for pool boiling studies, although the exact
is achieved by performing a stepwise regression analysis with variation depends on the individual coating characteristics.
multiple inputs in MATLAB [33], tailored to represent the Comparing the boiling curves for the different coating geometries,
input–output relationship in the present study. The data obtained it is observed that boiling incipience occurs for all of the different
from the boiling experiments and the coating characterization are coatings at a low superheat (0.2–1 K) and a heat flux of 5–9 kW/
divided into three primary categories, viz., geometric coating m2. As the heat flux is increased beyond incipience, the wall super-
properties, effective thermophysical coating properties, and boil- heat continues to increase. For both spherical and irregular particle
ing performance characteristics. First, the dependence of the coatings, it is observed that the heat transfer coefficient generally
effective thermophysical properties on the geometric properties is decreases with increase of coating porosity throughout the range of
evaluated (i.e., each effective thermophysical property is treated porosities investigated (40–80%). For the low-porosity coatings
as an output and the geometric properties are treated as inputs). (up to 55% for spherical and 51% for irregular), the heat transfer
Subsequently, the dependence of the boiling characteristics on the coefficient increases at a rapid rate after incipience, reaches a maxi-
individual geometric and thermophysical properties is determined mum value, and then decreases at higher heat fluxes (indicated by
(i.e., each boiling characteristic is treated as an output, while the the increasing and then decreasing slope of the boiling curve). How-
geometric and thermophysical properties are treated as inputs). ever, increasing the porosity above these porosities alters the nature
A stepwise regression is performed using the stepwiselm func- of the boiling curve. For these higher porosities, heat transfer coeffi-
tion in MATLAB for each input–output pair, using a quadratic poly- cient increases monotonically as the heat flux increases, as has been
nomial fit. For the quadratic polynomial fit chosen, there can be
either a linear or quadratic dependence between each input and
output. The stepwise regression uses forward and backward
regression to add or remove terms from the model, based on the
p-value of a term, thus determining whether the dependence for
each input–output pair is linear or quadratic. The R2 value of the
polynomial fit indicates how well the model fits the actual data
(better fit as R2 approaches unity), while a lower p-value indicates
a more significant input parameter term [34,35].
In order to study the combined effect of multiple input geomet-
ric properties, the most critical input geometric parameters are
then selected based on the lowest p-values for each output consid-
ered. The number of samples tested experimentally limits the
number of critical input parameters that can be assessed in combi-
nation (maximum of three). A direct regression is then performed
using the fitlm function in MATLAB with only the significant terms
for each input parameter. The p-value and R2 value of the result-
ing dependence are checked for goodness of fit of the significant
terms identified. This approach is taken because using a stepwise
regression for the final dependence of the combined variable could
only identify a smaller number of critical coating parameters,
which may not be unique, given the limited number of experimen-
tal data points. Hence, after identifying the critical input terms for
each output based on the number of experimental data points
available (i.e., three critical inputs), a direct regression forces the
final model to include all these critical input parameters so as to
obtain a combined dependence. It is noted that the p-value of the
final model should therefore not be compared to that of the indi-
vidual input–output pairs, but simply used as a stand-alone indica-
tor of the significance of the terms in the final model.

4 Results and Discussion

4.1 Boiling Curves. For porous coatings formed from par-
ticles of the same size (90–106 lm), each distinct particle mor-
phology and porosity creates pores of unique shapes, thereby
leading to differences in the boiling heat transfer performance.
Figure 7 shows the boiling curves obtained for the irregular and
spherical particle coatings described in Table 1.
Compared to a polished surface [9], these sintered surfaces
show boiling incipience wall superheat that is 16–80 times lower,
peak heat transfer coefficients that are 8–25 times higher, and
CHF that is higher by a factor of 1.2–2. The increased density of
active nucleation sites and enhancement of heat transfer area are
the primary contributors to the heat transfer coefficient enhance-
ment for such sintered surfaces. The overall trend of the boiling
curves is similar to that of typical sintered surfaces. At very low
heat fluxes (below 4 kW/m2, not shown on the graph), natural Fig. 7 Boiling curves of area-averaged heat flux versus wall
convection heat transfer occurs. As the heat flux is increased, superheat, DT (Ts 2 Tsat), for (a) spherical and (b) irregular parti-
nucleate boiling incipience occurs at very low wall superheat cle coatings. Occurrence of CHF is indicated by horizontal
(<1 K). With further increase in heat flux, the wall superheat arrows.

Journal of Heat Transfer FEBRUARY 2017, Vol. 139 / 021502-7

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jhtrao/935799/ on 04/24/2017 Terms of Use: http://www.asme.org/about

 4.2 Qualitative Image Analysis. It is observed from Fig. 2
that the structure of the open pores and the particle packing
arrangement in the sintered coatings depend on the particle mor-
phology and coating porosity. The pores in the irregular particle
coatings qualitatively appear smaller and more tortuous than in
the case of the spherical particles. Larger pores at similar porosity,
as seen in the spherical particle coatings, may be indicative of
lesser interfacial area between the liquid in the pores and the cop-
per particles, as well as reduced contact between individual cop-
per particles, thereby reducing the effective thermal conductivity
of the coating. It is also clear from the images that the spherical
particles have a notably smoother surface compared to the rough-
ness displayed by the irregular particles. This inherent surface
roughness likely leads to an enhancement of the interstitial surface
area available for heat transport during boiling with the irregular
particles at the same porosity. For a fixed particle type, Fig. 2
shows that pore diameters increase with porosity. The heat trans-
fer performance of the coatings is consistent with these qualitative
observations regarding the pore size and surface roughness; irreg-
ular particles have a higher heat transfer coefficient than spherical
particles, and higher porosity coatings have lower heat transfer
coefficients.
The pore size also has an effect on the CHF reached by the
coatings. Larger pores allow more efficient vapor escape at higher
heat fluxes, thereby leading to higher CHF for coatings with
higher porosities, consistent with findings from previous studies
[11,37]. Between the two particle types, spherical particle coat-
ings have slightly higher CHF at the same porosity compared to
irregular particle coatings, which is consistent with the larger pore
size observed in the images.
The pore structure may also explain the observed change in the
slope of the boiling curve for the different coating porosities. As
the heat flux is increased, the heat transfer coefficient initially
increases for all samples. However, as noted in Sec. 4.1, further
increases in heat flux reduce the heat transfer coefficient for coat-
ing porosities below 55% (Fig. 8). This behavior may be attributed
to inefficient vapor escape and liquid replenishment through these
coatings, compared to the behavior with the larger pores at higher
coating porosities.
While the SEM images provide qualitative visualization of the
pore structure of the coatings, the quantitative characterization
presented in Secs. 4.3 and 4.4 is essential to establishing the geo-
metric differences between the coatings, identifying critical
parameters that affect the boiling heat transfer performance of sin-
tered particle coatings, and allowing firm conclusions to be drawn
Fig. 8 Heat transfer coefficient versus area-averaged heat flux about the boiling curve trends on the basis of significant coating
for (a) spherical and (b) irregular particle coatings characteristics.

previously observed for sintered spherical coatings [8,36]. The irreg-

ular particle coatings show better heat transfer performance (i.e., a
lower superheat at a given heat flux, implying a higher heat transfer 4.3 Geometric and Effective Thermophysical Properties of
coefficient) compared to the spherical coatings at similar porosities the Coating. The geometric properties of the coatings are obtained
throughout the boiling curve. from the l-CT reconstructions as described in Sec. 3.3. Table 3 shows
The coating characteristics also have an effect on the CHF. the geometric coating parameters for each particle morphology and
Low-porosity particle coatings generally have a lower CHF than porosity investigated, viz., coating porosity (e), particle diameter
higher-porosity coatings, e.g., irregular particle coatings with (dpart) distribution, pore diameter (dpore) distribution, sphericity (/)
porosity 58% reach CHF at 135 6 2 kW/m2 and spherical par- distribution, unit interfacial area, (Sint) and unit necking area (Sneck).
ticle coatings with porosity 55% reach CHF at 159 6 4 kW/ The numerically simulated values of effective permeability (Kz) and
m2. Above these coating porosities, a sudden jump in CHF is effective thermal conductivity (kz) along the z-direction are also
observed (to 226 6 5 kW/m2 for irregular particle coatings and shown. The arrows at the bottom of each column indicate whether the
236 6 0.4 kW/m2 for spherical particle coatings). property increases or decreases with an increase of coating porosity.
This drastic change in the behavior of the boiling curve heat Figure 9 shows representative temperature and pressure con-
transfer coefficient and CHF at a critical porosity value suggests tours obtained from the numerical simulations for a spherical par-
the presence of a corresponding change in some governing coating ticle coating at 39% porosity and an irregular particle coating at
characteristics at this porosity. The general trends in boiling per- 66%. The temperature and pressure are observed to be stratified
formance are first qualitatively analyzed in light of the morphol- along the z-direction, as with past observations in the literature
ogy of the sintered particle coatings revealed by the SEM images [25]. The heat transfer coefficient and CHF for the boiling curves
in Fig. 2, followed by a quantitative analysis of the boiling charac- are also shown in Table 3. The heat transfer coefficient (hav) is an
teristics in terms of the coating properties evaluated from l-CT average value computed over the range of heat fluxes varying
scans. from 25% to 75% of CHF. This average heat transfer coefficient

021502-8 / Vol. 139, FEBRUARY 2017 Transactions of the ASME

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jhtrao/935799/ on 04/24/2017 Terms of Use: http://www.asme.org/about

 Table 3 Geometric properties, effective thermophysical properties, and boiling performance of the sintered particle coatings
investigated. The arrows at the bottom of each column indicate whether the property increases (upward arrow) or decreases
(downward arrow) with increase of porosity, for each type of coating.

Geometric properties Effective thermophysical properties Boiling performance

Particle shape e (%) dpore (lm) dpart (lm) / Sint (mm1) Sneck (mm1) Kz (105 mm2) kz (W/m K) hav (kW/m2 K) CHF (kW/m2)

Spherical 39 48 6 24 85 6 18 0.79 6 0.11 74 12.4 1.6 129 66.6 162

45 65 6 27 80 6 17 0.73 6 0.16 72 9.5 3.5 105 56.9 154
55 73 6 31 84 6 17 0.70 6 0.09 63 8.1 7.7 78.3 41.3 160
68 96 6 52 77 6 15 0.75 6 0.1 34 5.4 31.2 40.3 36.9 235
72 104 6 50 79 6 22 0.79 6 0.12 30 2.8 33.1 29.3 34.3 236
79 107 6 59 77 6 15 0.79 6 0.11 27 2.2 41.8 16.9 28.0 235
" " — — # # " # # "
Irregular 51 52 6 199 75 6 17 0.57 6 0.09 84 12.9 1.6 141 78.2 137
58 61 6 24 73 6 16 0.60 6 0.22 73 8.3 1.8 86.6 63.5 134
66 78 6 32 71 6 14 0.70 6 0.10 54 5.5 6.9 54.1 59.4 223
71 85 6 39 69 6 14 0.74 6 0.15 47 4.3 11.1 42.7 55.9 223
74 94 6 42 70 6 14 0.71 6 0.10 43 3.9 17.6 37.5 48.9 233
" " — — # # " # # "

Fig. 9 Representative contours of (a) pressure field in the pore domain, (b) temperature field
in the solid domain, and (c) temperature field in the pore domain for the spherical particle
coating at 39% porosity. The corresponding contours for irregular particle coatings at 66%
porosity are shown in (d–f). The contours are obtained from numerical simulation of the effec-
tive permeability and effective thermal conductivity.

is merely an indicative representation of the heat transfer coeffi- 50 lm to 100 lm over the porosity rage), as was observed
cient provided by the coating across much of the boiling curve. qualitatively from the SEM images. The average pore size (as
It may be noted here that while all coatings were fabricated well as the standard deviation) shows a gradual increase with
using loose particles sieved to the same size range (90–106 lm), porosity up to an intermediate value (55% for spherical and 58%
the average particle diameters are reduced upon sintering, and are for irregular particle coatings), followed by a sharp increase. This
observed to be of the same order, of 70–85 lm. The sphericity quantitative characterization is consistent with the drastic change
of the coatings is also observed to be fairly constant over the range in the trends of the boiling heat transfer coefficient and CHF noted
of porosities investigated. The irregular particle coatings show a at this intermediate porosity, as was described in Sec. 4.1. This
slightly smaller average sphericity compared to the spherical par- sharp increase may be attributed to the large increase in the per-
ticle coatings. The sphericity is indicative of the qualitative mor- centage of carbonate particles from 10% to 45% that was neces-
phological differences between the two types of coatings formed sary to achieve the desired higher porosity. While for low
by spherical and irregular particles. carbonate percentages (0–10%), the pore size may be dominated
For both spherical and irregular coatings, the average pore by the size of copper particles, the pores formed at higher carbon-
diameter increases with an increase in the coating porosity (from ate percentages may be determined by both the carbonate and the

Journal of Heat Transfer FEBRUARY 2017, Vol. 139 / 021502-9

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jhtrao/935799/ on 04/24/2017 Terms of Use: http://www.asme.org/about

 Table 4 Analysis of the relationship between geometric coating properties and the effective thermophysical properties. The criti-
cal input properties affecting each output are shown underlined and bold.

Spherical particle coating Irregular particle coating

Effective thermophysical Input geometric Variable p-value Variable p-value

property property dependence R2 (103) dependence R2 (103)

Permeability, Kz (mm2) e f(e2) 0.96 0.34 f(e2) 0.98 12.7

dpore f(d2pore) 0.95 0.55 f(d2pore) 1 5.4
Sint f(S2int) 0.99 0.86 f(S2int) 0.94 30.3
Sneck f(Sneck) 0.88 3.5 f(Sneck) 0.56 90.7
dpart f(dpart) 0.63 37 f(dpart) 0.56 89.8
rdpore f(rdpore) 0.99 0.02 f(rdpore) 0.87 12.9
U f(/) 0.04 333 f(/) 0.55 92.5
e, dpore, rdpore f(e2, d2pore, rdpore) 0.99 8.6 f(e2, d2pore, rdpore) 0.96 120
Thermal conductivity, kz (W/m K) e f(e) 0.99 0.02 f(e) 0.91 6.5
dpore f(d2pore) 0.98 0.91 f(d2pore) 0.96 19.2
Sint F(S2int) 0.95 0.54 f(S2int) 0.95 2.8
Sneck f(Sneck) 0.97 0.22 f(Sneck) 1 0.01
dpart f(dpart) 0.57 52 f(dpart) 0.90 8.6
rdpore f(rdpore) 0.92 1.4 f(rdpore) 0.84 18.6
U f(/) 0.16 599 f(/) 0.83 20.8
e, Sneck, Sint f(e, Sneck, S2int) 0.99 5.8 f(e, Sneck, S2int) 1 24.5

copper particle sizes, and hence show a wider size distribution. general trends of the boiling performance, it is insufficient to
Coatings with higher porosity and pore sizes are also shown to explain the finer differences such as the drastic change in CHF,
have higher permeability values in Table 3. Since permeability which is more clearly explained from the pore structure analysis.
indicates the ease with which fluid can pass through the coating, Hence, a high-fidelity characterization of the pore statistics is
higher permeability leads to more efficient vapor escape, resulting essential to interpreting the boiling performance of the coatings,
in a higher CHF. This is in accordance with the observations as is performed herein.
above, where coatings with a higher porosity or pore size show a The unit interfacial area (Sint) of the sintered particle coatings
drastic increase in CHF due to the associated higher permeability. denotes the surface area available for heat transfer between the
Irregular particle coatings which have smaller pore sizes also have solid and fluid domains during boiling, while the unit necking area
lower permeability, and hence, lower CHF than spherical particle (Sneck) indicates the contact area available for solid–solid conduc-
coatings at similar porosity. While measuring the bulk coating tion between individual copper particles. For both spherical and
characteristics such as porosity may provide insight into the irregular particle coatings, an increase in porosity is associated

Table 5 Analysis of the relationship between the geometric and effective thermophysical properties on boiling performance char-
acteristics. The critical input properties affecting each output are shown underlined and bold.

Spherical particle coating Irregular particle coating

Boiling Input Input geometric Variable p-value Variable p-value

performance variable type parameter dependence R2 (103) dependence R2 ( 103)

Average heat Geometric property e f(e) 0.96 0.31 f(e) 0.96 2.6
transfer coefficient, dpore f(dpore) 0.94 0.79 f(dpore) 0.91 7.6
hav (kW/m2 K) Sint f(S2int) 0.92 9.9 f(S2int) 0.92 42.5
Sneck f(Sneck) 0.92 1.5 f(Sneck) 0.98 1.0
dpart f(dpart) 0.56 88.4 f(dpart) 0.87 13.1
rdpore f(rdpore) 0.85 5.3 f(rdpore) 0.90 9.3
/ f(/) 0.21 738 f(/) 0.77 32.4
e, dpore, Sneck f(e, dpore, Sneck) 0.93 39.5 f(e, dpore, Sneck) 0.96 114
Effective thermo- Kz f(Kz) 0.84 6.8 f(Kz) 0.67 56.2
physical property kz f(kz) 0.98 0.13 f(kz) 0.97 1.20
Critical heat flux, Geometric property E f(e) 0.80 10.1 f(e) 0.81 23.6
CHF (kW/m2) dpore f(dpore) 0.80 9.8 f(dpore) 0.85 16.5
Sint f(Sint) 0.95 0.59 f(Sint) 0.88 11.4
Sneck f(Sneck) 0.74 18.2 f(Sneck) 0.71 47.6
dpart f(dpart) 0.58 48.6 f(dpart) 0.75 37.4
rdpore f(rdpore) 0.91 1.9 f(rdpore) 0.81 25
U f(/2) 0.10 539 f(/2) 0.86 71.8
e, dpore, Sint f(e, dpore, Sint) 0.99 2.4 f(e, dpore, Sint) 0.99 55.6
Effective thermo- Kz f(Kz2) 0.93 8.8 f(Kz2) 0.91 46
physical property kz f(kz) 0.79 10.8 f(kz) 0.70 49.2

021502-10 / Vol. 139, FEBRUARY 2017 Transactions of the ASME

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jhtrao/935799/ on 04/24/2017 Terms of Use: http://www.asme.org/about

 with a reduction in Sint as well as Sneck, as shown in Table 3. At metrics on the geometric and thermophysical coating properties.
similar porosities, spherical particle coatings have smaller Sint The regression analysis approach is extended to quantitatively
compared to irregular particle coatings, owing to the smoother par- establish this dependence of the heat transfer coefficient and CHF
ticle surfaces, as seen from an observation of the SEM images in on the coating properties, as shown in Table 5, with critical inputs
Sec. 4.2. Spherical particle coatings also have smaller Sneck, com- shown in bold. The variation of hav with the normalized critical
pared to irregular particle coatings. A larger interfacial area within input properties is shown in Fig. 10. Each input property is nor-
the coating layer can be more efficiently utilized when the necking malized based on its maximum value for each type of coating. It
area is large; this combination leads to a higher effective thermal is demonstrated that hav is most significantly affected by the geo-
conductivity and heat transfer coefficient. Across all the samples metric properties e, dpore, and Sneck (all linear-dependence rela-
considered, coatings with a lower porosity or with irregular par- tionships), as also observed qualitatively in Sec. 4.3 and indicated
ticles definitively exhibit higher heat transfer coefficients than in bold in Table 5. The effective thermal conductivity kz, which is
those with a higher porosity or with spherical particles (as also directly dependent on these geometric properties as discussed in
observed in Sec. 4.1), due to the larger interfacial and necking Sec. 4.3, is shown to be the most significant thermophysical prop-
areas (and higher thermal conductivities) associated. erty affecting the average heat transfer coefficient linearly.
As shown in Table 5, CHF is observed to vary linearly with the
key parameters e, dpore and Sint. This agrees with the discussion in
4.4 Quantitative Analysis of the Dependence of Boiling Sec. 4.3, where e and dpore were observed to affect the effective
Curves on Coating Properties. Section 4.3 described trends in permeability Kz (also shown to be the most significant thermo-
the boiling heat transfer coefficient and critical heat flux with the physical property affecting CHF of the coatings), which likely
geometric and effective thermophysical coating properties. While
the trends with respect to each property have been qualitatively
identified, a modified stepwise regression analysis is now per-
formed to quantify the relative dependence of the boiling perform-
ance on each property, and to identify the critical coating
properties to which the boiling performance is most sensitive.
To analyze the effect of geometric properties on the effective
thermophysical properties (permeability Kz and thermal conduc-
tivity kz) of the coatings, each geometric parameter is first treated
as a single input with the thermophysical property as an output.
The standard deviation of pore size differs significantly across the
coatings; hence, it is treated as an individual input parameter
(rdpore) that could affect the coating performance independent of
the pore size itself. Based on the stepwise regression analysis
described in Sec. 3.6, the best-match relationship (i.e., linear or
quadratic) between the inputs and output is identified, as are the
critical input parameters with lowest p-values (indicated in bold in
Table 4). The function “f” in Table 4 represents the dependence of
each output on the input parameters and takes a linear form (i.e.,
f(x) ¼ Ax þ B) or quadratic form (i.e., f(x) ¼ Cx2 þ Dx þ E),
depending on the results of the stepwise regression. For Kz, the
lowest p-values are observed for porosity, pore size, and standard
deviation of pore size, deemed to be the most significant factors in
determining the effective permeability. This is in accordance with
the literature, which shows a strong dependence of permeability
on coating porosity from the classical Kozeny–Carman relation,
Eq. (1), and several variations on this equation [38–40] that iden-
tify the pore structure characteristics (such as pore size distribu-
tion) as a critical factor affecting permeability. The permeability
Kz in the present work is observed to show quadratic variation
with e and dpore, and linear variation with rdpore for both particle
types. To determine the final dependence of the combined varia-
bles, a direct regression is then performed using only the critical
input terms e2, dpore
2
and rdpore, which results in a low p-value and
2
high R value as shown in Table 4. Following the same analysis,
the most significant factors in determining the effective thermal
conductivity are observed to be e and Sneck (linear relationship), as
well as Sint (quadratic relationship). The thermal conductivity of
coatings would be expected to vary with coating porosity [39–43]
as well as the unit necking area [25], thus supporting the validity
of the method used in the present study. The p-value and R2 value
of the resulting functions are shown in Table 4 and are indicative
of the goodness of fit of the significant terms. It is important to
note that, while the p-value demonstrates the relative significance
of the inputs within one dataset, the exact magnitudes cannot be
compared between two statistically different datasets, i.e., spheri-
cal and irregular particle coatings. Fig. 10 Heat transfer coefficient versus normalized critical
The dependence of effective permeability Kz and thermal con- coating properties for (a) spherical and (b) irregular particle
ductivity kz on the geometric coating properties is fairly well coatings. Each input property is normalized by its maximum
established in the literature [17,18,25]. However, there are no uni- value. The symbols indicate experimental data points, while the
fied models depicting the dependence of boiling performance dotted lines indicate a linear fit to the data.

Journal of Heat Transfer FEBRUARY 2017, Vol. 139 / 021502-11

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jhtrao/935799/ on 04/24/2017 Terms of Use: http://www.asme.org/about

 determines vapor escape and hence CHF of the coatings. While hav ¼ average heat transfer coefficient, kW/m2 K
the effect of Sint on CHF may not be intuitive, all the geometric hfg ¼ latent heat, kJ/kg
parameters are not truly independent. The value of Sint is depend- k¼ thermal conductivity, W/m K
ent on both e and dpore as observed qualitatively, and hence, the K¼ permeability, m2
cumulative effect likely enhances its influence on CHF. P¼ applied pressure, Pa
q00 ¼ wall heat flux, kW/m2
Sint,tot ¼ total interfacial area of unsegmented particles, mm2
5 Conclusion Sneck,tot ¼ total neck area between connected particles, mm2
Sint ¼ total interfacial area of unsegmented particles per unit
The effects of coating characteristics on pool boiling heat trans- volume, mm1
fer performance are investigated for porous coatings formed by Sneck ¼ total neck area between connected particles per unit
sintering copper particles of two different morphologies, viz., volume, mm1
irregular and spherical, for particle sizes of 90–106 lm over a T¼ temperature,  C
wide range of coating porosities (40–80%). Image analysis of the U¼ average fluid velocity
coatings is performed using SEM images and l-CT scans to V¼ volume of a particle, m3
obtain qualitative and quantitative characteristics of the coating
geometry. Numerical simulation of transport in the reconstructed
3D coatings is performed to obtain the effective thermophysical Greek Symbols
properties of the coatings. The dependence of the heat transfer c¼ surface tension, N/m
coefficient and CHF on the geometric and thermophysical coating d¼ coating thickness, m
properties is quantitatively determined using a modified stepwise e¼ porosity
regression analysis. l¼ dynamic viscosity, kg/m s
The boiling heat transfer performance for the coatings is q¼ density, kg/m3
observed to be strongly dependent on the observed differences in rdpore ¼ standard deviation in pore diameter, m
the structure of the interstitial pores formed in each coating. The /¼ sphericity
coating porosity, pore diameter, unit necking area, effective ther-
mal conductivity, and effective permeability are observed to be Subscripts
the most critical coating properties affecting the boiling heat
transfer coefficient of the coatings. The CHF of the coatings is s ¼ surface
most strongly affected by the coating porosity, pore diameter, unit sat ¼ saturation
interfacial area, effective permeability, and effective conductivity. z ¼ coordinate direction
The coatings with lower porosities or irregular particles show
larger heat transfer coefficients along the boiling curve compared References
to coatings with higher porosities or spherical particles. A sharp [1] Milton, R. M., 1968, “Heat Exchange System,” U.S. Patent No. 3384154.
[2] Liter, S. G., and Kaviany, M., 2001, “Pool-Boiling CHF Enhancement by
shift in the nature of the boiling curve and CHF is observed at an Modulated Porous-Layer Coating: Theory and Experiment,” Int. J. Heat Mass
intermediate porosity. The boiling curve characteristics depend Transfer, 44(22), pp. 4287–4311.
strongly on the particle morphology, coating pore structure, and [3] O’ Connor, J. P., and You, S. M., 1995, “A Painting Technique to Enhance
effective thermophysical properties of the coated layer, which can Pool Boiling Heat Transfer in Saturated FC-72,” ASME J. Heat Transfer,
117(2), p. 387.
vary widely even for coatings having similar porosity. The bulk [4] You, S. M., and Simon, T. W., 1992, “A Technique for Enhancing Boiling Heat
coating porosity is inadequate to explain this sharp change in boil- Transfer With Application to Cooling of Electronic Equipment,” IEEE Trans.
ing curve behavior, but this trend is clearly explained in terms of Compon. Hybrids Manuf. Technol., 15(5), pp. 823–831.
the coating pore size distribution, interfacial area, necking area, [5] El-Genk, M. S., and Parker, J. L., 2005, “Enhanced Boiling of HFE-7100
Dielectric Liquid on Porous Graphite,” Energy Convers. Manage., 46(15–16),
effective permeability, and effective thermal conductivity pp. 2455–2481.
obtained from high-fidelity characterization of the pore statistics [6] Parker, J. L., and El-Genk, M. S., 2005, “Enhanced Saturation and Subcooled
using l-CT image analysis. This characterization approach is Boiling of FC-72 Dielectric Liquid,” Int. J. Heat Mass Transfer, 48(18),
demonstrated to provide essential insight into the dependence of pp. 3736–3752.
[7] Nishikawa, K., Ito, T., and Tanaka, K., 1979, “Enhanced Heat Transfer by
boiling curve metrics on critical coating properties. Nucleate Boiling on a Sintered Metal Layer,” Heat Transfer—Jpn. Res., 8(2),
pp. 65–81.
[8] Ranjan, R., Murthy, J. Y., and Garimella, S. V., 2012, “Bubble Dynamics Dur-
Acknowledgment ing Capillary-Fed Nucleate Boiling in Porous Media,” 13th IEEE Intersociety
Funding for this work from the Cooling Technologies Research Conference on Thermal and Thermomechanical Phenomena in Electronic Sys-
tems (ITherm), San Diego, CA, pp. 1114–1126.
Center, an NSF I/UCRC at Purdue University, is gratefully [9] Sarangi, S., Weibel, J. A., and Garimella, S. V., 2015, “Effect of Particle Size
acknowledged. The authors thank Steve Titolo at Thermophysical on Surface-Coating Enhancement of Pool Boiling Heat Transfer,” Int. J. Heat
Properties Research Laboratory Inc. (TPRL) for assistance with Mass Transfer, 81, pp. 103–113.
fabrication of the sintered surfaces, Dr. Xuemei Chen and Kevin [10] Furberg, R., and Palm, B., 2011, “Boiling Heat Transfer on a Dendritic and
Micro-Porous Surface in R134a and FC-72,” Appl. Therm. Eng., 31(15),
Drummond for acquiring the SEM images in the Birck Nanotech- pp. 3595–3603.
nology Center at Purdue University, and Dr. James Williams at [11] O’ Hanley, H., Coyle, C., Buongiorno, J., McKrell, T., Rubner, M., and Cohen,
Indiana University School of Medicine for providing the l-CT R., 2013, “Separate Effects of Surface Roughness, Wettability, and Porosity on
scans. The authors also thank Simpleware Ltd. for providing the Boiling Critical Heat Flux,” Appl. Phys. Lett., 103(2), p. 024102.
[12] Li, C., and Peterson, G. P., 2008, “Experimental Studies on CHF of Pool Boil-
licensed software usage of the software, and specifically Tom ing on Horizontal Conductive Micro Porous Coated Surfaces,” AIP Conference
Spirka for his valuable guidance. Proceedings 969(12), Albuquerque, NM, Feb. 10–14, pp. 12–20.
[13] Chi, W., Sampath, S., and Wang, H., 2006, “Ambient and High-Temperature
Thermal Conductivity of Thermal Sprayed Coatings,” J. Therm. Spray Tech-
Nomenclature nol., 15(14), pp. 773–778.
[14] Dixon, A. G., 1988, “Wall and Particle-Shape Effects on Heat Transfer in
A¼ surface area of a particle, m2 Packed Beds,” Chem. Eng. Commun., 71(1), pp. 217–237.
C¼ shape factor [15] Deng, D., Tang, Y., Shao, H., Zeng, J., Zhou, W., and Liang, D., 2014, “Effects
cp ¼ specific heat capacity, J/kg K of Structural Parameters on Flow Boiling Performance of Reentrant Porous
d¼ particle diameter, m Microchannels,” J. Miromech. Microeng., 24(6), p. 065025.
[16] Lin, Y. J., and Hwang, K. S., 2009, “Effects of Powder Shape and Processing
dpart ¼ particle diameter after sintering, m Parameters on Heat Dissipation of Heat Pipes With Sintered Porous Wicks,”
dpore ¼ pore diameter after sintering, m Mater. Trans., 50(10), pp. 2427–2434.

021502-12 / Vol. 139, FEBRUARY 2017 Transactions of the ASME

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jhtrao/935799/ on 04/24/2017 Terms of Use: http://www.asme.org/about

 [17] Bear, J., 1988, Dynamics of Fluids in Porous Media, Dover Publications, [29] Zhao, Y. Y., Fung, T., Zhang, L. P., and Zhang, F. L., 2005, “Lost Carbonate
New York. Sintering Process for Manufacturing Metal Foams,” Scr. Mater., 52(4),
[18] Vafai, K., ed., 2015, Handbook of Porous Media, 3rd, ed., CRC Press, Taylor & pp. 295–298.
Francis Group, Boca Raton, FL. [30] Webb, R. L., 1981, “Nucleate Boiling on Porous Coated Surfaces,” Heat Trans-
[19] Ranjan, R., Murthy, J. Y., Garimella, S. V., and Vadakkan, U., 2011, “A fer Eng., 4(3–4), pp. 71–82.
Numerical Model for Transport in Flat Heat Pipes Considering Wick Micro- [31] 3M Corporation, 2000, “Fluorinert Electronic Liquid FC-72 Product Informa-
structure Effects,” Int. J. Heat Mass Transfer, 54(1–3), pp. 153–168. tion 98-0212-2308-0 (HB),” 3M Corporation, St. Paul, MN.
[20] Bodla, K. K., Weibel, J. A., and Garimella, S. V., 2013, “Advances in Fluid and [32] Simpleware, 2009, “ScanIP, ScanFE, and ScanCAD Tutorial Guide for SIMPLE-
Thermal Transport Property Analysis and Design of Sintered Porous Wick WARE 7.0,” Simpleware Ltd., Exeter, UK.
Microstructures,” ASME J. Heat Transfer, 135(6), p. 061202. [33] MathWorks, 2014, “MATLAB 8.3,” The MathWorks Inc., Natick, MA.
[21] Aziz, A. A., Saury, C., Xuan, V. B., and Young, P., 2005, “On the Material [34] Devore, J. L., 2009, Probability and Statistics for Engineering and the Sciences-
Characterization of a Composite Using Micro CT Image Based Finite Element Brooks/Cole Cengage Learning, Brooks/Cole Cengage Learning, Belmont, CA.
Modeling,” Proc. SPIE 6176, Nondestructive Evaluation and Health Monitoring [35] Esbensen, K. H., Guyot, D., Westad, F., and Lars, P., 2002, Multivariate Data
of Aerospace Materials, Composites, and Civil Infrastructure V, p. 617605. Analysis—in Practice: An Introduction to Multivariate Data Analysis and
[22] Watson, I. G., Lee, P. D., Dashwood, R. J., and Young, P., 2006, “Simulation of Experimental Design, CAMO, Oslo, Norway.
the Mechanical Properties of an Aluminum Matrix Composite Using X-Ray [36] McHale, J. P., Garimella, S. V., Fisher, T. S., and Powell, G. A., 2011, “Pool
Microtomography,” Metall. Mater. Trans. A, 37(3), pp. 551–558. Boiling Performance Comparison of Smooth and Sintered Copper Surfaces
[23] Tabor, G., Young, P. G., West, T. B., and Benattayallah, A., 2007, “Mesh Con- With and Without Carbon Nanotubes,” Nanoscale Microscale Thermophys.
struction From Medical Imaging for Multiphysics Simulation: Heat Transfer Eng., 15(3), pp. 133–150.
and Fluid Flow in Complex Geometries,” Eng. Appl. Comput. Fluid Mech., [37] Polezhaev, Y. V., and Kovalev, S. A., 1990, “Modeling Heat Transfer With
1(2), pp. 126–135. Boiling on Porous Structures,” Therm. Eng., 37(12), pp. 617–620.
[24] Bodla, K. K., Murthy, J. Y., and Garimella, S. V., 2010, “Microtomography- [38] Childs, E. C., and Collis-George, N., 1950, “The Permeability of Porous Mate-
Based Simulation of Transport Through Open-Cell Metal Foams,” Numer. Heat rials,” Proc. R. Soc. A: Math. Phys. Eng. Sci., 201(1066), pp. 392–405.
Transfer Part A, 58(7), pp. 527–544. [39] Millington, R. J., and Quirk, J. P., 1961, “Permeability of Porous Solids,” Trans.
[25] Bodla, K. K., Murthy, J. Y., and Garimella, S. V., 2012, “Direct Simulation of Faraday Soc., 57, pp. 1200–1207.
Thermal Transport Through Sintered Wick Microstructures,” ASME J. Heat [40] Marshall, T. J., 1958, “A Relation Between Permeability and Size Distribution
Transfer, 134(1), p. 012602. of Pores,” J. Soil Sci., 9(1), pp. 1–8.
[26] Lienhard, J. H., and Dhir, V. K., 1973, “Hydrodynamic Prediction of Peak [41] Hashin, Z., and Shtrikman, S., 1962, “A Variational Approach to the Theory of
Pool-Boiling Heat Fluxes From Finite Bodies,” ASME J. Heat Transfer, 95(2), the Effective Magnetic Permeability of Multiphase Materials,” J. Appl. Phys.,
pp. 152–158. 33(10), p. 3125.
[27] Davis, J. R., 2001, Copper and Copper Alloys (ASM Specialty Handbook), [42] Landauer, R., 1952, “The Electrical Resistance of Binary Metallic Mixtures,” J.
ASM International, Materials Park, OH, pp. 222–241. Appl. Phys., 23(7), p. 779.
[28] Upadhyaya, G. S., 2002, Powder Metallurgy Technology, Cambridge Interna- [43] Kirkpatrick, S., 1973, “Percolation and Conduction,” Rev. Mod. Phys., 45(4),
tional Science Publishing, Cambridge, UK, pp. 18–21. pp. 574–588.

Journal of Heat Transfer FEBRUARY 2017, Vol. 139 / 021502-13

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jhtrao/935799/ on 04/24/2017 Terms of Use: http://www.asme.org/about