Professional Documents
Culture Documents
Modeling and Optimization of Superhydrophobic Condensation
Modeling and Optimization of Superhydrophobic Condensation
Superhydrophobic Condensation
Superhydrophobic microlnanostructured sutfaces for dropwise condensation have recently
Nenad Miljkovic received significant attention due to their potential to enhance heat transfer peiformance by
Department of Mechanical Engineering, shedding water droplets via coalescence-induced droplet jumping at length scales below
Massacliusetts institute ot Technoiogy,
the capillary length. However, achieving optimal surface designs for such behavior requires
77 Massactiusetts Avenue,
capturing the details of transport processes that is currently lacking. While comprehensive
models have been developed for flat Hydrophobie surfaces, they cannot be directly applied
Cannbridge, MA 02139
for condensation on micro/nanostructured surfaces due to the dynamic droplet-structure
interactions. In this work, we developed a unified model for dropwise condensation on
Ryan Enright superhydrophobic structured surfaces by incorporating individuell droplet heat transfer,
Department oí Mectianicai Engineering, size distribution, and wetting morphology. Two droplet size distributions were developed,
Massactiusetts institute ot Tectinoiogy, which are valid for droplets undergoing coalescence-induced droplet jumping, and exhibit-
77 Massactiusetts Avenue, ing either a constant or variable contact angle droplet growth. Distinct emergent droplet
Cambridge, iViA 02139; wetting morphologies, Cassie jumping, Cassie nonjumping, or Wenzel, were determined by
Stoi<es Institute, coupling of the structure geometry with the nucleation density and considering local energy
University ot Limerici<, barriers to wetting. The model results suggest a specific range of geometries (0.5-2 ßm)
Limerick, ireland allowing for the formation of coalescence-induced jumping droplets with a 190% overall
surface heat flux enhancement over conventional flat dropwise condensing surfaces. Subse-
Evelyn N. Wang^ quently, the effects offour typical self-assembled monolayer promoter coatings on overall
Department ot Mectianicai Engineering, heatflta were investigated. Surfaces exhibiting coalescence-induced droplet jumping were
Massachusetts institute ot Technoiogy, not sensitive (<5%) to the coating wetting characteristics (contact angle hysteresis), which
77 Massachusetts Avenue, was in contrast to surfaces relying on gravitational droplet removal. Furthermore, flat
Cambridge, MA 02139 surfaces with low promoter coating contact angle hysteresis (<2 deg) outperformed struc-
e-mail: enwang@mit.edu tured superhydrophobic surfaces when the length scale of the structures was above a cer-
tain size (>2 ¡xm). This work provides a unified model for dropwise condensation on micro/
nanostructwed superhydrophobic surfaces and offers guidelines for the design of structured
surfaces to maximize heat transfer. [DOI: 10.1115/1.4024597]
1 Introduction number of small droplets [14] which efficiently transfer the major-
Dropwise condensation has received significant attention since ity of the heat from the surface [8,15-17]. In addition, gravity inde-
its discovery in 1930 by Schmidt et al. [1], due to the superior pendent droplet removal allows utilization of such surfaces in any
heat transfer performance over conventional filmwise condensa- orientation, unlike conventional passive condensing surfaces which
tion [2-7]. The high performance enhancement of passive drop- require gravity for droplet removal, a severe limitation for mobile
wise condensing surfaces is attributed to their ability to form electronics and space appfications [18]. Surface structuring may
discrete nonwetting droplets which, upon growing to a critical therefore be an ideal method to enhance condensation heat transfer.
size ('^lmm), can shed from the surface by gravity, thereby While several groups have demonstrated that properly designed
reducing the overall thermal resistance compared to filmwise con- surfaces can enable stable superhydrophobic condensation
densation. More recently, micro/nanostructured superhydrophobic [19-23], a unified modeling framework to optimize structure
surfaces have been investigated as potential dropwise condensing design that captures the dynamic phase change process is lacking
surfaces for energy conversion [8], atmospheric water harvesting [24]. The early dropwise condensation model of Le Fevre and
[9,10], and high heat fiux thermal management applications [11] Rose [25,26] combined individual droplet heat transfer with drop-
owing to the promise of further improvements in overall heat let size distribution theory. Advanced models have since followed
transfer performance compared to traditional dropwise condensa- this work by including more accurate expressions for the growth
tion. Indeed, a recent study showed that when small droplets of small droplets. Tanaka [27] used population balance theory to
(~10-100,um) merge on superhydrophobic structured surfaces, evaluate the local droplet size by taking into account the two
they can spontaneously jump via the release of excess surface mechanisms of growth: direct vapor accommodation onto the
energy independent of gravity [12]. This phenomenon is attributed droplet and coalescence with neighboring droplets. As a result,
to surface structuring, which enhances the hydrophobicity, and better predictions of the droplet size distribution for small noncoa-
thereby decreases droplet pinning to the surface [13]. Droplet re- lescing droplets were obtained. Abu-Orabi [28] further refined the
moval by this mechanism is highly desirable due to the increased dropwise condensation model by considering all possible thermal
resistances associated with the droplet. More recently, the work of
Kim and Kim [29] extended the previous models by determining
Corresponding author. more accurately the conduction resistance for droplets exhibiting
Contributed by the Heat Transfer Division of ASME for publication in the large contact angles (9 > 90 deg).
JOURNAL OP HEAT TRANSFER. Manuscript received March 31, 2012; final manuscript
received September 2, 2012; published online September 23, 2013. Assoc. Editor: Despite significant developments on dropwise condensation
Sujoy Kumar Saha. modeling, predictive models for condensation on micro/
Journal of Heat Transfer Copyright © 2013 by ASME NOVEMBER2013, Vol. 135 / 111004-1
nanostructured superhydrophobic surfaces are still lacking. Specifi- can exhibit three distinct wetting morphologies: S where con-
cally, three main inconsistencies arise when applying previous densed droplets sit on top of the micro/nanostructure (Eig. l(c))
models to condensation on structured surfaces: (1) Droplet wetting [30], PW where the droplets form a liquid bridge connecting the
morphology cannot be predicted (i.e., Cassie [30], Wenzel (W) base of the droplet (Eig. l{d)) [24], or W where droplets wet the
[31], suspended (S) [24], or partially wetting (PW) [24,32]), (2) cavities ofthe micro/nanostructure (Eig. l(e)) [31].
droplet contact angle can vary during droplet grovrth [14,32,33], To accurately predict the wetting morphology of a single drop-
and (3) droplet size distribution is not valid for droplets with non- let, a nonequilibrium thermodynamic energy criterion is used
constant contact angles and with surfaces undergoing coalescence- which emphasizes the role of contact line pinning on the resultant
induced droplet jumping as the main mode of droplet removal. droplet morphology [22]. By comparing the dimensionless energy
This work develops a model framework to predict dropwise of the advancing Wenzel (cos of = r cos öa) [31] and Cassie
condensation heat transfer for micro/nanostructured superhydro- (cos Ö™ = — 1) [36] droplet morphologies, the expected morphol-
phobic surfaces. Pillar arrays, spanning a wide range of dimen- ogy can be estimated by
sions (~10nm-10/^im), are utilized as the model structured
surface. The current model incorporates prediction capability for coso'iCB
the emergent droplet wetting morphology (Sec. 2), accounts for E* = (1)
COS0" r cos oa
nonconstant contact angle droplet growth (Sec. 3), and extends
the previously developed droplet size distribution theory to both When E*>1 the contact line can overcome the energy barrier to
constant and nonconstant contact angle droplets growing on surfa- depin and a W droplet is formed (Eig. l(e)). If £* < 1 complete
ces experiencing coalescence-induced droplet jumping [12,14] depinning is not possible and the droplet grows upward over
(Sec. 4). The results from the model are subsequently used to the top of the pillar array forming a PW Cassie droplet (Eigs. l(c)
study the effects of surface structure design and size scale (Sec. and l{d)).
6), surface inclination (Sec. 7), and contact angle hysteresis with Droplet coalescence introduces a further length scale depend-
different promoter coatings (Sec. 8) on overall surface heat trans- ency on the emergent droplet wetting morphology. When the av-
fer. Optimization of the structure geometry indicates that surfaces erage condensing droplet spacing (4) approaches that of the pillar
with nanometer length scales and high nucleation densities have spacing /((/c) ~ I), the formation of local pinning barriers is dis-
the potential to enhance overall condensation heat transfer per- rupted. While it may be energetically favorable to form Cassie
formance by 190% when compared to conventional flat dropwise droplets (E* < 1), this may not be possible due to droplet coales-
condensing surfaces. The outcomes of this work create important cence between two adjacent unit cells which bypasses the pinning
regime maps and design guidelines for highly efficient superhy- barriers associated with the pillar sides. This effect results in the
drophobic condensation surfaces. formation of thermodynamically unfavorable W droplets which
are highly pinned to the surface. To avoid formation of W droplets
on Cassie-Baxter (CB) stable surfaces, a critical droplet separa-
2 Droplet Wetting Morphology Model tion distance of at least {k) > 21 must be maintained [22].
During dropwise condensation on a structured superhydropho- Eor the purposes of this model, condensation on the structured
bic surface, droplets can depart either by coalescence-induced surface is assumed to be spatially random. The mean condensing
droplet jumping (if droplet/surface adhesion is small) or by grav- droplet spacing (4) can be related to the nucleation density Ns
ity (if droplet/surface adhesion is large). The individual droplet by [24]
contribution to the heat transfer process is highly sensitive to the
droplet wetting morphology and therefore is necessary to predict 1
the overall surface heat transfer by dropwise condensation.
(4)=- (2)
To study the effects of surface structure geometry on emergent
condensing droplet wetting morphology, we consider a model These two wetting criteria (Eqs. (1) and (2)) have been vali-
structured surface consisting of a micro/nanopillar array with dated experimentally for a wide variety of structured surfaces
heights h, diameters d, and pillar-to-pillar spacings / (solid frac- with a range of length scales and surface energies [2,4,22,24]. The
tion (p = nd^/AP' and roughness factor r = 1 -|- ndh/P) (Eigs. criteria form the basis for determining the emergent condensing
l(a) and 1(0)). To achieve superhydrophobicity, the model surface droplet morphology on the micro/nanostructured surface. How-
is assumed to be coated with a promoter film such as a self- ever, to gain an understanding of overall surface condensation
assembled-monolayer (SAM) having intrinsic advancing/receding performance, the individual droplet heat transfer for each mor-
contact angles of dJ9r [34,35]. During condensation on a struc- phology needs to be incorporated.
tured superhydrophobic surface, the condensing liquid droplets
3 Droplet Heat Transfer Model
To accurately model dropwise condensation on micro/nano-
structured surfaces, individual droplet growth rates and heat trans-
fer are needed. The growth behavior of each droplet morphology
can be obtained by modifying the Kim and Kim model [29] to
account for the pillar geometry and emergent droplet morphology
[24]. It is important to note that most previous models assume a
constant droplet contact angle, 9, dudng growth. This assumption
is appropriate for dropwise condensation on flat hydrophobic
surfaces [28,29,37]; however, it does not apply for structured
superhydrophobic surfaces, since droplets have been observed to
have variable contact angles during growth [14,18,24,32,33,38].
(c) (d) To resolve this discrepancy, a model was developed to predict the
droplet contact angle 0 as a function of the droplet radius R.
Fig. 1 Schematics of the modeled structured surface showing
(a) side view and (b) top view of the characteristic structure
dimensions h, d, and / representing the pillar height, diameter 3.1 Contact Angle Modeling. The droplet contact angle 6
and center-to-center spacing, respectively. Schematics show- was modeled as a function of the droplet wetting morphology (PW,
ing the (c) S, (d) PW, and (e) W morphologies. W, or S) and droplet radius R. When the droplet radius R is lower
Fig. 2 Time-lapse schematics of (a) S, (b) PW, and (c) W dropiet morphoiogies
during growtii on the structured surface. To the right of the schematics are corre-
sponding environmentai scanning electron microscopy (ESEiM) images of dropiets
with the different morphoiogies on a nanostructured surface (/7 = 6.1 ftm, t = 2fim,
d= 300 nm) [23,24]. Schematics not to scaie.
(7)
' /!i27l/?2(l -COSS)
where q is the heat transfer rate through the droplet and hi is the
condensation interfacial heat transfer coefficient given by [42,43]
Pillars
hi
(8)
(C) where /fg is the specific gas constant and Vg is the water vapor spe-
cific volume. The condensation coefficient a is the ratio of vapor
molecules that will be captured by the liquid phase to the total
number of vapor molecules reaching the liquid surface (ranging
from 0 t o i ) .
Once the vapor condenses on the droplet interface, the latent
heat must be conducted through the droplet to the substrate. This
resistance is modeled as a pure conduction resistance which leads
to a temperature drop (AT¡i) given by [29]
The droplet heat transfer is related to the droplet growth rate {dRI 4.1 W and Flat Hydrophobic Surface Droplet Size
dt) by the latent heat of phase change Distribution. Structured superhydrophobic surfaces exhibiting W
droplet growth have a droplet size distribution analogous to that
in Q)^ñih = h ~ of a flat hydrophobic surface. W droplets grow and merge until
'" dt reaching a size characteristic of the capillary length before being
swept off by gravity. Sweeping droplets roll down the surface and
= -p„hfc,— \{l-cos9)(2 + cos6)R^\ (14) remove all droplets in their path, cleaning the condensing surface
3 ai L J for new droplets to renucleate. The droplet sweeping mechanism
allows for small, more effective, droplets to populate the surface
Differentiating Eq. (14), we obtain an explicit term for dR/dt and thereby minimizes the condensation heat transfer resistance.
Due to the significant droplet/surface adhesion exhibited by W
q{R, 6) = ;ip^Afg/?2^ droplets, coalescence-induced droplet removal is not possible,
dt dR rather gravity assisted sweeping of droplets from the surface dom-
inates [12,14,22,24].
+ (1-cos0)^(2 +cos (15) For small W droplets undergoing noninteracting growth, the
population balance theory can be used to determine the droplet
size distribution [27-29]. The steady-state W droplet size distribu-
Equation (15) has been shown to have excellent agreement with
tion is determined from the conservation of number of droplets
experimental results for a variety of differing droplet morpholo-
entering a size range containing droplets with radii Ri to /?2, i.e.,
gies including PW, S, W, and droplets growing on flat surfaces
the number of droplets entering this size range must equal to
[2,24]. In order to determine the overall surface heat transfer per-
the number of droplets leaving. The W droplet growth rate is
formance, the individual droplet heat transfer must be combined
defined as
with the morphology dependent droplet size distribution.
AT--
G= (19)
- cos 0)^(2 + cos 0) 1 RB 1
2/zi ( 1 - cos 0) sin 0 in^ 0 [pnckp + hknc
where
AT
(21)
- cos ö)-(2 + cos 0)
(22)
, sin 0 10^'
0.1 1 10 50
1 R(nm)
2/îi(l-cos0)
Fig. 4 Droplet size distribution for a surface with droplet re-
moval by gravity (flat hydrophobic surfaces) as a function of
•-I-- (23) droplet radius R for various nucleation densities /Vs. The popu-
Â:HC sin
lation density is shown for small droplets (/7(R)) with color
curves and large coalescing droplets {N{PÍ¡) with the black
Solving Eq. (18) analytically, we obtain an expression for n(R) curve. Higher nucleation densities lead to earlier droplet coa-
which is identical for growth of noninteracting droplets on a fiat lescence and smaller coalescence lengths (l^ = 2Re). The popu-
hydrophobic surface [29]. Note that in this case, W droplets are lation of small (R<Re) noninteracting droplets is higher than
assumed to grow with constant apparent contact angles, making large (R> Re) droplets because large droplets experience coa-
the analysis analogous to droplet condensation on a flat hydropho- lescence in addition to being swept off the surface. Model pa-
bic surface. rameters: h = 10fim, l=4nm, d=300nm, \T=Tsat - 7's = 5K,
For large coalescing droplets, the droplet size distribution N(R) 0a/0r=121.6deg/86.1deg.
was established by Glicksman and Rose [25,26]
Where
-(2/3)
1
(24)
^?-Ä\„ r/f R) R^ I ^ ^
(31) 5x10"
' •
1.05
- cos^ 0) sm9—R
For structured surfaces showing the S droplet morphology, or
+ ( 1 - c o s 0)^(2-I-cos Ö) (38) coalescence-induced droplet departure with a static droplet con-
tact angle, the surface heat flux is given by
By relating Eq. (38) to the droplet heat transfer Eq. (19), we can
solve explicitly for Gpw q{R)n,{R)dR (45)
Rn
1 - For structured surfaces showing the PW droplet morphology, or
R (39) surfaces exhibiting nonconstant contact angle behavior, the sur-
A2R+A3 face heat flux is given by
Where
(46)
AT
(1 - cos 0)^(2 + cos 0) To compare the utility of the developed dynamic contact angle
model, we determined the predicted surface heat flux ratio of the
(40)
constant contact angle with the variable contact angle formula-
tions for a surface undergoing PW droplet growth. Figure 5{b)
Because Gp» is a function of R and 0, Eq. (37) is first expanded shows the results of the comparison, indicating that the static con-
and solved numerically. The differential equation to be solved is tact angle formulation is a good approximation for variable con-
given by tact angle droplets at departure lengths larger than approximately
8 /^m. However, at smaller departure lengths, the static formula-
1
(41) tion deviates significantly from the dynamic solution, with error in
dR dR excess of 50% at coalescence lengths below 2 ßm. Although the
error is large at small length scales {R<S /(m), the static approxi-
The boundary condition for this case is identical to that of the S mation works fairly well at larger coalescence lengths, showing
case, where the droplet size distribution a.t R=Re is equal to the the error to be within 10%.
Rose distribution [25] given by In order to gain a better understanding of structured surface
design, the dropwise condensation models developed were applied
-(2/3) to an optimization scheme to examine the interplay between sur-
(42) face structure, droplet morphology, removal mechanisms and
overall surface heat transfer performance.
where
6 Desigu and Optimization of Structured Surfaces
R - ' ' - (43) A number of recent works have fabricated superhydrophobic
structured surfaces for the purpose of enhanced dropwise
condensation via coalescence-induced droplet removal
Figure 5 (a) shows the droplet size distribution, n¿(R) as a func- [10,19-21,23,31,47-50]. These surfaces were designed to be
tion of droplet radius R for a variety of nucleation densities. The Cassie stable such that PW or S droplets are formed on micro/
three plotted curves (dashed lines) correspond to coalescence nanostructures [30] to have minimal contact line pinning. How-
lengths of /c = 10, 15, and 20 ßm. The dynamic droplet size distri- ever, the design methodology of these surfaces is not very well
bution varies from the "static" or S solution (Eq. (35)). The plot understood. In an effort to provide a rational basis for structured
indicates that inclusion of the droplet contact angle variability is surface design, prior to and post droplet departure, we utilized the
dll
lall
h=l /im) for surfaces with smaller scale structures due to the
reduced structure thermal resistance. In addition, reduction of the
pillar height acts to extend the peak performance of the surface
for a larger range of coalescence lengths {IJl).
A second important advantage of scaling down the micro/nano-
structure is seen by the broadening of the dll ratio for the PW
jumping regime. Figure 8 shows the structured surface steady-
state wetting morphology as a function of the pillar diameter to
center-to-center spacing ratio {dll) and the center-to-center spac-
ing to pillar height ratio (///;). As / is reduced for a fixed h, the
region of dll where PW jumping droplets are favored expanded.
This increase of the dll range for PW jumping stability has very
important implications for nanostructured surface design, since
the structure can consist of highly irregular pillar type protrusions
arising from nonuniform fabrication techniques such as oxidation
[4] or self-assembly [33]. The larger range of dll at smaller length
scales facilitates tolerance of structure irregularity and imperfec-
tion, allowing for less costly, scalable, and robust fabrication tech-
niques tobe used [4].
Although the heat flux increases as the structure scale is
decreased from micro to nano, the corresponding required coales-
cence lengths are also decreased. This effect may pose a problem
Fig. 7 Normalized overall steady-state surface heat flux in realistic structured surface design since the coalescence-
cf'lcfmax, as a function of coalescence length {IJI} and ratio of induced droplet jumping mechanism has been shown to have a
pillar diameter to center-to-center spacing (dtl) for (a) h = 5fim, minimum size limit [12].
\b) h = 2tim, and (c) h=^ /im. Scaling down the surface struc-
ture ((a) to (c)) enhances performance due to the reduced
micro/nanostructure thermal resistance. Regions favoring PW 7 Intrinsic Contact Angle
jumping droplet removal show peak heat fluxes for all three
cases ((a) to (c)). qfmax" was determined from examining the Dropwise condensation of water on metal/metal oxide surfaces
peak heat flux in all three cases, which occurred for the small- is rarely observed in natural conditions due to their high surface
est scale structure (c), t/max" = 342.12 kW/m^. Model parameters: energies (the exception being noble metals such as gold and plati-
h/l=2, A r = 5 K , 0a/0r = 121.6 deg/86.1 deg, /CHC = 0.2 W/mK, num due to adsorption of atmospheric hydrocarbons [51-53]). To
/fp = 150 W/mK, áHc = 1nm. Insets: emergent droplet morphol- promote dropwise condensation, the use of SAMs has emerged as
ogy schematics for each region. popular and robust technique to obtain reduced surface energies
[4,10,12,14,19,22-24,29,32,38,47,54,55]. The use of SAMs as
hold for different surfaces, since the morphology is strongly de- surface promoter coatings has three advantages: (1) The SAM
pendent on the solid fraction and surface roughness. thickness is on the order of a few nanometers, resulting in minimal
resistance to heat flow [29]; (2) a range of SAMs can form a cova-
lent bond with metal oxide surfaces greatly increasing robustness;
6.2 Effect of Micro/Nanostructure Scale. Scaling down of and (3) the highly nonwetting properties of fluorinated SAMs act
the structured surface has been shown to potentially increase the as a suitable promoter for dropwise condensation. The developed
overall performance due to the reduction in thermal resistance model is utilized to quantify the effect of SAM coating properties
between the base of the emerging droplet and the structured sur- for selecting an optimal dropwise condensation promoter.
face substrate [2,24]. To study the effects of structure scale on the Four chemistries for SAM formation were examined, with
overall surface heat transfer performance, the unified model was varying chain lengths and intrinsic contact angles on smooth
used to simulate condensation on a surface having h/l = 2 for a va- surfaces. The advancing and receding contact angles used
riety of different pillar heights h. Figures l{a)-l{c) show the nor- were: 0^/0^= 121.6 deg/86.1 deg, 0^/0^= 110.8 deg/85.8 deg, and
malized surface heat flux, q"lq"m¡,,,, for pillar heights oí h = 5 /im, 0a/0r = 103.8 deg/102.7 deg corresponding to typical wetting prop-
h = 2ßm, and h=\ /tm, respectively. The results show enhanced erties for deposited films of (tridecafluoro-l,l,2,2-tetrahydrooc-
heat transfer performance (up to 22.5% from h = 5ßm to tyl)-l-trichlorosilane (SAMl), octadecyltrichlorosilane (SAM2),
80
•
^E 60
20
SAM1
swa
THHDL
0
SAM1;a/6l=121.6786.
SAM2;Éya = 110.e785.8 0 10 20 30 40 50 60 70 80 90
THOL; 9 J f l = 121.17106.3
106
Acknowledgment
94 This work was supported as part of the MIT S3TEC Center, an
82 Fiai Surface
Energy Erontier Research Center funded by the U.S. Department of
70
Jumping Surfece Energy, Office of Science, Basic Energy Sciences under Award #
0 6 12 18 24 30 36 42 48 54 DE-FG02-09ER46577. R.E. acknowledges support from the Irish
Research Council for Science, Engineering, and Technology,
cofunded by Marie Curie Actions under EP7. This work was per-
Fig. 11 Structured (jumping) and flat (gravity shedding) surface formed in part at the Center for Nanoscale Systems (CNS), a mem-
steady-state heat flux </' as a function of intrinsic promoter coat- ber of the National Nanotechnology Infrastmcture Network (NNIN),
ing contact angle hysteresis A0 for three structured surfaces which is supported by the National Science Foundation under NSF
coated with the SAM1 promoter. As Ad decreases for the flat sur-
award number ECS-0335765. CNS is part of Harvard University.
face, cf' increases due to the lower droplet adhesion to the
surface and lower departure radii (inset). As a result, the flat
surfaces begin to show enhanced qf" compared to the structured Nomenclature
surfaces. Inset: Droplet departure diameter {R), as a function of
intrinsic flat surface contact angle hysteresis (Ad). Model param- A = area of condensing surface (m^)
eters: A r = 5 K , /c = 10/<m, /fHc = 0.2 W/mK, kp = 150 W/mK, d = pillar diameter (m)
áHc = 1 nm, and SAMl coating: 0a/ör = 121.6deg/86.1 deg. E* = wetting-state energy ratio