International Journal of Heat and Mass Transfer 100 (2016) 150–164

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer

journal homepage: www.elsevier.com/locate/ijhmt

Condensation of carbon dioxide in microchannels

Brian M. Fronk a, Srinivas Garimella b,⇑
a
School of Mechanical, Industrial and Manufacturing Engineering, Oregon State University, Corvallis, OR 97331, USA
b
George W. Woodruff School of Mechanical Engineering, Georgia Institute of Technology, Atlanta, GA 30332, USA

a r t i c l e i n f o a b s t r a c t

Article history: Heat transfer coefficients and pressure drops during condensation of carbon dioxide (CO2) are measured
Received 10 November 2015 in small quality increments in rectangular microchannels of 0.10 6 Dh 6 0.16 mm and aspect ratios from
Received in revised form 21 March 2016 1 to 4. Channels are fabricated on a copper substrate by electroforming copper onto a mask patterned by
Accepted 23 March 2016
X-ray lithography, and sealed by diffusion bonding. The test section is cooled by chilled water circulating
Available online 6 May 2016
at a high flow rate to ensure that the thermal resistance on the condensation side dominates. A conjugate
heat transfer analysis, in conjunction with local pressure drop profiles allows driving temperature differ-
Keywords:
ences, heat transfer rates, and condensation heat transfer coefficients to be determined accurately.
Carbon dioxide
Condensation
Condensation heat transfer coefficients and pressure drops are measured for G = 400, 600 and
Heat transfer 800 kg m
2 s
1, for 0 < x < 1, and saturation temperatures of 15, 20 and 25 °C (Pr = 0.69, 0.78 and 0.87).
Pressure drop The data are used to evaluate the applicability of correlations developed for larger hydraulic diameters
Experimental and different fluids for predicting condensation heat transfer and pressure drop of CO2.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction [2] relate the magnitudes of surface tension and gravitational

forces through the Confinement number (microchannel effects
The need for smaller, more efficient and environmentally exist when Co > 0.5), while Serizawa et al. [3] relate the same
friendly thermal energy systems continues to drive research on forces through the Laplace constant (L), proposing microchannel
two-phase flow condensation in mini- and microchannels. As effects exist when L > Dh. Both relations imply that microchannel
channel diameter is decreased, higher heat transfer coefficients effects are relevant in larger channels for fluids with larger surface
and heat fluxes can be obtained, leading to a reduction in heat tension. However, defining microchannels in this manner is
exchanger size, material use and inventory of refrigerant. Addition- complicated by the general inconsistency in usage of terminology
ally, small channels can more effectively contain the higher work- between academic literature and industry. For consistency
ing pressures characteristic of new environmentally friendly throughout this paper, channels with Dh < 1 mm are referred to
refrigerants including carbon dioxide. as microchannels, with minichannels encompassing a range of
A growing body of research characterizing two-phase flow 1 6 Dh 6 5 mm.
regimes, flow transitions, pressure drop and heat transfer for con- In an effort to better understand the increased role of surface
densing flows in mini- and microchannels is now available. Still, tension and other effects at this scale, the authors’ research group
the exact definitions of, and demarcation between macro-, mini- has conducted studies on flow visualization, heat transfer and
and microchannels remain nebulous. Rather than defining a hard pressure drop during condensation of a variety of refrigerants, over
cutoff between channel classifications, it can best be concluded a range of operating conditions in channels with Dh from 0.1 to
that the difference between mini- and microchannels and 5 mm, with a focus on synthetic refrigerants including R-134a,
macrochannels assumes importance when forces or phenomena R-404A and R-410A. Coleman and Garimella observed and classi-
generally not significant at the macro scale take on increasing fied the significant flow regime transitions in conventional round
importance. For two-phase flow, this is usually the increasing channels and mini- and microchannels of varying geometries using
importance of surface tension forces and the decreasing influence air–water mixtures [4] and condensing HFC refrigerants [5–7].
of gravitational forces as channel size decreases. Kew and Cornwell With an understanding of the important flow mechanisms of con-
densing refrigerants, Garimella et al. developed experimentally
⇑ Corresponding author at: Sustainable Thermal Systems Laboratory, Georgia validated multi-regime pressure drop models for circular [7] and
Institute of Technology, Atlanta, GA 30332, USA. noncircular [8] mini- and microchannels. The development of the
E-mail address: [email protected] (S. Garimella). novel thermal amplification technique [9] enabled determination

http://dx.doi.org/10.1016/j.ijheatmasstransfer.2016.03.083
0017-9310/Ó 2016 Elsevier Ltd. All rights reserved.
 B.M. Fronk, S. Garimella / International Journal of Heat and Mass Transfer 100 (2016) 150–164 151

Nomenclature

AAD absolute average deviation (%) Greek symbols

AD average deviation (%) a void fraction (
)
AR aspect ratio (W/H) (
) q density (kg m
3)
C constant correlated bypChisholm [1] (
)
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Co Confinement number r=ðgðqL
qV ÞÞ=Dh (
) Subscripts and superscripts
D diameter (circular channels) (mm) amb ambient
Dh hydraulic diameter (non-circular channels) (mm)
avg average
G mass flux (kg m
2 s
1) cond condensation
h heat transfer coefficient (kW m
2 K
1) crit critical
i specific enthalpy (kJ kg
1) exp experimental
L length (m)
f friction
m _ mass flow rate (kg s
1) in inlet
N number of data points (
) L liquid
P pressure (kPa) out outlet
Pr reduced pressure ðP=Pcrit Þ (
)
pre pre-heater
Q_ heat duty (kW) post post-heater
Re Reynolds number (
) sat saturated
W width (m)
V vapor
x quality (
)
X Martinelli parameter (
)

of condensation heat transfer coefficients at small quality incre- has a lower surface tension, viscosity and liquid/vapor density ratio
ments with high accuracy, and led to the development of heat than both air/water and R-134a, which impacts the relative super-
transfer models for circular [10] and non-circular mini- and ficial liquid and vapor velocities, and the interfacial shear between
microchannels [11]. Extensive studies on condensation and super- vapor and liquid in the annular flow regime. These combined
critical gas cooling heat transfer and pressure drop with newer effects may impact heat transfer and pressure drop in ways that
‘‘high pressure” refrigerants have been conducted at reduced pres- conventional models fail to predict accurately.
sures (Pr) from 0.8 to 1.2 in channels with 0.76 6 Dh 6 9.40 mm
[12–15]. A technique for determining heat transfer and pressure
drop in extremely small channels (0.10 6 Dh 6 0.16 mm) was 2. Prior work
developed for condensing flows of R-134a [16].
The present work extends the studies of Agarwal and Garimella Internal condensation heat transfer and pressure drop are inti-
[16] and Fronk and Garimella [17] to consider condensation of car- mately related to the local flow patterns and transition mecha-
bon dioxide (CO2) at 0.10 6 Dh 6 0.16 mm and aspect ratios (AR) nisms, which in turn are functions of quality, void fraction, mass
from 1 to 4. As a low cost, non-toxic, low global warming potential flux and channel geometry. Dobson and Chato [20] state that inter-
(GWP) fluid, CO2 has garnered much attention recently as a nal condensation in large horizontal tubes can be divided into
replacement for synthetic refrigerants [18,19]. With a low critical gravity controlled and vapor shear controlled regions. The gravity
temperature (31.1 °C), CO2 based heat pump systems typically controlled region is characterized by formation of condensate
reject heat through non-isothermal supercritical gas cooling rather around the top circumference, with liquid pooling and flowing axi-
than a typical condensation process. However, CO2 condensation is ally along the bottom. Heat transfer in this regime is generally
of interest in applications such as cascaded vapor compression sys- modeled through a Nusselt type analysis [21,22]. However, as
tems. The investigation of CO2 condensation represents several shown by Coleman and Garimella [6] in a visualization study of
notable differences from studies of conventional refrigerants. At condensing R-134a, the size (as seen on a G versus x flow regime
the temperatures of interest (0–25 °C), the saturation pressure map) of the stratified and wavy flow regimes decreases with
ranges from 3500 to 6435 kPa, significantly higher than the satura- decreasing diameters, and is completely absent at small diameters
tion pressures of interest for synthetic refrigerants. Thus, it is nec-
essary to design test channels, pumps, auxiliary heat exchangers
and all other experimental equipment to safely contain these high
pressures. Due to the low critical temperature, CO2 condensation at
the saturation temperatures of interest (0–25 °C) occurs at much
higher reduced pressures (0.47–0.87) than those of synthetic
refrigerants. In contrast, the reduced pressures of R-134a and
R-404A at the same saturation temperatures range from 0.07 to
0.16, and from 0.16 to 0.33, respectively. Fig. 1 shows the variation
of reduced pressure of R-134a, R-404A and CO2 as a function of
saturation temperature.
Additionally, CO2 condensation at high reduced pressures
results in thermophysical properties substantially different from
the conventional synthetic refrigerants and air/water mixtures that
have typically been investigated. Table 1 shows a comparison of
liquid and vapor densities, viscosities and surface tension for air/
water, R-134a and CO2 at typically investigated conditions. CO2 Fig. 1. Comparisons of refrigerant reduced pressures.
152 B.M. Fronk, S. Garimella / International Journal of Heat and Mass Transfer 100 (2016) 150–164

Table 1
Comparison of properties.

Fluid T (°C) Psat (kPa) Pr (
) r (N m
1) lL (kg m
1 s
1) lV (kg m
1 s
1) qL (kg m
3) qV (kg m
3)

2
4
5
Air/water 25 101 – 7.20  10 8.90  10 1.85  10 997 1.18
R-134a 50 1683 0.41 3.72  10
3 1.24  10
4 1.39  10
5 1053 87.5
CO2 20 5729 0.78 1.19  10
3 6.80  10
5 1.81  10
5 773 194

(Dh < 1 mm) at the mass fluxes of interest. Thus, for the hydraulic effects of channel diameter and mass flux on flow regime transition
diameters and mass fluxes in the present study, the flow is and void fraction in channels of size comparable to those investi-
expected to be primarily annular or intermittent, with heat trans- gated in the present study. They found that channels with
fer and pressure drop dominated by vapor shear. Dh > 0.10 mm exhibited similar flow regimes as minichannels
Shear dominated condensation in large diameter tubes has (i.e., bubbly, slug, slug-annular and annular flow, with the absence
received much attention in the past, resulting in theoretical, of any stratified or wavy flow regimes), while channels with
semi-empirical and purely empirical correlations. An example of Dh 6 0.10 mm exhibited only slug flow, which they attributed to
a correlation that has enjoyed widespread use is the purely empir- the greater viscous and surface tension effects. Furthermore, they
ical Shah [23] correlation and its subsequent modification [24]. The found negligible difference in flow behavior between rectangular
correlation is simple to apply and yields good predictions for a and circular channels. In the present study, caution should be used
range of tube sizes and fluids. In recent years, efforts have been when attempting to extrapolate this flow behavior in the channels
made to develop and refine accurate heat transfer and pressure of interest due to the higher order of magnitude of surface tension
drop models that are valid over a range of tube sizes, mass fluxes of water compared to that of R-134a or CO2.
and fluids, including experimental and analytical studies by Agarwal and Garimella [16] developed a technique for
Dobson and Chato [20], Cavallini et al. [25] and Thome et al. [26]. accurately measuring the heat transfer coefficient and pressure
In general, these studies have focused on round tubes with drop during condensation of R-134a at small quality increments
Dh > 3 mm. in rectangular microchannel test sections (0.10 < Dh < 0.16 mm,
Accurate measurement of condensation heat transfer and pres- AR = 1–4.) The test section design, experimental techniques and
sure drop in microchannels at temperatures relevant for heat rejec- data reduction methods used in their work form the basis for those
tion applications is needed to assess the applicability of these and used in the present study and are discussed in greater detail in the
other models at small Dh. Obtaining accurate, repeatable results is following sections. They found that condensation heat transfer and
made difficult by the small condensation heat duties, low mass pressure drop correlations developed for macro and mini channels
flow rates, and high heat transfer coefficients characteristic of underpredicted the data consistently. The correlation of Lockhart
microchannels and the inability to easily apply and measure a con- and Martinelli [36] with C = 12 (laminar film and turbulent vapor
stant heat flux. These difficulties are compounded when it is core) surprisingly made the most accurate prediction of pressure
desired to measure heat transfer coefficients in small quality incre- drop despite being developed for gas oil mixtures in much larger
ments at high reduced pressures; where the associated ifg and con- tubes.
densation heat duties are very small. Whereas evaporation Work on CO2 condensation has been extremely limited, regard-
experiments can accurately measure and control electric resistance less of channel size. Large diameter tube studies were conducted
heaters to determine evaporation heat duty, no similar device by Iqbal and Bansal [37] and Kang et al. [38] in horizontal round
exists for condensation. Some researchers [27] have used thermo- tubes with D = 6.52, and 5.15, respectively. Huai and Koyama
electric coolers (TEC) to provide a fixed cooling load, with opera- [39] investigated pressure drop and heat transfer during CO2 con-
tion being controlled by an electrical power input, similar to densation in a multi-port extruded aluminum test section with 10
electrical heaters. However, it has been shown that inconsistencies circular ports with D = 1.31 mm. They conducted experiments at
and high uncertainties in TEC performance make them unreliable high reduced pressure (0.88 < Pr < 0.99) with 25.4 6 Tsat 6 30.5 °C
for use as actual heat flux measurement devices [28]. Thus, the and 123 6 G 6 315 kg m
2 s
1. They reported heat transfer coeffi-
condensing refrigerant must be coupled with circulating water or cients between 2 and 6 kW m
2 K
1; however, their data showed
another coolant. However, it is desirable to maintain a high unusual fluctuations as quality varied from 1 to 0, which they
coolant mass flow rate and minimal difference in temperature attributed to poor thermal contact between the test section and
(Tsat–Tcoolant) across the test section to limit coolant-side thermal chilled copper block. Park and Hrnjak [40] conducted an experi-
resistance, which results in high uncertainty in measured heat mental investigation of CO2 condensation at reduced pressures of
duty. Other methods to determine the condensation heat duty 0.23 and 0.31 (Tsat =
25 °C and
15 °C) at a wall subcooling from
more accurately include the use of thermopiles as heat flux sensors 2 to 4 °C in circular channels with D = 0.89 mm and
[29] and the thermal amplification technique proposed by 200 6 G 6 800 kg m
2 s
1. They measured heat transfer coeffi-
Garimella and Bandhauer [9]. cients from 3 to over 10 kW m
2 K
1, with increasing heat transfer
An extensive review of two-phase flow regime mapping, void coefficients at higher qualities, higher mass fluxes and lower
fraction, heat transfer and pressure drop and experimental tech- reduced pressures. The effect of the different wall subcooling was
niques for condensation in mini- and micro channels appears in weak and well within the reported experimental uncertainty. They
[30–32] and is not replicated here. However, as noted by Agarwal found that heat transfer was best predicted by the correlation of
and Garimella [16] extremely limited research exists on condensa- Akers et al. [41], with an average absolute difference <20%. For con-
tion in channels with Dh < 0.20 mm. They concluded that there densation pressure drop, they found that the McAdams et al. [42]
were no reliable models for predicting condensation heat transfer correlation, based on a homogenous flow model, provided reason-
and pressure drop or flow regime at the small channel sizes of ably accurate results. Kim et al. [43] studied CO2 condensation in
interest here (0.10 < Dh < 0.16 mm). smooth tubes with D = 3.48 mm and microfinned tubes with a
Adiabatic N2/water visualization studies by Kawahara et al. [33] 3.51 mm melt-down diameter (i.e., the inner diameter of a theoret-
(Dh = 0.10 mm, rectangular), Chung and Kawaji [34] (0.05 < Dh < ical smooth tube with the same cross-sectional area as that of the
0.50 mm, rectangular) and Chung et al. [35] (D = 0.096 mm, microfin tube) at the same reduced pressures as the Park and
circular and Dh = 0.10 mm, rectangular) attempted to capture the Hrnjak [40] study. Heo et al. [44,45] investigated CO2 condensation
 B.M. Fronk, S. Garimella / International Journal of Heat and Mass Transfer 100 (2016) 150–164 153

in multiport rectangular tubes for hydraulic diameters ranging

from 0.68 to 1.5 mm, and saturation temperatures of
5 to 5 °C,
while Zhang et al. [46] explored CO2 condensation in a minichannel
(Dh = 0.9 mm) condenser. They both found the condensation model
of Thome et al. [26] to predict heat transfer best. Recently, Shah
[47] published an investigation comparing the published CO2 con-
densation data with multiple correlations from the literature. For
mass fluxes below 300 kg m
2 s
1, he found that the empirical
Shah [24] correlation had acceptable predictive ability. However,
no correlation was able to predict data for mass flux greater than
300 kg m
2 s
1 (the range of the present study), suggesting the
need for additional research. Investigations on CO2 condensation
at hydraulic diameters similar to those under consideration in
the present study (Dh < 0.20 mm) are not available.

3. Experimental approach

Heat transfer and pressure drop experiments for CO2 in

microchannels with 0.10 6 Dh 6 0.16 mm and 1 6 AR 6 4 were
conducted. The experimental methodology, microchannel test sec-
tion design, and data analysis procedures used in the present study
were described in detail in [48] and [16]. Thus, only a brief over-
view of the methodology, and the modifications necessary for
CO2 condensation experiments, is presented here.
Multiple parallel rectangular channels were formed on a copper
substrate by electroforming copper onto a mask patterned by X-ray
lithography. Fig. 2a shows the layout of seven separate test sec-
tions on a single copper wafer. Each test section has a channel
length of 40 mm, a depth of 0.10 mm, and a 4 mm inlet and outlet
header to allow for the requisite refrigerant distribution. The
aspect ratio of the test sections varies from AR = 1–4. The number
of channels in parallel varies from 15 to 40, depending on the
aspect ratio. Multiple parallel channels were fabricated so that
condensation heat duties and refrigerant mass flow rates were
large enough to yield accurate heat transfer coefficients at fine
quality increments. The uncertainty in the width and depth of
the channels is ±0.5 lm, with a surface roughness of 10–15 nm
as validated by SEM images of the channels.
The etched sheets were then diffusion bonded to a 6-mm cop-
per wafer to form closed channels. The resulting bond strength
was conservatively estimated to be 10 MPa by the manufacturer.
After bonding, the top wafer was thinned to 1 mm to reduce the Fig. 2. (a) Microchannel test section fabrication pattern, (b) completed microchan-
thermal resistance between the condensing refrigerant and the nel test section, and (c) microchannel test section with coolant blocks.

heat sink. Copper inlet and outlet tubes with a 3.175 mm OD were
soldered to mating holes drilled in the channels to allow for con-
nection to the experimental facility. A hydrostatic pressure test
up to 10.5 MPa was performed on the refrigerant test section to
ensure that it would safely work at CO2 saturation pressures. A
completed refrigerant channel assembly is shown in Fig. 2b.
Upper and lower coolant blocks were soldered to the assembled
refrigerant channels as shown in Fig. 2c. Water flows countercur-
rent to the refrigerant through five parallel 15-mm long channels
in each block (D = 0.794 mm). The completed assembly was then
installed in the experimental loop, shown schematically in Fig. 3.
Temperature measurements at the inlet and outlet of each com-
ponent shown are obtained using NIST traceable microprobes
(ThermoWorks, Inc.) with accuracy of ±0.1 °C. Pressure is measured
with Rosemount absolute pressure transducers with a set span of 0
to 10,000 kPa and an accuracy of ±0.065% of span at the inlet and
outlet of the pre- and post-heater assemblies.
Starting at state point 1 in Fig. 3, the CO2 is a subcooled liquid.
By measuring the temperature and pressure, the specific enthalpy
is determined using an equation of state proposed by Span and
Wagner [49]. Between state point 1 and 2, a specified amount of
heat is added with an 80 W electrical resistance heater controlled Fig. 3. Experimental facility schematic.
154 B.M. Fronk, S. Garimella / International Journal of Heat and Mass Transfer 100 (2016) 150–164

by Variac device. The electrical input to the heater (Q_ pre ) is mea- 4. Data analysis
sured using an Ohio Semitronics watt transducer (GW5-103E),
with an accuracy of ±0.2% of reading. The specific enthalpy of state A thorough description of the data analysis procedure is pre-
point 2 is then calculated from Eq. (1), where (Q_ amb ) is the heat sented in [48] and [16]; a representative data point for the
gain from the surroundings due to natural convection and radia- 0.30  0.10 mm (15 parallel channels) with G = 595 kg m
2 s
1
tion, calculated based on the physical geometry of the heater and xavg = 0.61 at a saturation temperature of 20 °C is described
assembly, insulation properties and the local fluid temperature as here to provide an overview of the analysis.
described in Garimella et al. [48].
4.1. Pressure drop analysis
Q_ pre ¼ mði
_ 2
i1 Þ
Q_ amb ð1Þ
The frictional pressure drop through the microchannels was
The thermodynamic quality at state point 2 is determined as a
determined from a combination of the measured differential pres-
function of the measured local pressure and calculated enthalpy.
sure, the calculated minor losses due to expansions, contractions
This specifies the quality at the test section inlet. From state point
and bends, the frictional pressure drop through the inlet and outlet
2 to 3, the CO2 flows through a microchannel test section. The test
tubes, and the pressure rise due to deceleration as shown in Eq. (3).
section is coupled with a cooling water stream, flowing at a suffi-
Fig. 4 shows a schematic representation of the test section with
ciently high rate to maintain a nearly isothermal wall temperature.
locations of the important sources of minor pressure ‘‘losses”
The condensation duty (Q_ cond ) is rejected from the CO2 and trans- indicated.
ferred to the coolant in the test section.
At state point 3, the CO2 is a two-phase mixture of liquid and DPf ¼ DPmeasured
DPothers; in
DPother; out þ DPdeccleration ð3Þ
vapor. From state 3 to 4, heat is added until the CO2 is a superheated The measured pressure drop for the representative case is
vapor at state 4. By measuring the local temperature and pressure, 25.1 kPa. The frictional pressure drop in the inlet (93 Pa) and outlet
the enthalpy, h4, is determined. Heat is added and measured as tubes (83 Pa) was determined from the multiple flow regime
before, using an 80 W resistance heater and watt transducer. The model of Garimella et al. [7]. Pressure drops from flow contractions
condensation heat duty is then found from Eq. (2). The calculated from the inlet fitting to the inlet tube (AA: 44 Pa), from inlet header
ambient heat/loss gains in the preheater and post heater were on to the microchannels (CC: 392 Pa), and from outlet header to the
the order of 0.5–1 W. This corresponded to an average of 1.4% of outlet tube (EE: 40 Pa) were calculated from the homogeneous
the calculated heat duty, with a maximum of 3.5%. Heat loss/gain model as recommended by Hewitt et al. [50]. Expansion DPs from
was minimal as experiments were conducted at saturation temper- inlet tube to inlet header (BB: 6 Pa), from the microchannels to the
atures near ambient conditions (15–25 °C). outlet header (DD: 143 Pa), and from the outlet tube to the outlet
    fitting (FF: 11 Pa) were calculated using the separated flow model
Q_ cond ¼ Q_ pre þ Q_ pre; amb þ Q_ post þ Q_ post; amb
mði
_ 4
i1 Þ ð2Þ
[50]. The DPs due to tube bends in the inlet and the outlet (6 and
The superheated vapor is condensed to a subcooled liquid state 4 Pa, respectively) and changing flow direction in the inlet and the
in a shell-and-tube heat exchanger, coupled with a circulating outlet headers (390 and 266 Pa, respectively) were estimated using
chilled water/glycol solution. The liquid refrigerant is circulated the homogeneous flow model [50].
through the test loop by a Micropump gear pump (GAH-X21) cap- The deceleration pressure gain (530 Pa) was found using the
able of a maximum differential pressure of 520 kPa and a maxi- separated flow model (Eq. (4)) as described in Carey [51]. The
mum system pressure of 34.5 MPa. The refrigerant mass flow equation was evaluated at the appropriate fluid properties, mass
rate is measured using a Rheonik Coriolis mass flow meter flux, inlet and outlet qualities, and inlet and outlet void fractions
(RHM015) capable of measuring flow rates from 0.004 to as determined from the Baroczy model [52].
" ! #
0.6 kg min
1 with an uncertainty of ±0.2% of rate. G2 x2out G2 ð1
xout Þ2
The system pressure is maintained at the desired saturated
DPdeceleration ¼ þ
qV aout qL ð1
aout Þ
pressure by adjusting the total CO2 charge in the system using an " ! #
attached liquid CO2 tank and a series of needle valves. For each G2 x2in G2 ð1
xin Þ2

þ ð4Þ
data point, the heat input to the pre-heater and post-heater, pump qV ain qL ð1
ain Þ
speed and refrigerant charge level were adjusted to achieve the
desired test conditions. The system ran until steady state was A conservative uncertainty of ±50% was assigned to each calcu-
achieved, after which data were recorded at a rate of 100 Hz for lated pressure drop. By summing all of the pressure losses and
a minimum of 5 min. gains, the frictional pressure drop in the channels was determined.

Fig. 4. Refrigerant flow path schematic for pressure drop analysis.

 B.M. Fronk, S. Garimella / International Journal of Heat and Mass Transfer 100 (2016) 150–164 155

For the representative case, the net pressure drop through the test conditions. Due to the high absolute pressure of CO2, expected
section inlet and outlet was 0.92 kPa and 0.24 kPa respectively, pressure drops from 10 to 125 kPa represent a change of less than
with a 0.53 kPa deceleration pressure gain. The resulting frictional 2% of the total pressure and correspond to a saturation tempera-
pressure drop was found to be 24.49 kPa, or 97% of the measured ture change of less than 0.9 °C at the highest pressure drop.
DP. Because the expected pressure drops resulted in a small change
in CO2 properties, it was assumed that the pressure decreased lin-
4.2. Heat transfer coefficient early through the test section. With all inputs specified, the system
of equations was computed iteratively to obtain the average
Garimella et al. [48] developed a segmented, iterative model condensation heat transfer coefficient. The resulting refrigerant-
using the Engineering Equation Solver (EES) [53] software platform side heat transfer coefficient for this representative case was
to calculate the average condensation heat transfer coefficient in 17.3 kW m
2 K
1.
the test section. Due to the high thermal conductivity of the copper An analysis of heat transfer and pressure drop uncertainty was
test sections, it was necessary to account for axial heat conduction conducted using a propagation of errors approach [55]. A summary
through the channel walls. Thus, the coupled conjugate effects of of the uncertainties in measured and estimated parameters is pre-
conduction through the wall and convection within the refrigerant sented in Table 2. The uncertainty in the measured differential
and coolant channels were addressed by dividing the test section pressure was ±0.25 kPa (25.21 ± 0.25 kPa), and a ±50% uncertainty
into discrete segments, as shown in Fig 5. While this model was was assigned to each calculated inlet/outlet DP and deceleration
originally developed for use with R-134a, it is not dependent on pressure gain described above. The resulting uncertainty in the
the working fluid, and is therefore used here for CO2 condensation frictional pressure drop was 24.58 ± 0.40 kPa or 1.6%.
data analysis with minimal changes. Considering the ±0.2% uncertainty in mass flow measurement
The resistance network equations were setup and the refriger- and ±0.5 lm uncertainty in channel dimension, the resulting mass
ant heat transfer coefficient was computed using the appropriate flux for the representative case was 807 ± 3.3 kg m
2 s
1 (±0.55%).
segment heat transfer areas and heat transfer coefficients. On the The uncertainty of the pre-heater inlet temperature (16.7 ± 0.1 °C)
refrigerant side, the heat transfer area was calculated assuming and pressure (5755 ± 6.5 kPa), combined with the uncertainty in
the channel walls are rectangular fins with adiabatic tips. For the the heat added by the pre-heater (35.2 ± 0.07 W) and the ambient
conditions studied, the calculated fin efficiency was very high heat gain with a conservative ±50% uncertainty yielded a test section
(gfin > 99%); thus, the refrigerant-side heat transfer area was essen- inlet quality of 0.79 ± 0.004 (0.5%). A similar analysis on the post-
tially the entire channel surface area. Closure to the system of heater yielded a test section outlet quality of 0.42 ± 0.003 (0.7%).
equations was provided by setting the sum of each segment heat The calculated condensation heat duty was 14.7 ± 0.22 W (1.5%).
duty to the measured condensation heat duty. The uncertainty of the channel inlet and outlet pressure
For the coolant channels, the heat transfer coefficient was resulted in a saturation temperature uncertainty of approximately
determined using the Churchill correlation [54], with the measured ±0.05 °C (0.25%). The calculated water-side heat transfer coefficient
mass flow rate and the properties evaluated at the average local was 42 kW m
2 K
1 with an assigned uncertainty of ±25%. Thus,
measured temperature. The water-side heat transfer coefficient
was assumed to be constant through the test section. The coolant Table 2
flow rate was high enough such that the DTcoolant < 0.1 °C. For the Summary of uncertainty of measured and estimated parameters.
representative case, the velocity through each channel was
Parameter Uncertainty
8.15 m s
1, with Re = 5784 and resulting hwater = 40 kW m
2 K
1.
Measured parameters Mass flow rate ±0.2%
A conservative uncertainty of ±25% was assigned to the calculated
CO2 temperature ±0.1 °C
value of hwater to determine the refrigerant heat transfer coefficient Ambient temperature ±2.0 °C
uncertainty. Channel dimensions ±0.5 lm
The conductive thermal resistance in each copper node was Absolute pressure ±6.5 kPa
evaluated from the thickness and thermal conductivity. The refrig- Differential pressure ±0.25 kPa
Pre/post heater power ±0.2%
erant temperature at each node was determined from the local sat-
uration pressure. Garimella et al. [48] developed empirical Calculated/estimated Ambient heat losses ±50%
parameters Water-side heat transfer ±25%
relations to account for the change in saturation temperature due coefficient
to the non-linear saturation pressure change through the test sec- Minor losses ±50%
tion for R-134a, which could be on the order of several °C for some

Fig. 5. Heat transfer nodes.

156 B.M. Fronk, S. Garimella / International Journal of Heat and Mass Transfer 100 (2016) 150–164

the resulting refrigerant heat transfer coefficient was determined coolant water temperature of 4 °C for all test conditions. A total
to be 17.4 ± 1.9 kW m
2 K
1, corresponding to an uncertainty of of 335 data points were obtained, 79 for Dh = 0.1 mm, 83 for
11.5%. Dh = 0.13 mm, 88 for Dh = 0.15 mm and 85 for Dh = 0.16 mm.
It should be emphasized that the present investigation was
5. Experimental results focused on multiple parallel microchannels. To mitigate the possi-
bility of maldistribution or instabilities, the design of the test sec-
Four test sections with different hydraulic diameter, aspect tion ensured that the mass flux and (and thus pressure drop)
ratio and number of parallel channels were investigated: through the microchannels is larger (mass flux 1.5–2 greater in
channels) than through the headers. Furthermore, as the data were
 0.10  0.10 mm  40 parallel channels (Dh = 0.10 mm) obtained at high reduced pressure, the liquid/vapor density ratio in
 0.20  0.10 mm  18 parallel channels (Dh = 0.13 mm) the present study is relatively low (3–5) compared to water/air
 0.30  0.10 mm  15 parallel channels (Dh = 0.15 mm) (>800). This should also help mitigate maldistribution concerns.
 0.40  0.10 mm  15 parallel channels (Dh = 0.16 mm) Finally, maldistribution and instabilities are expected to be more
prevalent in boiling flows, which are characterized by energy addi-
Experiments were conducted at nominal mass fluxes of G = 400, tion and rapid flow acceleration as bubbles nucleate and are con-
600 and 800 kg m
2 s
1 at saturation temperatures of Tsat = 15, 20 fined in the small geometries. In prior flow visualization studies
and 25 °C (Pr = 0.69, 0.78 and 0.87). An effort was made to maintain of condensation in multiple parallel mini- and microchannels, flow
a temperature difference between the saturated refrigerant and instabilities were not observed.

Fig. 6. Frictional pressure gradient experimental results.

 B.M. Fronk, S. Garimella / International Journal of Heat and Mass Transfer 100 (2016) 150–164 157

5.1. Pressure drop results independently. Fig. 7 shows pressure gradient data for each tube
plotted at Tsat = 15 and 25 °C and G = 600 and 800 kg m
2 s
1. For
The frictional pressure drop through the microchannel test sec- both saturation conditions, the Dh = 0.10 mm, AR = 1 test section
tions was calculated from the measured pressure drop as detailed has the highest pressure gradient at an equivalent mass flux and
in Section 4.1. The average uncertainty in frictional pressure drop quality. The Dh = 0.16 mm, AR = 4 test section has the second high-
for all data points was 1.8%. The combined deceleration pressure est gradient, followed by the Dh = 0.13, AR = 2 and Dh = 0.15 AR = 3
gain and minor losses for the four test sections was on an average test sections. For channels with a fixed aspect ratio, an increasing
only 2.6% of the measured pressure drop. In the present study, the pressure gradient with decreasing hydraulic diameter would be
mass flux through the inlet and outlet header was much smaller expected. In experiments with R-134a at the same channel dimen-
than the mass flux through the microchannels, resulting in very sions, Garimella et al. [48] also noted an aspect ratio effect, with
low flow velocities. Thus, the frictional pressure drop through the the highest gradients observed at the highest aspect ratio
microchannels dominated any losses in the inlet and outlet head- (AR = 4), followed by Dh = 0.10 mm, AR = 1, Dh = 0.13, AR = 2 and
ers. Additionally, it was desired to maintain a small quality change Dh = 0.15 AR = 3 in that order. Garimella et al. [48] reasoned that
(and the related void fraction change) through the test section. the annular film may be less stable at higher aspect ratios, result-
Therefore, the calculated pressure gain from deceleration of the ing in higher pressure drop and heat transfer. In the absence of a
flow was small (<1.5% of the measured DP) and on the same scale more detailed study in which hydraulic diameter and aspect ratio
as the calculated minor losses. effects are studied independently, or detailed flow visualizations
Fig. 6 shows the frictional pressure gradient as a function of test are performed to ascertain the impact of aspect ratio on flow
section average quality for each test section at saturation temper- regimes and transition mechanisms at channels of this scale, it is
atures of 15 and 25 °C, and mass fluxes of 400, 600 and difficult to draw further conclusions. However, it is clear that pres-
800 kg m
2 s
1. The pressure gradient was determined by dividing sure drop models that do not account for aspect ratio will fail to
the measured frictional pressure drop by the length of the predict the data accurately.
microchannels, which was 40 mm for all test sections. The results
at a saturation temperature of 20 °C are not presented here; how- 5.2. Heat transfer results
ever, they exhibit the same trends.
Several trends can be discerned from Fig. 6. For all diameters, The uncertainty of the calculated heat transfer coefficient is a
saturation temperatures and mass fluxes, pressure gradient is a function of the measured quantities, the uncertainty in channel
strong function of quality, with increasing gradient observed at dimensions and the assumed uncertainty in the water-side heat
higher qualities. This is as expected due to the higher void fractions transfer coefficient. The average uncertainties in the condensation
and associated high velocities of the lower density vapor at high heat transfer coefficient for the Dh = 0.16, 0.15, 0.13 and 0.10 mm
quality. Additionally, all diameters and saturation temperatures channels are ±16.0%, ±11.5%, ±11.3% and ±18.4%, respectively. The
show an increase in pressure gradient with increasing mass flux primary contribution to the uncertainty is the assumed ±25%
at a fixed quality. Comparison of the top four plots with the bottom uncertainty in water-side heat transfer coefficient. At conditions
four plots in Fig. 6 reveals an increase in pressure gradient with a where high refrigerant-side heat transfer coefficients are expected,
decrease in saturation temperature/reduced pressure. The liquid/ the thermal resistance ratio between the water and refrigerant-
vapor density ratio increases at lower reduced pressure, which side decreases, resulting in an increase in uncertainty. Thus, the
yields a larger difference between the vapor and liquid velocities highest experimental uncertainties are observed at high mass flux,
and an increase in interfacial shear stress. high quality and low saturation temperature, when the heat trans-
The effects of changing hydraulic diameter and aspect ratio fer coefficient is expected to be the highest. The overall uncertainty
on pressure gradient are less clear, as they are not varied can be further decreased by evaluating the refrigerant heat transfer

Fig. 7. Effect of channel geometry on frictional pressure gradient.

158 B.M. Fronk, S. Garimella / International Journal of Heat and Mass Transfer 100 (2016) 150–164

coefficient at larger quality differences, which would increase the pressures can be discerned as the saturation temperature is
condensation heat duty and minimize the uncertainty contribution increased from 15 to 25 °C (Pr from 0.69 to 0.87). This can be attrib-
of the heater power measurement, albeit at the cost of lower reso- uted to the same reasons for reduced pressure drop at higher
lution in vapor quality. reduced pressures, namely the reduction in liquid/vapor density
Fig. 8 shows the condensation heat transfer coefficient as a ratio resulting in decreased flow velocities and interfacial shear
function of test section average quality for each test section at sat- stress. For the saturation temperatures considered here, the
uration temperatures of 15 and 25 °C, and mass fluxes of 400, 600 liquid-to-vapor density ratio is 5.1, 4.0, and 2.9 for Tsat = 15, 20
and 800 kg m
2 s
1. and 25 °C, respectively. In addition, the enthalpy of condensation
As the two phenomena are related, the data exhibit trends that decreases as the temperature increases.
correspond to those seen for pressure drop in Fig. 6. As the quality As with pressure drop, an influence of aspect ratio is observed.
increases, the void fraction and the probability that annular flow Fig. 9 shows the data for each test section at Tsat = 15 and 25 °C and
(with its characteristic thin liquid layer) is present increases, G = 600 and 800 kg m
2 s
1 so that the combined effect of changing
resulting in increased heat transfer coefficient. Similarly, for the hydraulic diameter and aspect can be observed.
same channel and quality, increasing mass flux results in higher Much of the data at equivalent mass flux, saturation temperature
heat transfer coefficients. and quality for each tube lie within experimental error; however,
At equivalent channel dimensions, mass flux and quality, a some general trends can be extracted. For all conditions, over the
trend of decreasing heat transfer coefficient at higher reduced range of qualities, the highest heat transfer coefficients are observed

Fig. 8. Condensation heat transfer coefficient experimental results.

 B.M. Fronk, S. Garimella / International Journal of Heat and Mass Transfer 100 (2016) 150–164 159

Fig. 9. Effect of channel geometry on condensation heat transfer coefficient.

in the Dh = 0.10 mm, AR = 1 and Dh = 0.16 mm, AR = 4 test sections. correlation developed from condensing R-134a data in channels of
These are the test sections with the smallest hydraulic diameter the same hydraulic diameter as the present study. Of the correlations
and largest aspect ratio, respectively. At Tsat = 15 °C, for both mass investigated, four were developed from data for a single tube, two for
fluxes, the Dh = 0.13 mm, AR = 2 case has the third highest conden- data from multiple parallel channels, one from a mixed dataset, and
sation heat transfer coefficient over the range of qualities studied, the last was the homogenous model. None of the available correlations
followed by the Dh = 0.15 mm, AR = 3 test section. At Tsat = 25 °C were developed or verified with data encompassing both the diameter
(Pr = 0.87), the difference between these two sections is within the range and fluid properties of interest in the present study.
experimental error. These trends are again consistent with those The results predicted by these correlations for all of the data are
observed by Garimella et al. [48] for experiments with R-134a. He presented in Fig. 10. It can be seen that other than the Agarwal and
observed the highest heat transfer coefficient in the channels with Garimella [8] and Garimella et al. [59] correlations, the data were
the highest aspect ratio (Dh = 0.16 mm, AR = 4). Clearly, aspect ratio generally underpredicted. Table A1 in the Appendix shows a break-
has an effect on heat transfer coefficient. Through detailed analyti- down of the average deviation (AD, Eq. (5)) and absolute average
cal modeling of condensation in small channels, Wang and Rose [56] deviation (AAD, Eq. (6)) of each correlation for each test section.
showed a strong dependence of heat transfer coefficient on "PN  #
microchannel shape, in addition to hydraulic diameter. The i¼1 DP model;i
DP exp;i
AD ¼  100 ð5Þ
observed effects imply that there is an optimum combination of N
hydraulic diameter and aspect ratio to maximize heat transfer coef-
ficient and minimize pressure drop for a given operating condition. "PN  #
 
i¼1 DP model;i
DP exp;i
Further insights into the effect of changing hydraulic diameter and AAD ¼  100 ð6Þ
aspect ratio could be obtained by varying the two independently. N

6. Comparison with the literature Only the Garimella et al. [59] model accounted for channel
aspect ratio; thus, it is perhaps unsurprising that the data for the
Data from the present study were compared with the predic- test section with the highest aspect ratio (AR = 4) were in general
tions or several correlations from the literature developed for predicted least accurately. The pressure drop trend for each test
macro, mini- and microchannels for two-phase pressure drop section over the range of conditions was predicted well by homoge-
and condensation heat transfer coefficient. Many of the correla- neous model, although the absolute value of the pressure drop was
tions were developed for significantly different channel shapes systematically under predicted using this model. Triplet et al. [60]
and sizes, fluid types and reduced pressures. showed good agreement of air/water pressure drop measurements
in circular (D = 1.1 and 1.45 mm) and semi-triangular (Dh = 1.09
6.1. Pressure drop correlations and 1.49 mm) with the homogeneous model in the bubbly and slug
flow regimes. However, as shown previously, the very high surface
The data are compared with predictions of the homogeneous tension of water in air compared to CO2 may make the uniform
mixture model, the correlations of Lockhart and Martinelli [36] phase velocity assumption necessary for the homogeneous model
and Cavallini et al. [25] developed for macrochannels, the correla- more realistic for air–water mixtures than for the present case.
tions of Mishima and Hibiki [57] and Lee and Lee [58] developed Of the three other two-phase multiplier correlations not devel-
for air–water flow in mini- and microchannels, the Agarwal and oped specifically for microchannels, the Lockhart and Martinelli
Garimella [8] correlation for non-circular mini- and microchannels, [36] correlation predicts the data the best (AAD = 61.5%), followed
the Andresen et al. [13] correlation for pressure drop of high- by the correlation of Andresen et al. [13] (75.7%) and Cavallini et al.
reduced pressure refrigerants, and the Garimella et al. [59] [25] (AAD = 82.9%). All three correlations underpredict the data.
160 B.M. Fronk, S. Garimella / International Journal of Heat and Mass Transfer 100 (2016) 150–164

Fig. 10. Comparison of correlations with frictional pressure gradient data.

 
While showing the best agreement, the results of the Lockhart and C ¼ 21  1
e
0:319Dh ð7Þ
Martinelli [36] correlation exhibit larger scatter, most probably
due to the step change in the C parameter as the liquid or vapor Thus, at the small hydraulic diameters of interest in this study, C
film Reynolds number transitions between laminar and turbulent. tends towards zero, which effectively eliminates any contribution
The Andresen et al. [13] and Cavallini et al. [25] correlations were to frictional pressure drop of the liquid and vapor phase coupling,
both developed from data for refrigerants at intermediate and high which is unrealistic. The Lee and Lee [58] correlation, also based on
reduced pressures, and both include terms to account for the effect high aspect ratio channels with air/water mixtures, shows better
of surface tension through the Confinement and Weber number, agreement, but still significantly underpredicts the data from the
respectively; however, both still significantly underpredict data. present study. The inclusion of surface tension, slug velocity and
The correlations were developed primarily from data for round liquid Reynolds number probably extends the applicability of the
tubes with diameters significantly larger than the rectangular correlation to a greater range than that of Mishima and Hibiki
channels of interest here, where the flow regime transitions are [57], although the fluid properties and geometry in the present
anticipated to be different and surface tension forces are increas- study are quite different.
ingly important. The Agarwal et al. [8] (AAD = 26.6%) and Garimella et al. [59]
The two correlations developed for mini- and microchannel (AAD = 24.3%) correlations shows the best agreement of any
using air–water data under predicted the data for all conditions, correlations studied. Both of these approaches were developed
with the Mishima and Hibiki [57] correlation, developed for high- for condensing refrigerants in small channels. The Agarwal et al.
aspect rectangular channels and round tubes with air/water data, [8] correlation is based on flow visualization, heat transfer and
performing the worst. As seen in Eq. (7), they correlate the Chisholm pressure drop studies of R-134a at geometries reasonably close
parameter (C) parameter as a function only of hydraulic diameter. to those of interest in the present study (Dh = 0.424 mm, AR = 2),
 B.M. Fronk, S. Garimella / International Journal of Heat and Mass Transfer 100 (2016) 150–164 161

Fig. 11. Comparison of correlations with condensation heat transfer coefficient data.

and captures the data trend well despite being applied to a differ- empirical input for closure. Thus, the large difference in properties
ent fluid at a smaller diameter. The correlation accounts for surface between R-134a and CO2 likely explain the observed deviation
tension through the same capillary number as Lee and Lee [58], but between the data and the Garimella et al. [59] model. Further
although it was developed for rectangular channels, it does not refinement of the underlying inputs could extend its range of
explicitly capture the effects of aspect ratio. Thus, like the other applicability to the fluid properties of interest in the present study.
correlations, deviation is the highest at the highest aspect ratio.
The Garimella et al. [59] model was developed from condensing
R-134a data using the same microchannels of varying aspect ratio 6.2. Heat transfer correlations
as those investigated in the present study. The model considers
pressure loss mechanisms in the annular and intermittent flow Condensation heat transfer data were compared with eight heat
regimes, and uses a technique to provide a smooth transition transfer correlations; four developed from data for single large
between the regimes. This technique is referred to as the annular tubes, and four for multiple parallel mini- and microchannels. All
flow factor (AFF) concept, which quantifies the predominance of eight correlations were generally derived for shear-dominated
annular flow in the channels. As this model accounts for small condensation, which is expected to be most significant in the chan-
hydraulic diameter and aspect ratio effects, the overall agreement nels of interest. The approaches taken to predict heat transfer can
with the data is best. However, the AFF concept relies on an be divided into two categories: two-phase multiplier correlations
162 B.M. Fronk, S. Garimella / International Journal of Heat and Mass Transfer 100 (2016) 150–164

[20,23,29,61] and those that take a boundary-layer approach uncertainties in heat transfer coefficients and frictional pressure
[11,25,62,63]. drops were quantified using a propagation of errors approach.
The predictions of these correlations are compared with the The data showed an expected trend of increasing heat transfer
data from the present study in Fig. 11, with a breakdown of the coefficient and pressure drop at lower reduced pressure and
average and absolute average deviation shown in Table A2 in increasing quality. Hydraulic diameter and aspect ratio were not
the Appendix. independently varied; therefore, their effects were not explicitly
Unsurprisingly, the correlations based on the more accurate documented. However, it was found that the pressure gradients
pressure drop models (Lockhart and Martinelli [36] and Agarwal and heat transfer coefficients followed similar trends, with the
and Garimella [8]) provide the best results, with Traviss et al. highest h and DP for the 100  100 lm (Dh = 0.1 mm, AR = 1) chan-
[62] (AAD = 18%), Dobson and Chato [20] (AAD = 19%) and Agarwal nel, followed by Dh = 0.16 mm/AR = 4, Dh = 0.13 mm/AR = 2, and
et al. [11] (AAD = 20%) correlations predicting the heat transfer the Dh = 0.15 mm/AR = 3, in descending order. With further investiga-
best. Of these three models, Agarwal et al. [11] predicts the data tions that isolate the effects of these two variables, it may be pos-
with the least scatter. sible to develop an optimal combination of hydraulic diameter and
Of the two other macro channel correlations, the Shah [23] aspect ratio that yields low pressure drops and high heat transfer
overpredicts much of the data (AAD = 29%), while the Cavallini coefficients.
et al. [25] correlation underpredicts nearly all of the data The data were compared with the predictions of several heat
(AAD = 41%). The pressure drop correlation used in this model also transfer and pressure drop models developed for a range of work-
did not yield good predictions, significantly underpredicting all of ing fluids, operating conditions and channel geometries. None of
the data. However, the overall trends in the data appear to be pre- the correlations were developed for condensation of such high
dicted best by the Cavallini et al. [25] correlation, indicating that pressure refrigerants at comparable hydraulic diameters. Addition-
some of the flow mechanisms are being captured properly. ally, few of the correlations explicitly accounted for channel aspect
Of the mini- and microchannel correlations, the Wang et al. [61] ratio. Of the pressure drop models, the best results were obtained
was best (AAD = 26%), followed by the Cavallini et al. [63] from the Garimella et al. [59] model (AAD = 24%) developed for
(AAD = 38%) and the Koyama et al. [64] (AAD = 51%), respectively. condensing R-134a in channels with similar diameter and aspect
The order of agreement of these correlations is matched by that ratio to those in the present study. The heat transfer correlations
of the pressure drop model used. Clearly, accurately modeling [11,20,61,62] based on the more accurate pressure drop correla-
two-phase pressure drop allows for successful implementation of tions showed the best agreement with the data.
the heat and momentum transfer analogy and yields accurate heat Further studies should focus on developing techniques for accu-
transfer coefficient predictions. rately measuring heat transfer coefficients at low mass fluxes,
extending the range of available data. Flow visualization of CO2
and other refrigerants under various operating conditions would
7. Conclusions allow verification and quantification of the local void fraction
and transition mechanisms between annular and intermittent
Heat transfer coefficients and pressure drops during the con- flow, leading to more reliable and applicable flow-regime based
densation of carbon dioxide in rectangular channels with models for both pressure drop and heat transfer coefficient.
0.10 6 Dh 6 0.16 mm and 1 6 AR 6 4 were accurately measured
at reduced pressures of 0.69, 0.78 and 0.87 (Tsat = 15, 20 and Acknowledgments
25 °C) and mass fluxes of 400, 600 and 800 kg m
1 s
1. A seg-
mented model developed by Garimella et al. [48] for R-134a, which The authors acknowledge support from the Office of Naval
accounted for the conjugate conduction and convection heat trans- Research and Dr. Mark Spector for this research.
fer effects in the copper test section, was modified to obtain the
CO2 heat transfer coefficients at fine vapor quality increments. By
accounting for minor pressure losses/gains due to expansions, con- Appendix A
tractions and bends in the test section inlet and outlet headers, the
frictional pressure drop in the microchannels was determined. The See Tables A1 and A2.

Table A1
Pressure drop correlations summary.

Correlation Error Dh = 0.10 mm Dh = 0.13 mm Dh = 0.15 mm Dh = 0.16 mm

AR = 1 (%) AR = 2 (%) AR = 3 (%) AR = 4 (%)
Homogeneous model AD
69
71
68
80
AAD 69 71 68 80
Lockhart and Martinelli [36] AD
47
46
39
61
AAD 47 46 43 61
Cavallini et al.[25] AD
69
74
73
83
AAD 69 74 73 83
Andresen et al. [13] AD
57
63
60
76
AAD 57 63 60 76
Mishima and Hibiki [57] AD
82
85
82
89
AAD 82 85 82 89
Lee and Lee [58] AD
45
55
52
70
AAD 45 55 52 70
Agarwal and Garimella [8] AD 15 11 19
27
AAD 16 12 19 27
Garimella et al. [59] AD
22
7 39 19
AAD 22 11 39 22
 B.M. Fronk, S. Garimella / International Journal of Heat and Mass Transfer 100 (2016) 150–164 163

Table A2
Heat transfer correlation summary.

Correlation Error Dh = 0.10 mm Dh = 0.13 mm Dh = 0.15 mm Dh = 0.16 mm

AR = 1 AR = 2 (%) AR = 3 (%) AR = 4 (%)
Traviss et al. [62] AD 1 15 15 1
AAD 14 25 20 14
Shah [23] AD 14 31 33 14
AAD 22 38 35 23
Dobson and Chato [20] AD
2 13 14
1
AAD 15 25 20 14
Cavallini et al.[25] AD
42
39
40
46
AAD 42 39 40 46
Wang et al. [61] AD
13
5
3
18
AAD 25 31 24 26
Koyama et al. [29] AD
55
49
48
54
AAD 55 49 48 54
Cavallini et al.[63] AD
39
31
31
43
AAD 40 37 33 43
Agarwal et al. [11] AD 7 20 20 5
AAD 16 26 21 17

References [20] M.K. Dobson, J.C. Chato, Condensation in smooth horizontal tubes, J. Heat
Transfer 120 (1) (1998) 193–213.
[21] H. Jaster, P.G. Kosky, Condensation heat-transfer in a mixed flow regime, Int. J.
[1] D. Chisholm, A theoretical basis for the Lockhart–Martinelli correlation for
Heat Mass Transfer 19 (1) (1976) 95–99.
two-phase flow, Int. J. Heat Mass Transfer 10 (12) (1967) 1767–1778.
[22] J.C. Chato, Laminar film condensation inside horizontal and inclined tubes,
[2] P.A. Kew, K. Cornwell, Correlations for the prediction of boiling heat
ASHRAE J. 4 (1962) 52–60.
transfer in small-diameter channels, Appl. Therm. Eng. 17 (8–10) (1997)
[23] M.M. Shah, A general correlation for heat transfer during film condensation
705–715.
inside pipes, Int. J. Heat Mass Transfer 22 (4) (1979) 547–556.
[3] A. Serizawa, Z. Feng, Z. Kawara, Two-phase flow in microchannels, Exp. Therm.
[24] M.M. Shah, An improved and extended general correlation for heat transfer
Fluid Sci. 26 (6–7) (2002) 703.
during condensation in plain tubes, HVAC&R Res. 15 (5) (2009) 889–913.
[4] J.W. Coleman, S. Garimella, Characterization of two-phase flow patterns in
[25] A. Cavallini, G. Censi, D. Del Col, L. Doretti, G.A. Longo, L. Rossetto,
small diameter round and rectangular tubes, Int. J. Heat Mass Transfer 42 (15)
Condensation of halogenated refrigerants inside smooth tubes, HVAC&R Res.
(1999) 2869–2881.
8 (4) (2002) 429–451.
[5] J.W. Coleman, S. Garimella, Two-phase flow regime transitions in
[26] J.R. Thome, J. El Hajal, A. Cavallini, Condensation in horizontal tubes, Part 2:
microchannel tubes: the effect of hydraulic diameter, in: ASME Heat Transfer
New heat transfer model based on flow regimes, Int. J. Heat Mass Transfer 46
Division – 2000, American Society of Mechanical Engineers, Orlando, FL, 2000,
(18) (2003) 3365–3387.
pp. 71–83.
[27] J.R. Baird, D.F. Fletcher, B.S. Haynes, Local condensation heat transfer rates in
[6] J.W. Coleman, S. Garimella, Two-phase flow regimes in round, square and
fine passages, Int. J. Heat Mass Transfer 46 (23) (2003) 4453–4466.
rectangular tubes during condensation of refrigerant R134a, Int. J. Refrig. 26 (1)
[28] M.M. Derby, H.J. Lee, R.C. Craft, G.J. Michna, Y. Pelese, M.K. Jensen, Exploration
(2003) 117–128.
of experimental techniques to determine the condensation heat flux in
[7] S. Garimella, A. Agarwal, J.D. Killion, Condensation pressure drop in circular
microchannels and minichannels, in: 14th International Heat Transfer
microchannels, Heat Transfer Eng. 26 (3) (2005) 28–35.
Conference, ASME, Washington, DC, 2010.
[8] A. Agarwal, S. Garimella, Modeling of pressure drop during condensation in
[29] S. Koyama, K. Kuwahara, K. Nakashita, K. Yamamoto, An experimental study
circular and noncircular microchannels, J. Fluids Eng. 131 (1) (2009) 011302–
on condensation of refrigerant R134a in a multi-port extruded tube, Int. J.
011308.
Refrig. 26 (4) (2003) 425–432.
[9] S. Garimella, T.M. Bandhauer, Measurement of condensation heat transfer
[30] S. Garimella, Condensation in Minichannels and Microchannels, in: S.G.
coefficients in microchannel tubes, in: 2001 ASME International Mechanical
Kandlikar (Ed.), Heat Transfer and Fluid Flow in Minichannels and
Engineering Congress and Exposition, American Society of Mechanical
Micochannels, Elsevier, 2014.
Engineers, New York, NY, 2001, pp. 243–249.
[31] S. Garimella, B.M. Fronk, Condensation Heat Transfer, 1st ed., World Scientific
[10] T.M. Bandhauer, A. Agarwal, S. Garimella, Measurement and modeling of
Publishing, 2015.
condensation heat transfer coefficients in circular microchannels, J. Heat
[32] M.M. Awad, A.S. Dalkiliç, S. Wongwises, A Critical Review on Condensation
Transfer 128 (10) (2006) 1050–1059.
Heat Transfer in Microchannels and Minichannels, J. Nanotechnol. Eng. Med. 5
[11] A. Agarwal, T.M. Bandhauer, S. Garimella, Measurement and modeling of
(1) (2014) (010904-010904).
condensation heat transfer in non-circular microchannels, Int. J. Refrig. 33 (6)
[33] A. Kawahara, P.M.Y. Chung, M. Kawaji, Investigation of two-phase flow
(2010) 1169–1179.
pattern, void fraction and pressure drop in a microchannel, Int. J. Multiphase
[12] Y. Jiang, B. Mitra, S. Garimella, U.C. Andresen, Measurement of condensation
Flow 28 (9) (2002) 1411–1435.
heat transfer coefficients at near-critical pressures in refrigerant blends, J. Heat
[34] P.M.Y. Chung, M. Kawaji, The effect of channel diameter on adiabatic two-
Transfer 129 (8) (2007) 958–965.
phase flow characteristics in microchannels, Int. J. Multiphase Flow 30 (7–8)
[13] U.C. Andresen, S. Garimella, B. Mitra, Y. Jiang, B.M. Fronk, Pressure drop during
(2004) 735–761.
near-critical-pressure condensation of refrigerant blends, Int. J. Refrig. 59
[35] P.M.Y. Chung, M. Kawaji, A. Kawahara, Y. Shibata, Two-phase flow through
(2015) 1–13.
square and circular microchannels - Effects of channel geometry, J. Fluids Eng.
[14] S. Garimella, U.C. Andresen, B. Mitra, Y. Jiang, B.M. Fronk, Heat transfer during
126 (4) (2004) 546–552.
near-critical-pressure condensation of refrigerant blends, J. Heat Transfer 138
[36] R. Lockhart, R. Martinelli, Proposed correlation of data for isothermal two-
(5) (2015).
phase, two-component flow in pipes, Chem. Eng. Prog. 45 (1) (1949) 39–48.
[15] S. Garimella, B. Mitra, U.C. Andresen, Y. Jiang, B.M. Fronk, Heat transfer and
[37] O. Iqbal, P. Bansal, In-tube condensation heat transfer of CO2 at low
pressure drop during supercritical cooling of HFC refrigerant blends, Int. J.
temperatures in a horizontal smooth tube, Int. J. Refrig. 35 (2) (2012) 270–277.
Heat Mass Transfer 91 (2015) 477–493.
[38] P. Kang, J. Heo, R. Yun, Condensation heat transfer characteristics of CO2 in a
[16] A. Agarwal, S. Garimella, Representative results for condensation
horizontal tube, Int. J. Refrig. 36 (3) (2013) 1090–1097.
measurements at hydraulic diameters 100 microns, J. Heat Transfer 132 (4)
[39] X.L. Huai, S. Koyama, An experimental study of carbon dioxide condensation in
(2010) 1–12.
mini channels, J. Therm. Sci. 13 (4) (2004) 358–365.
[17] B.M. Fronk, S. Garimella, Measurement of heat transfer and pressure drop
[40] C.Y. Park, P. Hrnjak, CO2 flow condensation heat transfer and pressure drop in
during condensation of carbon dioxide in microscale geometries, in: 14th
multi-port microchannels at low temperatures, Int. J. Refrig. 32 (6) (2009)
International Heat Transfer Conference, ASME, Washington, DC, 2010.
1129–1139.
[18] G. Lorentzen, Revival of carbon dioxide as a refrigerant, Int. J. Refrig. 17 (5)
[41] W.W. Akers, H.A. Deans, O.K. Crosser, Condensation heat transfer within
(1994) 292–301.
horizontal tubes, Chemical Engineering Progress Symposium Series, 55, 1959,
[19] A. Cavallini, C. Zilio, Carbon dioxide as a natural refrigerant, Int. J. Low Carbon
pp. 171–176 (29).
Technol. 2 (3) (2007) 225–473.
164 B.M. Fronk, S. Garimella / International Journal of Heat and Mass Transfer 100 (2016) 150–164

[42] W.H. McAdams, W.K. Woods, R.L. Bryan, Vaporization inside horizontal tubes- [54] S.W. Churchill, Comprehensive correlating equations for heat, mass and
II-benzene-oil-mixtures, Trans. ASME 64 (1942). momentum-transfer in fully developed flow in smooth tubes, Ind. Eng. Chem.
[43] Y.J. Kim, J. Jang, P.S. Hrnjak, M.S. Kim, Condensation heat transfer of carbon Fundam. 16 (1) (1977) 109–116.
dioxide inside horizontal smooth and microfin tubes at low temperatures, J. [55] B.N. Taylor, C.E. Kuyatt, Guidelines for evaluating and expressing the
Heat Transfer 131 (2) (2009) 021501–021510. uncertainty of NIST measurement results, National Institute of Standards
[44] J. Heo, H. Park, R. Yun, Comparison of condensation heat transfer and pressure and Technology, Washington, DC, USA, 1994.
drop of CO2 in rectangular microchannels, Int. J. Heat Mass Transfer 65 (2013) [56] H.S. Wang, J.W. Rose, A theory of film condensation in horizontal noncircular
719–726. section microchannels, J. Heat Transfer 127 (10) (2005) 1096–1105.
[45] J. Heo, H. Park, R. Yun, Condensation heat transfer and pressure drop [57] K. Mishima, T. Hibiki, Some characteristics of air–water two-phase flow in
characteristics of CO2 in a microchannel, Int. J. Refrig. 36 (6) (2013) 1657– small diameter vertical tubes, Int. J. Multiphase Flow 22 (4) (1996) 703–712.
1668. [58] H.J. Lee, S.Y. Lee, Pressure drop correlations for two-phase flow within
[46] Z. Zhang, Z.L. Weng, T.X. Li, Z.C. Huang, X.H. Sun, Z.H. He, J. van Es, A. Pauw, E. horizontal rectangular channels with small heights, Int. J. Multiphase Flow 27
Laudi, R. Battiston, CO2 condensation heat transfer coefficient and pressure (5) (2001) 783–796.
drop in a mini-channel space condenser, Exp. Therm. Fluid Sci. 44 (2013) 356– [59] S. Garimella, A. Agarwal, B.M. Fronk, The Intermittent and annular flow
363. condensation continuum: pressure drops at the microscale, Int. J. Multiphase
[47] M.M. Shah, Prediction of heat transfer during condensation of carbon dioxide Flow (2016), http://dx.doi.org/10.1016/j.ijmultiphaseflow.2016.03.023.
in channels, Appl. Therm. Eng. 93 (2016) 192–199. [60] K. Triplett, S. Ghiaasiaan, S. Abdel-Khalik, A. LeMouel, B. McCord, Gas–liquid
[48] S. Garimella, A. Agarwal, B.M. Fronk, Condensation heat transfer in two-phase flow in microchannels: Part II: Void fraction and pressure drop, Int.
rectangular microscale geometries, Int. J. Heat Mass Transfer 100 (2016) 98– J. Multiphase Flow 25 (3) (1999) 395–410.
110. [61] W.-W. Wang, T.D. Radcliff, R.N. Christensen, A condensation heat transfer
[49] R. Span, W. Wagner, A new equation of state for carbon dioxide covering the correlation for millimeter-scale tubing with flow regime transition, Exp.
fluid region from the triple-point temperature to 1100 K at pressures up to Thermal Fluid Sci. 26 (5) (2002) 473–485.
800 MPa, J. Phys. Chem. 25 (6) (1996). [62] D.P. Traviss, W.M. Rohsenow, A.B. Baron, Forced-convection condensation
[50] G.F. Hewitt, G.L. Shires, T.R. Bott, Process Heat Transfer, CRC Press, New York, inside tubes: a heat transfer equation for condenser design, ASHRAE Trans. 79
1994. (Part 1) (1973) 157–165.
[51] V. Carey, Liquid–Vapor Phase-Change Phenomena, 2nd ed., Taylor & Francis [63] A. Cavallini, D. Del Col, L. Doretti, M. Matkovic, L. Rossetto, C. Zilio, A model for
Group, New York, 2008. condensation in minichannels, in: 2005 ASME Summer Heat Transfer
[52] C.J. Baroczy, Correlation of liquid fraction in two-phase flow with application Conference, ASME, San Francisco, 2005, pp. 297–304.
to liquid metals, Chemical Engineering Progress Symposium Series, 61, 1965, [64] S. Koyama, K. Kuwahara, K. Nakashita, Condensation of refrigerant in a multi-
pp. 179–191 (57). port channel, in: 1st International Conference on Microchannels and
[53] S.A. Klein, Engineering equation solver, in, F-Chart Software, 2008. Minichannels, ASME, Rochester, NY, USA, 2003.