i n t e r n a t i o n a l j o u r n a l o f r e f r i g e r a t i o n 3 4 ( 2 0 1 1 ) 1 7 e2 8

available at www.sciencedirect.com

w w w . i i fi i r . o r g

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

Water-coupled carbon dioxide microchannel gas cooler for

heat pump water heaters: Part II e Model development and
validation

Brian M. Fronk, Srinivas Garimella*

Sustainable Thermal Systems Laboratory, George W. Woodruff School of Mechanical Engineering, Georgia Institute of Technology,
Atlanta, GA 30332, USA

article info abstract

Article history: An experimental and analytical study on the performance of a compact, microchannel
Received 15 January 2010 water-carbon dioxide (CO2) gas cooler was conducted. The experimental results addressed
Received in revised form in Part I of this study are used here in Part II to develop an analytical model, utilizing
19 April 2010 a segmented approach to account for the steep gradients in the thermodynamic and
Accepted 19 May 2010 transport properties of supercritical CO2. The model predicted gas cooler heat duty with an
Available online 8 June 2010 average absolute deviation of 7.5% with varying refrigerant and water inlet conditions. The
segmented model reveals that near the pseudo-critical point, there is a significant local
Keywords: decrease in refrigerant-side thermal resistance, which yields a sharp increase in local heat
Heat exchanger duty. The impact of this spike on gas cooler performance is analyzed. Results from this
Gas cooler study can be used to predict the effect of changing geometric parameters of the heat
Carbon dioxide exchanger without the need for expensive prototype development and testing.
Transcritical cycle ª 2010 Elsevier Ltd and IIR. All rights reserved.
Microchannel
Heat pump
Heating
Hot water
Modelling
Simulation

Refroidisseur de gaz à microcanaux sur eau souterraine pour

des pompes à chaleur au dioxyde de carbone utilisées pour
chauffer de l’eau sanitaire : Partie II – Développement et
validation du modèle
Mots clés : échangeur de chaleur ; refroidisseur de gaz ; dioxyde de carbone ; cycle transcritique ; micro-canal ; pompe à chaleur ;
chauffage ; eau sanitaire ; modélisation - simulation

* Corresponding author.
E-mail address: sgarimella@gatech.edu (S. Garimella).
0140-7007/$ e see front matter ª 2010 Elsevier Ltd and IIR. All rights reserved.
doi:10.1016/j.ijrefrig.2010.05.012
18 i n t e r n a t i o n a l j o u r n a l o f r e f r i g e r a t i o n 3 4 ( 2 0 1 1 ) 1 7 e2 8

Nomenclature Q_ heat duty (kW)

2 R thermal resistance (K kW1)
A heat transfer area (m )
Re Reynolds number ()
Ac fin flow area (m2)
T temperature ( C)
Cmin minimum thermal capacitance rate (kW K1)
D tube diameter (m) Greek symbols
Dh fin hydraulic diameter (m) a fin geometric parameter ()
G mass flux (kg m2 s1) b fin geometric parameter ()
f friction factor () d fin geometric parameter ()
h heat transfer coefficient (kW m2 K1) 3 Effectiveness ()
j colburn factor () r Density (kg m3)
L length (m)
Subscripts
l fin length (m)
b bulk
m _ mass flow rate (g s1)
eff effective
Nu Nusselt number ()
ref refrigerant-side
P pressure (kPa)
seg local segment
Pr Prandtl number ()

1. Introduction reasonable accuracy. Cheng et al. (2008) present an extensive

review of supercritical CO2 correlations for heat transfer
A desire to utilize more environmentally friendly refrigerants coefficient and friction factor in macro and microchannels.
has spurred interest in the development of vapor compression Based on the experimental results of several researchers, they
systems with carbon dioxide (CO2) as the working fluid. CO2 has conclude that there is no consensus for a best correlation for
a low global warming potential (GWP ¼ 1), favorable transport predicting heat transfer and pressure drop, absent further
properties, low cost, high availability, and is nonflammable experimental data. Many authors (Yin et al., 2001; Cecchinato
and non-toxic. Unlike conventional refrigerants, CO2 has et al., 2005; Hwang et al., 2005) have predicted the heat
a relatively low critical temperature and pressure transfer coefficient of supercritical CO2 using a constant
(31.1  C/73.7 bar). Therefore, under most conditions of interest, property, single-phase model such as the Gnielinski (1976)
the system operates as a transcritical cycle. Rather than correlation. The Gnielinski correlation is valid for turbulent
a constant temperature condensation process, the supercrit- flows with 2300 < Re < 5  106 and 0.5 < Pr < 2000. For
ical CO2 exhibits a temperature glide through the component supercritical flows, this correlation has been shown to under
known as the gas cooler. predict the heat transfer coefficient, particularly near the
One of the most promising applications of CO2 heat pump pseudo-critical temperature (Wang and Hihara, 2002). At low
devices is the provision of domestic hot water, either as Re near the critical point and in heated upward flow, the
a stand-alone water heater, or in conjunction with a chiller property differences between the wall and bulk fluid become
function. The high temperature lifts required in water heating significant (Pettersen et al., 1998). Other correlations account
match well with the temperature glide exhibited by the for these effects. In a critical review of supercritical CO2 heat
supercritical CO2. The glide allows water delivery tempera- transfer coefficients, Pitla et al. (1998) discuss the correlation
tures of up to 90  C without significant degradation in system developed by Krasnoshchekov et al. (1970) for supercritical gas
efficiency (Kim et al., 2004). Attempting to heat water to this cooling in horizontal tubes. This correlation addresses the
temperature with a conventional system (e. g., R134a) could effects of the difference between bulk and wall temperatures
only be accomplished by raising the compressor discharge on the heat transfer coefficient. When the tube wall temper-
pressure to a high magnitude to eliminate temperature ature is below the critical temperature of the fluid, the pre-
pinches. In doing so, the available enthalpy difference across dicted heat transfer coefficient using the property corrections
the vapor-liquid dome decreases, and the compressor pres- is seen to increase compared to that of a constant property
sure ratio increases, drastically reducing system efficiency. single-phase fluid. An experimental investigation of super-
In addition to the temperature glide, CO2 thermodynamic critical CO2 cooling by Baskov et al. (1977) found the effect of
and transport properties vary significantly in the region near free convection to be negligible at high Re. They proposed
the critical point. These variations have a significant impact a correlation for cooling of supercritical CO2 in vertical tubes.
on local heat transfer and must be accounted for. Due to these Pitla et al. (1998) compared the Krasnoshchekov et al. (1970);
property variations, analyzing the gas cooler through an Baskov et al. (1977), and a numerically derived correlation
overall log mean temperature difference (LMTD) or effective- proposed by Petrov and Popov (1985) with the correlation by
ness-NTU method yields only approximate estimates, and Petukhov and Kirillov (1958). The correlations were plotted at
a segmented approach must be utilized to achieve accurate a mass flow rate of 30 g s1, pressure of 100 bar, refrigerant
performance predictions. temperature of 32e120  C and a heat sink temperature of
A method of evaluating the supercritical CO2 heat transfer 17e32  C. The tube under consideration had an ID of 4.572 mm
coefficient is required to develop gas cooler models of and an OD of 6.35 mm. The authors show that throughout the
 i n t e r n a t i o n a l j o u r n a l o f r e f r i g e r a t i o n 3 4 ( 2 0 1 1 ) 1 7 e2 8 19

range of carbon dioxide temperatures, the Baskov et al. (1977) friction factor correlation of Kuraeva and Protopopov (1974).
and Petrov and Popov (1985) correlations are in good agree- Additionally, entrance and compressibility effects were
ment. The Krasnoshchekov et al. (1970) correlation is in good considered. The individual model was validated with experi-
agreement when the carbon dioxide temperature is outside mental test data from Zhao et al. (2001), which featured
the pseudo-critical range. microchannel tubes with ten 1.0 mm diameter ports per tube.
Many researchers have modeled the performance of water- The model predicted heat duty within 3% of CO2-side
and air-coupled gas coolers for use in split-system AC/heat measurements and 5% of air-side measurements, with poorer
pumps, mobile air conditioning systems and water heating agreement at low refrigerant mass fluxes attributed to the
systems using a variety of heat transfer and pressure drop potential inapplicability of the modified Gnielinski
correlations. A majority of the modeling studies are utilized in correlation.
system level models for simulating overall performance. Most Kim et al. (2005) developed a model of a water-coupled gas
researchers employ a finite volume approach to account for cooler for use in a system model for evaluating the perfor-
the changing supercritical CO2 properties, particularly near mance of a transcritical system with a suction line heat
the pseudo-critical point. A brief overview of modeling tech- exchanger in water heating applications. The modeled gas
niques and comparisons with experimental data is provided cooler was a tube-in-tube counterflow geometry, with refrig-
here. Yin et al. (2001) developed a segmented model of an air- erant tube ID ¼ 7.5 mm, and water tube ID ¼ 14.9 mm. The gas
coupled, parallel-serpentine gas cooler consisting of three cooler was divided into equal length segments, with the local
passes of 13, 11 and 10 microchannel tubes with louvered fins. heat duty calculated using the LMTD method. Refrigerant heat
There were 11 circular ports per tube with ID ¼ 0.79 mm. transfer coefficient and pressure drop were calculated from
Refrigerant heat transfer coefficient in each segment was the Krasnoshchekov et al. (1970) and Petrov and Popov (1985)
calculated from the Gnielinski (1976) correlation and pressure correlations. The model was not independently validated;
drop from the Churchill (1977) correlation. Conduction however, they reported that the system model predicted
between tubes was neglected, and the incoming air was system heat duty with an absolute mean deviation of 3.56%.
assumed to have uniform velocity and temperature. Local Cecchinato et al. (2005) developed a similar segmented model
heat duty in each segment was calculated based on the log of a water-coupled, tube-in-tube gas cooler for use in a system
mean temperature difference (LMTD) method. The simulation model. The simulated refrigerant tube had an ID ¼ 6.5 mm and
model was validated with 358 data points corresponding to 48 the outer water tube ID varied from 15.0 to 30.1 mm,
different indoor/outdoor operating conditions. They showed depending on test condition. Simulations were carried out
agreement in predicted heat duty within 2%, while refrig- comparing CO2 transcritical cycle performance to R134a
erant pressure drop was systematically under predicted by performance for different operating conditions. No experi-
approximately a factor of 3, which was attributed to potential mental validation was presented in the study. Sarkar et al.
manufacturing defects in the microchannel tubes. Garimella (2006) reported a gas cooler component model that was part
(2002) presented a model for a near-counter flow air-coupled of a larger system model for predicting transcritical cycle
CO2 gas cooler using serpentine refrigerant tubes and louvered performance in simultaneous water heating and cooling
fins on the air-side. The geometry required the tracking of operations. Again, the gas cooler under consideration was
refrigerant and air temperature along both the length and a water-coupled counterflow tube-in-tube heat exchanger,
width of each refrigerant tube. The simulated gas cooler had with an inner tube OD of 6.35 mm and thickness of 0.8 mm,
36 tubes with six 1.905 mm diameter circular ports per tube. outer tube OD of 12 mm and thickness of 1 mm. The model
The effectiveness-NTU method was employed to obtain the was discretized and local heat duty found from the LMTD
local heat duty in each segment. The Krasnoshchekov et al. method. The correlation proposed by Pitla et al. (2002) for
(1970) correlation was used to predict the local heat transfer refrigerant-side heat transfer coefficient and the Petrov and
coefficient. With a volumetric air flow rate of 0.334 m3 s1 and Popov (1985) equation for friction factor were used. The gas
a refrigerant mass flow rate of 31 g s1, the model predicted cooler model was not individually validated using compo-
a heat duty of 6.97 kW and an approach temperature differ- nent-level measurements; however, the overall system model
ence of 5.3 K. Garimella (2003) extended this analysis to showed agreement with a 15% maximum deviation in pre-
include axial conduction losses due to heat transfer between dicted COP, with poorest agreement at low compressor
adjacent tubes through the louvered fins, and demonstrated discharge pressure (<80 bar) in data presented by Sarkar et al.
a reduction in heat duty of 13% at an unlouvered fin fraction of (2009).
30%. Ortiz et al. (2003) developed and incorporated Chang and Kim (2007) developed a simulation model of an
a segmented model of an air-coupled microchannel gas cooler air-coupled gas cooler consisting of two or three banks of
into a split-system simulation model. The gas cooler was serpentine tubes connected in series. Each bank consisted of
a cross flow, extruded aluminum design. The heat duty in between eight and twelve 7 mm OD microfin tubes with
each segment is solved in iterative fashion, where an initial louvered fins. The heat exchanger under investigation was
tube wall temperature is estimated, and then heat duty found a multi-pass cross flow configuration. They treated each tube
from the effectiveness-NTU method and from evaluating the in each bank as a cross flow heat exchanger, and assumed
local UADT. Refrigerant heat transfer coefficient is calcu- a constant CO2 heat transfer coefficient based on the single-
lated from the modified Gnielinski correlation (Pettersen et al., phase correlation of Han and Lee (2005) for microfin tubes. The
2000) at high mass flux (G > 350 kg m2 s1) and the Petrov and correlation was developed from data on water for tubes with
Popov (1985) correlation at lower mass fluxes. Refrigerant an OD from 5.1 to 9.52 mm. They theorized that since the air-
pressure drop was calculated in each segment using the side resistance dominated heat transfer, uncertainties in the
20 i n t e r n a t i o n a l j o u r n a l o f r e f r i g e r a t i o n 3 4 ( 2 0 1 1 ) 1 7 e2 8

refrigerant-side would have a negligible effect on heat

exchanger performance. At a refrigerant mass flow rate of
70 g s1, the predicted heat duty varied from 2 to 4 kW, with
variations from experimental data within 10%. Koyama et al.
(2009) developed a model of an air-coupled cross-counter flow
fin-tube heat exchanger. The ID of the simulated refrigerant
tubes was 4.75 mm. The heat exchanger was divided into
i  j  k segments to account for varying refrigerant and air
properties and temperature. Each segment was treated as
a cross flow heat exchanger, where the local heat duty was
calculated as a function of the segment UA and average
refrigerant and air temperature difference. The refrigerant
heat transfer coefficient was determined from the Dang and
Hihara (2004) correlation for supercritical CO2 and the fric-
tion factor from the Blasius equation. The component model Fig. 1 e Gas cooler photograph.
was incorporated into a system simulation model. The system
model was compared qualitatively with CO2 refrigeration
systems, however no explicit validation of the gas cooler
2008) platform. The model was then validated with data from
component model was conducted.
a water-coupled CO2 heat pump test facility presented in Part I.
Many researchers have employed a segmented approach to
Fig. 1 shows a photograph and Fig. 2 a cross-sectional
modeling CO2 gas coolers using various correlations for pre-
schematic of the gas cooler under investigation. Water enters
dicting heat transfer coefficient and friction factor. Much of
through one side of the heat exchanger and makes several
the work has focused on cross flow and near-counter flow air-
serpentine passes before exiting the heat exchanger. The
coupled gas coolers, and relatively simple tube-in-tube
water channels contain offset strip fins to enhance heat
counterflow water-coupled heat exchangers. There has been
transfer and provide structural stability. Refrigerant enters
comparatively less work on compact microchannel based
the inlet header as shown, is divided between 16 refrigerant
CO2-water gas coolers. The thermal resistance of the water-
tubes (each containing four 0.89 mm diameter circular chan-
side of a CO2-water gas cooler will be much less than that of
nels), and serpentines between each water channel before
the air-side in an air-coupled heat exchanger. Thus, accurate
exiting on the same side as the water inlet. Thus, the fluid
prediction of the heat transfer performance of the CO2 side
motion is generally counterflow, although the local heat
assumes much greater importance. Generally, investigators
transfer occurs in a cross flow configuration. Table 1 shows
have modeled the CO2-water gas cooler as a simple tube-in-
the primary dimensions of a seven-plate gas cooler, which is
tube geometry for use in cycle simulation models (Cecchinato
defined as a gas cooler with seven water passes and six
et al., 2005; Kim et al., 2005; Sarkar et al., 2006; Laipradit et al.,
refrigerant passes.
2008). While adequate for system modeling, a tube-in-tube
The heat exchanger was modeled using a segmented
heat exchanger is not an optimized geometry for an actual
element approach, in which the water passage and the
water heating system. A compact, counterflow gas cooler is
refrigerant tubes were divided into equal sized nodes. A
the key enabling component to minimize heat exchanger
schematic representation of the nodes is shown in Fig. 3.
material use and system footprint, while taking advantage of
Refrigerant nodes are shaded. Each node transfers heat with
the unique water heating capabilities of transcritical cycles.
In the present work, therefore, an analysis of the heat
transfer mechanisms of a water-coupled gas cooler with
a compact, multi-pass cross-counter flow design is conducted.
A segmented model for an aluminum brazed plate, micro-
channel CO2 gas cooler is developed, validated with experi-
mental data presented in companion paper Part I, and used to
parametrically evaluate the impacts of changing inlet water
and refrigerant conditions. The gas cooler is a multi-pass
cross-counterflow, water-coupled heat exchanger intended
for heating water. The heat exchanger is composed of several
finned plates that function as water passes and multiple
microchannel refrigerant tubes. Five-, seven- and a simulated
twelve-plate heat exchanger constructed by connecting the
two in series were tested and modeled.

2. Model development

A segmented model of a water-coupled CO2 gas cooler was

developed using the Engineering Equation Solver (EES) (Klein, Fig. 2 e Gas cooler cross section schematic.
 i n t e r n a t i o n a l j o u r n a l o f r e f r i g e r a t i o n 3 4 ( 2 0 1 1 ) 1 7 e2 8 21

effectiveness was determined as a function of the NTU,

Table 1 e Seven-plate gas cooler dimensions.
assuming that the segment could be approximated as a cross
Overall dimensions flow heat exchanger with both fluids unmixed. Local heat duty
Number of refrigerant tubes 16 was then determined from Eq. (1), using the water and
Number of water passes 7 refrigerant segment inlet temperatures.
Gas cooler length 191 mm


Gas cooler width 54.0 mm Q_ seg ¼ 3$Cmin Tref;in  Twater;in (1)
Gas cooler height 84.0 mm
To determine the UA value used in the 3-NTU calculations, the
Refrigerant side
local heat transfer area and heat transfer coefficients, and
Total refrigerant tube length 516 mm thermal resistance of the aluminum tube wall were deter-
Refrigerant pass tube length 81.0 mm
mined. The resulting formulation of UA is shown in Eq. (2).
Number of channels per tube 4
The effective water-side heat transfer coefficient was a func-
Channel diameter 0.89 mm
Tube width 6.35 mm tion of the local fin efficiency, which was generally predicted
Tube height 1.65 mm to be between 50 and 75%, depending on the water flow rate.
Tube wall thickness 0.38 mm
1
Tube web thickness 0.64 mm UA ¼ (2)
92,292 mm2 1 1
Refrigerant-side heat transfer area þ Rw þ
href Aref hwater Aeff;water
Water side
The water flow was assumed to be uniform through each
Fin height 6.41 mm
segment and fin section. The local heat transfer coefficient
Fin space 2.23 mm
Fin thickness 0.31 mm was calculated using the Manglik and Bergles correlation
Fin length 3.18 mm (1995) shown in Eq. (4). This correlation was developed for
Fin pitch 4.4 fins per cm determining the Colburn factor, j, for rectangular offset strip
Water-side heat transfer area 385,140 mm2 fin geometries such as that utilized in the gas cooler design
under consideration. The empirically based correlation
the two neighboring nodes. Water nodes on the two exterior considers data in the laminar, transition and turbulent
plates were assumed to have an adiabatic boundary on one regime, and utilizes an asymptotic matching technique,
side. Local heat transfer in each segmented node was solved resulting in one equation valid (Eq. (4)) for all flow regimes.
using the effectiveness-NTU approach for cross flow Thus, for the expected water-side Reynolds numbers in the
segments. The resulting system of coupled equations was current study (Re < 1000), the model is expected to be appli-
solved iteratively. The segment length was chosen in cable. The water-side Reynolds number is based on the
a manner in which property variation across the segment was hydraulic diameter as defined by Manglik and Bergles (1995),
not significant. For a seven-plate gas cooler, 30 segments per shown in Eq. (3).
tube were used, which yielded less than a 0.4% difference in 4Ac
predicted capacity when compared to a tube with 60 Dh ¼ (3)
A=l
segments.
The local heat duty was calculated using the effectiveness-
j ¼ 0:6522Re0:5403
water a
0:1541 0:1499 0:0678
d g
NTU approach, with NTU based on UA and the minimum  
5 1:340 0:504 0:456 1:055 0:1
thermal capacitance rate, Cmin of each segment. The local  1 þ 5:269  10 Rewater a d g ð4Þ

Fig. 3 e Segmented schematic.

22 i n t e r n a t i o n a l j o u r n a l o f r e f r i g e r a t i o n 3 4 ( 2 0 1 1 ) 1 7 e2 8

The refrigerant-side local heat transfer coefficient was calcu-

lated using the Gnielinski (1976) correlation (Eq. (5)), using
properties evaluated at the average segment temperature and
inlet pressure. The local Fanning friction factor ( f ) was found
using the Filonenko (1954) correlation (Eq. (6)).

ðf =8ÞðReD  1000ÞPrb
NuD ¼   (5)
1=2
1 þ 12:7ðf =8Þ Prb2=3  1

2
f ¼ ð0:79$lnðReÞ  1:64Þ (6)

The refrigerant Reynolds number is based on the local mass

flux, viscosity and circular channel diameter. The refrigerant
Reynolds numbers under consideration ranged from 3000 to
25,000, which are within the range of applicability of the
Gnielinski (1976) correlation. The refrigerant flow is not
expected to be purely laminar for any of the mass flow rates
investigated. The local refrigerant heat transfer coefficient
ranged from 1.5 to over 10 kW m2 K1, with the peak heat
transfer occurring at high mass fluxes in the pseudo-critical
region. The local heat duty was used to determine the outlet
enthalpy of the water and refrigerant for each segment. The
coupled equations were solved simultaneously using the Fig. 4 e Test facility schematic.
nonlinear solver included with EES.
Refrigerant and water pressure drop predictions were also
incorporated into the model. The supercritical CO2 was using a laboratory chilled water supply and an immersion
treated as a single-phase fluid for calculation of pressure drop electric heater bank. Compression was provided by one or two
as shown in Eq. (7). The friction factor in each segment was reciprocating compressors, depending on the desired mass
calculated using the Churchill (1977) correlation (Darcy flow rate. The compressors were lubricated by a 25 mL charge
version), which is applicable for the laminar, transitional and of polyolester (POE) oil. Thermocouples and pressure trans-
turbulent regimes. A tube roughness of 5 mm was assumed. ducers at the inlet and outlet of each component allowed the
determination of the thermodynamic state. For the gas cooler,
fG2 L
DP ¼ (7) the water and refrigerant inlet and outlet enthalpies were
2rD
determined from the measured temperature and pressure,
Minor losses, including microchannel tube bends, entrance using the equation of state developed by Span and Wagner
effects and exit effects were modeled based on the local flow (1996) for CO2. The heat duties for the water and refrigerant
conditions and the loss coefficient K as presented in White sides were determined from the enthalpy difference and the
(2003). Most of these coefficients were based on experiments measured mass flow rate of each fluid. The average of the
with larger tube diameters, and thus may not be applicable to refrigerant- and water-side heat duties was compared with
the geometry of the heat exchanger under investigation. the model predictions.
Water-side pressure drop was calculated using the Fanning Over 125 data points were obtained from the three different
friction factor correlation of Manglik and Bergles (1995) shown gas cooler sizes. Experiments were conducted at refrigerant
in Eq. (8), which is a function of the same dimensionless fin mass flow rates from 8 g s1 to 25 g s1, with the inlet
geometry parameters used to determine the Colburn factor for temperature varying from 85 to 115  C. High-side pressure
heat transfer predictions. ranged from 7900 to 11,000 kPa. The gas cooler water flow rate
was varied from 0.93 L min1 to 5.68 L min1, with inlet
f ¼ 9:6243Re0:7422
water a
0:1856 0:3053 0:2659
d g
  temperatures ranging from 5 to 20  C. Measured heat duties
 1 þ 7:669  108 Re4:429
water a 0:920 3:767 0:236
d g ð8Þ ranged from 2.0 to 6.5 kW, with an uncertainty of 30
to 140 W. Further description of the experimental setup and
results can be found in companion paper Part I.
3. Model validation and analysis The measured gas cooler refrigerant inlet temperature,
pressure and mass flow rate, and water inlet temperature and
The model described above was validated using experiments flow rate were used as model inputs for each data point. The
on the CO2 heat pump facility shown schematically in Fig. 4, predicted heat duty was compared with the measured
and described in greater detail in companion paper Part I. capacity to validate the model. A plot of predicted versus
Experiments were conducted with a five-, seven- and simu- measured capacity is shown in Fig. 5.
lated twelve-plate gas cooler. The twelve-plate gas cooler was The absolute average difference between measured and
formed by connecting the five- and seven-plate gas coolers in predicted heat duty was 7.5%. Agreement was poorest at the
series. The gas cooler and evaporator were coupled to separate lowest water flow rate (0.93 L min1), with the low water flow
closed water loops; with the water temperature regulated rates in the five-plate gas cooler showing the poorest
 i n t e r n a t i o n a l j o u r n a l o f r e f r i g e r a t i o n 3 4 ( 2 0 1 1 ) 1 7 e2 8 23

Fig. 6 e Heat duty vs. position.

Fig. 5 e Model results.

agreement with the model. Errors at the low water flow rate a locally varying heat transfer coefficient, the driving
may be attributed to possible water maldistribution and temperature difference between the water and refrigerant
uncertainty in the water flow rate measurement. Refrigerant varies through the heat exchanger, which results in a large
pressure drop was systematically under predicted by the variation in heat duty through the gas cooler. Fig. 6 shows the
model by an average factor of 3.5. This was consistent across variation in local heat duty and refrigerant temperature as
all water and refrigerant inlet conditions. This can potentially a function of position for a gas cooler with a water flow rate of
be attributed to the presence of compressor lubricant, which 0.93 L min1 and an inlet temperature of 5  C, and a refrigerant
increases the pressure drop. The ASHRAE standard (2006) for flow rate of 12 and 24 g s1, with an inlet temperature and
measuring proportion of lubricant in liquid refrigerant calls pressure of 100  C and 9000 kPa, respectively. The variation is
for a lubricant/refrigerant mixture to be drawn out of the plotted along a single representative serpentine refrigerant
system into an evacuated cylinder of known volume, where- tube (tube 8), which is a good approximation for the trends of
upon the refrigerant is boiled off and the remaining lubricant the entire heat exchanger.
is weighed. The high operating pressure of the experimental In Fig. 6, refrigerant segment 1 represents the refrigerant
system would require the use of a thick walled cylinder, and inlet, and segment 30 is the refrigerant outlet. Since the
require a large volume of refrigerant/lubricant to be removed refrigerant/water flow is in a generally counterflow orienta-
to resolve the lubricant weight with acceptable accuracy. Due tion, the temperature of the refrigerant is highest at the inlet,
to the small refrigerant charge in the water-coupled system, and lowest at the outlet, whereas the water is coolest at the
this was not attempted. However, based on operational refrigerant outlet and hottest at the refrigerant inlet. In the
experience at Modine (Hoehne, 2007), a lubricant circulation seven-plate gas cooler, the refrigerant tube makes six passes.
rate of 5% by mass was assumed to explore its effect on These passes are evident in Fig. 6, occurring every five
pressure drop. The two-phase Lockhart and Martinelli (1949) segments. The jump in heat duty between each pass can be
pressure drop correlation showed some promise in attributed to the different water-side boundary conditions of
improving the agreement between measured and estimated the refrigerant tube in each pass.
values with this lubricant mass fraction. However, without The observed spike in heat duty corresponds to the spike in
accurate measurement of the lubricant circulation rate and an local refrigerant heat transfer coefficient. The spike can be
appropriate pressure drop correlation developed for super- observed between segments 18e22 for the 12 g s1 mass flow
critical CO2 and POE lubricant, accurate pressure drop and segments 25e30 for the 24 g s1 cases. As the CO2
modeling remains a challenge.
The segmented model, validated with experimental
results, can be used as a design tool for predicting gas cooler
performance with varying refrigerant and water inlet condi-
tions. The analysis presented here considers a seven-plate gas
cooler, with refrigerant inlet temperature varying from 70 to
100  C, refrigerant flow rates of 12e24 g s1, refrigerant inlet
pressures of 9000e12,000 kPa, water inlet temperatures from 5
to 65  C and water flow rates of 0.93e5.68 L min1. The
dimensions of the modeled gas cooler are the same as those
used in the experimental study, Part I (Table 1).
The segmented model evaluates the local heat transfer
coefficient and heat transfer area for the water and refrigerant
sides. In a supercritical gas cooler, the properties of the fluid
change drastically near the pseudo-critical point as the fluid
transitions from a “vapor-like” to “liquid-like” fluid, resulting Fig. 7 e Refrigerant heat transfer coefficient vs.
in an enhanced heat transfer coefficient. In addition to temperature.
24 i n t e r n a t i o n a l j o u r n a l o f r e f r i g e r a t i o n 3 4 ( 2 0 1 1 ) 1 7 e2 8

temperature approaches the pseudo-critical temperature, the

specific heat and other transport properties change drasti-
cally, yielding the spike in heat transfer coefficient. The
temperature corresponding to the pseudo-critical tempera-
ture increases as the refrigerant pressure is increased. These
effects are illustrated in Fig. 7, which shows a plot of CO2 heat
transfer coefficient at a mass flux of 500 kg m2 s1, in
a circular channel with ID ¼ 0.89 mm, at three different
pressures and temperatures ranging from 0 to 100  C. The heat
transfer coefficient is calculated using the Gnielinski (1976)
correlation. As seen in Fig. 7, at a given pressure, the heat
transfer coefficient spike is a function of the refrigerant
temperature. Thus, as mass flow rate increases, the temper-
Fig. 9 e Refrigerant/water resistance ratio.
ature change across the heat exchanger decreases and the
spike shifts towards the refrigerant outlet. Further increasing
the mass flow (and decreasing the temperature glide) would
result in this spike not being observed, as the refrigerant and water-side thermal resistances are defined as a function
would not cool into the pseudo-critical region. of heat transfer coefficient and effective heat transfer area as
For the water-coupled CO2 gas cooler under consideration, shown in Eqs. (9) and (10).
the refrigerant-side resistance is generally the limiting factor.
This is illustrated in Figs. 8 and 9. Fig. 8 shows the diminishing 1
Rref ¼ (9)
returns in overall UA with increasing water flow rate for href Aref
refrigerant mass flow rates of 12 g s1 and 24 g s1. The
reported UA is the summation of the UA of each segment in 1
Rwater ¼ (10)
hwater Awater
the gas cooler model. The water flow rate is varied from 0.93 to
9.5 L min1 at an inlet temperature of 5  C. Increasing the Fig. 9 further confirms the conclusions from Fig. 8 that the
water flow rate to 150% from 0.93 L min1 to 2.38 L min1 refrigerant-side thermal resistance in general limits the heat
results in a 28% increase in UA for the 12 g s1 mass flow rate, transfer process for both mass flows examined. While Fig. 9
and 31% for the 24 g s1 mass flow rate. However, increasing shows the resistance ratio at a water flow of 0.93 L min1,
the water flow 66%, from 5.68 to 9.5 L min1, results in an Fig. 8 shows that the ratio will only become increasingly
increase in UA of only 1 and 4% for the 12 g s1 and 24 g s1 limited by the refrigerant-side resistance as water flow rate
refrigerant flow rates, respectively. For a fixed water flow rate, increases. Thus, increasing the water flow rate would yield
doubling the refrigerant mass flow rate from 12 g s1 to 24 g s1 little improvement in heat transfer performance. For the
results in an increase in UA of 41, 58, 87 and 92% at flow rates 12 g s1 refrigerant flow rate, a dip in refrigerant thermal
of 0.93, 2.38, 5.68 and 9.5 L min1 respectively. It can be seen resistance is observed when the refrigerant temperature is
that at water flow rates greater than 0.93 L min1, doubling the near the pseudo-critical temperature of 40  C (segments
refrigerant mass flow rate results in large percentage 15e20). This is attributed to the spike in refrigerant heat
increases in UA, whereas increasing the water flow rate has transfer coefficient, resulting from the rapidly changing
diminishing returns. This indicates that the refrigerant-side thermodynamic and transport properties. For the 24 g s1 flow
presents the limiting resistance in the heat transfer process. rate, the refrigerant temperature remains above the pseudo-
Fig. 9 tracks the refrigerant-to-water thermal resistance critical temperature until near the outlet of the gas cooler, and
ratio and refrigerant temperature through the gas cooler with the dip in thermal resistance ratio is not observed until this
a water flow rate of 0.93 L min1 and 5  C inlet temperature, region (segments 25e30). These changes in resistances
and refrigerant flow rates of 12 g s1 and 24 g s1 at an inlet correspond to the increase in local heat duty observed in Fig. 6.
temperature of 100  C and pressure of 9000 kPa. Refrigerant The implications of this observation are as follows: (1) if
refrigerant temperature is not reduced below the pseudo-
critical temperature, either due to an undersized heat
exchanger or a high water inlet temperature, the enhanced
heat transfer in the pseudo-critical region will not be realized;
and (2) increasing refrigerant pressure generally raises the
pseudo-critical temperature, which dampens the spike in
refrigerant heat transfer coefficient as seen in Fig. 7, but
expands the temperature range in which heat transfer is
enhanced. Thus, CO2 gas cooler performance will be sensitive
to water inlet temperature, water flow rate and refrigerant
inlet pressure, as shown in Figs. 10 and 11. Heat duty (Fig. 10)
and the approach temperature difference (Fig. 11) are plotted
as a function of water inlet temperature for a refrigerant inlet
temperature of 100  C, refrigerant mass flow rate of 16 g s1,
Fig. 8 e UA vs. water flow rate. refrigerant pressures of 9000, 10500 and 12,000 kPa, and water
 i n t e r n a t i o n a l j o u r n a l o f r e f r i g e r a t i o n 3 4 ( 2 0 1 1 ) 1 7 e2 8 25

Fig. 10 e Heat duty vs. water inlet temperature.

Fig. 12 e Carbon dioxide Peh diagram.

flow rates of 0.93 L min1 and 5.68 L min1. Approach
temperature difference is defined in Eq. (11).
difference curve for the 0.93 L min1 water flow case at
DTapproach ¼ Tref;out  Twater;in (11)
9000 kPa.
With all other parameters fixed, the heat duty is seen to Increasing the water flow rate to 5.68 L min1 results in
increase with increasing refrigerant inlet pressure for the some different trends. Fig. 11 shows that for the higher water
0.93 L min1 flow rate case at each water inlet temperature. flow rate, approach temperature difference is low (<5 K) for all
As was shown in Fig. 7, for these cases, as the refrigerant water inlet temperatures. In this case, with the refrigerant
pressure increases, the temperature band corresponding to inlet temperature and mass flow rates fixed, the heat duty in
the improved heat transfer coefficient widens and shifts to Fig. 10 decreases with increasing pressure. As a comparison,
a higher temperature. Thus, the improved local refrigerant- in a conventional refrigerant system, increasing the high-side
side heat transfer coefficient corresponds to a locally higher pressure will raise the saturation temperature while
driving temperature difference for a longer portion of the decreasing the enthalpy of vaporization. This allows higher
CO2 temperature glide. Examination of Fig. 11 for the water outlet temperature, but also reduces the maximum
0.93 L min1 case indicates that the seven-plate gas cooler potential heat duty, with all other parameters fixed. For
is undersized for nearly all inlet refrigerant pressures and supercritical CO2, the temperature and pressure can vary
water inlet temperatures. An ideally sized gas cooler would independently. Examination of a CO2 Peh diagram (Fig. 12)
have an approach temperature difference approaching zero. shows that above the critical point, with the same refrigerant
For the 0.93 L min1 water flow rate, the approach inlet and outlet temperatures, increasing the pressure will
temperature differences are as high as 27 K at a water inlet decrease the available enthalpy difference due to the shape of
of 5  C. As the water inlet temperature increases, the heat the isotherms. Thus, for a gas cooler designed in a manner
duty and the approach temperature both decrease. Since that Tref,out / Twater,in, increasing the pressure while main-
the refrigerant outlet can only be driven down to a temper- taining the same refrigerant inlet temperature will decrease
ature equal to the water inlet, the enthalpy difference heat duty, since the limiting factor is the available enthalpy
between the refrigerant inlet and outlet decreases. This difference. For a gas cooler with a large approach temperature
decrease is not a linear relationship, and varies dramatically difference, increasing the pressure at the same refrigerant
near the pseudo-critical temperature and with pressure. As inlet temperature will result in an increase in heat duty, as the
the water outlet temperature rises above this, the available extended temperature range of elevated specific heat is
refrigerant enthalpy difference decreases. The effect of this utilized, and the heat duty is not limited by the total available
can be seen in the shape of the approach temperature enthalpy difference.

Fig. 11 e Approach temperature vs. water inlet

temperature. Fig. 13 e CO2 Enthalpy difference vs. outlet temperature.
26 i n t e r n a t i o n a l j o u r n a l o f r e f r i g e r a t i o n 3 4 ( 2 0 1 1 ) 1 7 e2 8

of 22% for a water flow of 0.93 L min1 and 5.8% for

5.68 L min1. At a water temperature near the pseudo-critical
regime (45  C), an increase in pressure of 33% yields an
increase in heat duty of 104% for the 0.93 L min1 water and
117% for the 5.68 L min1 water flow cases. In general, as
discharge pressure and compression ratio increase,
compressor volumetric efficiency and isentropic efficiency
tend to decrease. Thus, in reality, designing a system with
higher gas cooler refrigerant pressure may not be as beneficial
as theoretically possible.

Fig. 14 e Heat duty vs. water inlet temperature (isentropic 4. Conclusions

compression).
The results presented here show that accurate modeling of gas
cooler performance is critical to designing an efficient tran-
The inflection points observed in Fig. 10 for the 5.68 L min1 scritical CO2-water heating system. In a water-coupled gas
water flow rate at 9000 kPa can be attributed to the same cooler, the predicted performance is much more sensitive to the
inherent CO2 properties that caused the inflection points in refrigerant heat transfer coefficient, compared to air-coupled
the approach temperature difference curve discussed earlier. systems dominated by the air-side resistance. Additionally,
Fig. 13 shows a plot of CO2 enthalpy difference assuming an performance is sensitive to water inlet temperature. Increasing
inlet temperature of 100  C, and an outlet from 5 to 65  C at the water inlet temperature not only decreases the driving
three different refrigerant pressures. This is a good approxi- temperature differential, but as the temperature crosses the
mation of the predicted gas cooler behavior observed in Fig. 10 pseudo-critical temperature at a given pressure, causes
for the 5.68 L min1 water flow rate. In Fig. 13, at a refrigerant a sudden change in available enthalpy difference. This effect is
pressure of 9000 kPa, an inflection in the available enthalpy more pronounced at pressures closer to the critical pressure.
difference can be observed in the pseudo-critical regime. Many CO2 vapor compression system models in the litera-
These inflections dampen out and shift to higher tempera- ture show that for a given compressor and gas cooler size, there
tures as pressure is increased. This illustrates the sensitivity is an optimum high-side pressure at which COP is maximized.
of gas cooler design to water inlet temperature, particularly at As shown in Figs. 10 and 11, this is true when the gas cooler is
lower refrigerant pressures and at temperatures around the not designed in a manner in which approach temperature
critical temperature. difference / 0. The analysis presented here shows that by
The information presented in Figs. 10e13 shows that for sizing a CO2 gas cooler in a segmented fashion, the approach
a given gas cooler with a fixed refrigerant inlet and mass flow temperature difference can be minimized, and the high-side
rate, as approach temperature difference tends towards zero, system pressure can be reduced. An optimized compressor
increasing refrigerant pressure results in decreased heat duty and system can then be designed and built around the gas
and system COP. In reality, the refrigerant inlet temperature is cooler. The advantages of this are maximized COP for given
a function of many variables including evaporator tempera- operating conditions, a reduced pressure ratio, increased
ture, compressor efficiency, charge level and superheat. While compressor efficiency and improved compressor reliability.
supercritical CO2 temperature and pressure are independent Commercial CO2-water heating systems will need to rely
variables, in an actual system, increasing high-side pressure on the high heat transfer coefficients, large heat transfer area-
with a fixed compressor and evaporation temperature will to-volume ratio, and ability to withstand high pressures that
generally result in an increased high-side temperature. This are characteristic of microchannels. The heat exchanger
increases the driving temperature difference, and mitigates geometry modeled in this study offers better heat transfer
the decreasing enthalpy difference effects with fixed refrig- performance with a smaller footprint than a tube-in-tube
erant and water inlet temperatures observed in Figs. 10 and geometry.
11. Assuming an evaporation temperature of 0  C, isentropic While the present study has resulted in an accurate model
compression and high-side pressures of 9000, 10,500 and for predicting the heat transfer performance of a water-
12,000 kPa, the corresponding refrigerant inlet temperatures coupled microchannel gas cooler, there are several areas in
are 72.5, 85.5, and 96.8  C. Fig. 14 shows the model results which additional work is needed:
using these refrigerant inlet temperatures and corresponding
pressures, a fixed refrigerant mass flow rate of 16 g s1, water  The model developed here should be incorporated into
flow rate of 0.93 L min1 and 5.68 L min1 and water inlet a water-coupled CO2 vapor compression system model to
temperature from 5 to 65  C. predict the effects of changing water inlet temperature,
Assuming isentropic compression from a fixed evaporation high-side pressure, and evaporation temperature and
temperature, with identical mass flow, increasing the pres- compressor characteristics.
sure (and the corresponding inlet temperature) will result in  The effect of compressor lubricant on supercritical carbon
an increase in heat duty, as seen in Fig. 14. With a refrigerant dioxide heat transfer coefficient during cooling in micro-
flow rate of 16 g s1 and a water inlet temperature of 5  C, channels is not well understood. Investigations into this
increasing the pressure 33% results in an increased heat duty area may yield an even more accurate heat transfer model.
 i n t e r n a t i o n a l j o u r n a l o f r e f r i g e r a t i o n 3 4 ( 2 0 1 1 ) 1 7 e2 8 27

 Measurement of lubricant circulation rates in the test loop Hwang, Y., Jin, D.-H., Radermacher, R., Hutchins, J.W., 2005.
would provide a more accurate estimate of lubricant flow performance measurement of CO2 heat exchangers. ASHRAE
rates in the gas cooler, which would assist in more accurate Transactions 111 (2), 306e316.
Kim, M.-H., Pettersen, J., Bullard, C.W., 2004. fundamental process
computation of heat transfer coefficients and pressure
and system design issues in CO2 vapor compression systems.
drops. Progress in Energy and Combustion Science 30 (2), 119e174.
Kim, S.G., Kim, Y.J., Lee, G., Kim, M.S., 2005. The performance of
a transcritical CO2 cycle with an internal heat exchanger for
hot water heating. International Journal of Refrigeration 28 (7),
Acknowledgments
1064e1072.
Klein, S.A., 2008. Engineering Equation Solver, F-Chart Software.
The authors would like to acknowledge assistance from Mr. Koyama, S., Xue, J., Kuwahara, K., Takata, N., Yanagisawa, T.,
Christopher Goodman in design, fabrication, testing, analysis 2009. Simulation of a CO2 transcritical cycle for air
and modeling of the subject CO2 components. Financial conditioning. 3rd IIR Conference on Thermophysical
support from the US Army Ft. BelVoir office through Properties and Transfer Processes of Refrigerants, Boulder, CO.
Krasnoshchekov, E.A., Kuraeva, I.V., Protopopov, V.S., 1970. Local
a subcontract from Modine Manufacturing Company is also
heat transfer of carbon dioxide at supercritical pressure under
gratefully acknowledged. The authors also thank Modine
cooling conditions. Teplofizika Vysokikh Temperatur 7 (5),
Manufacturing Company for supplying several of the test heat 922e930.
exchangers and associated components. Assistance from, and Kuraeva, I.V., Protopopov, V.S., 1974. Mean friction coefficents for
insightful discussions with, Mr. John Manzione, Dr. Stephen B. turbulent flow of a liquid at supercritical pressure in
Memory, Mr. David Garski, Mr. Sam Collier, and Mr. Mark horizontal circular tubes. Teplofizika Vysokikh Temperatur 12
Hoehne for conducting this research are also acknowledged. (1), 218e220.
Laipradit, P., Tiansuwan, J., Kiatsiriroat, T., Aye, L., 2008.
Theoretical performance analysis of heat pump water heaters
using carbon dioxide as refrigerant. International Journal of
references Energy Research 32 (4), 356e366.
Lockhart, R.W., Martinelli, R.C., 1949. Proposed correlation of data
for isothermal two-phase, two-component flow in pipes.
ASHRAE Standard, 2006. BRS/ASHRAE Standard 41.4-1996 (Ra Chemical Engineering Progress 45, 39e48.
2006) Standard Method for Measurment of Proportion of Manglik, R.M., Bergles, A.E., 1995. Heat transfer and pressure drop
Lubricant in Liquid Refrigerant. ASHRAE, Atlanta, GA. correlations for the rectangular offset strip fin compact heat
Baskov, V.L., Kuraeva, I.V., Protopopov, V.S., 1977. Heat transfer exchanger. Experimental Thermal and Fluid Science 10 (2),
with the turbulent flow of a liquid at supercritical pressure in 171e180.
tubes under cooling conditions. Teplofizika Vysokikh Ortiz, T.M., Li, D., Groll, E.A., 2003. Evaluation of the Performance
Temperatur 15 (1), 96e102. Potential of CO2 as a Refrigerant in Air-to-air Air Conditioners
Cecchinato, L., Corradi, M., Fornasieri, E., Zamboni, L., 2005. and Heat Pumps: System Modeling and Analysis ARTI no.
Carbon dioxide as refrigerant for tap water heat pumps: 21CR/610e10030.
a comparison with the traditional solution. International Petrov, V.P., Popov, V.N., 1985. Heat transfer and resistance of
Journal of Refrigeration 28 (8), 1250e1258. carbon dioxide being cooled in the supercritical region.
Chang, Y.-S., Kim, M.S., 2007. Modeling and Performance Thermal Engineering 32 (3), 131e134.
Simulation of a Gas Cooler for a CO2 Heat Pump System, vol. Pettersen, J., Hafner, A., Skaugen, G., Rekstad, H., 1998. Development
13. HVAC&R Research, American Society of Heating, of compact heat exchangers for CO2 air-conditioning systems.
Refrigerating & Air-Conditioning Engineers, Inc., pp. 445e456. International Journal of Refrigeration 21 (3), 180e193.
Cheng, L., Ribatski, G., Thome, J.R., 2008. Analysis of supercritical Pettersen, J., Rieberer, R., Munkejord, S.T., 2000. Heat Transfer and
co2 cooling in macro- and micro-channels. International Pressure Drop for Flow of Supercritical CO2 in Microchannel
Journal of Refrigeration 31 (8), 1301e1316. Tubes. SINTEF, Trondheim, Norway, TR, A5127 pp.
Churchill, S., 1977. Frictionefactor equation spans all fluid flow Petukhov, B.S., Kirillov, V.V., 1958. Heat exchange for turbulent
regimes. Chemical Engineering 7, 91e92. flow of liquid in tubes. Teploenergetika 5 (4), 63e68.
Dang, C., Hihara, E., 2004. In-tube cooling heat transfer of Pitla, S.S., Groll, E.A., Ramadhyani, S., 2002. New correlation to
supercritical carbon dioxide. Part 1. Experimental predict the heat transfer coefficient during in-tube cooling of
measurement. International Journal of Refrigeration 27 (7), turbulent supercritical CO2. International Journal of
736e747. Refrigeration 25 (7), 887e895.
Filonenko, G.K., 1954. Hydraulic resistance of the pipelines. Pitla, S.S., Robinson, D.M., Groll, E.A., Ramadhyani, S., 1998. Heat
Thermal Engineering (4), 40e44. transfer from supercritical carbon dioxide in tube flow:
Garimella, S., 2002. Microchannel gas coolers for carbon dioxide a critical review. HVAC&R Research 4 (3), 281e301.
air-conditioning systems. ASHRAE Transactions 108 (1), Sarkar, J., Bhattacharyya, S., Gopal, M.R., 2006. Simulation of
492e499. a transcritical CO2 heat pump cycle for simultaneous cooling
Garimella, S., 2003. Conduction Effects in Cross-counterflow and heating applications. International Journal of
Supercritical Gas Coolers for Natural Refrigerants. Refrigeration 29 (5), 735e743.
International Congress of Refrigeration, Washington, DC. Sarkar, J., Bhattacharyya, S., Ramgopal, M., 2009. A Transcritical
Gnielinski, 1976. New Equations for heat and mass transfer in CO2 heat pump for simultaneous water cooling and heating:
turbulent pipe and channel flow. International Chemical test results and model validation. International Journal of
Engineering 16, 10. Energy Research 33 (1), 100e109.
Han, D.H., Lee, K.-J., 2005. Single-phase heat transfer and flow Span, R., Wagner, W., 1996. A new equation of state for carbon
characteristics of micro-fin tubes. Applied Thermal dioxide covering the fluid region from the triple-point
Engineering 25 (11e12), 1657e1669. temperature to 1100 K at pressures up to 800 Mpa. Journal of
Hoehne, M., 2007. Modine Manufacturing, Co. Physical Chemistry 25 (6).
28 i n t e r n a t i o n a l j o u r n a l o f r e f r i g e r a t i o n 3 4 ( 2 0 1 1 ) 1 7 e2 8

Wang, J., Hihara, E., 2002. Study on carbon dioxide gas cooler heat Yin, J.M., Bullard, C.W., Hrnjak, P.S., 2001. R-744 gas cooler model
transfer process under supercritical pressures. International development and validation. International Journal of
Journal of Energy Research 26 (14), 1237e1251. Refrigeration 24 (7), 692e701.
White, F.M., 2003. Fluid Mechanics, fifth ed.. McGraw-Hill Higher Zhao, Y., Ohadi, M.M., Radermacher, R., 2001. Micochannel Heat
Education, New York. Exchangers with Carbon Dioxide ARTI no. 21CR/604-10020-01.